Richard Andersen, Neuroscientist and Leading Researcher in Brain-Machine Interfaces

What if you could control movement only by using your thoughts? Once the sole domain of science fiction, Richard Andersen's breakthrough research is now enabling individuals who have suffered catastrophic spinal cord injuries to complete tasks such as using a computer or drinking a cup of coffee. The principle is as simple as the execution is complex: drawing on his long-term research that has established the role of intent in the region of the brain that dictates movement, and taking advantage of advances in robotics and implantation electronics that can externalize the brain's commands to artificial limbs, Andersen is leading the way in brain-machine interfaces to restore an element of independence and control to people paralyzed from the neck down. It is such an incredible, futuristic achievement, that one needs to see it to believe it.

In the discussions below, Andersen recounts his family's New York roots and his childhood in Louisiana and California. He describes the political tumult during his undergraduate years at the University of California Davis, and the value conferred by his focus on engineering during college. It was at the University of California San Francisco where Andersen became enamored of the burgeoning field of neuroscience, so new that the term was just beginning to be adopted. For his postdoctoral appointment Andersen worked at the Johns Hopkins University, then and now one of the world's largest and most distinguished centers of neuroscience research. He explains the early adoption of computers by neuroscientists and the numerous ways computation expanded possibilities both in theoretical neurobiology and in animal experiments.

As a junior faculty member at the Salk Institute in La Jolla, Andersen benefited from the focused research environment and unique research culture, and he recounts his interactions with Francis Crick, one of the founders of molecular biology who had shifted his interest to neuroscience in the 1960s and 1970s. Despite his enjoyment researching at Salk and living on the San Diego coast, the exciting developments at MIT proved too great of a pull to resist. In his role as a senior faculty member in the Department of Brain and Cognitive Sciences, Andersen deepened his facilities and collaborations in neurocomputation, and he began to focus in earnest on microelectrode technologies and implantation techniques for performing brain surgery on monkeys.

Andersen then received an offer to join the faculty at Caltech, whose leadership in neuroscience he had long admired. In addition to continuing on in the research he had pioneered at MIT (ensured only by the successful relocation of his large laboratory), Andersen began to work on neuroprosthetics, which, as the name suggests, examines the most feasible ways of connecting external artificial limbs to brain directives. As he explains, Andersen was always interested in human neuroscience, and one day, in helping people, but it was only at Caltech where the technology, experimental techniques, and administrative acumen converged to make this leap a reality. Andersen narrates the dramatic moment when the first patient successfully controlled a robotic limb through thought control, and the enormous media attention this breakthrough garnered. He also explains how these developments caught the attention of the philanthropists Chrissy Luo and Tianqiao Chen, who were so taken by Andersen's work - and the fearless research culture that generally defines Caltech - that initial conversations ultimately developed into the creation of the Tianqiao and Chrissy Chen Institute for Neuroscience at Caltech, thereby solidifying the Institute's world-leading role in this field.

The conversations end on a reflective note. By demonstrating the capacity to extend thoughts beyond the confines of the human body, a new range of questions must arise relating to ethical, legal, and philosophical implications. Andersen, a scientist's scientist, is laser focused on his research group and always on improving technological and experimental capabilities. But he emphasizes that wherever these capabilities lead us, there can always and should always be strict lines of demarcation between humans and machines. And specifically, we should not think of machines as a "Plan B" in preserving humanity if our self-destructive tendencies make "natural humanity" no longer possible. In other words, technology is here to help us, not to replace us.

DAVID ZIERLER: This is David Zierler, director of the Caltech Heritage Project. It is Friday, September 15th, 2023. I am delighted to be here with Professor Richard A. Andersen. Richard, it is so nice to be with you. Thank you so much for joining me.

RICHARD A. ANDERSEN: Thank you for asking me to be interviewed.

ZIERLER: Wonderful. Richard, to start, please tell me your titles and affiliations here at Caltech.

ANDERSEN: I am the James G. Boswell Professor of Neuroscience, and I am in the Division of Biology and Biological Engineering. I have the Leadership Chair in Brain-Machine Interface with the T&C Chen Neuroscience Institute. And I am the Director of the T&C Chen Brain-Machine Interface Center here. I am also Director of Swartz Theoretical Neurobiology at Caltech. That's a postdoc program. How much do you want me to go into?

ZIERLER: Are those all of your titles? You can keep going with the titles. We'll go into each of them one by one.

ANDERSEN: No, that's enough. [laughs]

ZIERLER: [laughs] Let's start with the named chair in the honor of Boswell. Who is or was Boswell, and is there any connection to your work?

ANDERSEN: He was the biggest farmer in California. I believe his uncle started farming here in the Central Valley, and they have several hundred thousand acres. J.G. Boswell was on the board of directors at the Huntington Memorial Research Institute. His company is based here in Pasadena, although the farm is up in Corcoran. He was also a trustee of Caltech. He endowed a chair at Caltech. When I came here, Seymour Benzer had that chair. That was at a time when faculty had to retire at 70, so he made a special chair for Seymour, resulting in two Boswell chairs at the time. I met Jim Boswell here at Caltech. His son actually lives locally, and so I see him and his family from time to time. Their kids went to the same school as our kids. The other Boswell chairs are at other Universities—I met them once at a meeting at the Boswell ranch in Corcoran—they are all agriculture-related, so this was the only neuroscience-related chair. I think that's partly because Seymour was a very famous geneticist, and at the Boswell farm, they're very science-oriented, so they specialize in genetics of crops and other advanced technologies. I was lucky enough to get the chair.

ZIERLER: Seymour Benzer, of course, was very interested in behavior—

ANDERSEN: Yes.

ZIERLER: —the obvious connections to your work. Does that suggest that Boswell himself had an interest in behavior and the biological origins of it?

ANDERSEN: His son, Jim Boswell, and his grandsons visit from time to time. One of Jim's sons, and one of his executives in the company's sons, worked in the lab as summer interns. Another of his sons, Cameron, is very interested in AI and brain implants, so that was kind of an added plus that came later. They have been very interested and supportive. It's not their main thing—which is more about advancing agriculture—but I feel very fortunate they are interested in our work and continue to be.

ZIERLER: The Brain-Machine Interface Center, does that precede the Chen Institute, or is that part of the creation of the Chen Institute?

ANDERSEN: It was part of the creation. Initially, we had a Science paper in 2015, and it got a lot of publicity.

ZIERLER: What was the paper?

ANDERSEN: It was the first implant outside of motor cortex. It was in the posterior parietal cortex of a human with Tyson Aflalo, Spencer Kellis and other colleagues. The reason for implanting posterior parietal cortex is that it's an association part of the cortex, so there's many more and varied signals—in other words it is more cognitive than motor cortex—being involve in high level aspects of action planning. Tianqiao Chen, and his wife Chrissy Luo saw me on TV in Singapore, so they sent me an email and then it was arranged for Tianqiao to come visit. He was interested in philanthropy and was also looking at Harvard, and he wanted a proposal. Both Tianqiao and Chrissy came out about a month later for a second visit. I wrote up a proposal for a brain-machine interface center, because he was very interested in that technology. At this point, I was intending to tell my chair, because this was going to be a big proposal.

ZIERLER: This was Steve Mayo at the time?

ANDERSEN: That's right. But he had already heard of it through a Caltech trustee who was also a trustee at Harvard. The trustee heard from Harvard people that they were thinking of potentially supporting me. At that point, our administration took it over [laughs] and—

ZIERLER: The Chens' first contact at Caltech was through your research?

ANDERSEN: That's right.

ZIERLER: Before they knew David Anderson, Tom Rosenbaum, Ed Stolper, Steve Mayo, any of them, it was your research that piqued their interest?

ANDERSEN: Right. Of course, in the end, the gift wouldn't have been successful without the administration [laughs]. At that point, part of the gift went to the building, and part of it to the executive, David Anderson, and then the BMI Center, and there's a humanities center. There's also a molecular biology center, an education center, and a new center for AI that has come from a more recent donation. There was also another endowed center, which was for Systems Neuroscience, that Doris Tsao directed, but she left, so Betty Hong is now the current acting director...

ZIERLER: Administratively, how big is BMI, the Brain-Machine Interface enterprise, as a proportion of the overall size of the Chen Institute?

ANDERSEN: I haven't calculated [laughs] how much of it.

ZIERLER: Generally, is it a small part of it, or does it really take up a significant amount?

ANDERSEN: It's significant, but not a major share of it.

ZIERLER: Did you give any consideration to moving into the new building, or you're quite comfortable here?

ANDERSEN: After Doris left, I thought of going over to the space she was to occupy in the Chen Building. I have a lab here in Beckman Behavioral Biology and one over in Broad, and a lot of space, and it took a lot of time to set things up. Also, it is kind of convenient—it has a lot of little offices; I thought for the students, they like that, they can hide away and concentrate. I guess that's different from the popular design now, which is to have everyone in rows of desks so they can interact the most. Maybe my preference comes from my background in neurophysiology, which tends to have enclosed rooms for doing experiments. Now, for us, it's becoming more of a human lab. I thought about moving since Doris had set things up for monkey work, and we continue to do monkey work, but it was too fine-tuned for her particular interests. It would have cost a lot to renovate, so in the end I stayed here.

Clinical and Theoretical Neurobiology

ZIERLER: The title Leadership Chair, is that an honorific, or are there administrative responsibilities that come with that?

ANDERSEN: It is to support researchers to do brain-machine interface studies, so we give out seed grants. We also collaborate with a number of labs here. Doing this kind of research is very complex, and so we're also involved with hospitals and rehab centers, and we collaborate with physicians for the surgical procedures and health care, and with engineers for machine learning. It's a big project. As a result of our collaborations, we can make neural data available for colleagues at Caltech. Originally the brain-machine interface with humans—the type we do, which is array recording of populations of neurons, came out of monkey labs, because it was very much like what you do with monkeys. More or less, in the late 1990s, early 2000s, independently several groups started doing it. Right now, because it's complicated, there's a big group called BrainGate that includes Harvard, Brown, and Stanford. The Stanford group was started by Krishna Shenoy who was a former postdoctoral fellow with me. There's another group that includes Pittsburgh, the University of Chicago and Johns Hopkins. Then there's our little group [laughs]. We include UCLA, USC, a couple big rehab centers, and we're starting up a site as well in Denver with the University of Colorado. A couple former students are doing that. Because it's so complicated—the FDA, IRB oversight; you're working with very delicate subjects, so they need a lot of monitoring of their healthcare—it's not like conventional research where you just go out and start a lab, get a few students and a colony of mice, and go for it. It's a much more complicated enterprise.

ZIERLER: Because there's an obvious clinical aspect to it?

ANDERSEN: Yeah, and also it merges a lot of different disciplines. You have neurophysiology, , machine learning, theoretical neurobiology. It's much more than any one person could possibly be trained to do. Like obviously, since I am not a neurosurgeon, I can't [laughs] do the surgeries. It's a highly collaborative, large effort.

ZIERLER: Tell me about the Swartz Theoretical Neurobiology program.

ANDERSEN: That started when I first got here. We put in a grant application for theoretical neurobiology with the Sloan Foundation. At that time, the Sloan grants were pretty big. There were five universities that were awarded Centers for Theoretical Neurobiology in 1994. The centers were training programs and out of these centers came a lot of the current generation of computational neuroscientists, and so it was highly, highly successful. Sloan then, as planned, pulled back after getting the field rolling. Next Jerry Swartz funded the effort. It went from five big centers to 11 smaller Swartz Centers.

ZIERLER: Who is Jerry Swartz?

ANDERSEN: He invented the barcode. He is an engineer by background, and he was interested in the brain. Hersh Cohen was the vice president at Sloan and originated the Sloan Theoretical Neurobiology program; he knew Jerry and the program continued as the Swartz Centers for Theoretical Neurobiology.

ZIERLER: It's mostly to support postdocs?

ANDERSEN: That's right. It is small compared to what it used to be [laughs], but it has been going a very long time, almost 30 years, so it's amazing it's still here.

ZIERLER: Let's look at some of your collaborations across campus. What are some of the groups that you have done work with?

ANDERSEN: For a long time, we worked with Joel Burdick. He is a mechanical engineering professor in Caltech's Division of Engineering and Applied Science (EAS). . Initially I thought we'd probably work on robotics, but he got interested in the fact that the students were always advancing electrodes manually, and it would take a few hours while they recorded neural activity. He thought a robot could do that. So he designed, with his students, an automated version of an electrode advancer, which is great. We tried to get some microelectrode companies to pick it up; no one did. [laughs] Such is life in science. Then he worked extensively with us on neural decoding. At the time, we were doing the animal studies for BMI with non-human primates, because of course, animal studies were required first before going to humans. Yu-Chong Tai, an electrical engineering professor in EAS, worked with us on making a new generation of microelectrodes. Since we've had the Chen BMI Center, we've worked with Mikhail Shapiro in the Division of Chemistry and Chemical Engineering developing an ultrasound BMI. With Azita Emami from EAS we have worked on a new feature extraction component of the decoder, and that has enabled us to extend our recordings in one participant, for a couple extra years. We have projects with Anima Anandkumar's group on machine learning. We're starting to work with Richard Murray, because we've been doing brain control of driving, and semi-autonomous driving is his field. I'm sure there are many more. [laughs]

ZIERLER: The theme, it sounds like, is that a lot of your collaborations are outside of BBE, that you work with engineers to a large degree.

ANDERSEN: That's true, it is a big component of collaboration. There's also clinicians, computer scientists, and the neuroscientists. We collaborated with neuroscientist Ueli Rutishauser, who is a visiting associate in BBE, with his main faculty position at Cedars.

ZIERLER: What about the interface of social science and neuroscience, all of the things that are happening in HHS, for example, and their work in neuroscience? Are you involved in that at all?

ANDERSEN: Yes, we had a publication with Colin Camerer, and I was one of the PIs for the Conte Center for Social Decision Making, which was a 10-year grant from NIH. Ralph Adolphs was the PI for that. Its theme was social neuroscience. Our major component was studies of the observation of other actions through electrophysiological recording in humans. Since we're in the high-level part of the brain, it not only makes movements and plans, but it also interprets others' movements, and thus is very social. In the course of our work with the Conte Center we interacted extensively with Ralph Adolphs, John O'Doherty and Antonio Rangel. Wolfram Schultz, who is from Cambridge, visited the Conte Center often. He is very famous for discovering that reward signals in the brain arise from the dopamine neurons. I got to know the neuroeconomists very well. [laughs]

The Arrival of Neuroscience

ZIERLER: Let's move on to some terminology now. Your home discipline, by your training, by what you do, is neuroscience the most inclusive term for everything that you work on?

ANDERSEN: I've been around a long time [laughs], and when I first got my PhD, it was in physiology. Some from UCSF claim that I was their first neuroscience student, but I wasn't, actually. I looked at my degree, and it's physiology. [laughs]

ZIERLER: Was the term "neuroscience" in use when you were in grad school? Do you remember it?

ANDERSEN: Not really, no. After I left UCSF began a program in neuroscience. Harvard for a long time had a program in neurobiology, which is similar. When I was in college, biopsychology was closest to neuroscience.

ZIERLER: Neurobiology, is that a subset of neuroscience? Is it the biological component of neuroscience, or is it something else?

ANDERSEN: It's pretty much the same thing. Its similar to dual names that appear in our field, people talk about brain-computer interfaces or brain-machine interfaces. They are pretty much equivalent.

ZIERLER: Since your career develops in tandem with the rise of neuroscience as a discrete field, what were some of the advances—technologically, theoretically, experimentally—that allowed for this new field called neuroscience? What was happening around that time?

ANDERSEN: Advancements were being made rapidly at the systems, cellular, and molecular levels, and they converged into the broader field of neuroscience. My experience was in systems neurophysiology. In college, I got to work with a vision neurophysiologist, Robert Scobey. He had come from Johns Hopkins. Hopkins and NIH were two of the big centers for systems neurophysiology He had been a student of one of the famous professors there, Gian Poggio. Another student that came from the Hopkins group was Mike Merzenich. He was an assistant professor at UCSF at the time, so I went to graduate school there. He was my thesis advisor and I worked on the auditory system. Merzenich was one of the co-discoverers of the cochlear prosthetic. I didn't think it would work at the time. [laughs] Of course, who knew that such crude signals could be learned by the brain, because it stimulates the auditory nerve in just a few places, but the brain is able to learn to decipher this simple signal. At the time, anatomy got a boost technically by having tracer methods and I combined them with neural recording for my thesis.

ZIERLER: What are tracer methods?

ANDERSEN: You could inject a compound that would either go anterogradely, along an axon, and so you could tell where an area projected to, or retrogradely, so you could tell what was sending signals to an area. A fellow graduate student had the idea of mixing the anterograde and retrograde tracers together, and we found every connection in the cortexes was reciprocal, as well as with the thalamus, and the reciprocity was very precise.

ZIERLER: What is the message there? The reciprocity, what does that tell you?

ANDERSEN: It has been a fascination and research endeavor of many labs. It's usually thought of as feedback; that higher level areas make sense of what they're getting and fine-tune what's coming into them. It's thought to be a structure for predictive encoding of signals, so you can predict on past experience, at these higher-level areas, what the lower-level areas should be saying. It is thought to be a mechanism for attention. In neural networks reciprocity exists for learning and for short term memories., We worked with recurrent neural networks a lot when I was an assistant professor. They're very powerful, particularly for sequential actions.

The Rise of Computational Neuroscience

ZIERLER: What about computers in the late 1970s and early 1980s? Were there computational developments that were relevant for the growth of neuroscience?

ANDERSEN: When I went for a postdoc, I wanted to study more cognitive things, and there were two labs at the time that had just developed recording from awake behaving monkeys, one at NIH and one at Johns Hopkins. At Hopkins there was a renowned neurophysiologist, Vernon Mountcastle, and I was fortunate to do postdoctoral study with him. At the time, the computers, they were Programmed Data Processors (PDPs) made by the Digital Equipment Corporation. They had very little memory, and so you had to use machine-code to program them, and it was a chore. There were only a few students doing behaving monkey experiments at the time, and when we went out to start our own labs, we wondered if the field would survive, because these machines were necessary but so underpowered. Then IBM came out with office computers with much more memory, and we learned to use those instead. Now it's not an issue at all, of course. We use personal computers for recording, data analysis and real-time data analysis. The real-time analysis allows us to feedback to the subject their brain signals. As a result, they can control devices with their thoughts, because we have decoding algorithms in the pipeline that can interpret the recorded brain signals. We'd be lost without the computers. If computers hadn't advanced when they did, the field probably still would have emerged, but it would have taken awhile.

ZIERLER: We've covered the term theoretical neurobiology. Of course, biology is generally understood to be an experimental discipline. Physics, for example, there's obvious delineations between experiment and theory. What does experiment and theory mean in neuroscience? What are theoretical projects to do in neuroscience, and how do you validate them experimentally? What does that look like?

ANDERSEN: There are two flavors of theoretical neuroscience, often referred to as computational neuroscience. One is signal processing and data analysis. That's kind of nuts-and-bolts stuff, and very useful for an experimentalist like myself. The second is more pure theory, having an idea and then testing it to see if it's valid. I found that super useful in setting a framework around the experiments that we do; an intellectual scaffolding. Like when I was at the Salk as an assistant professor, across the road at UCSD, were the early neural network founders, Geoff Hinton, David Rumelhart and others, so that was very fortunate. We were getting these signals out of the brain. We called them gain fields, because we thought we'd find receptive fields out in space. When you move your eyes, the world doesn't seem to shift, so we thought the receptive fields would not move with the eyes. But it turned out that in most areas, the receptive fields still move with the eyes, but they're modulated by where the eyes are looking. It was a weird thing, at the time, this kind of multiplicative convergence. Then we trained neural networks to do the transformation from retinal to spatial receptive fields and we found the hidden layer used these same gain fields. David Zipser and I had an article in Nature in 1988 that was the first study that I am aware of comparing a neural network model to neural data. Now, you see in almost every paper [laughs], they have a neural network model of the data. That was great, because I could then understand how a population of neurons solving coordinate transformations would use gain fields.

ZIERLER: In the way that theoretical physicists are looking to unify their theories, to have a grand explanation for the universe, for particles, for things like that, is there a similar push in neuroscience to unify theories, to simplify all of this work, for a grand overview understanding of the brain?

ANDERSEN: Yes, sure. This particular finding, gain fields, it turned out was used for all sorts of computations throughout the brain. In that sense, it was actually a discovery of how the brain computes, and it does have a wide range of applications. But I think what you're really asking—at the time, physicists, when I was at the Salk, were coming into the field, and they thought, "Oh, these dumb neuroscientists. [laughs] They're not sophisticated. We're going to handle it like a physics problem and, say, understand consciousness from first principles." But neurobiology is so complicated, so it hasn't panned out that way. There are certain rules, like from [Santiago Ramón y] Cajal, the idea of synapse, and signal direction and transmission of information. Francis Crick was at the Salk at the time, and he was hoping that since he had solved life, that he would now solve consciousness. So, he'd bring a lot of computational people from MIT and other places to visit. He had a Helmholtz Club that I was a member of. That was great.

ZIERLER: I've heard about the Helmholtz Club. What was it?

ANDERSEN: I am not sure it still exists. It was started by V.S. Ramachandran, who is a psychophysicist at UCSD, and Gordon Shaw who was a physicist at Irvine, but the real star that attracted everyone was Francis. There were about 20 neuroscientists from Caltech, UCLA, Irvine, Salk, and UCSD that once a month gathered at UC Irvine, since it was the closest—or furthest—for everyone [laughs]. We'd invite a couple people from around the world, and they would give about a two-hour talk each, and then there would be dinner. There were some really dramatic arguments and talks and very lively. When I gave talks there, I sometimes wouldn't get beyond my first slide [laughs].

ZIERLER: You'd get peppered with questions?

ANDERSEN: Yeah. It was really a nice thing. Then when I moved to MIT, I thought, oh, they would have clubs like this, too. But, in Boston, there are so many neuroscientists that there wasn't really a great motivation to have a club like this. There are groups there, but nothing like the Helmholtz Club. Terry Sejnowski, who is at the Salk—actually when I left, he was my replacement—he kept the Helmholtz Club going for many years. After a while, it kind of, at least for me, petered off. First, I was in Boston [laughs]. Then when I came back to southern California, I went to a couple meetings but it was a long drive and I got awfully busy.

From Animal to Human Research

ZIERLER: We've been talking about the brain, an abstract term. We haven't been talking about the body that encases it. I'm thinking about the great experimental animals in biology—drosophila, zebrafish, mouse, monkeys, humans. What brains do you work on? What organisms are in your lab to work on?

ANDERSEN: Originally, I wanted to work with humans, because I was interested in human cognition, intention—

ZIERLER: By originally, you mean like from graduate school?

ANDERSEN: From graduate school. I started working with a Ben Libet, who was collaborating with Bert Feinstein, Dianne Feinstein's husband. He was a neurosurgeon. He died, unfortunately, and so that project stopped. Libet went on to become famous for an experiment where he found that there's neural activity in the electroencephalogram that predicts what you're going to do before you're aware you're going to do it. (We've recently redone that experiment in the human with recordings from single neurons.) After graduate school I worked with non-human primates, which are the closest you can get to humans as an experimental animal.

ZIERLER: Is that to say that drosophila and mice are too lower-order to be relevant to the kinds of questions you're posing?

ANDERSEN: Yeah. For instance, we've recently discovered that we can decode internal speech, as well as vocalized speech.

ZIERLER: What's the difference between internal speech and speech?

ANDERSEN: That's when you're talking to yourself.

ZIERLER: The inner monologue.

ANDERSEN: Yes. I doubt that flies are doing that, too, and even if they did—

ZIERLER: They wouldn't have any way of telling us.

ANDERSEN: Yeah, and crickets chirp. I have always been interested in higher-order cognitive functions. The human, and non-human primate—with the human, you can train a task in a day, whereas with monkeys, it takes months. So you can do a wide variety of experiments in a short period of time. Then, like you say, you can ask them what they experienced or how they're doing the task. That's very nice. But obviously you can't do things that inhibit pathways, or in developing new recording tools like our functional ultrasound prosthetic, we have to try that in experimental animals first. So, it's nice having both, and they are synergetic. It depends on the problems you're looking at.

ZIERLER: What species of monkeys do you work with?

ANDERSEN: Rhesus, Macaca mulatta.

ZIERLER: What is unique about rhesus or why are they relevant for you?

ANDERSEN: They're an Old World monkey. They're really good workers. You encounter them if you go to India, where they're in the wild. They live in a very hierarchical structure. We hope that you'd become alpha with your monkey, but not always. [laughs] They have their own personalities. They work in the lab for several years, and then we retire them to animal sanctuaries. We pay to build enclosures at these sanctuaries, so the animals live out their lives in retirement.

We were one of the first labs to start doing that. Because you get kind of attached to them. They get to know you. There are other monkeys people use. Now researchers are beginning to use marmosets, because rhesus monkeys, it takes years to breed and grow, if you want to do transgene experiments. Marmosets breed much faster. They're kind of hyper, hard to train. On the other hand, they're highly social, so I think if one were to use them, you'd ask questions about social neurobiology. There are other non-human primates that people have used. Part of it I think is just so many studies have used rhesus monkeys that it helps to keep everything constant.

ZIERLER: Have some of the moral qualms about primate research affected your laboratory? Does it affect operations? Does it affect what you can do or what you even want to do?

ANDERSEN: No. There has always been this cloud of having animal rights extremists breaking into your lab or that kind of thing. We regulate the work very carefully so the animals are healthy and treated humanely. Also I have been on the IACUC for years, and have been the chair of the IACUC.

ZIERLER: What is that organization?

ANDERSEN: It's the Institutional Animal Care and Use Committee.

ZIERLER: Which ensures humane conditions, that kind of thing?

ANDERSEN: Yeah, for all vertebrate animals. Each university has one. We follow guidelines set by NIH. Also, Caltech is accredited with what's called AAALAC (Association for Assessment and Accreditation of Laboratory Animal Care), which is a worldwide organization of veterinarians that approve institutions every three years. So, it's highly, highly regulated with a focus on the health and care of the animals.,

ZIERLER: Is the sequence of your research that you always start with monkeys and then move on to humans, or are there some aspects of your research where you go right to humans, or it never progresses from monkeys to humans?

ANDERSEN: As we've used humans more and more, we've been using monkeys less and less, but still I'm glad to have both, because of the synergy that exists in certain research paths.

ZIERLER: What is the big trendline that explains how your research has gone more and more into human research?

ANDERSEN: Working in humans initially required showing that the technology would work using non-human primates. In non-human primates we've primarily studied posterior parietal cortex. In humans, besides posterior parietal cortex, we have been able to expand our clinical studies to motor cortex and somatosensory cortex. The human subjects are paralyzed but also lose somatosensation, so we can write in sensory signals through electrical stimulation. We can place sensors on the robotic hand that controls the electrical stimulation to improve dexterity. But also, it's nice to return the sense of the hand that is lost after spinal cord injury.

ZIERLER: Physiologically, how similar are monkey brains and human brains?

ANDERSEN: Parts are nearly identical, like the early parts of the visual system. But then if you look at the monkey brain, they have very small frontal lobes and very small posterior parietal cortex. In the human, it's just enormously expanded. Some things we find in humans that we found in monkeys, like persistent activity for planning, and some things, like mixed selectivity, seem more common and complicated in humans. This latter finding may be because we can do more types of experiments in humans due to the ease of training. Now we've gone back to the monkey and we're finding some of the same principles of mixed coding. So the comparison of the species has helped in the monkey research as well.

Mapping the Brain

ZIERLER: Let's take now a verbal tour of the brain. What are the regions of the brain that you're most interested in, and where are they located?

ANDERSEN: We are looking at sensorimotor transformation. That's how you take in sensory signals, process them to make a plan, and then you send out an execution signal. You start with sensory parts of the brain, you go through association areas like posterior parietal, and then you go over to premotor areas that elaborate some of the planning, and then to motor cortex where you send the signals out to move. We're studying this whole pathway. So far, we implanted the two main lobules in posterior parietal cortex. We've implanted premotor cortex, and somatosensory cortex for the writing in touch signals. Motor cortex we compare with posterior parietal. We've just received FDA approval to look at prefrontal cortex. In a monkey, prefrontal cortex has been proposed to function for executive control, regulating other parts of the cortex. Studying prefrontal cortex will be interesting, particularly for neuroprosthetic applications. The posterior parietal cortex is in the dorsal stream, a pathway important for visualmotor transformation. The ventral stream starts with sensory areas, particularly for vision, and then goes through a series of association areas to arrive at the inferior temporal cortex, where the shapes of objects are processed. Then between the two is speech cortex, although we're now finding some forms of speech in posterior parietal cortex, too. We think speech is more distributed and many cortical areas are involved in the transformation from semantics to vocalization.

ZIERLER: This is to say that the effort to map the brain in terms of functions is an ongoing enterprise?

ANDERSEN: Right.

ZIERLER: It's not yet completely resolved what aspect of the brain is responsible for what activity? This is still being studied?

ANDERSEN: Right.

ZIERLER: What are some of the big open questions in that regard?

ANDERSEN: It's much more distributed than I thought it would be. You have areas—we use these tiny little arrays, they're four by four millimeters, 100 electrodes, and we can decode almost anything out of them that has to do with planning in posterior parietal cortex from, observation of movement of others, the planning of movements of self, movement of all parts of the body. I didn't expect that. That's mixed coding, and now we're looking at how it might be structured. It may be structured as building blocks—in high-dimensional spaces, these building blocks can be combined in different ways to do different functions. That kind of theoretical framework, to try to explain this mixing, and how it accounts for so many functions, is one challenge. Speech—we have a paper in revision on the internal speech—others haven't found robust internal speech yet. They've been focusing on speech in the premotor cortex Broca's area in the frontal lobe, these areas produce the motor output for speech articulation. Current efforts include using large language models after the decoding stage to predict sentence structure. We are planning to use large language models as well, but using decoded language thoughts. Using language thoughts will be particularly helpful for participants who cannot speak from stroke or amyotrophic lateral sclerosis (ALS),

ZIERLER: Conveyed how? Through word processing?

ANDERSEN: The early parts of the pipeline are similar to a motor prosthetic, where features of the neural activity are extracted, for example action potentials, and then machine learning is used to train a model to classify words. Language models can also be incorporated in the decoding model. The output of the decoder can then control any of a variety of devices that produce written words or synthetic speech. Besides decoding words, we are now trying phonemes which would reduce the number of classifications required for a large vocabulary.

ZIERLER: The binary of fundamental research and translational research, does everything in your lab exist somewhere on that continuum? Is there anything that you do that is what we might say strictly curiosity driven? Is it all ultimately related to translation to help people? How does that look in your lab?

ANDERSEN: I think for BMI labs, we're kind of at the science end, but the eventual goal is the more you understand about the brain, the more advanced translation can be. Other labs are more at the engineering end, . I have a basic interest in how the brain works, but it's pretty amazing when—I mean, it's amazing it works at all, these BMIs.

ZIERLER: [laughs]

ANDERSEN: Whenever I go to a session, it's just—seeing someone that can drive a car—we have a Ford project where one of our subjects can, with his brain signals, drive a car in Michigan from here in southern California, over the internet. That kind of thing, you just—"How does this work?" In the beginning, we weren't even completely sure it would work. So that is an example of the exploratory component you refer to.

Clinical Trials and Regulatory Frameworks

ZIERLER: I wonder if you can talk about the regulatory framework, how you deal with human subjects. What are the limitations and how do you work within them?

ANDERSEN: We have an IDE, Institutional Device Exemption. The IDE is approved by the FDA and we are registered as a clinical trial. Then we have our protocol, how the study is performed, the informed consent document the participant signs, etc. reviewed and approved by Institutional Review Boards at the various institutions involved in the clinical trial. There are annual reports and renewal processes for both the FDA and IRBs. If there are any adverse events, you report them, especially unanticipated adverse events which are reported immediately.

ZIERLER: What would be an example? What could be an adverse event?

ANDERSEN: An adverse event could be an infection around the pedestal, and although rare that would not be unexpected. An unexpected event would be if the pedestal came off. That's the recording connector. That hasn't happened. In fact, in clinical terms, the implants have functioned very well, without infection, and remaining viable for many years. There's a lot of oversight with the non-human primate, but even more so with humans. Of course, that's a good thing. [laughs]

ZIERLER: Do you interface directly with patients, or are you always working through a physician? Do you work directly with people?

ANDERSEN: I and our lab members work directly with the participants.

ZIERLER: In what settings? Where do you go? Hospitals?

ANDERSEN: There are about three sessions a week that last three or four hours. We can do the sessions at the rehab centers, we can do the sessions in the participants' homes, and we can do sessions here at Caltech. In the beginning we just did sessions in rehab centers. One of my colleagues counted worldwide there have been about 30 participants implanted, so it's still pretty rare, but as it has become increasingly appreciated that it's a safe procedure. So now we can do the sessions in home.

ZIERLER: The subjects you're working with, do they all suffer from some malady, or do you work with healthy people, also?

ANDERSEN: [laughs] No, no. Couldn't do that.

Injury and Independence

ZIERLER: It's somebody who wants either to help themselves or at least they are committed to the idea that their condition will help the research and ultimately help others?

ANDERSEN: Right. We work with tetraplegics. They have a high spinal cord injury, so they're paralyzed in all limbs, and they cannot feel their limbs.

ZIERLER: This comes from a catastrophic accident, or are some of these congenital issues?

ANDERSEN: From catastrophic incidents. Other labs work with amyotrophic lateral sclerosis (ALS) subjects.

ZIERLER: What are the neurophysiological implications of an ALS diagnosis? Where would you come in for that?

ANDERSEN: The labs that work with ALS participants generally do their studies prior to the participants becoming completely paralyzed. I would like to emphasize that participants in BMI studies are very brave and heroic. We are very clear that the studies are not going to benefit them directly but it will for future patients. On the other hand, it seems to be a great experience for them.

ZIERLER: It provides meaning, to their troubles.

ANDERSEN: Right. Paralyzed people don't get out that much. People get depressed. They have a job. We pay them for their time. It's not exorbitant so it's not coercive, but they are doing work. They also interact with media, outreach programs, and other participants. One good thing about being in clinical trials, for any clinical trial, is you have greater access to physicians.

ZIERLER: Someone is paralyzed, and then there's something that is severed between the brain and the body?

ANDERSEN: Right.

ZIERLER: What exactly is being severed?

ANDERSEN: It's the spinal cord, the highway of the brain to control voluntary movements. It's sending down signals to the muscles. It's also the path for somatic sensation—that's proprioceptive end touch signals—going back up—so you end up with someone who is completely paralyzed for voluntary action and cannot feel there body below the lesion. Some patients need to be on a respirator, although none of our participants so far has required one. . They're very delicate subjects. Actually, you'd have to talk to a physician to know all the issues they have, but imagine being paralyzed from the neck down and not being able to feel anything below the lesion.

ZIERLER: From this research, what are some potential therapies or clinical benefits that could come as a result of the research? What would that look like?

ANDERSEN: Part of it is communication. They could send emails, write, control a computer. I mentioned driving. Patients would like to drive, especially if they're not completely paralyzed. The upper limb control is very important, so they would be able to use robotics so that they could again feed themselves or perform other dexterous tasks that make them more independent. The payoff would be huge. Also, it would require less care from caregivers, because they could perform many behaviors themselves.

ZIERLER: The theme here is independence. That seems to be what people want.

ANDERSEN: Right.

ZIERLER: Have you gotten involved in the startup space? Have you seen opportunity to start companies or advise companies that could make these technologies possible?

ANDERSEN: I haven't. A couple of former postdocs are trying to start up one with functional ultrasound for depression diagnosis. I haven't really tried to start a company, because I've got so much going on [laughs]. It would have to be someone from the lab that would do it.

ZIERLER: These are all devices and programs. There's stuff here to build, ultimately, to help these people.

ANDERSEN: Right. Also in that space, there are a number of startups already. One is Blackrock, which we use their technology—we do have patents for a lot of things, and those will probably get used, licensed at different stages. Another big one is Elon Musk has a company [laughs] to do this, as part of his goal of enhancing humans so they could—well, he says—to compete with AI. That's Neuralink, a self-funded company in the Bay Area. There are other companies that use different technologies. But the ones that are using single unit activity are probably the most sophisticated and gives the highest resolution signals that you can get. I would expect in the next few years, this technology will be available for clinical use. You can't go to a hospital now and say, "I want an implant."

ZIERLER: None of these things are actually out to market yet?

ANDERSEN: That's right.

ZIERLER: What are you most bullish on? What's going to achieve market performance first? What kind of devices or programs?

ANDERSEN: It will probably be an implant in the motor cortex. I'm not sure quite what the killer apps would be. One might be a communication application. Or more generally, using applications on a computer.

ZIERLER: What exactly is the implant doing? What is it touching, to start?

ANDERSEN: The cortex. It is a subdivision of the brain that is like a a sheet, and it's all crumpled up, to fit this sheet into the skull.

ZIERLER: That's to say that you could flatten it out and it would take up a lot of space?

ANDERSEN: Yes, I think about .4 by .6 meters.

ZIERLER: Wow.

ANDERSEN: There are two hemispheres. And it is thin varying from about a millimeter and a half to two millimeters, and the neurons are arranged in layers. For the electrodes that are currently used, there's a little pneumatic device that punches the electrode arrays into the brain. The tips of the electrodes sit next to neurons. The electric field around the neurons are picked up on the electrodes.

ZIERLER: Then what happens with those signals? Where do they go?

ANDERSEN: Ideally, they would be wirelessly transmitted, but right now they go to a plug connector on the skull. Let me see if I have one. [pause] These are two damaged ones that can be held. They come to us like this, and if you look in there, that's the array.

ZIERLER: Okay. Is that a wire I see?

Implantation and Planning

ANDERSEN: Maybe it helps to look at this first. [laughs] All the electrodes have been knocked off. It's a little chip.

ZIERLER: This is implanted into the skull?

ANDERSEN: Yes, that's right. It's screwed into the skull. This one has some electrodes, still. If you look, they are tiny—like a little bed of nails.

ZIERLER: Yeah. And this is actually what's touching the cortex?

ANDERSEN: In the cortex, yeah. The top part is not, but the little bed of nails is. If you look at this again, you can see the little nails.

ZIERLER: Oh, yeah. Then what does this attach to? A computer?

ANDERSEN: That's right. First it goes to amplifiers, and then to a computer.

ZIERLER: Then the patient will give you a thumbs up or some way to acknowledge that what is coming out on the screen is an accurate representation of what's going on in their brain?

ANDERSEN: After an implant, we wait about two weeks, and then we hook it up, and yeah, on a screen, we see the action potentials from all the channels. Last year we implanted arrays in six different locations, and a couple months ago we did four locations, so you get a few hundred neurons whose activity you can watch on the screen. Then we can ask the subject to do something, like imagine they're reaching for an object. Imagination is somewhat correlated with actual—because they can't move anymore—with what it could be if they could move. That partially trains the decoder, and then—

ZIERLER: Is that because when you go to pick up a cup of coffee, you first have to, even if it's subconscious, you have to imagine yourself doing it? Is that the idea?

ANDERSEN: You have to plan it. And for some reason—that's another surprise we found—is that these areas are so dominated by imagination, and it seems to play out on the same pathways that are involved for actual planning. Then when they start doing things like moving a cursor or moving a robot hand, then based on how well they're doing, we can further train the decoder. In our first implant, we showed the action potentials to the subject, and he could turn them on and off. This had been shown in monkeys, in the late 1950s, and it was thought to be so amazing, that you could go in and turn on and off individual neurons. But as we explored it more, you're turning on and off a lot of neurons [laughs], and so you just explore the right thoughts that turns them on and off. Subjects can do that. They can watch one of their neurons, they can make it fire or not fire, depending on what they are imagining.

ZIERLER: You're already seeing positive progress here? You're seeing a trajectory where these are devices that will come to market and they will absolutely help people?

ANDERSEN: Yeah, absolutely.

ZIERLER: What does the cost look like? Are insurance companies going to pay for these things? Is this going to be only for the domain of the wealthy? How accessible will these devices be?

ANDERSEN: That's a good question. That's what neuroethicists raise. One of the robot limbs we use is by a Canadian company. It gets screwed onto a wheelchair, and paraplegics or tetraplegics with partial spinal cord injury, can use a little joystick to control it. That is one of the robot limbs we use and its relatively inexpensive, around $50,000. The Modular Prosthetic Limb we initially used with a DARPA project was very similar to a human hand and arm and was around $500,000. Many tasks can be done with just a computer and a screen. You can play music, do video games, use Photoshop—

ZIERLER: Are there advocacy groups for the paralyzed that lobby the government for research support and funding?

ANDERSEN: Yeah. There are, and we interact with them to try to get participants, but they're more therapy groups, I would say. I think veterans' groups lobby, but in a general way. One of our postdoctoral fellows, David Bjanes, received a fellowship from the Craig Neilsen Foundation. There is the Christopher and Dana Reeve Foundation; we've had funding from them. Advocacy may evolve as BMIs becomes more commonplace.

ZIERLER: Some questions about your professor's life—do you interact with undergraduates? Do you teach undergraduate classes?

ANDERSEN: Yeah, I teach. [laughs] In principle, I teach graduate classes, but undergraduates take them.

ZIERLER: That was my question. You interact with undergraduates who are taking graduate-level classes?

ANDERSEN: Yeah, and occasionally we have summer undergraduate students who do research in the lab.

ZIERLER: Just a generational question—twentysomethings now, they're so comfortable with computers and video games and all of that—in what way is that an asset for you, their skillsets, just based on how young they are and the technology that they've grown up with?

ANDERSEN: Oh yeah, it's wonderful. [laughs] It's daunting, in a way, but yeah. The Caltech undergrads are phenomenally talented, so they can pick up the programming tasks right away.

ZIERLER: How many graduate students do you normally have at any given time?

ANDERSEN: Right now I have five, but I think I'm going to be adding two more. Currently there are nine postdocs, but two are going to be leaving to jobs this year.

ZIERLER: How eclectic is everyone's research? Is there always a unifying theme? Is it all related to the brain interface issue and devices, or is it more diverse than that?

ANDERSEN: Originally it was just recording from areas outside of motor cortex to see what high-level signals we'd get, so that was the unifying theme. Then we added micro-stimulation. We can stimulate through these same electrodes with tiny amounts of current. We have a grant for that, as well as a grant for comparing motor cortex with posterior parietal cortex. So we have two subgroups now. We have two IDEs. One is primarily working with UCLA and the other with USC. Then a third subgroup has bubbled up in the last few years with developing a functional ultrasound BMI, for imaging changes in blood volume that are linked to neural activity. It's like an MRI but it's much higher resolution, much cheaper, and much more versatile, transportable.

ZIERLER: Your graduate students and postdocs, are they generally interested in academic careers? Do some of them want to go into industry or even into policy?

ANDERSEN: It used to be almost everyone went on an academic track, and only occasionally would they go to industry. Now it's about half and half, I would say.

ZIERLER: What's the takeaway there? What are the big trendlines for you?

ANDERSEN: Either is fine. When I started out, there were many more positions, I would think, in academia. It's harder to get positions now, although my students still do, so that's great. But the expansion of tech industries offers a broader variety of careers. Industry I think pays a lot more. It has its good points, and academia has its good points.

ZIERLER: Just as a snapshot in time, what are you currently working on, circa September 2023?

ANDERSEN: I always realize when people ask that, that I'm working on too many things. [laughs]

ZIERLER: [laughs] If you could set aside all the stuff that you don't have to work on, and you can just concentrate on what's most exciting to you right now, what is it?

ANDERSEN: The silent speech decoding is exciting. The human functional ultrasound imaging, we're doing that with people that have part of their skull replaced after traumatic brain injury. The stimulation, we've gotten approval, so now we can stimulate through all the arrays in different places, so who knows what we'll find. We have this kind of theory of how the brain—this mixed coding, how it's really organized, it's a compositional code. This table here has legs and the flat surface; you put that together, so that way, you don't have to have neurons for everything; you can have parts, and you can combine them. You could do that with movement and sensation. Then there are a million other little projects. [laughs]

Considerations of Evolution and AI

ZIERLER: Richard, for the last part of our talk today I want to ask some questions that might even border on the philosophical, since you work with the brain. First, is evolution a useful intellectual framework for you? Given that you know so much about the difference between monkey brains and human brains, is the concept of evolution relevant for your work?

ANDERSEN: Yeah, of course, and looking back at other species. Probably the most dramatic thing has been the expansion of cortex and the encephalization of function. For instance, Markus Meister here at Caltech and others have found that removing the motor cortex of mice, they're fine [laughs], except for more complex motor tasks.. A cat, likewise, they can continue to walk without cortex. But a human has a stroke in motor cortex and they're devastated. The cortex has gotten larger, and it has invaded all the other structures in the brain through neural pathways, so it has kind of taken over in some ways. How that happened in evolution is interesting. Perhaps more cortex resulted from some gene mutations which then found a use through evolutionary pressure. That would certainly be helpful. [laughs]

ZIERLER: Of course machine learning and artificial intelligence, everyone is talking about these days. First, just as a research tool, have you already embraced it? Is AI a force multiplier for the kinds of questions you can ask and answer?

ANDERSEN: Yeah, sure. These decoders we use are examples of machine learning. I hadn't realized in some ways that I was doing that kind of research, but yeah, the decoders learn, and we train them. More recently we have been using convolutional neural networks for feature extraction. We plan to use large language models, that will really be an advance, because we'll be able to directly read out inner dialog.

ZIERLER: For all of the big questions we're wondering about where AI is going—some people think that artificial intelligence and the human mind are going to continue on in separate tracks, and others see a convergence, where at some point in the way distant future, there's going to be points of contact where questions about where the machine begins and where the human brain ends are going to become murkier than now. In the grand sweep of history, are you contributing to those kinds of questions, where things might be headed?

ANDERSEN: You mean would someone come back from the future and whack me? [laughs]

ZIERLER: [laughs] I hope that doesn't happen! [laughs]

ANDERSEN: Well, a convergence is happening.

ZIERLER: It is?

ANDERSEN: Yeah, the decoders are machine learning, and also the devices we control, like the cars and the robot limbs, they're smart, so they use their own AI. So it's going to be assistive. What I do is medical research. There are whole areas of philosophy and neuroethics that imagine we're going to enhance the brain in healthy people, but that's not really my motivation. It's fun to speculate, and it has been in science fiction books for a very long time. But in terms of medical applications, it's already here. It's already happening.

ZIERLER: In the popular press, the brain is talked about as the most complex thing in the known universe. Does that resonate with you? Do you have an enhanced sense, a sense of awe, for just how complex the human brain is?

ANDERSEN: It depends on what you mean by complex.

ZIERLER: Do you also see simplicity in the brain?

ANDERSEN: [laughs] In certain domains, that's true. There is a branch of the study of consciousness that somehow you can measure the amount of complexity in things, and then the idea is that if something reaches a certain stage of complexity, it's conscious. So measures have been proposed. There are colloquial statements, like, "You use 10% of your brain." You use all of it. It's complex in that we don't understand a lot of it.

ZIERLER: On that point, what is the frontier for you in terms of understanding the brain? Where can you articulate, "We don't know now, but I can imagine a scenario where we will understand this aspect of the brain"? What seems in that regard to be the frontier for you?

ANDERSEN: One thing we've taken advantage of is the fact that the human participants can report their conscious experience. So we repeated a version of the Libet experiment and showed that actually what was going on in posterior parietal cortex is preconscious, and it must bubble up to consciousness somewhere else. So we can begin to ask questions about awareness and consciousness, those sorts of things, by using this particular paradigm. I think that will be interesting. And, we should find, now that we can stimulate through electrode arrays, we can manipulate things as well, so I think we're going to make inroads into questions of awareness, at least, if not consciousness. Speech can be a great approach too. For instance, in bilingual participants we can show that there is a phonetic component in posterior parietal cortex because word in different languages evoke different patterns of activity. But then it's also semantic. In experiments in which we use homonyms, so they sound the same, we can also decode what the semantics are. So it's a combination of both, which may be expected from a sensory-motor brain area. Getting into these really deep, high-level cognitive topics I think will be the future. [laughs]

ZIERLER: On that note, last question for today, maybe it's a bit of a personal question, but because you think about these things, do you consider yourself a materialist, strictly speaking? In other words, concepts of a soul or spirituality that may or may not exist, or metaphysical aspects to the human mind, is that all gobbledygook to you? Do you accept the potentiality of such things?

ANDERSEN: When I was young, like nine years old, I was beginning to doubt—I was raised Lutheran, and I was thinking, "This doesn't make any sense." [laughs]

ZIERLER: [laughs] There must be a better explanation out there?

ANDERSEN: Yeah. I asked my father about it, and he said, "Well, maybe you're right, but cover all bets." [laughs]

ZIERLER: [laughs] It's not bad advice!

ANDERSEN: [laughs] Yeah, yeah! But I guess I'm agnostic in that I think there are a lot of things we just can't be aware of.

ZIERLER: Is that a statement for our current level of understanding, or is that more of a total statement, that there are some things that are beyond the realm of what we're capable of understanding? If it's that, maybe that's the opening point for people who do think that there's such a thing as a soul or spirituality?

ANDERSEN: Yeah, although it's always changing, right? As science progresses, the domains of spirituality get smaller. Actually, you should probably be talking to physicists, since they have the really way-out ideas. [laughs]

ZIERLER: You already demonstrated that at the Salk Institute, they didn't have all the answers then either.

ANDERSEN: [laughs] Right, right.

ZIERLER: This has been a wonderful initial conversation. Next time we'll go back, we'll establish your family background and childhood, we'll take the story from there. Thank you so much.

ANDERSEN: Yes, thank you.

ZIERLER: Wonderful.

[End of Recording]

ZIERLER: This is David Zierler, Director of the Caltech Heritage Project. It is Friday, September 22nd, 2023. It's wonderful to be back with Professor Richard Andersen.

ANDERSEN: Great to see you.

ZIERLER: It's so nice to be with you. Thank you for joining me again.

ANDERSEN: Thank you.

ZIERLER: In our first conversation, we had a great tour of your approach to neuroscience, all of the big questions in the field. Let's go back and establish some personal history now. We'll start with your family background. How many generations back did you know your family? Did you know any of your great grandparents or your grandparents?

ANDERSEN: I knew my grandparents.

ZIERLER: Tell me about them. Where were they from?

ANDERSEN: On my mother Norma's side, they were Americans. They lived in Queens. On my father John's side, my grandfather was Danish and my grandmother was German, and they lived in Brooklyn.

ZIERLER: You're a New Yorker by roots!

ANDERSEN: [laughs] Yeah, I guess so!

ZIERLER: Andersen with an "e" would be the Danish variant?

ANDERSEN: That's right.

ZIERLER: How did they all get to this country? Do you know any of their immigration stories?

ANDERSEN: My grandfather on my father's side was a Merchant Marine, and when he came to the U.S., he just didn't get back on the ship, and traveled around the country for a while. He got citizenship at some point. My grandmother and her sister came to this country probably about the same time. I'm not sure where they met; I think in New York, though. My grandfather was a house painter, and grandmother stayed at home. On my mother's side, her father was a milk delivery man and her mother was a stay at home mom. He fought in World War I and was injured, but he lived until age 97, so he survived to a ripe old age.

New York Roots

ZIERLER: Where did your parents grow up?

ANDERSEN: My mother in Queens, my father in Brooklyn.

ZIERLER: Where did they meet?

ANDERSEN: Through mutual friends in New York.

ZIERLER: Did they stay in New York? Did they leave to start a family or for jobs?

ANDERSEN: My father was a chemical engineer, and so they moved near Pittsburgh, where I was born, and then to Ardsley, New York, for a brief period, then to Buffalo, then to the Bay Area, then to Louisiana, then back to the Bay Area.

ZIERLER: Oh, my!

ANDERSEN: A lot of moving, yeah.

ZIERLER: A lot of moving. All for new career opportunities?

ANDERSEN: Yeah. Then after I went to college, they moved to Cleveland. After that my father retired. At the time, I was at MIT, so they moved to Cape Cod. When I moved out here, they followed us and moved to Palm Springs.

ZIERLER: What was your father's education? How far did he go?

ANDERSEN: He had a bachelor's degree in chemical engineering from Brooklyn Polytechnic, which is now part of NYU.

ZIERLER: What kinds of jobs did he have? What was his career path?

ANDERSEN: He was mostly an executive. In the beginning, he designed chemical plants and then got them started. He was a bit hands-on for a while, but in a sort of executive capacity. He worked for Stauffer first—they made chlorine—and then he worked for Kaiser Aluminum and Chemical Corporation. Around the time Kaiser Permanente was started, the healthcare organization. It's the only part that survived. [laughs] I was in on the ground floor as a patient when we lived in California. In Louisiana, my father made chemicals and aluminum. They got bauxite for the aluminum from Jamaica.

ZIERLER: Did your father involve you in his work at all? Did you grow up having a sense of how chemistry and industry interacted?

ANDERSEN: Yeah. He talked to me about it all the time, even when I was very small. [laughs] Usually I didn't understand what he was talking about, but through osmosis I kind of got a feel for chemistry. He had a lot of stories. He had an amazing memory.

ZIERLER: Let's trace the chronology of your childhood through all of these moves. You said you were born in Pittsburgh.

ANDERSEN: New Kensington. It's a suburb of Pittsburgh.

ZIERLER: What year was that?

ANDERSEN: 1950.

ZIERLER: How long were you there for?

ANDERSEN: I think four or five months. [laughs]

ZIERLER: Oh, you don't remember it.

ANDERSEN: Yeah. [laughs]

ZIERLER: Then what's the next stop?

ANDERSEN: Ardsley. That was just a few months. Then Buffalo for four years.

ZIERLER: So your first memories probably would have been in Buffalo?

ANDERSEN: That's right., I remember it was fun. It was cold. We lived on Grand Island, which is an island just before the Niagara Falls. It was fun to play in the snow. My oldest memories are happy ones. Then we moved to Walnut Creek, California. Kaiser had its central office in Oakland. I was in first grade there. That was nice, too. Then I moved to Louisiana and was there for 10 years.

From Louisiana to California

ZIERLER: Louisiana is sort of the bulk of your childhood?

ANDERSEN: From six to sixteen, yeah.

ZIERLER: Where in Louisiana? New Orleans?

ANDERSEN: Baton Rouge. I went to elementary school. They didn't have middle school, just high school, so high school was six years.

ZIERLER: What were race relations like? Do you remember in Baton Rouge? Were the schools integrated?

ANDERSEN: I was there in the late fifties and early sixties. They were not good. There were still segregated facilities and things like that, in the beginning. That was during the whole period of the Civil Rights movement. There were not many people at that time from the north or west, so I was considered a Yankee [laughs] and sometimes the contrarian. That was all happening when I was in school. We moved in 1966 to Moraga, California, which is near Berkeley. It's just outside the Caldecott Tunnel. I finished my junior and senior years of high school in Moraga.

ZIERLER: You went to public schools in Louisiana?

ANDERSEN: Yes, and in California. All public.

ZIERLER: The schools in Louisiana, were they good schools?

ANDERSEN: Yeah, surprisingly, the particular school I was in was located in a well-off part of Baton Rouge. When I got to California, I actually found that in science and math the California high school as a bit behind, so that was a bit of a surprise to me.

ZIERLER: Wow. You wouldn't think that.

ANDERSEN: No, you wouldn't. [laughs] Yeah, it was all public schools, all the way.

ZIERLER: Growing up, were you always more on the math and science track?

ANDERSEN: I had thought I was a good student, but when my wife and I found some of my old report cards, apparently I wasn't. [laughs] This is in Louisiana. I got a lot of poor grades in citizenship, which had to do with my correcting teachers about their ideas of what it's like in the North, because people didn't travel much back then.

ZIERLER: Did you have any sort of protean ideas about the brain or about consciousness as a kid? Were those kinds of things interesting to you, or that developed later on?

ANDERSEN: In high school in California, I did a couple of science projects. one was on pollution and another was building a fuel cell. Then I had a summer job with Dow Chemical for computing, back then it was big stacks of cards. And, I worked on this environmental study out in Marin County looking at the effects of dredging in the San Francisco Bay. At that point I had a science bent. Then when I went to college, I was an engineering major, did that for a couple years, but—

ZIERLER: What colleges did you look at? Where did you apply?

ANDERSEN: I think I only applied to one. It was different in those days.

ZIERLER: UC Davis? That was it?

ANDERSEN: Yeah. I wanted to not live at home but I had a girlfriend in Orinda near Moraga, so I wanted it to be close. Of course that didn't last long, but Davis was very nice.

ZIERLER: Being in California—I'm always curious about this—had you heard of Caltech?

ANDERSEN: I wonder when the first time I heard of Caltech—I certainly didn't apply to it. [laughs] It probably wasn't much on my radar. Probably I became aware of it in college. I didn't apply here to graduate school. I only applied to Rockefeller and UCSF.

Engineering at UC Davis

ZIERLER: You started college in 1969?

ANDERSEN: 1968.

ZIERLER: You graduated in 1973? Did you take an extra year?

ANDERSEN: Yeah, I did. That was during the antiwar period, and I was—

ZIERLER: You went to college in the middle of all of that.

ANDERSEN: [laughs] Yeah, and I had a little fun and distraction in the middle. In the last two years I thought I'd better get serious, so at that point I got all A's, and I had to take a lot of courses.

ZIERLER: Was UC Davis a political hotbed? Were there marches and demonstrations and things like that?

ANDERSEN: There were some, but it was mostly Berkeley. Davis had a radio station, and I was the news director. I also did music shows. I would go to the demonstrations in San Francisco, Berkeley and Santa Barbara as the press [laughs]. So, I was aware of it. Davis was a bit sheltered, which was probably good for study, at the time.

ZIERLER: Probably a bit of a more conservative environment than Berkeley would have been?

ANDERSEN: Not really. It was just that Berkeley was—my wife went to Berkeley, and they had People's Park, and even before that, the Free Speech Movement. It had really been one of the major centers of the student movement at that time.

ZIERLER: Was the draft something you needed to deal with?

ANDERSEN: Yes. I had a student deferment, being at college. Then there was the lottery. I think that kind of broke the student movement, because some people had high numbers, some low. I had a high number, meaning I was less likely to get drafted. But that was kind of toward the end of the Vietnam War.

ZIERLER: You mentioned your course of study was engineering at UC Davis. Was that influenced by some of the work you did before college?

ANDERSEN: Right, yeah, probably, and also my father was an engineer. Which is good for what I do now.

ZIERLER: Oh, that's interesting. In what way?

ANDERSEN: It gave me some quantitative background. Of course things are much more advanced now [laughs]. I guess I found the engineering courses were a bit dry. In the end, I knew I wanted to do neuroscience, but there wasn't such a thing as neuroscience. The closest thing was psychobiology. So I pieced together various courses to get a biochemistry degree.

ZIERLER: As an undergraduate?

ANDERSEN: As an undergraduate, right.

ZIERLER: What was your initial connection to neuroscience? Did you read an article? Was there a professor there that you connected with?

ANDERSEN: Yeah. I actually worked my way through college even though I didn't have to. The first year, I worked in the Peace Corps dining facility over the summer. The Peace Corps group was housed outside of Davis and the volunteers were destined for Nepal. The two summers, I worked in the Hunt's cannery——and I was a Teamster. The pay was very good. It canned tomato products like canned tomatoes, ketchup, tomato sauce, tomato juice, and I was quality control. It was very intense. You worked many hours when the season was on. I would always make enough money to pay my way through the next year of college, but the experience made me realize that I never wanted to have a job like that [laughs], because even though I had a more technical job there, it was so boring, and you'd watch the clock. That I think influenced me to get serious about school. The next two summers, I worked in a neurophysiology lab at the Davis Medical School.

ZIERLER: Davis has a medical school?

ANDERSEN: Yeah. Part of it may have moved to Sacramento since then, or at least has a hospital there, but yeah, it has a medical school. There was a neurophysiologist on the faculty, Robert Scobey, and he had trained at Johns Hopkins with a famous visual neuroscientist, Gian Poggio. We did recordings from the retina, single-cell recordings, looking at the effect of displacements of visual stimuli within the receptive fields. That was a great experience, and I got paid for it, too.

ZIERLER: You were hooked? This experience drew you in?

ANDERSEN: Yeah, because certainly although I was interested in the psychology of consciousness, awareness, these sorts of things, like most college kids, I thought psychology wasn't going to give the answers; it would have to be more biological. That's how I got into it.

Neuroplasticity at UC San Francisco

ZIERLER: Of course this influenced where you applied to graduate school?

ANDERSEN: Yeah. Again, in those days you didn't apply many places. I applied to UCSF and Rockefeller, and I got into both. I went to UCSF, because at that time—

ZIERLER: Were you married at this point?

ANDERSEN: No, I wasn't. At that time, it had a very good neuroscience faculty, a lot of young, dynamic faculty. Again, there wasn't a field of neuroscience, but a lot of them were hired out of Harvard Neurobiology. There was a faculty member, Ben Libet, who studied awareness with a neurosurgeon, so that was my first rotation. Then the surgeon passed away, so I went Mike Merznich's lab. Scobey had been from Hopkins, and one of the leaders of neurophysiology, Vernon Mountcastle, was there, and he was a part of that group of neurophysiology faculty. Merzenich had been a student of Mountcastle's, so it was kind of a Hopkins connection.

ZIERLER: What was Merzenich's area of research?

ANDERSEN: He has two main areas, and he was working on both when I was in the lab. One was the auditory system. He was one of the developers of the cochlear prosthetic. Then he also was very well known for neuroplasticity. He showed, by grafting nerves and other experiments, a high degree of plasticity within somatosensory cortex. Then he went on to start one of the first companies of video training for brain improvement video programs. First he addressed dyslexia, but then he branched out into all kinds of neurological disorders..

ZIERLER: What did his lab look like? What were the instruments?

ANDERSEN: For UCSF, at that time, it was all at the hospital on Parnassus hill. They didn't have the Mission campus. The labs were generally pretty small. His lab was bigger than many, but since he had so much going on it was quite crowded. There were about 50 people working in around 2,000 square feet. I mostly worked at night so I could use the equipment and do the experiments, because during the day it was kind of chaos.

ZIERLER: Were there human subjects in the lab or would you go to the hospital?

ANDERSEN: The humans for the cochlear implant surgeries were in the hospital, not the lab. Besides Merzenich, only human studies in the department that I recall were in Libet's lab. At that time, he was looking at, and later became famous for, the timing of awareness.

ZIERLER: What does that mean, timing of awareness?

ANDERSEN: He found that there is neural activity for planned movements even before you are aware you are going to move.

ZIERLER: There's a physiological component before there's a neurological—?

ANDERSEN: Yeah, awareness—so if you wanted to turn off your recorder, you would already be programming that without being aware of it. He had an experimental design where he could look back in time before movement execution. He found you become aware of the intent to move before you move, but you are not aware of the plan to move before movement execution. That sparked a lot of—

ZIERLER: Does that go back to like lizard brain, fight or flight reflex? Is it connected there?

ANDERSEN: No. It's more about whether you aware of your intentions.

ZIERLER: It's higher order?

ANDERSEN: Yeah. It touches on a lot of philosophical and legal issues. It has been suggested to be evidence for a lack of free will. Legally, intent is an important element in criminal courts. Tyson Aflalo and I actually recently did some experiments with human participants with implants in posterior parietal cortex. The planning activity again precedes awareness. This is true even for sequences of movements. It appears that a lot of what we do, we're not aware of. It's implicit, like driving a car while thinking about your next lecture. But then there are other parts of the brain that perhaps sample this implicit planning activity. It may be areas identified as active for theory-of-mind tasks or self-awareness, and that those areas may tap into areas that we study. That being said, there is also a bit of evidence that posterior parietal cortex is important for awareness of intent. Electrical stimulation of the posterior parietal cortex in awake patients during brain surgery produces an urge to move, but they don't actually move, and when premotor cortex is stimulated, they move but think they didn't do it. Of course posterior parietal cortex is a big area, so it may be that implicit and explicit intent are represented in different parts of posterior parietal cortex.

ZIERLER: I can't help but ask, being in San Francisco in the early 1970s and working on things like awareness and consciousness, psychedelics were very big at this time. Were people thinking about that? Were people experimenting with psychedelics?

ANDERSEN: I moved to San Francisco in 1973, so that was after the Summer of Love, which was 1967, although I lived in the Haight-Ashbury because it was right next to the UCSF campus. The Haight-Ashbury had historically been a blue-collar neighborhood with a diverse population, but then there was the influx of hippies from the Summer of Love that mixed in to the area. By the time I moved to the Haight, it had gotten kind of seedy along Haight Street. Also, it was around that time that California had reduced the number of patients in psychiatric hospitals. So a lot of former psychiatric patients people moved into the area. Doing your laundry at the laundromat could be quite a challenge. [laughs] But it was by and large a peaceful and interesting area; I really enjoyed living there.

ZIERLER: Did any of those cultural elements seep into the laboratory? Were people experimenting with drugs? Were they thinking about them? Were they thinking about the connection between altering your mind with psychedelics and what this might mean for laboratory experiments?

ANDERSEN: As far as I know, that's all new.

ZIERLER: It was all verboten, essentially? It wasn't a line to be crossed?

ANDERSEN: Well, it was just that a lot of people developed drug problems during the 1960s. LSD was legal early on, but it was illegal at that point. Yeah, that is curious, that there wasn't more research on psychedelics. In fact I was surprised when it came back. [laughs] Recently I asked one student who worked in one of the labs on the East Coast studying psychedelics, "Well, surely it wasn't the doses that people were taking on the street," but he told me they do take large doses. But I don't recall any research at UCSF on psychoactive drugs during the time I was there.

ZIERLER: Were computers in use at all in the lab?

ANDERSEN: Yeah, especially when I went to Hopkins. That was the Digital Equipment Corporation PDP mini-computers. It was one of the early computers that was designed for scientific use. They had 64K of memory, much less than a scientific pocket calculator [laughs]. We had to learn machine language so that we could conserve on the memory. You couldn't just write code like you do today. The computers, about a third of the time, were not working. They were wire-wrapped and everything. There were a few of us that were working with monkeys that graduated at that time, and many have gone on to become leaders at Stanford and MIT and many other universities. But at that time some of us doubted whether the field would continue because of the limitations of the computers. They just weren't powerful enough and easy enough to use. Then IBM started producing business computers that had much larger memory, so we'd use those to do the computations. At the time, there were just a few people doing monkey research. Then we all trained lots of people, and it became a big, thriving enterprise.

ZIERLER: Was monkey research part of the work in San Francisco, or that's later on?

ANDERSEN: That was later on.

Cats and Cortex Reciprocity

ZIERLER: Tell me about developing your thesis research in San Francisco.

ANDERSEN: I worked on anatomy of the cat auditory system. At that time, there were a couple of new techniques for tracing pathways in the brain. I think we talked about this before, that we found everything reciprocal.

ZIERLER: Remind me, what does reciprocity mean here?

ANDERSEN: It means an area in cortex that projects to another cortical area that gets a projection back from that target area. Likewise, the thalamus, which is kind of the gatekeeper to most information coming up into cortex, shows reciprocity with cortex. The thalamus is made up of nuclei and if a nucleus projects to a particular area of cortex, it gets a projection back from that cortical target.

ZIERLER: I wonder if you could narrate, how do you discover this? What does that look like?

ANDERSEN: It was really simple, and this friend Steve Colwell's idea. You just mix the tracer that goes forward with the tracer that goes back and inject it into cortex. It turns out you get near identical labeling of the forward and backward tracers with every area the injected cortex connects, the connections are all reciprocal.

ZIERLER: What is a tracer? Is it a chemical?

ANDERSEN: Yeah. One is horseradish peroxidase. It gets transported retrogradely, so if you inject it somewhere, it gets taken up by the synapses and transported back to the cell bodies. The other is a tritiated amino acid. It gets made into proteins and is transported the other direction from the cell body to the synapse. In that way, you could look both at what's coming and what's going.

ZIERLER: What's the instrument that you use to track the tracing?

ANDERSEN: The horseradish peroxidase, you'd use a chemical to react with it, and you could see it under a microscope, these cell bodies full of the reaction product. With the tritiated amino acid, it was radioactive, so you'd put an emulsion on the slide, and then like photography, you would develop the slide and see under the microscope where the radiation exposed the emulsion.

ZIERLER: This is vivisection? This is live subjects that you're working on?

ANDERSEN: They're not alive when they're on the slide. You make these injections and you euthanize the animals and make sections of the brain to follow the labelled pathways.

ZIERLER: What is the bigger takeaway of learning about reciprocity? What does that tell you about neurobiology?

ANDERSEN: It's a couple of things. One is that every pathway that goes somewhere gets feedback. People have been interested in that for a while. One idea is that you feed forward information to higher centers and they have gathered through learning and experience a prediction of the input, so they feed back to mold the activity in the lower areas—because you have an ill-posed problem that the higher area helps to solve. You can, say, in vision, improve what you're seeing. People have examined it with neural networks, trying to model cortical pathways, and the reciprocal pathways help out the computations. In AI, it's commonly called RNNs, recurrent neural networks. They're very valuable for understanding dynamics. Rather than just showing static data from a layer of a neural network, the recurrence shows sequential computations and is also useful in holding memories. All that being said, that's all kind of conjecture because [laughs] it's hard to prove. One could knock out the feedback pathway—probably someone has tried that experiment—but yeah, I think it's still somewhat of an open question.

ZIERLER: Was the going assumption when you were in graduate school that what you were learning in non-human subjects ultimately would be relevant for understanding the human brain?

ANDERSEN: Absolutely, and these new tracing methods afforded a detailed accounting of brain connections. David Van Essen is a prominent neuroanatomist who at the time went on to become a department chair at Washington University, and I was his replacement. This was his lab space where we are sitting now. One of the things he has done is to collect tract tracing data from different labs and put it in a database and to work out all the connections of all the cortical areas. Since then, he went on to work on that with functional magnetic resonance imaging and the Connectome Project in humans, and that has been a valuable repository of information, too.

ZIERLER: Where were you looking for your postdocs?

ANDERSEN: The anatomy was fine, the results were clear, but it was not so interesting. UCSF at the time had a really great library, probably does still, in the hospital. I'd go there and read books and papers by Luria, Holmes, Balint and Geschwind and other famous neurologists—describing really interesting deficits that you get from lesions to the brain.

ZIERLER: What does deficits mean here?

ANDERSEN: Oh, so like with the posterior parietal lesions, you have mis-reaching. You have what's called neglect, where you can't focus attention to the opposite side of an object or the visual field. You can't navigate. A lot of these deficits helped to design experiments. The neglect we thought was attention but it seemed more like intention, unable to plan to look or move in a particular area of space. It's an area that is very important for spatial awareness. That's why we started looking at coordinate frames and how space is represented in the brain. Now we're doing navigation experiments guided by the navigation deficits when posterior parietal cortex is damaged.

ZIERLER: These questions were informing where you would look for your postdoc?

ANDERSEN: Yeah. At the time, just a couple of labs had developed behaving monkey experiments so that they could record single cells in awake monkeys. Before that, they were anesthetized.

ZIERLER: Behaving monkeys meaning that you can watch what they're doing?

ANDERSEN: Right. You train them to perform a task and when they are doing it—you record the activity of single neurons at the same time, so you can correlate the activity of the neurons in a particular area with the properties of the task that you've designed. Vernon Mountcastle had done a lot of work in the somatosensory system, and he was interested in how you went from somatosensation over to awareness, to decision-making, and higher-order cognition. He thought since the connections of the somatosensory cortex went to the posterior parietal cortex, that he'd examine this interesting area in behaving monkeys. That was great. It fit perfectly with what I wanted to do. I wrote him, and since I was recommended by Merzenich, who was one of his students, he took me into his lab, so I went out to Baltimore.

Johns Hopkins and the Arrival of Neuroscience

ZIERLER: Was the term "neuroscience" already coming into fashion by the time you got to Hopkins?

ANDERSEN: It had started, yeah. Vernon was one of the first presidents of the Society for Neuroscience just a few years before I got to Hopkins. Initially it had 1400 attendees. Now it typically has 20,000 to 30,000 attendees.

ZIERLER: Either in real time or looking back, was there a convergence in the field that allowed for this umbrella discipline of neuroscience to combine things that were happening in a disparate manner up until that point?

ANDERSEN: Yes. There was a lot of work on development, and that tended to be molecular and anatomical. There was a lot of work on how axons are guided during development. The monkey work was an important component. Human research, not so much back then that I can recall, but now it's a big component as well. Even in the beginning, it was multidisciplinary. At times, people have wondered whether it has gotten too big, but it kind of works out. There's a lot of little meetings within a big meeting. And it is invigorating to see so many people, especially young people, interested in the field.

ZIERLER: Tell me about the lab at Hopkins. Was it a big operation?

ANDERSEN: No, it was very small. It was in one way intellectually big, because they had Mountcastle, as well as Apostolos Georgopoulos, who became a famous motor control person and helped start the field of brain-machine interfaces. They had Gian Poggio who was a visual neuroscientist and Mahlon DeLong, who has become famous for basal ganglia disorders and Parkinson's Disease. It was probably the best collection of neurophysiologists at the time. There was also two groups at NIH with Bob Wirtz working on vision, and Ed Evarts, working on motor control.

ZIERLER: Did you consider NIH for a postdoc?

ANDERSEN: No. Again, this was [laughs]—you just didn't try multiple places. But the labs themselves, like with Mountcastle, he did his own experiments. He worked with me directly and with another postdoc, Brad Motter. Georgopolous had a couple of postdocs that have now become famous, John Kalaska and Roberto Caminiti. Usually it was one senior investigator and a couple postdocs, so it was small in that regard.

ZIERLER: What was new for you at Hopkins and what was a continuation from what you had been doing in San Francisco?

ANDERSEN: What was new, of course, were the computers. At the time that I got there, there had been two programmers—one was a faculty member—and Vernon fired both of them right after I arrived [laughs], and so we had to start programming ourselves.

ZIERLER: Was the idea that he wanted scientists doing the programming?

ANDERSEN: No, he thought it was beneath us. So, we would do it at night [laughs] when he wasn't around.

ZIERLER: Undercover. [laughs]

ANDERSEN: Yeah, somehow the computer programs just magically appeared. Also we started applying statistical analysis to the data. At that time, there was some suspicion about statistics. Vernon would say that if you don't see it with your own eyes it's not there. Now, statistics are broadly applied in neurophysiology.

ZIERLER: You're saying intuitively you recognized the value of programming yourself? There was something about that experience that was important for your development?

ANDERSEN: Well, we couldn't do anything without it. We couldn't analyze the data. We couldn't change the paradigms we were running. I think it was probably obvious to Vernon, too, but he just didn't want us working on it while he was around. So yeah, you had to do it; it was no great insight.

ZIERLER: What were the big research questions in the lab? What were you after, for your postdoc?

ANDERSEN: Vernon had this idea of a command hypothesis, which those of us in the lab thought he should temper a bit.

Investigating Intentionality

ZIERLER: What does that mean, command hypothesis?

ANDERSEN: He found cells in posterior parietal cortex that were active for fixating, for smooth pursuit eye movements, for saccades, for reaching, and for hand manipulation of objects. He thought based on these results, and the lesion literature, that posterior parietal cortex made general commands for behaviors that were elaborated in downstream areas. Unfortunately he picked a bad term, because a command neuron, in a crustacean, is just this big escape neuron, that drives the animals backwards. Mountcastle proposed a "command hypothesis," and that movement commands were coming from posterior parietal cortex.

ZIERLER: As opposed to where? Where else could they be coming from?

ANDERSEN: True command would be very motor and at the point of executing the movement. There were two investigators at NIH, Michael Goldberg and David Lee Robinson, who also started studying posterior parietal in behaving monkey, and they hypothesized the opposite, that it was important for attention, and that all of Mountcastle's findings could be explained by attention. Like when you reach, you attend to the object you are reaching to. When you make an eye movement, you attend to the target of the eye movement. There were sensory visual responses that both groups reported. The role of posterior parietal cortex became a contentious question: Is it attention or is it motor?

ZIERLER: Is there a third option, or it has to be one of them?

ANDERSEN: Well, that's what happened to me, yeah. [laughs] I thought, well, it's pretty abstract, but it is related to planning movements. There is attention in sensory systems and then you plan movements. The posterior parietal cortex represents the world around you and your self in the world, and taking into account many factors to formulate plans for movements. Then those plans go to the frontal lobes, where the movements are executed. My idea was more of a sensorimotor theory in-between sensation and movement. But it was daunting, because when I started my lab, I was pursuing this theme, which was not popular with these giants in the field, especially those attributing only attention to the posterior parietal cortex.

ZIERLER: What convinced you to develop an alternative approach?

ANDERSEN: Well, because I thought it was right. [laughs]

ZIERLER: But why? What compelled you that this approach was right?

ANDERSEN: In experiments I did with Jim Gnadt we found this persistent activity for planned eye movements that was there even if you shifted your eyes, and it would be recomputed to form the spatially correct movement plan. We had controls to show that the activity reflected planning and was not linked to vision or memory. That was one bit of evidence. The next that I think nailed it for us was Larry Snyder, Aaron Batista and I used different effectors, so we'd use either eye movements or reaches, and we discovered one part of the cortex, which is famous now, and people work on it a lot, called the lateral intraparietal area, or LIP, would be active for planning saccades but not reaches. Then we found on the other bank of the intraparietal sulcus a reach region which was active for making reaches but not saccades.

ZIERLER: What are these terms, reaches and saccades?

ANDERSEN: Reach is reaching for something. [laughs] The monkeys reach out, touch—

ZIERLER: Which is a form of intentionality? It shows what they want to do?

ANDERSEN: Right. And you can withhold an action—you can flash a target for reach, say, if it's red, it indicates the monkey should reach to it. But then he has to withhold a response and then when he gets the cue to go, he reaches to that location. And so the activity in between is the motor plan, which he's holding. Then if you, say, have a green target appear at the same location, he has to plan a saccade instead. Attention would say it didn't matter whether you were reaching or saccading; the recorded neuron would be active, whereas we found one cortical area was active for reaching and another was active for saccading. That kind of nailed it for planning.

ZIERLER: They're separate? Are they talking to each other, these regions?

ANDERSEN: They do communicate, yeah, because saccades and reaches are highly coordinated, but they don't always go in the same direction. We could also do things like have a reach and a saccade in opposite directions, and you'd see the reach area light up for the one direction—because they're direction-tuned, the neurons—and the saccade area would light up for the movement in the opposite direction. Again, that shows activity for movement planning and not attention to a particular location in space.

ZIERLER: Was NIH supporting most of the lab? Was that where the funding was coming from?

ANDERSEN: Yes, but there's no conflict there, because NIH has its own internal—or they call it intramural—research programs, whereas I was getting funding from their external programs. A majority of NIH funding is external.

ZIERLER: Your findings as a postdoc, were you publishing? Were you presenting at conferences at lot?

ANDERSEN: Yeah.

ZIERLER: How were your ideas received, given the fact that you were in contention with the giants in the field?

ANDERSEN: Well, there was some debate. Actually, they were a pretty nice group. It was not contentious in a personal way. It led to, though, debate, which probably helped me, because it focused—

ZIERLER: Gets you attention.

ANDERSEN: [laughs] Right, yeah. But it was scary sometimes, because I'd be the next one up to talk after the previous person said something—same data but different interpretation.

ZIERLER: It was a great experience in Baltimore? It was very productive for you?

ANDERSEN: Yeah, I learned a lot. Mountcastle was very rigorous.

ZIERLER: Has Hopkins maintained its leading position in neurobiology over the years?

ANDERSEN: Yeah, the neurophysiology component has shifted out of the medical school to the Homewood campus. There is in addition a very large neuroscience contingent in the medical school. These days, though, often, like at UCLA, I went there once to advise on their neuroscience program, and they have 250 neuroscientists. [laughs] The scale is much, much larger now. The same is true at Hopkins and other universities.

Joining the Faculty at the Salk Institute

ZIERLER: Did you consider a second postdoc or were you looking at faculty positions at that point?

ANDERSEN: I only had a two-and-a-half-year postdoc but I learned a lot. My wife was fourth generation, and fifth generation, Californian.

ZIERLER: When did you get married? While you were a postdoc or in graduate school?

ANDERSEN: While I was a postdoc. We met while I was in graduate school at UCSF. Her name is Carol Ahern. She then went to do a master's in audiology at Hopkins. But she was not a big fan of the East Coast, and her family was all mostly in Oakland and the San Joaquin Valley.

ZIERLER: The pull of California was strong at that point?

ANDERSEN: Yes. I told Mountcastle that I needed to go back to California, and he thought—I guess he was half joking, but only half—he was worried that I would end up with beads and [laughs] a Hawaiian shirt and consulting for Hollywood [laughs]. He was fairly conservative. This job came up at the Salk Institute, and he actually directed me toward it. One of his colleagues, Max Cowan, was looking to fill a faculty position within his unit there.

ZIERLER: What year would this have been when you arrived at Salk?

ANDERSEN: The end of 1981.

ZIERLER: How established was Salk at that point? When did it start?

ANDERSEN: It had been around for a while, since 1963. It was started with funding from NSF and the March of Dimes. Jonas Salk founded the institute after he had developed the Salk vaccine. He was still there when I arrived.

ZIERLER: Did you get to meet him?

ANDERSEN: Oh, yeah.

ZIERLER: What was he like? He had moved on to very different ideas from virology, right?

ANDERSEN: Right. He also had some controversial ideas, like that evolution could also be social and not biological. He certainly liked to interact with the young people, and his son worked there as well. He had this nice house in La Jolla Farms, right next to the Institute. Yeah, he had certainly taken a different track at that time.

ZIERLER: What was the job for you? Was it a professorship? Was it a staff scientist position?

ANDERSEN: It was an assistant professorship. At the time the Salk didn't have undergraduates or graduate students, so we'd get graduate students from UCSD. It only at that time had about 60 professors.

ZIERLER: The Salk Institute really relied on UCSD?

ANDERSEN: I had mostly postdocs. At the time, the deal was I would teach at UCSD so I could be an adjunct faculty member in order to get graduate students. In the end I only got one, before I moved. I mostly had postdocs.

ZIERLER: Was Salk known as a leading center of neuroscience at that point? Were you part of the generation of building it up?

ANDERSEN: It was known. It had, at that time, a bit of a hierarchical structure. There was Cowan, who was a famous neuroanatomist who went on to be the chief scientific officer of the Howard Hughes Medical Institute, so a big influence. There was Floyd Bloom, who was a neuropharmacologist, very famous. There was Jim Patrick and Stephen Heineman who were molecular neuroscientists and had big labs. There was Roger Guillemin, who received a Nobel Prize for neuroendocrinology. So it had these big giants. Then depending on the lab you were in determined your degree of independence—in my case, Cowan let me do what I wanted. He was suspicious of the computers too, [laughs] and he couldn't believe that I bought this $33,000 computer with my startup money. So he said I'd have to share it with everyone in his lab. But of course, that never happened.

Talking Neuroscience with Francis Crick

ZIERLER: Was Salk a wealthy institution? Were your startup funds generous?

ANDERSEN: The startup funds were generous for the time. Actually, Mountcastle had advised me—because the same had happened in his career; he was trained as a neurosurgeon—that he had a few years where he didn't teach. He thought the Salk would be good because I wouldn't have much teaching or administrative responsibilities.

ZIERLER: Sort of like being a glorified postdoc?

ANDERSEN: Yeah, very much. [laughs] It was great. I made some major discoveries at that time. Francis Crick was there, and he was very much into computational neuroscience.

ZIERLER: Did you interact with Crick?

ANDERSEN: Yeah, quite a bit.

ZIERLER: What was Crick like?

ANDERSEN: He was amazing. Brilliant. He liked young people. He was inquisitive. Actually, a very nice guy. When I first met him—because I had read The Double Helix by Jim Watson—the first sentence being "I have never seen Francis Crick in a modest mood" [laughs]—I was a bit fearful of meeting him. My apprehension was unwarranted. He just came in my office one day, introduced himself and we shook hands. He influenced a lot of people at that time in the San Diego area, a quasi-mentor for a number of us.

ZIERLER: In building up your lab, before we get to the discoveries that you made while you were at Salk, what were the big questions that you were concentrating on at that point? How did that influence the way you put your lab together?

ANDERSEN: I thought it might be a good idea to do two tracks. One was spatial transformations, and the other was motion perception.

ZIERLER: This was a continuation from the postdoc research?

ANDERSEN: I had discovered the gain fields when I was with Mountcastle.

The Ubiquity of Gain Fields

ZIERLER: What is that, gain fields?

ANDERSEN: This area is supposed to be about space, and when people get lesions to it, they only see the opposite part of space.

ZIERLER: Space in one's field of vision?

ANDERSEN: Right. A hypothesis was that when you look at receptive fields in visual cortex, they are mapped to the eye, but that in posterior parietal cortex, they'd be mapped to the world, and so you could move your eyes around and still always know where objects are in space. I tried to see how space is represented in the posterior parietal cortex and found that it was still in the coordinates of the eyes, it still moved with the eyes, but information about where the eyes were looking was present and multiplied by the visual response. That was weird. So, it was a combination of two variables. I pursued that. That was when the original neural network people were centered at UCSD, and so we made models for transforming coordinates, and you needed these gain fields, so it made sense suddenly, so that was great. The other thing I was studying was visual motion, because that's the way you navigate. You can tell where you are headed from the focus of expansion of the visual scene—kind of like Star Wars——But that's complicated by the fact that if you make eye movements, you add motions due to the eyes moving which adds to the expansion and shifts the focus of expansion. Yet you still can drive and move your eyes without going off the road.. There was evidence that the posterior parietal cortex was also involved in this perception of self-motion I thought I'd work on that in case the examination of coordinate transformations didn't work out. They both worked out, so that was great.

ZIERLER: Let's define working out. What does that mean? We'll start with the first. How do you define success in the experiment?

ANDERSEN: Originally at Hopkins, we just noticed these gain fields, so I worked on them in more detail and how vision and eye position interacted. That was a paper in Science.

ZIERLER: That's a big deal, to get published in Science.

ANDERSEN: Yeah.

ZIERLER: How was it received, the paper?

ANDERSEN: Oh, great. Then other people started looking elsewhere in the brain, and it turned out that gain fields were everywhere. Eye movements and eye position are globally distributed and are used for calculating objects, how they piece together, or navigation, or—

ZIERLER: Did you have a sense of its universality, or it required other researchers to put this together?

ANDERSEN: Yeah, we did a series of experiments like—okay, if it's the eye, what about the head? Yes, gain fields for head position. What about when you reach? Yes, gain fields for arm position. All these things converge to compute positions in space. So, a network looking at the visual image, modulated by all these gain fields, can very easily transform where you should go, reach, and look. It would be possible to even send the signals to the muscles without ever explicitly having receptive fields anchored to space. We are now looking at spatial representations in human cortex, but it appears that the brain does do the transformation to absolute space, so you can find receptive fields anchored in space, as well. That was a big deal, and also got me into the neurocomputation component, because it showed how the brain computes the transformations, not just represents them. The visual motion field was a very hot at the time, and so we could show how, again, in this case, moving your eye when you're moving through space can influence how you interpret the motion field and the world, and you can use that to—

ZIERLER: What does that mean, the motion field and the world?

ANDERSEN: If you have this expansion going on, and your eyes are still, then that focus is where you're going.

ZIERLER: What's the expansion?

ANDERSEN: It's kind of like Star Wars, when they accelerate—or is that Star Trek? Anyway—[laughs].

ZIERLER: Warp speed?

ANDERSEN: Yeah, when they are at warp speed, everything goes fshhh! The same thing happens when you're driving. If you're looking ahead, everything is expanding from the point you're heading. Now, if you start tracking someone, a pedestrian walking on the side of the road, while you're moving forward, that generates this shear motion that goes in the opposite direction, and the two motions add together and displace the focus of expansion, which is no longer the direction you're going. Yet, you don't crash, or run into the person [laughs]—what we found was that there's a pursuit signal, an eye movement signal, that influences the receptive fields. Working with David Bradley we showed these cells are tuned to the location of the focus of expansion, and when the eye moves, again a gain field shifts the focus tuning of the neurons to correctly indicate the true direction of heading. We worked with a psychophysicist, Marty Banks at Berkeley who showed perceptually humans also compensate for eye movements. You perceive where you're going, and these are the internal computations it takes to get there.

ZIERLER: Was this also a Science paper?

ANDERSEN: Yeah, a Science paper, although done here at Caltech. The other thing we studied in motion was the structure for motion. If you paint a cylinder with dots, and you see it rotating, you see a 3D shape. We had a Nature paper at the Salk on that with Ralph Siegel.

ZIERLER: The $30,000 computer that you insisted on, you were vindicated?

ANDERSEN: [laughs] Yeah.

ZIERLER: It really worked out.

ANDERSEN: Oh, yeah.

ZIERLER: More generally, you're saying, you're getting more involved in the field of neurocomputation.

ANDERSEN: Right.

ZIERLER: Were you part of its founding? Is there a neurocomputation before you got involved?

ANDERSEN: It was forming at that time. Before that, the eye movement community, which was at that time a very popular area of research—there were engineers at Hopkins—David Robinson—who modeled eye movements and how they are processed in the brain stem. Then at MIT there was a school of AI, but they were more symbolic and around language, and it's not what is meant by AI today. Then at San Diego, they had the parallel distributed processing group, which were the neural network researchers. They had trained networks to speak, too. So there was kind of competition. Then there were young faculty coming up in the field of computational neuroscience—we had meetings at Woods Hole often t. There was Christof Koch here and Jim Bower, Terry Sejnowski who was at Hopkins at the time, several others. They were young then but are now big stars in this field. It's shifting a bit now, though, more toward machine learning. Neural networks died out for a while, but then has really come on now that computers are powerful enough to take in massive amounts of data for training.

ZIERLER: Looking back, your first faculty appointment at Salk, did you ever think it was risky not going to a bigger, more traditional university?

ANDERSEN: Yeah, it was, somewhat, because you had to raise your own salary, so it was soft money. But then it was easier to get funding.

ZIERLER: Even for tenured—? What was the tenure track like? Was everybody on soft money?

ANDERSEN: Pretty much, yeah. You had to go out and raise your own salary and support for your lab—it's sort of like that everywhere [laughs], but particularly at the Salk when I was there. If you lost your funding, I don't know what would have happened at the Salk, but people didn't.

ZIERLER: And they didn't because it attracted top people?

ANDERSEN: Yeah, certainly. And then, the pay line for NIH grants was much higher. My first grant I wrote, I got. Now sometimes junior people need to try several times before they get a grant. So it wasn't that risky.

ZIERLER: How long did you end up staying ultimately at Salk?

ANDERSEN: Six years.

ZIERLER: Did you achieve tenure within the six years, or you were just about to come up?

ANDERSEN: I did, yeah, I was five years assistant professor and one year an associate professor with tenure. At that time, I got the offer from MIT. I liked the Salk a lot, but MIT was leading the field at the time in neuroscience and computation, so it was, career-wise, a good move. Carol didn't like it very much. [laughs]

ZIERLER: The East Coast! You can't get much more east than Cambridge, Massachusetts.

ANDERSEN: Yeah.

ZIERLER: I think that's a great place in terms of a narrative turn for the next chapter. We'll pick up next time when you move to MIT. But last question for today, just to put a cap on your time at the Salk, what were your main contributions to the field? Obviously, your research was gaining a lot of attention. It was a great place for you academically, intellectually. What were your contributions to this burgeoning field of neuroscience at the time, would you say?

ANDERSEN: Also at that time we discovered that the neural activity in posterior parietal cortex was actually intent, intention—

ZIERLER: As opposed to what? What were the competing ideas?

ANDERSEN: Attention. Yeah, and the gain fields for coordinate frames, and the structure from motion, which was at the time also being studied a lot at MIT. I'd say those were the three main—oh, and doing the neural network, that was a Nature paper. That was the first time showing that a neural network could explain neural data, and also the first time to show that the gain fields could perform neural computations.

ZIERLER: To clarify, your use of the term "neural network" is happening within one brain? You're not talking about a neural network like how starlings coordinate their flight?

ANDERSEN: Yes, within one brain. Then, they were pretty simple, so they were layers of synthetic neurons. There was an input layer, a middle layer, and an output layer. The middle layer did the transformation. Now usually when people talk of neural networks, they have many, many layers. Also, they have fancy bells and whistles to improve learning in different ways. So neural network, in that sense that I'm using it, is something in a computer. But of course, there are neural networks in our brains, too, and in a way this research showed that these artificial neural networks could behave like real neural networks.

ZIERLER: Did the Salk Institute try to retain you? Did you give them an offer to counter?

ANDERSEN: Yeah. They would have had me direct a vision group there, so I'd have my own independent center.

ZIERLER: Which might have been compelling but not compelling enough?

ANDERSEN: It was compelling, yeah. I had a hard time sitting there looking at the Pacific Ocean, thinking—[laughs] so it was not an easy decision.

ZIERLER: With a happy wife.

ANDERSEN: [laughs] Yeah. It was not an easy decision.

ZIERLER: We'll pick up next time when you move to MIT. Wonderful.

[End of Recording]

ZIERLER: This is David Zierler, Director of the Caltech Heritage Project. It is Wednesday, September 27th, 2023. It is wonderful to be back with Professor Richard Andersen. Richard, as always, great to be with you. Thank you for having me.

ANDERSEN: Thank you for hearing my story. [laughs]

MIT and Neurocomputation

ZIERLER: We're going to pick up when you make the decision to leave La Jolla and join MIT. I just want to clarify, it was the Department of Brain and Cognitive Sciences?

ANDERSEN: That's right.

ZIERLER: That's what it's called now. Was that what it was called then as well?

ANDERSEN: Yeah, it was called that then, as well.

ZIERLER: Let's do a little history of this department. With your hire, looking back, was your hire part of a building mode of the department? Was it already well established at that point?

ANDERSEN: It was well established, and I was a target of opportunity.

ZIERLER: Do you know how far back the department goes or what its origin story is?

ANDERSEN: Hans-Lukas Teuber, an eminent neuropsychologist, founded the Department of Psychology at MIT in 1960. In 1986 the subsequent chair, Richard Held, a prominent perceptual psychologist, expanded the Department of Psychology by including neuroscience and computational neuroscience. This was the new Department of Brain and Cognitive Sciences with a new chair, Emilio Bizzi, who was one of the founders of the field of motor neurophysiology.

ZIERLER: What can be read into the name of the title, Brain and Cognitive Science? Did it really aspire to be sort of an umbrella kind of department?

ANDERSEN: It was great, a revolutionary mix of brain scientists across many subfields and one of the first departments of its kind.

ZIERLER: Tell me about starting up the lab there. What elements did you bring with you from La Jolla, and where was there opportunity for new instrumentation, new research questions, just because you're in a new environment?

ANDERSEN: It was a bigger lab at MIT than at the Salk. A graduate student that had just joined my lab came with me, Martyn Bracewell. At the Salk I had a very small lab; the others had graduated. It was a nice space. In terms of instrumentation, I was able to get more recording setups, but we were doing many of the same behaving monkey experiments we were doing at the Salk. I think at MIT, the real plus was that I began collaborating with people in the AI/neuralcomputation group.

ZIERLER: Were people using the term "AI," artificial intelligence, back then?

ANDERSEN: That's right, yeah, and they meant something totally different. [laughs]

ZIERLER: That was my next question. What did they mean? What did that mean back then?

ANDERSEN: They meant symbolic. The brain was thought to be like a computer, and cognition was the software and by understand the software you could understand cognition. There was also a computational component exploring motor control. As I mentioned, the head of the department was Emilio Bizzi, a very famous motor neurophysiologist, and he attracted theorists in motor behavior. On the West Coast, where I was, at UCSD, they were the early developers of neural networks. They were referred to as "connectionists" and the more symbolic theorists were the "artificial intelligence—AI" group. There was a debate around language. The AI school thought that it should be handled symbolically.

ZIERLER: As opposed to what?

ANDERSEN: As opposed to the approach from the UCSD connectionists, and also Terry Sejnowski, who was at Hopkins at the time but then went to the Salk. They trained neural networks, that began with no intrinsic structure, and these networks could learn the rules of language. MIT was the center of speech for a long time, notably having Noam Chomsky on the faculty. The controversy was whether your brain just innately learns the grammatic structure due to prewiring, or through learning. The psycholinguist Steve Pinker was there at MIT at the time—and he proposed that language in humans was an innate instinct.

ZIERLER: Did you jump into that debate?

ANDERSEN: No. [laughs] I understood it. A test case in English was the past tense of verbs. Pinker argued that children learn the rule for regular verbs, and they memorize the irregular verbs. With the success of large language models there is less concern now about whether models are neuromorphic, i.e., use the same general principles to those used by the brain. One roadblock to using neural networks to model brain function is you often don't know what the networks are actually doing. When we did the network simulation of coordinate transformations and found gain fields, we could figure out how the neural network was using a general principle for computing the transformation. But for more complex tasks, it's kind of a black box. Using neural networks as a tool to understand the brain can have shortcomings. That being said, a lot of the decoders that we use for neural prosthetics, which use machine learning techniques for training, have neurally realistic components, and that may be why they work well.

ZIERLER: For all of the reasons that might have compelled you to stay at the Salk Institute—it was a great environment, you were doing great work—you seemed to suggest that when you got to MIT, your lab was able to scale up, that you were able to just get better. That really begs the question, why was that necessary for your research agenda? Why did the lab need to grow?

Focus on Visual Systems

ANDERSEN: I went to MIT because at the time it had all these systems neuroscientists and computational neuroscientists. It also had Peter Schiller. I was doing vision, then. He was a leader in that field.

ZIERLER: Tell me about Schiller's work. Why was that relevant for you?

ANDERSEN: He works on the visual system, and like the motion work we were doing, is very visual system oriented. There was Bizzi, working on motor control. They were two leaders in those fields, so that was pretty good. Or, very good. [laughs] The cognitive component, they had Sue Corkin, who worked with H.M., the patient who couldn't form long term memories There is a famous neuroanatomist, Ann Graybiel.. It was a real powerhouse. It was much more diverse in faculty interests than the Salk. Not to put the Salk down in any way; it's just smaller and was more oriented toward, at the time, molecular biology, and not necessarily systems neuroscience.

ZIERLER: Administratively, could you have grown at the Salk? Could you have scaled up the way that it was so easy to at MIT?

ANDERSEN: Yeah, I think so, because they were going to have me be the director of a vision group.

ZIERLER: Did that happen without you? Was there a successor that did actualize that?

ANDERSEN: Yeah, Terry Sejnowski, who came from Hopkins. He's a famous brain theorist. I guess, too, it's interesting to have a change. [laughs] The East Coast no longer became that mysterious, although I had been at Hopkins already.

ZIERLER: Speaking of change, you're going from this tiny, academic community to the behemoth that is MIT. What did that mean for you as a new professor at MIT? How did things change for you?

ANDERSEN: Not too much. I certainly had more collaborators, and great students—at the Salk, they didn't have graduate students. Also it was an attractive place for getting postdocs. It's very important to be at a major research university or center, because then you get good students and postdocs. They are attracted both to you but also to the place.

ZIERLER: Last time we concluded where you provided a very nice summation of all of the things you had done at the Salk up until that time. What stayed with you? What were ongoing projects that you continued that you brought with you to MIT?

ANDERSEN: I brought the motion work. A lot of that, at that stage, was psychophysical, meaning it's designing stimuli that people look at, and that you can measure what they're seeing, so it's perceptual neuroscience. We collaborated with Ellen Hildreth, who was a theorist there, and with graduate students Stefan Treue and Masud Husain.

ZIERLER: What do those experiments look like?

ANDERSEN: I should say that Stefan is now the head of the Primate Center in Germany, and Masud is at Oxford, so they became very successful. The experiments consisted of seeing structure from motion. So, there are just a few dots on a screen, and they're moving—like two surfaces, and there's no depth cue. But your brain uses that motion signal to see a rotating cylinder, if you make the dots slow at both ends and fast in the middle. It's ambiguous, kind of like a Necker cube, so it flips back and forth.

ZIERLER: The subject is narrating what they're seeing? That's how you get the data?

ANDERSEN: Right. What we did was design experiments to look at perceptual thresholds or other kinds of psychophysical measures. One thing we found was that these dots form perceived surfaces. You could reduce the number of dots quite a bit and you could still see the surfaces. You could introduce new dots and they'd become part of the surface. That led to a series of monkey experiments. Collaborating with Ted Adelson, who is a a theorist at MIT, and postdoc Ning Qian, we showed that the first visual area V1 doesn't construct the perception of depth from motion. But a subsequent area, the middle temporal area (MT) does. The nice feature is you can measure people's perception by which way the cylinder rotates, and you can bias it by adding disparity to the dots. Because you can look at these stimuli with one eye and it still looks 3D, but if additionally, the dots have disparity cues and are viewed with two eyes, then the direction of rotation is unambiguous. We found that V1 only extracted the direction of motion. MT had inhibition to push the planes apart. This research continued here at Caltech so the structure from motion experiments kept going across institutions. At the Salk we had just discovered this intent signal, so we worked on that. In fact, we discovered this area, the lateral intraparietal area (LIP) that appeared to be specialized for saccades, which are rapid eye movements. At MIT we examined LIP in great detail to establish that it was specialized for saccades. We examined how LIP processed eye movements using recordings, saccade deficits with its inactivation, and stimulation to produce eye movements.

Microelectrodes to Measure Planning

ZIERLER: How do you stimulate it?

ANDERSEN: In experiments with Peter Their, we stimulated with low currents through tiny microelectrodes. We thought then that the area was for saccade planning, in other words, intent to make eye movements. Of course at that time most people were thinking about attention, so we began to do experiments to separate intention from attention and kept those experiments going here as well.

ZIERLER: How did the experiments look different to separate the ideas?

ANDERSEN: The most basic one is just to determine effector specificity. In the course of experiments at Caltech, we found that the other bank of the intraparietal sulcus had a reach area, and so we could show that it was only active when reaching to a target, whereas the eye movement area—LIP--was only active when making an eye movement to a target. If it was just attention, it would be the same, because you attend to where you reach, and you attend to where you look. Then with EunJung Hwang we inhibited the reach area, and it affected reaching but not saccades. In a series of experiments with Vasilios Christopoulos, Igor Kagan, and Melanie Wilke, we found the reverse is true for LIP inactivation. I'd say pretty much every experiment led to the conclusion of a strong representation of intention in PPC. Now, with the human work, it's very obvious, since there's so much effector specificity and intermixing of effector specificities, so there's no focus on attention with our current human studies.

ZIERLER: What's the bigger takeaway? What does this tell us?

ANDERSEN: I think it says that we plan things earlier than we think. There is a motor theory of attention which posits that attention is really motor planning. However, I don't think all attention is motor planning. I began reading the attention literature when I was at Hopkins because we were doing some attention experiments. William James wrote in 1890, "Everyone knows what attention is." But actually, it meant different things to different investigators. Most early studies defined it as a filter for further processing of locations or features of interest, highlighting this information from the immense amount of information coming into the brain. That's true for vision, audition and somatosensation, so it's very sensory. Later works began to expand the domain of attention to include motor components. Attention became so popular that researchers were including motor plans as attention, so-called motor attention, and even eye movements were conflated with attention, which obviously eye movements are motor. [laughs] I guess the takeaway is that it's hard sometimes to take psychological terms and apply them directly—even though we do—to complicated neural mechanisms.

ZIERLER: Absent psychological terms, what terms would be more appropriate or would be more descriptive of what you're looking at?

ANDERSEN: I'd say the intent was planning. We could show that the planning could appear and then be cancelled. I'm sure there is attention in the parietal complex, because lesion patients have attentional deficits. I just haven't studied it. Others have. [laughs] It's kind of a sensorimotor area, so it will have both sensory and motor related activity.

ZIERLER: What were other projects that you carried with you from the Salk to MIT?

ANDERSEN: Those were the major ones, I think. Oh, we did some more modeling—at the time, it was believed that backpropagation, which was used to train neural networks, was not biological. We worked with Michael Jordan—not the basketball player, but he was a theory professor at MIT—and he had worked with reinforcement learning, and we could show that reinforcement could train just like backprop. So, you could make it biological.

ZIERLER: What does that mean to make reinforcement learning biological? What are the other options?

ANDERSEN: Reinforcement learning examines how you learn by trial and error. It's one of the mechanisms the brain uses for learning, even though you can put it in computers.

ZIERLER: But just to call it biological means what exactly?

ANDERSEN: That it's biologically plausible. The way backprop worked, you had to send error signals back through the network, and there's no easy to do that with neurons. But you could send reward signals back by a separate pathway that would change the strength of synapses. It's a broader signal, so it was thought to take longer for the network to converge on a solution. A lot of the projects that I brought here with me were partially done at MIT and continued at Caltech. I think I've covered it all.

ZIERLER: The opportunity to work with top-flight graduate students at MIT, this is obviously a new opportunity for you. What were some of the benefits there? Being able to attract excellent graduate students, what did that allow you to do that might not otherwise be possible?

ANDERSEN: Well, you can't do everything yourself. Also, they come in with new ideas and it provides a good intellectual environment.

Monkey Brain Surgery

ZIERLER: What were some of the new ideas? How did the graduate students change the direction of the lab?

ANDERSEN: I think we kept the lab going in the same direction, but there were aspects of day-to-day experiments that changed, as well as forks in the road based on results. Also, at MIT, I was doing all the surgeries. That was wiping me out after a while. So, when I got here, the lab got big enough that other people could do the surgeries.

ZIERLER: The surgeries means what, exactly?

ANDERSEN: It means putting on caps, recording chambers, eye coils for measuring eye position. Those are the main surgeries we do.

ZIERLER: Why is it called surgery?

ANDERSEN: Because it's a surgery. [laughs] The animals are anesthetized. You use all the same sterile techniques you use with humans. For the recording chambers we do a craniotomy, and then also you put on a head cap, which is a way of stabilizing the animal when you're recording.

ZIERLER: You must have truly felt like Dr. Andersen when you were doing these surgeries. [laughs]

ANDERSEN: Right, yeah. [laughs] But now I watch the human surgeries, for our participants, and actually, sometimes I feel almost like I could jump in. But then I know that would be a disaster. [laughs]

ZIERLER: Is there formal training to conduct the surgeries? Are you self-taught? Obviously you don't go to medical school. Where do you pick up these skills?

ANDERSEN: I picked it up at Hopkins. Mountcastle was trained as a neurosurgeon, and so that was good training. Then at the Salk and MIT, I would train people. When I got here, I did some of it in the beginning, but then less and less. OLAR—Office of Laboratory and Animal Research—started to grow here, and now they and my lab manager—do most of that training for the surgical procedures.

ZIERLER: Presumably there's more protocols than there were 20 or 30 years ago?

ANDERSEN: There's more regulation, but I don't think the actual procedures have changed much. We have to be sure the animal was healthy and didn't get an infection, so pretty much we always use the same standards as humans. Actually, one or two of the human neurosurgeons who have helped us with more complicated procedures have remarked that monkey surgery is very difficult because the animals are so small. So in some ways it's challenging and maybe more so in terms of physical scale compared to humans.

ZIERLER: Is the goal to get the animal subjects back to a healthy state, or are they euthanized after the surgery, after the experiments are done?

ANDERSEN: No, no. They are usually here five or six, seven years, and then we take the pedestals off and we retire them to animal sanctuaries. We give funds to the sanctuaries and they build enclosures for the animals and they are outdoors most of the time. In our case, for our monkey experiments, there is no need to euthanize the animals. Because we use non-invasive imaging, we don't have to anatomically reconstruct the recording sites.

ZIERLER: You mentioned some important collaborations you were able to pursue at MIT with some of the professors there. What were some of the key collaborations you did at MIT?

ANDERSEN: There was this series of papers on transparency in motion with Ted Adelson. There was the structure for motion with Ellen Hildreth.

ZIERLER: Let's start first. Let's go one by one. Transparency in motion, what was the research there?

ANDERSEN: It's kind of like how if you walk by a car and you see the reflection, but you can look in, too, so you can separate the two. That was in part the structure for motion experiments, but we began simple with just two transparent planes. The computational component was—Ted Adelson had developed this motion-energy model that was very influential, so he applied that to this idea of a two-stage visual processing stream. The structure from motion was the interpolation of surfaces. Prominent theorists like Shimon Ullman and Tommy Poggio, and others, were around, and so we were always within that community so it rubbed off on us a bit.

ZIERLER: The department was a really dynamic place. It was exciting to be there?

ANDERSEN: Yeah, yeah.

ZIERLER: Just on a personal level, how did you enjoy the relocation to Massachusetts?

ANDERSEN: It was fine for me. I mentioned my wife; she passed away six years ago now, but she was a Californian, so if there was an opening that came up in California, she wanted me to pursue it. [laughs]

ZIERLER: [laughs] I see. So you kind of went to MIT not thinking that this might be a permanent proposition in any regard?

ANDERSEN: I didn't think about it. I think what triggered the possibility of moving was, back then—unlike today, where people get tenure quite a bit, at MIT maybe one third of people got tenure—and it may still be strict. Harvard was another example. The idea was junior faculty could always go out and get a job somewhere else if they didn't make it at the most prestigious universities. So it was routine that you would go out and get—this may be true today, too—that you go out and get other job offers to improve your chance of getting tenured at your current university. I got an offer from Rockefeller and one at Caltech.

ZIERLER: This was around the early 1990s?

ANDERSEN: Yeah, and Harvard was also interested, but I didn't pursue it.

ZIERLER: Were there collaborations with Harvard just being so close when you were at MIT? Were they doing interesting things for you?

ANDERSEN: Yeah. I had friends there both at the medical school and on the campus. On the campus, Steve Kosslyn and Ken Nakayama. Steve was a cognitive scientist and Ken was a psychophysicist. At the medical school, there was Marge Livingstone and David Hubel. Hubel, very famous, won a Nobel Prize for working on the structure and plasticity of the visual cortex. We had one direct collaboration with Roger Tootell on the anatomical structure of the middle temporal area for motion tuning.

ZIERLER: What were some of the computational advances in the 1980s and into the early 1990s? What were some of the things you were able to start doing with computers that you couldn't do before?

ANDERSEN: This would be totally from my perspective, which is not as a computational neuroscience or theory guy. In the 1980s was this beginning of neural networks, and for me, it allowed me to understand that a single neuron's response properties is just one little piece of the puzzle, and that you have to look at a population of neurons to understand what they are doing. Then, of course, I mentioned before that just from a practical point of view, you needed computers to run the experiments. We've collaborated a lot with computational people, and there was a phase where they helped greatly with data analysis. They brought applied math to our own data, so that we could see it in different ways and quantify it better. Also, there is machine learning. I didn't realize it until recently, that it has become so popular, but we have been working on decoders for a long time that use machine learning.

ZIERLER: What does that mean here, decoders?

ANDERSEN: You train, in a sense, computers to recognize patterns in neural activity, and those are tagged with particular behaviors. You use machine learning to do this. A lot of the terms and techniques that are in machine learning, we use; we being the community of brain-machine interface researchers. But I'd say in the early 2000s the decoders were very simple., Then they got more complex, sometimes too complex. [laughs] Today we use techniques for improving and extracting signals. Also, we've been looking at hybrid systems—like you have a smart robot that takes signals from the human and interprets them so the human doesn't have to do everything. We have a driving project where our participant, located in Southern California driving a car over the internet in Michigan [laughs]. In that case, you're getting into semi-autonomous driving. We collaborate with two companies, Blackrock Neurotech and Ford Motor Company.

ZIERLER: When you were at MIT, was Cambridge, was Boston, already becoming a biotechnology hub? Was that already in train at that point?

ANDERSEN: Yes, yes. I noticed all these buildings going up. [laughs]

ZIERLER: Was that relevant at all for you? Were there collaborations to pursue?

ANDERSEN: What I was doing at the time was basic research that did not have an easy path to commercialization.

Caltech and the Pull Back to California

ZIERLER: As prelude to your decision to move to Caltech, from Hopkins, or from the Salk, where did Caltech neuroscience, neurobiology loom in your professional world? Did you think about Caltech? Did you track some of the work that was happening here? Did it not register because there really wasn't much happening here? I wonder if you can just give the backstory there.

ANDERSEN: Caltech was one of the hotbeds for neuroscience. It had a number of prominent neuroscientists.

ZIERLER: Who were some of the big names in your field, before you arrived?

ANDERSEN: David Van Essen, we're currently sitting in what was his lab. John Allman, who is still here. David Anderson. There was Jack Pettigrew, who went to the University of Queensland, Jim Hudspeth who is now at Rockefeller. Gilles Laurent and Erin Schuman, who are in Frankfurt and Christof Koch, who is now at the Allen Institute. People move around a bit. But it was quite a concentration. It was very similar to MIT. I think that's why it was appealing, because it was like MIT but on the West Coast.

ZIERLER: Doesn't get any better than that! [laughs]

ANDERSEN: It was a hard decision.

ZIERLER: Your late wife, was she working the phones in the background? Was she looking at job announcements?

ANDERSEN: No, not really. I guess she just trusted me, that [laughs] I'd do the right thing.

ZIERLER: What was the initial point of contact? Who made the overture to you from Caltech?

ANDERSEN: Let's see. I think it had to do with Van Essen leaving. I think I mentioned it to friends here, and that's how it got started.

ZIERLER: Tell me about his lab. What was he involved in?

ANDERSEN: He is a neuroanatomist and neurophysiologist, and he is also organizing data from many labs into an open database. He has since also included brain imaging into a human connectome. Van Essen is known for determining a lot of the connectivity of the visual pathway, so he was a visual neuroscientist and neuroanatomist.

ZIERLER: Had you spent time here? Did you ever come here to collaborate, either from La Jolla or from MIT?

ANDERSEN: Yeah. It was when I was at the Salk. Of course we had this Helmholtz Club. I did some research with John Allman, When I was at the Salk, I gave a couple seminars here. So, I was pretty familiar with Caltech.

ZIERLER: Tell me about your work with Allman. Was this his spider monkey research?

ANDERSEN: No animal studies, primarily discussion. Just I recall informally trying out a new technology that he had built here, which could produce a variety of visual stimuli including random dot displays that could separate into figure and ground depending on depth and motion direction. Also he came down to visit me in La Jolla, too, when I was at the Salk. He went on to—but not with me—using his new stimulus device to investigate what he called non-classic surrounds. It's the fact that we usually think of visual receptive fields as being small with an excitatory core and inhibitory surround. He showed that stimuli far outside of the classical receptive field could modulate cortical neural responses. He termed these non-classical receptive fields.

ZIERLER: It was an easy decision for you when the offer came through?

ANDERSEN: Yeah. Then MIT made a counteroffer, which made it a little more difficult, but for Carol it was [laughs]—not difficult for her.

ZIERLER: Tell me about starting up at Caltech. What was that like for you? Did you basically transport your entire lab? Did you bring most of your graduate students with you?

ANDERSEN: Yes, I brought a big group at that time. When I came from the Salk, I think I just had Martyn Bracewell. I didn't think many people would come to Caltech from MIT. That was around the time of the fires up here, the first bad fires, and the Rodney King riot I thought, "Oh, no."

ZIERLER: It was not a pretty moment for Los Angeles.

ANDERSEN: Yeah, so I was a little worried, but essentially the whole lab came over.

ZIERLER: You came with the named professorship? That was part of your arrival package?

ANDERSEN: Right. Also with the startup funds I did get to go in new directions. I got a vestibular chair so I could study the vestibular system.

ZIERLER: What is the vestibular system?

ANDERSEN: It's your sense of balance and motion of your head in space. If you're chronically dizzy, there might be something wrong with it. There are two subsystems. The semicircular canals are fluid filled, each with a sensory organ called the cupula. The semicircular canals sense the rotation of the head. Then there are the otolith organs, which sense linear acceleration. We had a chair that could rotate a monkey in three dimensions, or a human in two, so we did some monkey and some human experiments with the vestibular chair. As part of the setup we also got an anechoic chamber. That's a kind of chamber where the walls absorb all the echoes. So sounds made in the anechoic chamber are like sounds generated in an open field. Since there are no reflections, this kind of chamber is good for sound localization experiments. Earlier I forgot to mention Mark Konishi who was a neuroscience faculty colleague and whose lab was on the same floor as my lab. He was an auditory expert, so that was very appealing. I was able to buy these very expensive pieces of equipment with the move that enabled us to expand our research to the vestibular and auditory systems. Also, it was a large laboratory and we had six monkey setups. The lab got to a fairly big size at one point. [laughs]

ZIERLER: What was the zenith? How many students did you have at your biggest?

ANDERSEN: I don't know, maybe 25 or 30.

ZIERLER: Wow. Especially by Caltech standards, that's big. That's a big lab.

ANDERSEN: Yeah, but there are labs like that here. It certainly is a challenge to follow everything that's going on.

ZIERLER: With Caltech not having a medical school, did that give you pause? Was that a concern for you?

ANDERSEN: Not so much. It probably should have [laughs], but MIT doesn't have a medical school, either, and it was easy enough to get to know people at MGH and at Harvard Medical School. Herer we collaborate with UCLA and USC.

ZIERLER: You were able to forge those connections?

ANDERSEN: Right.

Neuroprosthetics and the Origins of BMI

ZIERLER: Did your research questions change, working with new hospitals? Was it the same thing that you were after at MIT?

ANDERSEN: Although I thought of neuroprosthetics as a new direction, it wasn't until I got to Caltech that I started pursuing it in earnest.

ZIERLER: What were your ideas? How did you go about pursuing this?

ANDERSEN: We started developing population recording. We developed an electrode array ourselves in collaboration with a small company, Microprobes for Life Science, and they worked pretty well. They are called Floating Microwire Arrays (FMAs) and you can still buy them. A postdoc in the lab, Sam Musallam, did the main development from our side. Also this other company, Blackrock Neurotech, was producing arrays originally designed for visual cortex, so the spacings are very tight between the electrodes. We started to then train monkeys to do brain control. They'd be reaching, and then after a while, we'd decode their intent to reach. We'd reward them based on their neural signals, and they would learn, pretty quickly, that they didn't have to actually reach, just think about it. For both regulatory and scientific purposes it was important to show that we had animal studies, and that this technology would work. At the time, we were targeting an early stage in the sensorimotor cortical pathway, the posterior parietal cortex where the plans for movements are made. We showed that that would be a promising place to record brain signals for neuroprosthetics.

ZIERLER: Administratively, the Brain-Machine Interface Center, this is not part of the beginning of your time at Caltech? This develops later?

ANDERSEN: That was 2015, much later.

ZIERLER: What are the antecedents? What is the work that you're doing that naturally flows into the formalization of the Center?

ANDERSEN: Late in the 1990s, there were three or four monkey labs that had the same idea of developing a cortical neuroprosthetic.

ZIERLER: Beyond Caltech?

ANDERSEN: Beyond Caltech. Immediately I had competitors. [laughs]

ZIERLER: What were the ideas?

ANDERSEN: The idea was that, if you were paralyzed, you could think about moving, and decode that intention to generate control signals to operate computers, robotics, and other assistive devices.

ZIERLER: This is sort of like a multiple independent discovery kind of idea?

ANDERSEN: I'd say so.

ZIERLER: What is happening, what's out there in the ether, that is making these different labs come up with the same idea?

ANDERSEN: Good question. Maybe it was obvious.

ZIERLER: But there needed to be conceptual breakthroughs, technological breakthroughs, that allowed the similar ideas to develop independently?

ANDERSEN: Certainly the population recording was important. I think the first out of the box was Nicolelis at Duke, using animals and population recording. Also at Brown, John Donoghue started a company for neural prosthetics that later became Blackrock. In the early 2000s, a few high-profile papers came out from the Brown group, first in monkey and then human. Andy Schwartz at the University of Pittsburgh showed monkeys could control robotic limbs with a BMI. We demonstrated brain control in monkeys using PPC signals. Then during the 2000s, there were a few more human implants, not many. We started around 2013 with human implants. That's when Tianqiao and Chrissy first became aware of our lab when we published the results of brain control using human PPC implants. In the beginning, around 2000, these were not the typical studies funded by NIH—now, we are supported primarily by NIH.

DARPA and Robotic Limbs

ZIERLER: What was it then?

ANDERSEN: Then it was DARPA. They supported several of us. Because it was one of those high-risk, high-reward [laughs]—studies.

ZIERLER: High risk, high reward.

ANDERSEN: [laughs] Right. They kind of kickstarted it.

ZIERLER: Was there a national security component to this, or this is just DARPA funding basic science because they can? The latter?

ANDERSEN: The initial contract was to fund multidisciplinary groups that meld electronics, biology, and theory. There was no particular agenda outside of the multidisciplinary requirement, but three of the six teams that were funded were BMI groups. The next round was called Revolutionizing Prosthetics. The rationale was—this was during the Iraq War—that people were coming back with amputations and neurological injuries. A component of the program was to develop a human-like robotic upper limb and brain control for using the limb. Gradually—as is the DARPA style, they start a new type of technology and then move on to something else. By that time BMI research was established enough to be attractive for support from NSF, NIH, and private foundations.

ZIERLER: When you came to Caltech, this was right around the time when Carver Mead was getting interested in neuroscience and neurobiology. Did that register with you? Did you see opportunities to collaborate with somebody coming from a physics, an electrical engineering perspective?

ANDERSEN: Sure. Carver was another attraction to moving to Caltech. He was doing neuromorphic engineering. Some of his former students organized a summer school in Telluride that I would attend and give lectures.

ZIERLER: Did you ever talk to Carver, how he got interested in the neuromorphic research, where that came from for him?

ANDERSEN: Yeah, early on, he would have these lunches at the Athenaeum, so several of us would go to that. Humans were performing much better than machines, so the idea was to look to humans—I can't tell you what he was thinking [laughs], but that was a motivation for some in the field. How do humans do it, maybe we can engineer what the brain is doing for improving performance? Now we're beginning to wonder if machines might be better than humans for a lot of tasks.

ZIERLER: Oh, interesting. Sort of crossing the Rubicon.

ANDERSEN: [laughs] Yeah.

ZIERLER: Did that get you more involved in instrument-building? Did your lab take on more of an engineering component to it?

ANDERSEN: We collaborate with engineers, but I'm not an engineer.

ZIERLER: Was the division BBE already, or that came later?

ANDERSEN: That came later. But I find that a lot of my collaborations are with engineers, and that has become very important for this kind of work. In the beginning, we collaborated with Joel Burdick and Yu-Chong Tai for some of the recording technology.

ZIERLER: What recording technology? What were you recording?

ANDERSEN: We were trying to develop electrodes, and Joel was automating the electrode advancers. We are currently collaborating with Mikhail Shapiro on a functional ultrasound BMI, That has been a recent engineering effort and has been very successful.

ZIERLER: I wonder if your perspective is part of the larger decision to go from the Biology Division to Biology and Biological Engineering. What is your perspective on that not just name change, but substantively, what were the trends that prompted the Division to change its name even if its identity?

ANDERSEN: That didn't involve me. The best person to really ask is Steve Mayo. As far as I understand, there was a group of biologically-oriented engineers, in Engineering, and they moved over to the Biology Division. That was great; we picked up several new colleagues.

ZIERLER: Obviously you're not an engineer, but your research did become more dependent on engineering-type collaborations, over the years?

ANDERSEN: Yeah.

ZIERLER: What's the idea there? That you need more sophisticated instruments?

ANDERSEN: Yeah. The BMI is a medical device. From my end, I am looking at the science, and that can then inform what sort of technology is required. In the lab I have a medical engineering student, a couple of bioengineering students, always had a few students with engineering backgrounds. It's kind of a synergy where you can learn new things about the brain, and then you can design technologies to harness the new findings. Probably our next steps will be—since we've been beginning to work on language— to meld language related activity with large language models. Oh, and self-driving. Richard Murray and I share a student working on BMI driving.

ZIERLER: I want to return to this competition beyond Caltech. How did your lab fare? What did the competition look like? Who got there first?

ANDERSEN: There's competition everywhere. [laughs]

ZIERLER: But on this one particular topic we were discussing.

ANDERSEN: Oh, the BMI? Well, that was kind of early on. There were only a few places doing it. Right now I think still only about 30 people have had implants worldwide, so it's still early, at this stage. Different groups are known for different things. We all have niches.

Higher Cognition and Machines

ZIERLER: What were your original research questions? What prompted you to get into the implant world?

ANDERSEN: It was studying posterior parietal cortex. I thought I could go for higher cognitive signals from there, which turned out to be true [laughs] We can decode many, many things.

ZIERLER: What was the theory or even the hunch that prompted you to think that higher-order signals were attainable?

ANDERSEN: From the monkey work. We could decode their planning. We could decode rewards they expected. Those were two of the main higher-order things we could record. And from studies of cortical damage in humans we know PPC would be important for spatial computations for motor behaviors and for navigation. These results suggest that PPC would be an important area for what is called internal models. They model what's you, and your relation to the world outside, so you form these models based on sensory inflow and motor outflow, to keep yourself updated about current conditions. It's an exquisite area, since you can tap into high level cognitive and motor functions. It seemed like a fascinating place to go.

ZIERLER: Were you an early adopter to the implant technology? Did you have colleagues or peers that were using this before your time?

ANDERSEN: Originally it was developed by Richard Normann at the University of Utah. That's why it's often called the Utah array.

ZIERLER: What's the chronology? How far back does this go?

ANDERSEN: Normann began developing the array in the mid 1980s. In some ways, it's an old technology, but it's FDA-approved. [laughs]

ZIERLER: Are these specialized implants? Are they used in other applications?

ANDERSEN: Yeah. The group at Brown acquired Normann's company. It was then for prosthetics applications, but they were also selling the arrays to animal researchers, so it fills both those needs.

ZIERLER: What can the data tell you? You do the surgery, you make these implantations; what does the experiment look like? What is the data telling you?

ANDERSEN: It's pretty fascinating, just to be there and see people controlling things by thinking about it. [laughs]

ZIERLER: Oh, so now we're not talking about monkeys; we're talking about human subjects.

ANDERSEN: Yeah, humans.

ZIERLER: Is this when you start to get involved with people who have suffered catastrophic injuries?

ZIERLER: That's the beginning of this research here.

ANDERSEN: Well, no, the beginning of the research, like you said, was with monkeys.

ZIERLER: What was the transformation to you? What prompted you to say, "We can take what we've done with monkeys and now go to human subjects"?

ANDERSEN: It was a gradual sort of thing. Well, maybe not so gradual. I thought it had occurred after I had given a talk in 1998, when reporters asked me, "Well, what's your work good for?" I said, "Oh, you could make a neural prosthetic." But then talking to friends, I realized that I had actually been thinking about it much earlier than that.

ZIERLER: That's what I'm trying to draw out of you now, the origin story here. What is that? How far back were you thinking about this, if only notionally?

ANDERSEN: As a researcher, you're always imagining what you can do.

ZIERLER: It's the next paper, essentially.

ANDERSEN: Yeah. Also, you can think about audacious things that could be possible. The fact that this works is amazing. [laughs] Also the idea itself has been around for a long time in science fiction literature.

The Path to Helping People

ZIERLER: The lines are always fuzzy between fundamental research and translational research, but can you draw, in your intellectual trajectory, in your research trajectory, when your motivations—maybe shifted is too dramatic a word, but for so much of your career, you probably gave zero to very little thought about how this might ultimately benefit humans or society. When did you start thinking that what you were doing in a basic science context really could ultimately be put to concrete use that helped people?

ANDERSEN: I always wanted to study humans. That goes back to graduate school.

ZIERLER: But studying humans is one thing; to do things that help humans is still a leap.

ANDERSEN: Yeah. I guess partly it was I had been studying the posterior parietal cortex of monkeys for so long, as a model for humans, so this was a chance to transition to humans. I was motivated by my father having polio, and then he had the post-polio syndrome.

ZIERLER: What did that mean? Was he not ambulatory?

ANDERSEN: Toward the end, right, yeah, his legs were weak. So, I had personal contact with the issue of a close family member having difficulty ambulating. Also, I thought it was partly I could show intention in humans, because they could tell me what they were doing. So many of our experiments were trying to indirectly find out what was the monkey was thinking of doing.

ZIERLER: But is there always an element of doubt, because obviously the monkey can't verbalize, or were you not concerned about that?

ANDERSEN: We designed experiments where you could guess the inner thoughts.

ZIERLER: You could make strong inferences?

ANDERSEN: Yeah, strong inferences. People make very clever designs for inferring perception and cognition. But being interested in awareness, consciousness, these sorts of topics, having someone tell you what they are thinking or experiencing—of course, a new problem I've come to find is that when people tell you something, you still don't know quite what—

ZIERLER: Sure, because there's biases, there's misapprehensions, there's different perspectives, so it's never a pure transmission of information from one person to the next, right?

ANDERSEN: A good case in point is silent language, silent speech. Sarah Wandelt and David Bjanes, who are the lead authors on that study; Sarah tells me she usually thinks in pure thoughts; she doesn't think in language. I am constantly thinking to myself in language.

ZIERLER: What are your other options? What else do we have?

ANDERSEN: She asked around, and there's a quite a bit of variability. It's only one aspect of one's inner life. Obviously you have perceptual awareness, that might be more basic across people. But this covert speech I think is probably different for different people. How is it for you?

ZIERLER: I'm not even sure I could verbalize it. [laughs]

ANDERSEN: [laughs] When you wake up in the morning, and you're thinking about what you're going to do during the day, are you talking to yourself?

ZIERLER: Yeah, and my wife sometimes catches me talking to myself, too. They say that's a sign of intelligence. I don't know; that might just be self-flattery. But I'm using language to talk to myself, right?

ANDERSEN: The non-verbal—not verbalizing it, but still talking to yourself in your head—

ZIERLER: Yeah, but I think that dialogue happens in English. I think.

ANDERSEN: It does for me, and it happens all the time. I wonder if it's different for people who are multilingual. Maybe they have—

ZIERLER: Right. That's fascinating.

ANDERSEN: Anyway, we can decode it. [laughs] We can ask people to have internal speech and decode what they're saying.

ZIERLER: I want to clarify. When you get involved with the human subjects, these are only people who have been injured? You're never dealing with healthy people?

ANDERSEN: No, no.

ZIERLER: My question is, are you ever dealing with injured monkeys?

ANDERSEN: No. That could have been a problem, right? [laughs]

ZIERLER: I'm just trying to understand, the transition from you saw how well these experiments were going for monkeys, and so you wanted to translate this to human subjects. But the disconnect here is healthy monkeys, injured humans, right?

ANDERSEN: Right.

ZIERLER: Conceptually, how do you translate the experiments so that you're continuing to make progress along a similar track?

ANDERSEN: We worried about that, if injury would change the brain. Merzenich, my mentor, and his colleagues showed that the somatosensory somatotopic map is very plastic in monkeys. So we were worried that—our first subject was ten years post lesion; did his brain reorganize? Our idea of a cognitive neural prosthetic was the brain was not that plastic and one should implant areas that normally have the specific functions one wishes to read out with a BMI. Early on in the BMI field, there was a question of whether you could implant any part of the brain and train it, because the cortex is plastic.

We're finding actually—when we first put the implant in, the two weeks we were waiting for the subject to recover, and record, we were thinking, "Oh, we hope this works." [laughs] And it did. It was fine. Immediately our participant could use a robot arm with brain control. Off the bat, he was able to shake a student's hand with the robot hand. So there's a lot of innate structure that remains even years after the accident. So that was great news. The same with our stimulation experiments where the participants have felt nothing for years in their limbs. We have stimulated in the part of the brain that interprets touch, the somatosensory cortex, and they feel again. Moreover, the map of the body in somatosensory cortex is not altered. These findings fit, very much, then, with this cognitive prosthetic idea. For example, you wouldn't want to go to visual cortex for language; you'd want to go to a language area. Because the brain still maintains a lot of its intrinsic connectivity.

ZIERLER: Even if there isn't anything for it to talk to?

ANDERSEN: Yeah. People that have had deficits for years can right away use this technology. In that case, it's maybe good the brain is not so plastic [laughs] in an adult. It may be different in adolescents.

ZIERLER: Richard, we're coming up on our time today, so my last question, I want to make it sort of an institutional question for Caltech. Next time, we'll take this research to see, first of all, how Tianqiao and Chrissy became aware of it, and how this blossomed into this amazing Institute that we now have at Caltech. But for you, when you made this decision to transfer or to move into human subjects—Caltech's institutional character is so fundamental, but in recent years it has become so much more involved in the betterment of society and things like that. By the time you arrived, for biology, was that transition already well underway? Were there biologists who were really thinking about translational work and thinking about applying basic science to society? Or do you think that what you were doing was part of that larger trendline to what we see today?

ANDERSEN: There always have been faculty doing that. An example would be Lee Hood.

ZIERLER: Although Lee Hood left, famously. That's sort of the basis of my question.

ANDERSEN: Yeah, I didn't know him.

ZIERLER: He left a few years before you arrived.

ANDERSEN: But certainly, a lot of the engineers have companies. Some of the biologists have companies. For me, the big difference was, when I got here, almost no one was working on humans—like the Caltech IRB, they had to kind of start up again when I arrived. It was new, working with humans in clinical trials coming out of Caltech and I'm really fortunate they let me do it. [laughs] I think in part because it is a good culture, and people were supportive of human studies.

ZIERLER: Who do you credit with that, when you made this transition? Who was supportive for you?

ANDERSEN: Paul Jennings.

ZIERLER: He was provost at the time?

ANDERSEN: Yeah, provost. Then in the regulatory part, there was Grace Fisher-Adams and Chantal Morgan D'Apuzzo. They were attorneys with the Office of General Council at the time and were very supportive.

ZIERLER: Really, this was a new direction not just for you but for the Institute?

ANDERSEN: Right. I think right now, we have the only clinical trials going. We have two. I imagine there will be more.

ZIERLER: I hope so! [laughs] Richard on that note, we'll pick up next time when you actually start with the human subjects. We'll see where the story goes. Wonderful.

[End of Recording]

ZIERLER: This is David Zierler, Director of the Caltech Heritage Project. It is Thursday, October 12th, 2023. It's wonderful to be back with Professor Richard Andersen. Richard, once again, thank you so much for having me.

ANDERSEN: Thank you.

ZIERLER: We're going to pick up just where we left off last time, when you've made the commitment to work with human subjects. If you could orient me in the chronology, also the pace of when you first thought about this to when you actually did this—what was the year, roughly, when you really committed to this new avenue of research, of actually working with human subjects?

ANDERSEN: I'd say it was around 1998.

ZIERLER: When was the first patient that you worked with? What was that lag?

ANDERSEN: It was 2013 [laughs], so it took a while.

ZIERLER: Wow, 15 years!

ANDERSEN: Yes.

ZIERLER: Let's cover what took so long. Perhaps let's start with the regulatory framework. What might have been difficult or longer than you had anticipated from an FDA or a government perspective?

ANDERSEN: We of course began looking into regulations early on. It was a long and sometimes circuitous path and the regulatory component was just one part. We had to find a hospital to collaborate with. We were also doing animal studies to show that the neural prosthetic would actually work by seeing if monkeys could use brain control to move computer cursors. These were healthy animals. Then of course there was the need to find funding for this kind of work. The first hospital that we engaged with was a high-end community hospital where a friend of mine is a neurosurgeon. However, it was going to be expensive and there was a lack of infrastructure for research. It was becoming apparent that we'd need to go with a major research hospital.

ZIERLER: From the Caltech perspective, had anyone worked with human subjects before at Caltech?

ANDERSEN: Yes, people had worked with human subjects before.. One example is functional imaging. Roger Sperry worked with split brain participants. A number of Caltech faculty in engineering and biology have developed either medical devices or drugs, but my understanding is they've done it through clinical studies at medical schools or companies.

ZIERLER: Whereas you were basically doing this on your own?

ANDERSEN: Not on our own, but in this case, it was the first time that Caltech had supported a clinical study itself, rather than, say, working through another medical school or other institution. That was a first for Caltech.

ZIERLER: Let's say you had stayed at MIT, for example. Do you think it would have taken as long as it did, just from the infrastructure, the research infrastructure at MIT? In other words, what aspects of your work at Caltech were you really inventing the wheel because Caltech had never been involved in anything like this before, where MIT might have, for example?

ANDERSEN: I don't think it would be any different. Neuroprosthetic research is currently being done out of the Massachusetts General Hospital, so that would have been a possibility. Although Caltech doesn't have a medical school, a lot of faculty collaborate with medical schools here in Los Angeles, and we have MD-PhD programs as well. There were three surgeons that we ended up working with, two from USC and one from UCLA. I had an MD-PhD student, Brian Lee, who then went to USC and is now an associate professor there in neurosurgery. At the time, he was doing a residency at USC, and his attending mentor was Charles Liu. He's a very high-powered research-oriented neurosurgeon, Also then someone who was new on the faculty at UCLA, Nader Pouratian, contacted me and came and visited me here, who was also, in the beginning, interested in doing research with recordings from patients with Parkinson's disease. With Charles and Brian, we went more directly to the BMI. While we were doing the monkey studies, we got funding from DARPA, because they fund things that are high risk but high reward.

ZIERLER: I want to ask, on the funding question, the late 1990s, early 2000s, there's so many converging areas of investment opportunity. There's biotechnology. There's the dot-com boom. Just the idea of heralding research where you can control things physically with your mind, were there a lot of people in the startup and investment space who were interested in this?

ANDERSEN: That's true. People contacted me even beginning around early 2000. But it was too far off. Usually you begin a startup when you are close to a product—

ZIERLER: The incubation period was too long in terms of what you thought was possible? That means government funding is really most appropriate?

ANDERSEN: Right. There was a company that started around 2002, Cyberkinetics. It was out of Brown and one of the founders, John Donoghue, was a monkey neurophysiologist researching motor control. They were able to publish the first implant in humans in 2006. That was a big deal.

ZIERLER: Did you feel like you were in a bit of a race at this point?

ANDERSEN: It's competitive, yeah. There was, early 2000s, the lab at Duke that was implanting monkeys. They never got to humans. There was one at Pittsburgh, who was again led by a monkey neurophysiologist, Andy Schwartz. The Pittsburgh group is still going, although I think he is more doing non-human primate work again. Yeah, there were certain goalposts to be achieved, and each one sort of guaranteed a Science or Nature paper. The company that Donoghue founded, Cyberkinetics, ran out of money. Looking back at it, I think they were still doing research, which is great, but that doesn't work too well in a business model. But they then became what's called BrainGate, which is a big consortium of several groups across the country. We implanted monkeys in posterior parietal and premotor cortex, published in 2004 in Science. Until then most groups implanted motor cortex. We showed that the monkeys could tell the value of targets and we could decode their intent before they moved, what they planned. These findings established the idea of a cognitive neural prosthetic, in which you could read out more complicated things than what you would find in motor control.

ZIERLER: Would you say that this was the pivotal proof of concept that put you on the path to working with human subjects?

ANDERSEN: Yeah. Our approach was, although we do implant motor control, too—use it often to compare with other areas—it put us on a path to implant other high-level cortical areas in the brain, and to look at cognition and how decoding and understanding cognition could be used to improve prosthetic applications.

ZIERLER: Your lab at this point, are you geared toward ultimately working with people? Would you have been doing this research even if that wasn't the end goal? From 2004, thinking toward 2013, is everything you're doing in this interregnum all about working ultimately toward that goal, or was there research that you would have pursued even if that was not your goal?

ANDERSEN: We had also a lot of basic science research going on in animals. With He Cui, Hans Scherberger and others, we were studying decision-making. Also we studied the local field potential with Bijan Pesaran and Partha Mitra, which is composed of the summed local activity of hundreds or thousands of neurons. At the time, it wasn't too popular, but it turned out there was a lot of information in the LFP.

ZIERLER: I'm curious if the ADA, the Americans with Disabilities Association, did they take an interest? Were they supportive of what you were doing?

ANDERSEN: No. We've gotten funding from various disability foundations, and of course NIH eventually—although I can't knock NIH; we didn't try until we could show it would work.

ZIERLER: Is that the strategy generally with NIH, that you don't go to them until you have a proof of concept?

ANDERSEN: It's a strategy. They now have funding mechanisms where you can go with a concept and not have to provide preliminary data to show that it's going to work. They've changed their model a bit. Really a lot of it is the reviewers who want to see preliminary data.

ZIERLER: Why DARPA? Is there a disabled veteran kind of component? Was that the idea there?

ANDERSEN: It was, yeah. Our first funding from them was in 2000, and the goal of that grant was to have collaborative units of theory and electronics and biology. We had those three components like which naturally worked with brain-machine interfaces. Of the six funded groups, two others were BMI groups like ours. Then around 2004, we got funding for what was called the Revolutionizing Prosthetics program. A lot of injured soldiers were coming back from the Iraq War and there was high survivability, but many had come back with missing limbs, or concussions, severe brain damage. The goal of the DARPA program was to help at least those that had lost motor control of their upper limbs.

ZIERLER: Did you get to work with any veterans?

ANDERSEN: Certainly at meetings, we met them a lot. That funding went on until about 2014, at which time they shifted their—which they do [laughs], on a dime—interest to other things. At that point, since we had two subjects implanted, we had to quickly look for other sources to keep it going. We got an NSF grant that tied us over. Then we started getting NIH grants, later the Chen donation, so that has carried us up to today. It seems trivial talking about funding—

ZIERLER: It's what makes research possible.

ANDERSEN: Yeah, even some very famous scientists from centuries ago needed benefactors.

ZIERLER: An evergreen challenge for science.

ANDERSEN: Yes.

Human Subjects and the USC Connection

ZIERLER: You mentioned 2013. What was it about 2013? What was the timing there that allowed you finally to get involved with human subjects?

ANDERSEN: This collaboration with USC and with Charles Liu—he also is highly involved with Rancho Los Amigos, so he identified a participant there, through one of the occupational therapists. That's another difficult challenge, recruiting subjects, because they have to be willing to work for years in the study, and to undergo a surgery, even after already having damage to their spinal cord. They have to be healthy enough to participate in the studies.

ZIERLER: They have to have a moral framework and a generosity, understanding that what they're doing is not for them; it's for the benefit of the future.

ANDERSEN: Right. This is often when we lose people, when we begin to recruit, when we inform them that the study will not help them. We say this is for the future and it won't have any direct benefit. Those that have participated in the study are real heroes. I think there's a lot of benefit socially and they get a lot of attention from doctors that take care of them.

ZIERLER: And it can give meaning to their tragedy.

ANDERSEN: Right.

ZIERLER: What was it that finally allowed you to work with people? Was it a regulatory approval? Was it the funding? Was it a combination? What did it?

ANDERSEN: We had the funding ongoing. The regulatory, we needed Caltech to agree to allow this, because they had never done it before. They did, which was great.

ZIERLER: I imagine the lawyers must have gotten involved, thinking about liability issues and things like that.

ANDERSEN: Yeah. Apparently it was a high-up decision. [laughs] Also, we were originally looking at ways of handling the regulatory component. There are companies called CROs, clinical research organizations, and they will handle all the paperwork and regulatory components, but they want hundreds of thousands of dollars, I think in part because they're used to working with big drug companies and large clinical trials. We looked into what they actually did and realized, for what we were doing, that we could do it ourselves. We worked with the FDA first to get what's called an Investigational Device Exemption. It allows you to extend a study for more than 30 days. This was using the Blackrock arrays. Having that, we got IRB approval here, USC, and Rancho. We had a participant, we had approval, we had a good neurosurgeon, we had good people in the lab, so that's when it all started.

ZIERLER: It all came together. Potential patients, did you identify them through the neurosurgeons you were already working with?

ANDERSEN: The first one, yes. Often they come through either Rancho or Casa Colina, the two rehab centers we work with. Recently we have one who participant who was friends with a former student here and she contacted me, so I reached out to him. Then one participant was looking around and found us. So, the participants come through various channels.

ZIERLER: Obviously we're not going to talk about anyone at an individual level, but I wonder if you can provide a composite sketch of the ideal patient? What is the kind of injury or accident that happened, what is the actual effect to their body, what have they lost? What is the ideal patient, based on what they went through and their current condition, that makes them perfect for the kind of work you're now doing?

ANDERSEN: Of course, they have to be paralyzed.

ZIERLER: Paralyzed means both from the waist down and from the neck down?

ANDERSEN: Yes from the neck down.

ZIERLER: Either is appropriate?

ANDERSEN: Well, no, we don't work with paraplegics which would be from the waist down. We work with tetraplegics who are paralyzed in all limbs.

ZIERLER: From the neck down.

ANDERSEN: From the neck down, right.

ZIERLER: Because it's all important to work with your hands, or where your hands would have been?

ANDERSEN: Yes. Originally, we were looking at people with very high-level cervical spinal lesions, so that would be where in fact they only have a bit of shoulder movement. Recently we've been able to work with patients that have some limited upper limb movement but still can't use their hands.

ZIERLER: Is it important for patients to have limbs? What about amputees? What about an accident where you were paralyzed and you've lost your arms as well? Is that a suitable patient to work with, or they need to still physically retain their limbs?

ANDERSEN: With amputation the control signals are taken from the remaining nerves. Other groups investigate ALS subjects, but that requires a great deal of medical work and follow-up, because they eventually become totally paralyzed and die. There are other forms of paralysis that would be aided by BMIs. There are locked-in patients. They've had a brain stem stroke, so that cuts off even the innervation for movement in the face. Sometimes they just have eye movements. Often they can't talk. ALS subjects and locked-in subjects are very medically intensive,—not that tetraplegia is any picnic [laughs]—

ZIERLER: Of course, of course.

ANDERSEN: —but at least—

ZIERLER: Those are new layers of challenges.

ANDERSEN: Right, yes, so we concentrated on spinal cord lesions that produce tetraplegia.

ZIERLER: What was it like for you, working with your first patient? It's obviously not a monkey; it's a human being. Were you nervous? Did you have new feelings of empathy? What was that like for you?

ANDERSEN: It's great that you can interact with the subjects. It really puts a human side to everything you're doing. They're interested and become knowledgeable about the science. It's a real partnership. Every time, it's great to see it working, especially right after surgery, the first time you try to use the recordings for brain control.

From Surgery to Thought Control

ZIERLER: Let's talk about the surgery. What does it look like? How invasive is it?

ANDERSEN: Neurosurgeons say it's a very simple surgery. But, they're neurosurgeons. [laughs]

ZIERLER: [laughs] You have to be a brain surgeon to say that!

ANDERSEN: Yeah. I guess it's not much different than the surgeries we do on non-human primates, and they recover easily from that. It requires, in a human, taking out a piece of the skull, and making a dural flap, the dura c, is this fibrous membrane around the brain. We've done MRIs and functional MRIs to locate good sites for implantation. We ask the subjects to imagine—this is weeks before the surgery—in an MRI machine, imagine grasping, imagine reaching, and—

ZIERLER: You can see where that lights up on the image?

ANDERSEN: Right. It's amazing that a lot of the same areas that produce motor outputs also are active when you imagine the movements. It's just until the very end that some gate opens, and you go. It's important for posterior parietal cortex—motor control is very easy to identify by the folds in the brain, as well as somatosensory cortex, but posterior parietal cortex is kind of like a fingerprint; it's different in everyone, so you really need a way to identify its subdivisions. Once the flap is open, we look at our MRI images, compare them to what we're seeing on the surface of the brain. Then there's a little pneumatic device, once we have selected the location, that the surgeon moves over the area of implant, and it punches the array into cortex. Since the array is like a bed of nails, a slow insertion produces dimpling in the cortex without insertion.

ZIERLER: To punch it in?

ANDERSEN: —to punch it in. It doesn't produce much damage at all as assessed by our ability to record from many neurons. To record single neurons requires that the neurons be within 100 microns of the tip of the electrode, so damage is not extensive at all.

ZIERLER: Then you patch the skull back up?

ANDERSEN: Yes, of course. [laughs]

ZIERLER: [laughs] Is it a wireless connection? How are you interfacing with the implant?

ANDERSEN: After putting in the arrays, then one of the most challenging tasks is to route the cabling because the cables are stiff. You want to have a bit of slack so that when the brain moves, the electrode array moves with the brain. The cable goes to a pedestal that is attached to the skull. That pedestal is the one that is the connector to connect it to electronics. It's not wireless at this point.

ZIERLER: The pedestal is external?

ANDERSEN: Right.

ZIERLER: There's a connection from the inside of the brain to the outside of the brain where it's mounted?

ANDERSEN: That's right.

ZIERLER: What's connected to the pedestal?

ANDERSEN: The pedestal itself is the size of a quarter, and then it goes down to a little neck, so it's actually pretty small. It has a little cap over it to protect it. When we record, we take the cap off and attach a headstage which contains amplifiers. The signal coming out is in the tens of microvolts range, 50 or 100 microvolts, up to maybe a millivolt. It needs to be amplified near the electrodes or otherwise it's very hard to recover the signals from noise. So there's this headstage that amplifies the signal by several thousand times. Then that goes to another set of amplifiers and electronics that save the information to memory, as well as running decoder algorithms for online analysis.

ZIERLER: Is the technology not there for it to be wireless? Would that be preferable if there was an advance?

ANDERSEN: Yeah, it would be preferable, and some companies like Blackrock and Neuralink are working on wireless versions.

ZIERLER: In surgery, are you there? Are you in scrubs, right there with the patient?

ANDERSEN: Yeah. [laughs]

ZIERLER: Are you there just sort of out of curiosity, or are you playing a scientific role at that point?

ANDERSEN: It's largely when it comes to deciding where to make the implant. We're at the nexus between the imaging studies that took place and the surgeons that are making the implants. Of course they have looked at the anatomy as well. Often we consult one another and decide, yes, that's the right spot.

ZIERLER: The surgeries that you were doing on monkeys, they're functionally identical to the surgeries on people?

ANDERSEN: That's right, yes.

ZIERLER: Although of course you would have had to go to medical school if you were going to be qualified to do the surgery yourself, I take it.

ANDERSEN: On people, yes.

ZIERLER: Being a Caltech professor is not a relevant qualification! [laughs]

ANDERSEN: No, I could end up in jail! [laughs]

ZIERLER: Okay, so the patient comes to. How soon thereafter are you working with them?

ANDERSEN: There's recovery of a couple days in intensive care, and maybe a week in the hospital, but it varies. Some people come right out of intensive care and go home. We wait a couple weeks before we try the first recording session. The first session is usually seeing how well the arrays are picking up neurons.

From Concept to Proof

ZIERLER: Are you working with them in their homes? Are they coming to a doctor's office? Are they coming to your lab here?

ANDERSEN: The first participants, came to the rehab centers. Then we remodeled the lab so they can work here too. We have been working in their homes too over the last few years. We've done a combination of all of those.

ZIERLER: That first dramatic moment, that first patient that you had, they're recovered, you meet them in the rehab center, what is that moment like, when you say to yourself, "Oh my gosh, I hope this works"? What's the setup and how do you know if it works or not?

ANDERSEN: I meet the participants when they come out of the anesthesia, and they sometimes ask—"How did the time go by so fast?"—and we tell them things went well. Because they always have. Then we come visit them in the hospital. After two weeks or so—the first time, we thought in principle it should work, but our results were based on recordings from healthy monkeys. We knew from the Brown group that motor control was still working after paralysis, so there was a high probability that it would work well.

ZIERLER: What does that tell us about the cortex, that it was still working?

ANDERSEN: It turns out it's not as plastic in adults as we might think, which is a good thing. It's also good for the idea of a cognitive neural prosthetic. You can't implant anywhere and hope to learn to use the BMI. We proposed, no, you have go to the right spot. If you want language, you go to the language spot. If you want hand movements, to a motor area. At that point, we didn't know—especially in posterior parietal cortex; no one had implanted it and so we didn't know whether it would be degraded. Also we didn't know how successful our implant was from a technical perspective. So it was two very excruciating weeks of wondering—

ZIERLER: Then if you can narrate that dramatic moment when the patient is ready to start working with the array, what does that look like? What are you looking for?

ANDERSEN: The first participant, we connected the arrays and we only saw a couple neurons, and we thought, "Oh. Uh-oh."

ZIERLER: Versus what? What were you hoping to see?

ANDERSEN: Many neurons. But it turns out that—and it's common—that two things are happening. One is you're getting better at setting up the electronics, taking out ground loops and reducing the noise. The other is—what other groups have noticed too—is over the first two to four weeks, the signals begin to come in and get better. So the next time, we had many more neurons. It was a few days later. In the beginning we asked he participant, to imagine different movements to see if neurons could be activated. Some neurons he could activate when he imagined rotating his wrist.

ZIERLER: Like turning a doorknob?

ANDERSEN: Right, yeah. We had a robotic limb there, so we connected it, and he was able to gesture to shake hands with one of the students. He was thrilled. It was the first time in 10 years he was able to, in a sense, use a limb.

ZIERLER: That must have been an extraordinary moment, both emotionally and scientifically.

ANDERSEN: Yeah.

ZIERLER: What were you feeling at that moment? What was that like for you?

ANDERSEN: I missed that session! [laughs]

ZIERLER: Oh, no! Oh, no! [laughs]

ANDERSEN: But they taped it [laughs], which I use in talks. But yeah, it was really great, because it meant also it wasn't going to require a lot of retraining.

ZIERLER: This success story happened in the very first patient?

ANDERSEN: Right, and the others have followed suit.

ZIERLER: Were you waiting for that success before moving onto the next patient? Is there a concern that maybe there was some issue with this patient and we're going to try again with somebody else? What's the thinking there?

ANDERSEN: Well, in the planning, and also in our regulatory applications, we planned for five subjects.

ZIERLER: Regardless of how the first one did, you were going to keep trying with two through five?

ANDERSEN: I guess the assumption was, it is going to work.

ZIERLER: And it did.

ANDERSEN: And it did.

Breakthrough and Spotlight Attention

ZIERLER: What was the news coverage like, with this breakthrough? What kind of attention did this garner?

ANDERSEN: It was huge. We published it in Science in 2015, and Caltech did a really good job, Deborah Williams-Hedges in the Media Relations did a fantastic job of getting press and also film and interviews and all that. When it came out, it was covered by most of the major news outlets. It was covered around the world. That's how the Chens picked up on it.

ZIERLER: How did you deal with all of this attention? You're a mild-mannered person. Did you step up to the occasion, do you think?

ANDERSEN: Yeah, I had to. [laughs] Yes, there were a lot of interviews. The week before, a Wall Street Journal reporter was here, just to follow the process through.

ZIERLER: The news media is famous, they always want to hype up the story, right? Did you feel like you needed to tamp down expectations? Did you feel like what you had achieved, you could just talk about, and that was big news all by itself without making it even more excitable than the news media might have wanted to make it?

ANDERSEN: They are journalists, so they ask all kinds of questions. Our participant, which it's okay to use his name—Erik Sorto—he did interviews, too. Often, I would say—and I think it's well deserved—the journalists are really interested in the participants, so they are often a focus of the discussion. For us, it's all kinds of questions like you're asking now. [laughs] Things that we are careful not to overhype is that you can't get one of these now. It's going to be years before a regular medical device is approved. Also, there are always the questions about implanting healthy people so they can enhance themselves. I usually try to avoid that, because what we're doing is strictly medical.

ZIERLER: You're not in the superhero business?

ANDERSEN: Yeah, and we don't want to scare people, or alternatively maybe some people would like implants.

ZIERLER: With all of this media attention, did you get a flood of requests from doctors, from patients, like, "Hey, do me next"? Was that something that you had to contend with?

ANDERSEN: I got people who wanted to collaborate with us. Some potential participants contacted us. That's one of the really difficult parts, is to find good subjects. We did get people that would contact us, from around the country, actually other countries too. Usually, though, they wanted to specifically be cured for something, which we couldn't do. The physicians handled most of these inquiries. They know how to reduce expectations. The next, participant was Nancy Smith. She came through the UCLA group and Casa Colina. I think the media coverage helps, because all the participants we've had have read everything, and are up to date on—

ZIERLER: They're ready to go.

ANDERSEN: Yeah, and when they talk to us, they say, "How's this group doing?" And that group. They're very knowledgeable, and the media has really helped with that.

Meeting the Chens and Building an Institute

ZIERLER: We touched on it briefly in our very first discussion, but now that we're in the chronology, it bears some explanation here. When the Chens first learned about this research, did they get in touch with you directly? Did you get an email from them? How did that work?

ANDERSEN: Yeah, I got an email from Tianqiao. He also sent information on himself. I looked him up on the internet, and sure enough, he is this famous billionaire [laughs]. About that time he was thinking of beginning in philanthropy. His company produced video gaming, so he was very technically savvy and aware of the potential of this kind of work. He had taken a real interest in brain-machine interfaces. He had planned to make donations to universities here in the United States. At the time, he was in discussions with Harvard and then with me.

ZIERLER: Did you have any inkling that all of this was leading to a massively enormous and generous gift that would enable the new building, or was this more like small scale, as far as you could tell?

ANDERSEN: The first meeting, it was just fun to talk about things. The second meeting, he came with his wife, Chrissy Luo, about a month later. There was talk of philanthropy. I did give him a proposal—which didn't have a building in it.

ZIERLER: Was the proposal more expansive than your lab? Would it be something that would be broadly relevant to Caltech?

ANDERSEN: Right, yeah. It was similar to the neuroscience institute, but more focused on medical devices and brain-machine interfaces.

ZIERLER: Are you coordinating at all with the Medical Engineering Department, like YC Tai, for example? Is this something that goes even beyond BBE?

ANDERSEN: Yeah, I've collaborated with many of them.

ZIERLER: In terms of coordinating, when you begin to articulate how the Chens could support Caltech in a way that's more expansive than your laboratory?

ANDERSEN: I had been collaborating with, at the time, Joel Burdick and Yu-Chong Tai. Of course I wrote in other possible collaborations with people we were collaborating with at that time. Some of them were outside of Caltech, because we had a lot of collaborations. I was getting ready to go to Steve Mayo, our chair, and tell him, because I knew, "I can't do this on my own." [laughs] That would be almost as bad as doing the human surgeries on my own. [laughs] But Steve had heard about it from Harvard, so he actually contacted me first. I said, "Oh, yeah, I meant to tell you about this."

ZIERLER: From Harvard meaning that he knew that the Chens were considering Harvard also?

ANDERSEN: Yeah. Well, considering us, and he hadn't heard of it, at that point. Then the administration and the rest of the apparatus here at Caltech got involved. Obviously, that was really necessary. Because at that time, then, the Chens—word got out, all the universities were contacting them and giving them proposals, and they were going around the country meeting people. In the end, he chose Caltech, which was really great.

ZIERLER: Steve Mayo, Ed Stolper, Tom Rosenbaum, they could all narrate their perspective on this. But because the Chens were considering philanthropy so widely, what was your role? What was your specific role in articulating, "Caltech is where you want to invest"? What case did you make for this?

ANDERSEN: It was a group of us. For a while, it involved a group of neuroscientists, and we met with Development, and we formed this pitch. It became more expansive, so BMIs were just a part of it.

ZIERLER: Reputationally, Caltech is such a fundamental place, right? It has that tradition that goes back. Do you think your research, that you had already demonstrated the success in a patient, do you think that that was specifically crucial because the Chens long term have translational interests? Obviously they appreciate that so much work in neuroscience needs to be fundamental, there's so much that we need to understand before we can think about applications and therapies and medicines, but do you think the fact that you were sort of already there, that you were already demonstrating translational capacity, was satisfactory to them in convincing them that ultimately this is where Caltech institutionally is headed?

ANDERSEN: Yes. The Chens are of course familiar, and currently are very familiar, with the field. At the time there were just a few groups doing neuroprosthetic research in humans—this is still the case. The Chens have a very deep and genuine commitment toward developing science and technology for the good of humanity.

ZIERLER: And here you are, already doing good for humanity.

ANDERSEN: [laughs] Tianqiao is a Buddhist, so he has the kind of philosophy that goes with humanitarian work.

ZIERLER: Was there any consideration or did you immediately swat away any possibility that you would be the director of the Chen Institute? Was that anything that you were interested in or considered?

ANDERSEN: No. I think right away, David Anderson was identified. He had been for a while trying to get a neuroscience institute here to make Caltech more visible in neuroscience. I was certainly fine with that. [laughs] That was not an issue.

ZIERLER: How much were you able to retain the basic research, the fundamental science component of your lab, once you're now working with these patients? In other words, how much care do you have to give to not sort of get on this path where, now that you have demonstrated this capability, you just devote yourself full-time to this, because of its inherent goodness? You're helping people, right? But you're a scientist at the end of the day. Once you saw that this was so important, that it established momentum, how did you achieve that balance of continuing in the basic research but also giving as much as you could to helping these people?

ANDERSEN: That's a good question. I think part of it was that we had this underlying concept of cognitive neural prosthetics. A lot is known about the human cortex but not at the scale we were looking. So, we can implant new areas, ask new questions. We can do basic science and then see how that could translate. We've very much had a science pathway, but of course it works very well with translation. In terms of engineering, I've usually left that to engineering collaborations, although we get into tech, to a degree.

Clinical Advances and Basic Science

ZIERLER: From that first patient to the patients you're working with today, what are the advances? Why is it not simply a repeat of what you were doing in 2013?

ANDERSEN: In 2013, we could first demonstrate motor components that were related to intent, so we could do the various controls to show that it was actually planned movements. The other thing we could show is that the area we implanted is a presumed grasp area that is also in non-human primates, and so we could show that with a robot program, the subject could control different grasps and grasp different objects. That was the main findings from that study. Then we have gone on to do all kinds of things. With Nancy, we noticed she had played piano before her lesion. She could do that on a computer keyboard. But she had this unusual personal experience that she'd just think about notes and it would play. We thought, wow, that's pretty high level.

ZIERLER: High level meaning because it's so abstract? It's so different than picking up a pencil?

ANDERSEN: Yeah, she's not thinking, "I'm moving this finger or that finger." Like when you're typing, you don't really think that way. You're thinking about the words being typed.

ZIERLER: Was this surprising to you? Was there a theoretical basis where this made sense?

ANDERSEN: Yeah, I think intuitively, just that you could have these high-level conscious thoughts and yet, in posterior parietal cortex, we could see not only the current movement but also the future one. It had the whole motor plan in it. We did a series of experiments that has to do with this issue of free will and awareness of intent, and we found that most of the finger movement activity was automatic for her, like what she reported to us. It wasn't conscious. Presumably, other areas could interrogate posterior parietal and see what's going on, and have an awareness of it, or control it. Although we could decode the intent, it was implicit intent, so it's like when sometimes I get in my car, and I forget that I'm going to the gas station, and I wind up in front of the lab, because I'm thinking of something else. You can do an awful lot of behaviors subconsciously. These experiments showed that a lot of what we were looking at was subconscious.

ZIERLER: That really anticipates my next question, where, even though you're working in this translational environment, this is enormously valuable from a basic science perspective, for what you're bringing back to the lab. You're gathering all kinds of data and insight that will inform sort of like the next stage of research for you. It's a two-way street.

ANDERSEN: Definitely. And with this implicit intent, for instance, Nancy felt the piano was sometimes playing itself, that somehow the decoder was running ahead of her.

ZIERLER: How could that be? How could a decoder anticipate before the brain actually—?

ANDERSEN: Because it already had a plan for the next movement, and she was concentrating on the current movement.

ZIERLER: What's the takeaway there? What does that mean?

ANDERSEN: It meant that we needed to adjust our decoders to take into account that mismatch by delaying the decoder output to make the piano playing seem more embodied.

ZIERLER: I wonder if the big takeaway here is that the brain continues to be able to do what it does regardless of whether there are limbs to carry out its commands. Is that at the end of the day what this is all about?

ANDERSEN: Yes, but the brain likes feedback too. We were worried that if we implanted the somatosensory cortex that it might have rewired after being deafferented by the spinal cord lesion. But right away, immediately, when we stimulate, the subject feels that again and the somatotopic map remains intact. So, even years later, all that circuitry for feeling touch is still there.

ZIERLER: Does that suggest that one day there might be a restorative aspect to this research? Is it possible that this research contributes to regaining functionality?

ANDERSEN: We've wondered about that, and we have noticed in some subjects some gain, but we're not sure if it was because of us, or—

ZIERLER: A causation/correlation issue.

ANDERSEN: Right. But there are others that have speculated that these kinds of implants can be used for recovery from stroke, for instance.

ZIERLER: That might be the next frontier.

ANDERSEN: One of them. [laughs]

ZIERLER: One of so many.

ANDERSEN: Other things we've done that are new—we found, as I told you, this mixture of so many things in posterior parietal cortex, that the participant could control intended movements of all the body parts. Sensory modalities of vision, sound, and touch are all present. That led to this idea that was actually coming out of monkey research too, of how things are represented in a mixed form. In other words single neurons respond to more than one variable and this is how so much information can be packed in a small number of neurons. From that, we've further shown that although variables seem randomly mixed, it's not. You have building blocks within the population. For instance, we have found building blocks for touch type, body location, and self/versus other. We're now looking at building blocks for motor activities. This approach, referred to as compositional encoding, is a way you can get a vast variety of sensations, cognitive states, and complicated motor movements out of populations of neurons. It's a good theory.

ZIERLER: When COVID hit, did you have to stop working directly with patients, or were you able to keep that up?

ANDERSEN: For about a year, yeah, but we had a lot of data to analyze, and of course everyone was struggling. [laughs] We were able to get it back up maybe quicker than other kinds of experiments because there were a lot of doctors involved and a lot of knowledge about disease.

ZIERLER: There may have been even some value in that pause to concentrate on the data?

ANDERSEN: I suppose, although it was disruptive. Probably the most disruptive is so many Zoom calls instead of direct interactions.

ZIERLER: What did the data tell you, when you could take a slower moment to look at it all? What were some of the top-line things you were getting from what you were seeing?

ANDERSEN: At that time, we were looking at individual finger movements, so we hadn't expected that. We found that both in posterior parietal and motor cortex. That of course will be very useful. As we came out of the pandemic, we found the encoding of language and silent language in the posterior parietal cortex. It gave us the chance to write up manuscripts.

Autonomy and New Avenues

ZIERLER: We'll bring the story right up to the present. It's 2023. You started this with people in 2013. So, a ten-year review. How different is your lab now, working with all of these patients? What are the things that you continue doing seamlessly, and what have been really new areas for you over these past ten years?

ANDERSEN: We have more participants. We now are able to stimulate not only somatosensory cortex but other parts of the brain. The finding of language encoding will allow us to begin using large language models to improve decoding. We're extracting phonemes, so that should greatly expand the vocabulary. We had this driving project that it looks like it will continue, that we started with Ford. Paralyzed people have told me they would really like to drive.

ZIERLER: That's autonomy. In our culture, that's what autonomy means.

ANDERSEN: Especially L.A.! [laughs]

ZIERLER: That's right! [laughs] Does that happen in parallel with self-driving cars?

ANDERSEN: It will, yeah. We have a collaboration with Richard Murray. He works on autonomous driving. That project will include error correction and path prediction. Actually the subjects are faster, because the signals come from the brain. I guess I'll be learning about autonomous driving! We've been working on new algorithms for extracting features from the brain. We've developed this entirely new modality for BMI, which uses ultrasound, which is less invasive. That's through a collaboration with Mikhail Shapiro here—he's the engineer in this case—and with the inventor of functional ultrasound, Mickael Tanter, who is in Paris. It has been a very powerful trio of labs, for this project.

ZIERLER: Things sound like they are exciting as ever.

ANDERSEN: Yeah, yeah.

ZIERLER: On that basis, Richard, for the last part of our talk, I'd like to ask a few retrospective questions, and then we'll end looking to the future. There is an inherent science fiction aspect to all of this; you ask a man on the street, "You can control things just by thinking about them." For your own experiences, if you can think back to graduate school when you started thinking about these things in a rigorous way, if you can imagine, if you can go back in time and ask yourself these questions, knowing what you know now, circa 2023, what would have seemed like science fiction to you, like, "No way, we couldn't do that," and what aspects of what you've been able to achieve do you really see on a continuum, that first you did this, and then you did this, and very logically it yielded this? I wonder if you could reflect on that duality.

ANDERSEN: When I started out as a graduate student, I was just thinking of being able to understand cognition, awareness, consciousness, decision-making, planning. But right from the beginning—well, I started in one lab that was doing awareness of intent in humans, so that's what I wanted to do. [laughs] And that's what I'm doing! That has been a common theme throughout. At the time, I hadn't really thought of the brain control aspect, although it existed in science fiction literature. As we were doing these monkey experiments and teasing out that it was really their intent that we were looking at, that kind of seeped in that we could use this in humans.

ZIERLER: As a biologist, not a technologist, what have been the game-changing technology developments for you? What has allowed you to do what you've accomplished just from advances in technology?

ANDERSEN: Being able to record from populations of neurons. We've always known that the real understanding of the brain would come from looking at population activity, but you have to look at it at the level of populations of single neurons. You couldn't do it with other recording technologies—like EEG was too non-specific. That's recording the averaged activity from millions of neurons, where you can't really tease out what they're doing. AI and machine learning has been important. I hadn't realized that we were actually working with AI and machine learning, since obviously the decoders learned to interpret the neural activity. I think we will be taking advantage of AI in decoders more in the future.

ZIERLER: As these capabilities rapidly expanding?

ANDERSEN: Right, yeah. Those would probably be the two foremost technologies, array recording and neural decoding. Also intracortical micro-stimulation, which also uses arrays of electrodes.

ZIERLER: A counterfactual question—you couldn't know, because it would have to take place in a parallel universe—what aspects of your achievements do you really associate with being at Caltech? Do you think if you had stayed at MIT you would more or less be on the same track, or just by virtue of being here, the uniqueness of Caltech, the collaborations, it was possible only because of where you were, and where you are?

ANDERSEN: You can never really know, right? In principle it would have evolved in a somewhat similar manner at MIT, although coming to Caltech, it was a real advantage being able to construct a more advanced laboratory. That helped a lot. And of course at a small place, I could get to know and work with people in different divisions.

ZIERLER: Low hurdles in administration for collaboration?

ANDERSEN: Yeah, exactly. Like I had one very successful postdoc who had been a Physics PhD here. Similar to MIT, the students are tremendous, and having the strong emphasis on computation, that has is very important and exists at both institutions. Having a happy wife, that [laughs] helped quite a bit. Also, Caltech is socially very welcoming, so she participated in the Women's Club and been president a couple times, and we got to know a wide range of people here, not just scientists but also administrators, and made a lot of friends that way. It's a warm and embracing—not that MIT isn't; it's just very big. My partner, Theodora Page, she is also a widow and her husband worked for a while at Caltech and we met through Caltech friends. Small world!

ZIERLER: Finally, Richard, just some questions looking to the future. You mentioned that potential patients want to know, when can they get these enabling technologies themselves? What's the holdup? Why not now? What's the timeline for when they can?

ANDERSEN: There are some hurdles. They're not large in terms of science and engineering but they're expensive. They would be, like you mentioned, being wireless, being potentially less invasive, so more development of the surgeries. Companies will want to make profit. Insurance will need to cover it.

ZIERLER: But ultimately this is the goal? This is what you're working towards? This is not a theoretical proposition. You want people to have access to this.

ANDERSEN: Right, of course. Also that it's rather autonomous, so it doesn't have to have a student present fixing things.

ZIERLER: Right, that kind of defeats the whole purpose of what their desires are.

ANDERSEN: Yeah. And it's very difficult to say how long that will take.

ZIERLER: Just in terms of hope, is this like—you should live a long and healthy life—within your lifetime kind of thing?

ANDERSEN: Yeah, I would think so. Did I already tell you this? It has been so many interviews. But when I was at MIT, a reporter from Time magazine asked me when we'd figure out the brain. They did a special issue on the brain. I said, "Oh, about ten years." [laughs] My wife said, "Don't ever [laughs] say that again."

ZIERLER: [laughs]

ANDERSEN: It was kind of the last line of the article. [laughs]

ZIERLER: Famous last words.

ANDERSEN: Yeah, so I've always avoided these predictions.

ZIERLER: But these are not scientific questions. We're talking about political and financial and administrative questions. The technology is there. It's a matter of making it feasible on a wide scale.

ANDERSEN: Right. I think improvements in the electrodes, too, making them flexible. But you're right; there is nothing scientifically or physically impossible.

ZIERLER: This is a matter of scaling up, and the wherewithal to do it.

ANDERSEN: I mean, it's amazing it works! [laughs]

ZIERLER: As you articulated, looking to the future, there are so many exciting new avenues for your group to focus on. Time is as limiting resource. Money is a limiting resource. Whenever you find it is the right time to retire is something to consider. What is most important to you, if you have to sort of make those narrowing decisions, for however long you want to be active in this world? What's most important for you to see a particular avenue of research either through to completion or at a solid enough stage where you've enabled the next generation to pick this up? What do you prioritize?

ANDERSEN: I probably would have been thinking about things like that if my wife hadn't died six years ago. That kind of stirred everything up.

ZIERLER: This has taken up a bigger part of your life, because—?

ANDERSEN: Yeah, so it has been good to shift to work, because yeah, it's a good, not distraction, but something to focus on. So, I haven't really thought of retirement or those sorts of things. And as I mentioned my partner, Theodora, has made life wonderful again. I have had a lot of students, many of them very prominent now. I've had, I think I told you, with the current ones, and ones already through the lab, about 100. Also, things are moving fast now. Hopefully, through the Chen BMI Center, we'll be able to spread out research to a lot of other labs.

ZIERLER: You mean both within and beyond Caltech?

ANDERSEN: The Chen Institute is focused on Caltech, but yeah, we also have collaborations with a lot of other universities. We're starting a clinical site at the University of Colorado. We're collaborating with many of the major medical institutions in the L.A. area. Hopefully that will catch on and progress. I can't imagine how far it will go.

ZIERLER: But there is a momentum that has already taken place?

ANDERSEN: Oh, huge, yeah. It used to be at Society for Neuroscience, our main meeting, they'd have these posters for motor control and motor systems; now it's all BMI, and it's huge. I just got back from Amsterdam. In the Netherlands, they have many groups working on BMIs. Same in Switzerland. China. China is extremely interested in BMI and AI. It's hard for me to predict where it's going to be, say, in 20 years. It may be like 20 years ago I couldn't predict the success of BMIs.

ZIERLER: But it's only getting bigger and you want to be a part of it for as long as you can, is what it sounds like.

ANDERSEN: Yeah. It was very good to be able to transition to humans. It opened up many doors, it seemed like everything became possible.

Maintaining Our Humanity

ZIERLER: Last question—it will be a very philosophical question. Unfortunately, humans have proved themselves to be very good at killing each other, ruining our environment, and the possibility of human extinction is always a threat, always seems real. Maybe it's even getting more real, year to year. From a science fiction perspective, we talk about, how do we preserve humanity even if physically that's no longer sustainable? Do you see at some point in the future your research contributing to an ongoing form of humanity that exists in machines?

ANDERSEN: This is sort of the script of many [laughs] movies, books, whether you could upload your brain into a machine, for instance. In principle it may be possible but "Would that be you?" I'm not sure. Things do look bleak in a lot of ways. But again, I'm hopeful. Things always look bleak.

ZIERLER: You were around in 1968, for example.

ANDERSEN: Yeah, yeah, yeah. In elementary school, hiding under our desks for alerts, preparing for a nuclear attack. They're still there [laughs]. It seems like problems just pile up. But I'm hopeful. Hopefully the stuff we're working on can be used for good, and that scientists also, rather than just saying, "Oh, that's up to politicians" or something, really have a say in it. This is not the only area of biology that potentially is really impactful, like CRISPR and genetics.

ZIERLER: You're emphasizing hope, that we won't get there, to that point. Is that to say that regardless of the feasibility of uploading humanity, or our minds, or whatever that means, to a machine, in order to perpetuate the human race, it sounds like what you really want to say is, if we've gotten that far, we've already lost? Is that the idea?

ANDERSEN: Yeah. This calls for a lot of speculation! [laughs] But yeah, hopefully we can keep the biological body [laughs] and brain in the future.

ZIERLER: Richard, this has been an extraordinary series of conversations. Thank you so much for doing this.

ANDERSEN: Thank you, too. Before we had these, I would think, what am I going to talk about? But then with these interviews, I just figured David would guide me through it. [laughs]

ZIERLER: Wonderful, wonderful, thank you.

[END]