The Grand Quest To Simulate Life - EP 57 Ed Boyden

Suppose you did simulate a brain, you know, does it have subjective experiences? Would it be conscious? And of course, the core issue in some ways is that we can't really measure consciousness. Like even now, you don't know for sure if I'm conscious, right? There's no meter you aim at my head and aha, you have 3 out of 4 or whatever units of consciousness right now. Yeah, so in the lab, we are starting to wonder, is there a way to measure consciousness? And a couple ideas. One is that if you have a simulation of a brain, Right now in neuroscience, we mostly focus on behavior outwardly manifested in the end movements. I move my mouth to speak, I walk out the door, there's these observables that are governed. But if I'm sitting here daydreaming all morning and I never tell anybody about it, I never change my behavior based on that, there's no way to know that that happened. So one of my deep hopes is that if we have the simulation of the brain, we can look under the hood and if there's some pattern of activity that doesn't really link to behavior in any way, but we can watch it and simulate it and, you know, probe at it, we can ask ourselves what kind of information processing is happening in that circuit.

Ed Boyden.

Hi, how are you? I'm good.

How are you?

Good, good, good. Welcome.

Thank you. I wanted to talk to you for so many years. I feel like I run around the world talking to all these interesting people, and very often someone, the phrase that comes out of their mouth is, I worked in the Ed Boyden lab for some of the most interesting people I run into. And it seems to happen the most with, you you and George Church's. Oh yeah.

Wow. Interesting.

Yeah. You've got this litany of fascinating students and then obviously I followed your work.

Oh, great. Great.

For people who don't know your research and your work, I'll just frame it a little bit and then please correct me if I get anything wrong. But yeah, I'm so pleased to be here. You're a legendary scientist, researcher. You've done such incredible work, especially around the brain and how we get to know about the brain or what we know about the brain. We are at the McGovern Brain Institute. You're also a professor at MIT, and then you are affiliated with or have co-founded numerous startups. Is that fair so far?

Yeah.

Anything I missed?

No, it sounds good. Okay.

And this is funny. Pat McGovern, he was my first boss.

Oh, wow. Small World.

I worked at IDG back in the day.

Wow.

I was a cub reporter.

So cool.

Yeah. Did you meet Pat?

I did. Yeah, yeah, yeah. Pat and Laurie were very involved. I met them many times in the early days. And of course, Pat sadly passed away, but Laurie still comes to the meetings.

And yeah, yeah, he used to. So I worked at IDG, the publishing side of what he did, I guess, how he made his money, really. And every Christmas, Pat would come by and hand you your Christmas bonus in person.

Wow.

To like all, I don't know how many thousands of employees he had.

Incredible.

Yeah, he would travel the whole world. And then it was obviously a very nice gesture. It got slightly comical at times because somebody would brief him about who he was about to meet. And I was on the wrong end one time when he crossed wires. He confused two of us. And he would kind of recount your achievements for the year. And I just had to nod as he was like about somebody else than— Wow. Yeah.

Well, it's so great that he did that. I mean, it's all about the people, Dan, and I think that's really awesome that he did that.

I didn't know that. It is. It is. Like, obviously I'm still talking about it, so it leaves an impression. Yeah, that's great.

That's great.

Okay, well, there's a million things I want to talk to you about, Pat McGovern related. If you— you know, when I do my research for these shows, I mean, again and again, the thing that kept getting called out was, was that you co-invented optogenetics, which helped us understand how the brain works in, in profound new ways that we couldn't before, and that you developed expansion microscopy, which is kind of swelling parts of, of the body and the brain so that we can understand them, them better.

I would—

I was curious though, you know, when you're reflecting on your work.

Yeah.

What are the things that really stand out to you during your career?

Yeah. Well, I guess for me, these are all puzzle pieces, but there is an ultimate goal. And the goal is really— I trained as a physicist and engineer before becoming a biologist. You know, can we understand a biological system the same way that an engineer would completely understand a system that they built. Of course, the problem with biology is we didn't build it, and it's also really messy. But I take great inspiration from physics. So in physics, they've hit what you might call the ground truth, the list of the parts and how they work together. And on the surface of the Earth, quantum mechanics is more or less that, that gives you lasers and microchips and cell phones and the internet and all sorts of stuff, right? Can we have sort of the microchip moment for biology? Now, in contrast to physics, where there's like a couple things and a couple ways they work together, in biology there's thousands of things, right? Genes, gene products, biomolecules. And so the overarching vision is really, can we just understand all of that? Can we map all the parts and how they work together? And my dream is that we can make a simulation in a computer of a living thing. You know, could you see all the parts working together and maybe even pinpoint where in that network you could treat a disease. Or, you know, maybe someday, this is obviously in the future, could you run a clinical trial on a computer? And so you mentioned optogenetics, where we can control signals with light. And then we have expansion, where we can map the building blocks of life. And there's a third branch of technology we've recently been putting out, which is how to image many things at the same time in a living system. But the way I think about it, these are three legs of a tripod.. And my hope is that we can integrate them and collect the right kind of data that you can use it to make simulations of living things.

And that's what you feel like you've been charging after your entire career? Did you set out with this goal in mind?

Yeah, in the end, when I switched from physics, I guess I always was very intrigued by the intersection of science and philosophy. And so my first lab experience was in an origins of life lab. They were trying to create DNA. Out of clay, which of course didn't work or you would have heard about it, but it was still a very educational time. I left home really young. I was only 14 when I started that job. And so eventually I came around to the brain as a sort of philosophical, practical intersection. But my dream is that we could really understand the human condition. Like, why do we do what we do and feel what we feel? But the practical angle that pays the bills and keeps things going and is also very important to me is can we understand these complex biological systems in terms of their parts?

We will pause the genius that I am no doubt interviewing right now to tell you briefly about Brex. Brex is obviously the flagship sponsor of the Core Memory Podcast and our video series, but more than that, we actually use Brex to run our company. They have returned countless hours to all the members on our team by automating our expenses, automating some of our accounting, and they have returned money to us by using their high-yield savings accounts. It's made a huge difference to our company, and it will make a huge difference to your company too. Go to brex.com/corememory to learn more. 35,000+ companies are running on Brex. Give it a try. You will not regret it. I don't regret it. Since you brought it up, I was going to save it for a little bit, but you were talking about when you were 14 working on the origins of life. I mean, when anyone reads about your biography, one of these things that stands out is you were, for all appearance, it looks like a prodigy as a kid and started at MIT, I think it's at 16 and got a triple degree plus a master's degree by the time you were 19, I think. What was it? Electrical engineering.

I did two bachelor's in electrical and physics, and then I did a master's degree in electrical engineering.

Okay. And you grew up in Plano, in Dallas?

That's right. Just north of Dallas. Okay.

And so I'm always fascinated by people that excel at such an early age. I mean, I don't know. I mean, what was childhood? Did you have some sense that you were a bit different from the other kids pretty early on?

Well, I had a, I think, a very thoughtful childhood. My mother stayed at home and we would talk about science. We would run little experiments in the kitchen. My dad, he's retired. They're both, you know, he was a management consultant. And so it was a very sort of thoughtful home environment where we talked a lot about problems and how to solve them and things like that. I was probably 12 years old. I went to a science fair and I had my project, which I think was something like microwaving a bean in the microwave or something, and what would happen to it. But then at the science fair, other people had like real projects. They were looking at cancer genes and all sorts of stuff. And I was like, wow, I need to get to a place where I can do this. And so I just got really hardworking and ambitious. And then I started taking summer courses at a local community college and The state of Texas has this really quirky and interesting and amazing program called TAMS, the Texas Academy of Math and Science. And people skip the last 2 years of high school and go right to college. And so I ended up skipping the last, I guess, 4 years, I guess, more or less, and went at 14. But it's on the University of North Texas campus and you get to do real scientific work, lab work. You're side by side with college students and it was a lot of fun.

So you were homeschooled before you did that?

No, no, I went to the regular public school.

Oh, you did? You did? Okay. Okay. And then, I mean, you mentioned the science fair at 12. Didn't— I mean, didn't you end up winning the science fair fairly quickly?

Oh, there was a later year where I did a math project because, you know, math, you don't need a lab to do it. So I was very intrigued by geometry and the rules of geometry and did a project that I don't think With the current lens, I would be proud of, but as a 12 or 13-year-old, and I said 13 by then, I was pretty excited that you could do mathematics. And so math was sort of one of my first loves was that you could analyze things and think about them. But then quickly the question became, what can you do about reality, right? Tangible things, the world, can we make it a better place?

And you left home at 14 to go to the school.

Yeah, so 14, I went to this boarding school. Which was on the University of North Texas campus. And they pick 150 kids from the whole state and they all move there. And it's a very interesting program.

Were you the— you were the youngest one?

I think up till that point I was the youngest person. Yeah.

I mean, that has— so yeah. Is that strange or exciting? All of the above?

I thought it was a lot of fun. Yeah. Yeah. I had a great time there. I made lots of good friends there. It was a fantastic environment for doing science and it was only half an hour from home. I mean, Texas is a big state, so some people had a long commute, but, you know, it was not so far away from home that I would be, you know, too distant from the family.

Yeah.

Yeah.

No, I just— I always picture, you know, just being a young kid off on your own, especially when you came to MIT at 16.

Yeah.

Did you live in the dorms or did your parents move up there?

No, no, I lived in Baker House, same floor with— had roommates, the whole nine yards.

Yeah.

Yeah.

And that was That wasn't strange hanging out with all these?

Well, it's hard to know what's strange or not strange because I have no other reference than my own life, I guess. But for me, I was in a state of constant curiosity. I guess for a long time I had this frame of mind that I was much younger than everybody, so I was here to learn from them. And so when I came to MIT, I was going to continue. I was actually studying, focused on chemistry when I was back in Texas. And I thought, all right, I'm going to continue majoring in chemistry. And then this was 1995 and like the whole internet thing was coming out then, you know, web browsers were just exploding that very year. And so everybody on my floor was majoring in computer science. So I was like, okay, I guess I'll go along with that. So I ended up switching to physics and electrical engineering and computer science, which I don't regret because, you know, that turned out to be so useful as you know.

Yeah. And had you coded or anything before that?

I did. Yeah. Yeah. I guess I was very interested in mathematics beforehand, but I often use computer code to solve math problems. And so that was my entry point into computer science.

Okay. Okay. So if we go back to your central quest to understand the body almost as this engineering problem, I mean, there's a bunch of things that I've reported on where we've made stabs at doing this. The Connectome is this effort to image the brain on this level where you can see individual neurons and synapses. And, you know, it's been a struggle. I mean, yeah, we've done a worm. We just finished the fly. It seems like it could take many years and ridiculous amount of computing power to start doing a human brain on the same level. We've— I think the most we've done is like a cubic centimeter or something like that of the human brain so far.

Well, this spring though, a group in Austria showed that this expansion method that our group invented can be used to get connectomic information. And importantly, because it's a light microscopic technique, you could also get molecular information. So the way that I think about the brain is, so the connectome of course is the wiring diagram of the brain, the shapes of the cells, if you will. But in the end, we're made of biomolecules and biomolecules work by touching. And when two molecules touch, they can cut, bind, or change shape. And of course, there's, for the case of the brain, electric fields as well. But so I want to map more than just the connectome. I also want the molecules, right? The molecules that make the cells do what they do, and that in disease states are the things that go wrong. And so this expansion method that we built looks like it can give you not just connectomics, but also the molecules along those wires and so forth. And one of our former students, Andrew Payne, is running this nonprofit that's called FRO, Focused Research Organization, called E11, where they're adding in barcoding as well, basically a way of telling the cells apart. But the point I want to make is I'm really obsessed with the idea of ground truth. What are the fundamental building blocks and how do they work together? And in physics, of course, you have the particles and forces. In biology, we have the biomolecules and how they interact. So if we can get that, we should be able to make a simulation of life. Okay.

I mean, this is my question where I was going, and Andrew is actually— he's coming on the show in a couple of weeks. I've been to E11 and it's fascinating.

Yeah.

So you mentioned they barcode— they sort of inject brains with the virus to put all these barcodes in so you can map the wires almost automatically, I guess, for lack of a better term. But yeah. Okay. So, you know, we've been on this quest to, to map the brain, whether you pick a neuron or a molecule. And it's been this long, arduous path. It's been expensive. We've had like entire nation states trying to fund these efforts. Where would you say we're at? And if we really want to get this complete picture of the mind, I don't know, you know, when do you think this might happen?

Yeah. Well, so this expansion method that we developed—

Can you explain how it works?

Sure. Yeah. Yeah. So I guess for literally 300 years, the way biologists have imaged cells is with some kind of lens, right? So the microscope was invented hundreds of years ago. That's how cells were discovered in the first place. About 150 years ago, neurons were discovered in the brain. But light has a finite size or wavelength, and so you can't see really tiny things. So many people have tried to come up with tricks. Super-resolution microscopes use clever tricks to break that. So-called diffraction limit. The electron microscopy, X-ray microscopy, you can use particles that have smaller wavelength basically, but they're all tricky in some way. X-ray and electron imaging can be quite expensive and they also struggle to give you molecular info. And light microscopy methods that break that resolution limit can be quite slow and difficult to do 3D imaging. So in our group we thought, We have lots of critical thinking and creativity skills that I like to think of as our learnable and teachable skills. And so we started thinking, what's the most opposite thing we can do rather than zoom in? And we thought, let's just make it bigger. So we could take a brain or any biological specimen for that matter, and we immerse it in a solution of what are called monomers, basically little building blocks, and they self-assemble into polymers, basically a dense spider web-like mesh. And that's basically a mesh of polymer that's a lot like the stuff in baby diapers, except that it winds its way through the biospecimen. So if you do it just right and you process the sample just right, like softening it up a little bit, when you add water, the baby diaper material swells, but it makes the brain or whatever it's embedded in bigger. And what I think is kind of the surprising discovery we made is that that expansion process is even. In fact, with some colleagues in Germany, who we've been collaborating with, it looks like that the expansion process might be precise nearly down to atomic scale. And so the idea of mapping all the building blocks of life throughout an entire system is no longer science fiction. It sounds like it might be within grasp in the near future.

Yeah, I mean, because I kind of cut you off to explain that technique. But when you say in the near future, because even if I talk to E11, any of the connectomics people, they still present it as something of a journey. And I'm not looking for like a precise year, but I mean, you think where this is within what, 5, 10, 15, 20 years or?

Well, I mean, let's take a single cell. If we were to map all the proteins and nucleic acids and the major macromolecules of a single cell, and if we could simulate that cell, basically if some biomolecules aren't touching, they're diffusing around, and we know the mathematics of that. And if two biomolecules are touching, then with AlphaFold and other methods of simulating molecular contacts, you can imagine making a model of that. You might be able to make a model of a cell. So these numbers that I'm going to say now are just made up, but I think it's very intriguing to think of whether you could try to make a model of a cell and maybe a small brain like that of the worm C. elegans. In the next couple years, let's say 3 to 5 years would be ambitious but interesting to shoot for. Now, if you show that that can be done, then what happens, of course, is that more and more people will adopt the technology. We all share all of our tools freely, and Optogenetics and Expansion are being used by thousands of groups each. So if we have a way of converting a biological system into a computer simulation, I'm sure there'll be a similar kind of growth afterwards that can be hard to predict. But my hope is that if we can show a proof of concept with a small brain or a small cell, or maybe a bit of both, and we do also some work on small fish as well.

Zebrafish?

Yeah, larval zebrafish and the worm C. elegans. And then also very small cells are interesting as well. If we show that's possible, then I think it'll be irresistible for many of us in the community to give it a try for different purposes. If you can simulate an immune cell, could you understand how it works? Can you simulate a cancer cell and find its weakest spot? That kind of thing.

Okay. But what you're— okay, I feel like we've been on the same path with the central connectome work from the C. elegans to the fly to the human. And it has still been this slow process. Like, I feel like you're arguing that you're more optimistic that things maybe go faster with these new techniques. Yeah.

Yeah. So two thoughts. One, of course, is that with electromicroscopy, which is how the fly connectome and the worm connectome and so forth were done. Again, you see the shapes of the cells in exquisite detail, but it's really hard to identify the specific molecules, right? And you want the molecules. The way that brain cells work, they generate these electrical pulses. You have ion channels, you have receptors, you have transmitters. You really want to know where they are. And so my hope is that with our method of expansion, because it lets you label those, we can help pinpoint them.

And you're going to get—

And the second is scale. So when you expand something, now you can use much, much less expensive optics to image it. And in fact, one tantalizing thought is if you expand something big enough, could you chop it into sections and then use dirt cheap optics, maybe even modified kinds of optics, the kind that you find on a cell phone to image the brain. So if you can make sort of an asymptotically zero-cost microscope, And then you use commodity chemicals to do the magnification. Maybe we could democratize the ability to do extremely high-volume scalable nanoimaging.

Okay. Because with the Connectome, we've been doing these slices of the brain and each slice has to be imaged with pretty much the best imaging techniques we have. And it's, it's this manual labor part of this that's part of the reason it's so slow and difficult.

I think it's part of it. Yeah. Yeah. We did a collaborative paper many years ago with a couple other groups where basically it took a hacked webcam and then we expanded some bacteria. We showed that we could image bacteria with the hacked webcam. So this is not— this is not out of the realm of what has already been envisioned.

Okay. And you hit on it a little bit in the very beginning, but okay, with the Connectome, one downside has been that it's— you're getting this image of the brain in a fixed moment in time. It's not a living brain. It's just a snapshot of whatever was happening. And so there's There's people who would argue you can get tons of information even from that snapshot. And it's fascinating. And then there's some people who are kind of down on how valuable this could be because you don't have the activity. If we use your techniques to get this much more detailed picture of a brain, what do you think this unlocks?

Yeah. Well, so if you have a map of the static organization of molecules of a brain, of course, One can then look at the molecules and then using the principles of chemistry, which are well known, you could try to simulate what happens next. So again, molecules that are diffusing around, you can simulate that. If they're bound in this post-AlphaFold world, you could try to simulate that interface. But knowing the ongoing activity is also really helpful. Like right now, there must be millions of these MacBooks that you have all over the earth, but right now they're probably all doing somewhat different things. So knowing the ongoing activity is important because all these MacBooks have the same wiring more or less, but their ongoing dynamics can be quite different. And so the other two technologies we've been building, which are the imaging of signals in living cells and then the optogenetic control of living cells, I think are still very important. So my dream experiment would be starting with a small brain like a fish or a worm, let's image all the activity, as many molecular signals as we can, use optogenetics to perturb those cells and look at what happens. And then at the end of the day, preserve the specimen, expand it, and make a map of the molecules. Those three datasets, I think, could be combined to make models even better than if we use the technologies alone from each other.

When you say this is your dream experiment, is it something— is that how you're spending your days trying to create that experiment?

Yeah, yeah. I think that Yeah, we've got two goals. One is to advance each of the individual technologies to bring them to their limits of performance. And then the other is to integrate them so we can get these kinds of integrated structure and function datasets.

Okay.

Yeah.

And then, I mean, my question about what you get at the end of this, I guess it's sort of a philosophical one. I mean, on one hand, you're getting just this immensely useful tool to understand how our body works and to study it. We're also at this era of AI. Hundreds of billions, trillions of dollars are being funneled into creating an artificial intelligence. I keep seeing overlap between these fields. Well, you know, this just brings up a million philosophical questions about simulating artificial brains while we're trying to understand how our brain works. And I mean, one of the things I was most interested in talking to you was your take on the convergence of these worlds and the interplay between all this.

Oh, yeah. Yeah. Oh, it's so intriguing. I guess a couple of thoughts. One, of course, is that if we can simulate a brain, would you be able to develop interesting kinds of AI that act more like brains? For large language models, for example, they're very good at language, but Every week there seems to be some example showing that they're not really thinking the way that we do because they'll make some error that a 5-year-old would not make. And so maybe they kind of solve language, but do they solve thought? Probably not, it seems. And so if we can understand how the brain works though, in enough detail that lets you simulate it, maybe that does help you understand what thinking or feeling or other things are. But that brings us to a second big question, which is almost philosophical, which is, Suppose you did simulate a brain, you know, does it have subjective experiences? Would it be conscious? And of course, the core issue in some ways is that we can't really measure consciousness. Like even now, you don't know for sure if I'm conscious, right? There's no meter you aim at my head and aha, you have 3 out of 4 or whatever units of consciousness right now. Yeah. So in the lab, we are starting to wonder, is there a way to measure consciousness? And a couple ideas. One is that if you have a simulation of a brain, Right now in neuroscience, we mostly focus on behavior outwardly manifested in the end movements. I move my mouth to speak, I walk out the door, there's these observables that are governed. But if I'm sitting here daydreaming all morning and I never tell anybody about it, I never change my behavior based on that, there's no way to know that that happened. So one of my deep hopes is that if we have the simulation of the brain, we can look under the hood and if there's some pattern of activity that doesn't really link to behavior in any way, but we can watch it and simulate it and probe at it, we can ask ourselves what kind of information processing is happening in that circuit. And that's—

sorry, go ahead.

And then maybe a theory emerges out of what a subjective experience might be. In the end, of course, in the current stage of history, you still have to test it with informed consent human volunteers who maybe will there's ways of stimulating the brain with electric fields or other modalities that let you activate brain cells, even non-invasively in people. And so maybe one could then ask a question about whether that pattern that we saw in the model, does it actually change subjective experience?

So the simulation you're talking about is the one that's based off this new era of brain data that you're hoping to get. And so you would be creating a replica of the brain. Yeah. And, and so, and then getting, getting it to a level where we can watch its— we can watch its genuine activity and like probe a brain in ways we can't because people don't want us running around with electrodes out of their head all day.

Yeah, yeah.

And then how excited or not, you know, do you get with the progress you're using in the AI field? There's people who would want to match, yeah, the, the software brain versus the flesh brain and look for overlap, or have the— have our AI models based even more on, on our more complex models of the brain to make them sort of more aligned with humans. I don't know if this is stuff that you get interested in at all.

I have talked to people about it. Yeah, because a lot of people in AI are concerned about things like AI safety and, you know, what if you create something that you cannot control and so forth. Yeah, I do think that there are some interesting arguments for simulating brains in support of that. One, of course, is that if you understand what ethics is, like what is our brain doing when we make an ethical judgment or an empathetic judgment or a socially beneficial judgment, Could we recapitulate that in some way? The second thing, of course, is problem solving. So again, LLMs are pretty good at certain things, but if you think about generating entirely new ideas that solve a problem, humans, I think, still have vastly the upper hand on that for a lot of real-world problems anyway. Maybe not contrived problems or specific problems that are defined in competition space and so forth, but But if we want to have creativity, if we want to have effective problem solving, there are things that the human brain does that we might not even understand at all. I mean, somebody's walking down the street and an idea pops into their head. Where did it come from? We don't know. And the history of science is full of people, Einstein daydreaming, or the chemist who thought up the benzene ring that appeared to him in a dream when he thought of a snake swallowing its tail. There's all these sort of irrational subconscious things that are so important for human creativity. And yet maybe we haven't understood any of them completely, maybe even partially in terms of how the processes occur in our brains. And then finally, I think there is a hope amongst some that rather than having AIs over here and humans over here, we'll find ways to work together. And I think for that to happen, we kind of have to understand a bit about how our brains work, right? Because if we enter information in in a way that is completely incompatible or has unpredictable effects, right? And every neural therapy out there, to my knowledge, has a decent chance of causing side effects, at least in some people, right? Because the brain is a network. You stimulate one part of the brain, it's going to activate other parts of the brain. And so if we really want to have machines and brains talk to each other in a really meaningful way, I think we need to learn more about the brain. I mean, the hard part of neuroengineering, I think, is often the neuro part. We built, as a community, electrodes and non-invasive machines to deliver magnetic fields and ultrasound. The list goes on and on. But where do you aim that energy? Where do you record information? What do you do with it? Those are open questions.

You seem not terribly impressed with LLMs.

Well, it's maybe because I'm a scientist looking for in my own life, I guess, how we could generate really creative ideas. So if you think about really big ideas in biology, for example, PCR, CRISPR, things that really changed the game. Has an LLM built anything of that caliber? And maybe I'm just setting too high a bar, but I'm very intrigued about how we can make creativity more learnable and teachable and to boost it. And so if we could build AIs that have that capability, it'd be very exciting. But don't get me wrong, I mean, I do see LLMs doing amazing things in lots of other domains.

Hello, geniuses. Let me tell you about E1 Ventures. They are a venture capital firm in Silicon Valley, a longtime supporter of this podcast and a longtime supporter of big, fantastic, world-changing ideas. If you have such an idea, hit up E1 Ventures or send me a note and I'll put you in touch with E1 Ventures. It's, uh, it's quite a, quite a deal for listening to this podcast. Thank you again, as always, to UON Ventures for their support. Do you spend a fair amount of time using AI and/or digging into how they work?

Let's see. So using AI, I mean, yeah, we use AI routinely for things that, you know, I think it's good at. So if we want to, you know, analyze something in terms of physical principles or equations, it's pretty good at mature sciences like physics. We can run Hey, here's this material, what's it going to do? And it can look up coefficients and make calculations and so forth. When it comes to chemistry, which is a bit more nebulous, then I have asked AIs to help me design chemical reactions and so forth. And there I think now we're already in a much more nebulous space. I think humans might still have the upper hand for doing really creative chemistry for sure.

That's what I was kind of curious about.

And then biology, part of the problem with biology is that the literature has a lot of hidden structure, a lot of which is not even written down sometimes, right? Like maybe, you know, a group bought this chemical from a company and the company went out of business 5 years later. And now 25 years later, somebody wants to replicate that study, but you can't buy that chemical anymore. And, you know, that kind of thing won't be written down in the scientific literature. Who knows, maybe somebody even bought the brand of that company, rebooted it, and they're selling something with the same name, but it's very different, right? And so I think the problem with biology is that there's a lot of unwritten metadata, a lot of unwritten wisdom. You know, one group did it in, you know, their lab had the air conditioning blasting and the temperature is a little bit lower. Another group in maybe a country where they have less air conditioning, you know, had it, their room was a higher temperature. Maybe the reaction went differently. But again, people don't write down the temperature of their room typically, right? Yeah. A scientific paper.

Yeah. Do you have like ChatGPT or Claude open on your laptop on a day-to-day basis?

I think I've been using just Google Gemini a lot because it's able to— I feel like it's very integrated with different Google search functions that are just familiar to how I think. Yeah. And but MIT now does have a ChatGPT subscription for faculty where it won't use the information to train the model. And so I'm playing with that as well.

Yeah, I guess. I was curious because I keep running across these 20-year-olds when I'm reporting on stories, and they are— I feel like they're sitting with their laptop while I'm talking to them, and they're just using it in ways that— like, I use AI, you know, a bit, and, and every day for some purpose. But I mean, they have like 7 or 8 AI apps open. They're like talking to their laptop. They— it's just, it's, it's It reminds me of— you were talking about the rise of the consumer internet, sort of like one era to the next, and now you're just using your computer in a different way than the generation before. And I don't know, I was trying to put myself in your shoes a bit because you're doing such high-level science. And yeah, yeah, I was kind of curious if you had adopted where I see sort of where I feel like this next generation is going with this stuff.

Yeah, well, one of my former students, Sam Rodrikus, he started a thing called Future House, another FRO. And they are trying to build AI scientists and to be able to analyze the literature and draw conclusions and design experiments and so forth. And other entities are also trying to design AI scientists of different kinds. So it is a very active area of research. But it seems like they have made great advances, but also said there's a long way to go in terms of really revolutionizing science, like a drastically out-of-left-field idea. And maybe it's because of the way that machine learning works now, where they get a corpus of data and are trained on it, and they're great at kind of interpolating and extrapolating, but doing really out-of-distribution predictions, and it seems like no current AI does that super, super, super well.

Do you, last AI question, I don't know, do you, You know, where do you fall on the, um, going off what you just said? There's tons of people. I live in Silicon Valley. You know, if you talk to like Larry Page or, you know, people who are knowledgeable about this, right in the thick of it, there's— it's easy to find ones who would say those limitations you talk about in the next 3 years are blown away for sure. Superintelligence is here. And then those same people usually aren't the most optimistic, I find, about where this leads. This doesn't strike me as something that's like consuming you as something you're thinking about. But do you have a take on where this is heading in the relatively near future?

Well, it's really not my core expertise since I'm much more of a biologist, but It does seem like people are trying to scale to larger and larger models. People are trying to tweak the architectures and build networks of agents of them and so forth. But again, it's not my core expertise. I haven't seen a radically different architecture proposed. That's what got me thinking, maybe a little bit of brain data could go a long way. And if you take a simple motif and scale it up, Google or OpenAI or whatever scale, now it starts to exhibit interesting emergent behaviors. Well, what if we learn more about the brain and we could then try to scale that up?

Yeah.

I think it's worth a try.

This is why I wanted to ask you. I just feel like you have a better handle on the complexity of the body and the brain and are obviously clever enough to compare that against what the LLMs are. I just feel like you'd have a better handle maybe on some of it because Clearly one of the arguments sometimes is, as you mentioned, you just throw enough GPUs, maybe consciousness just arrives out of, out of enough complexity from throwing enough compute at something and connections. And, and I don't know, I felt like you might have an intuitive sense as to whether you believe something like that or not.

I don't know. I tend to be very data-driven. Like, I would love to measure consciousness. I want to be able to model it. I think we need to be very quantitative and detailed in our characterizations.

Okay.

And if you look at the history of other sciences, that's how it has been too, right? You know, centuries of astronomical data, and then Kepler made Kepler's laws. And then Newton came along and then distilled them further and made comparisons of things on Earth and so forth. And then we got Newton's laws. But, and people love to talk about how theory-driven physics is, but It started with centuries of data. So if that's physics, what hope do we have to do biology without lots of data? And the right kind of data too. That's why this idea of ground truth, the fundamental building blocks and how they work together, I think has been such a driving force for our research.

I will note for listeners of the podcast, if they hear some noise, I think people are washing lab equipment above us.

I think, yeah, there is a dishwasher room overhead, I believe.

So if that's what you hear, it's just science.

Taking place because you do this relaxing waterfall sound.

I want to shift gears a little bit to brain-computer interface stuff, which is an area that I'm just deeply into. I know. I remember when I was reporting on Kernel in the early— I think it was the first time I ever talked to you. That's right.

Yeah.

You were an advisor, I think, to Brian.

Yeah, I was one of the science advisors.

Right. So they were trying to build that kind of external. They started internal and then ended up at an external BCI. You also have— you have a startup that— or you're affiliated with a startup I talked to the other day that does an implant that kind of sits in the bone matter of the skull.

Inner Cosmos.

Right, right, right. Yeah, yeah, right.

And so, yeah, I helped get that company going as well.

And so that's not fully going into the brain, but it's sitting in this layer of the skull and it's for mental health. It's modulating I don't know, maybe I'm sure you'll do a better job explaining it.

Yeah, yeah. Well, the core idea is, you know, you can have non-invasive devices, but they're outside your scalp, and that can be quite inconvenient. And then you can put wires in the brain, but that can also damage or displace brain tissue. And so our idea was, what if you had a sort of a hybrid? You had a minimally invasive device, so it would never enter the brain or displace or damage brain tissue, but it also wouldn't be sticking out. And so you can imagine this sort of inside the skull, so to speak, implant, and then you could deliver electric fields. And yeah, so there's a surgical team and an engineering team that is exploring topics like, yeah, could you go after depression or other conditions?

Yeah. And so, you know, in this field you have so many different approaches. I mean, Neuralink is in some ways the most drastic, I guess, being the most invasive. And then all the way up to things that are totally external. You've got things that are in blood vessels and— yeah, I mean, when you— and this is— this has moved pretty rapidly. You have the Utah Array, which was the main tool of— for researchers doing this type of work for 20, 25 years. And it was an invasive device, but it was very hard to use. You had to be in a hospital or a research setting. And then, you know, over these last 12 years, all this venture capital money is poured into this. We have all these different approaches. There's, you know, each one has pros and cons. This must be kind of fascinating to you to see all this activity. And then do you tend to favor one approach over the other? I was— I'm curious what you make when you look at the BCI field.

Well, since there's no ultimate theory of the brain, what matters most is data. Is it working? Our group has put out a couple inventions which are being used in human patients or human subjects. So optogenetics, where we borrow molecules from plants, put them into brain cells, the molecules convert light to electricity, and then you can use light to control the brain. Very widespread in use in animal neuroscience to study the brain, but A few summers ago, a brave team in Europe did put one of our molecules into the human eye for blindness. And there are now, I think, a couple dozen people walking around. It's not approved by any regulatory body, but with at least a partial restoration of functional vision. So they lost—

Explain more how it works.

Sure. Yeah. So blindness specifically or optogenetics in general?

Let's explain optogenetics through the blindness example. Okay. Yeah.

All right. Yeah, so millions of people have lost the photoreceptors. These are cells in the eye that capture light and translate that into signals the brain can understand. So electrical signals. If somebody has this, then there's not a whole lot you can do for most of the patients, unfortunately. So with optogenetics, we take molecules from single-celled algae, microbes, and so forth, And what these microbes have are proteins. And these proteins convert sunlight into electrical current, like little solar panels, basically. And what we did with optogenetics is we took the gene that encodes for this protein, put the gene into a brain cell. And now if you shine light on the brain cell, you can activate the brain cell too, because that little solar panel converts your light from your LED or laser into electrical activity. So back to the eye. The photoreceptor cells have died off, but the rest of the eye is still there, right? You could take the gene that encodes for one of these light-activated proteins, put it in with gene therapy vectors, of which there are many, including many that are already used in people. The gene therapy vector will deliver the gene for light-activated protein to one of the spared cells of the eye in the retina at the back. Those cells become light sensitive. They were not beforehand. And now you basically have, biologically speaking, installed a camera in your eye. So these people wear goggles that take light from the world and project it onto the area of the eye that has the molecule. But they can recognize, you know, lines of a crosswalk or doors on a hallway. They can see household objects. It's not perfect vision, but it's a pretty significant restoration of functional vision.

Is this the stuff that Science Corporation, the French company they acquired, or it's different?

I don't know what the status is on it business-wise. There is a French company called GenSight, which was the company that licensed the molecule from us. And they were the first to put these molecules into people. If they got acquired, I don't know.

My memory is failing me. I feel like it was called Prisma maybe or something.

There may be multiple companies. Yeah. So GenSight was the first. And now there are a bunch of other ones that are going after different targets or using different molecules. But there was a French team that did, summer 2022, I think, was the first person who had these opsins put into the eye.

Well, and then in the BCI field writ large, all of this technology right now is being used for people who are suffering from usually quite dire conditions, whether it's blindness, ALS, paralysis.

Yeah. Yeah.

I keep getting both like excited about where all this is going and then sometimes a bit muted in that you almost want to see more. Yeah. And so I guess this was part of my— when you're looking, I don't know, you could pick Neuralink if you want, but, but or Or any— like, how, you know, okay. In terms of really restoring somebody to something like an ALS patient being able to get their powers of speech back to where they could communicate like they could before they had their illness. How excited are you about the power of this technology It's very interesting.

I mean, I've seen a couple examples recently where by reading neural activity out, even with older technologies, but with enough machine learning to try to interpret it, you can make very interesting predictions of what spoken words the person was intending and so forth. So on the one hand, I think at the phenomenological level, with enough data and machine learning, It's quite possible to have a lot of restoration of different kinds of functions. I think if we really want to understand the depths of cognition and emotion and be able to build brain-machine interfaces that have arbitrarily useful functions, then that's where the understanding of the brain would be wonderful to greatly increase. And that's one reason why we've Although we spent the first decade of the group building lots of brain-machine interfaces, we also developed a technology for non-invasive focusing of electric fields into the brain, which also got spun out as a company. Is it cochleto?

No.

This is called temporal interference. And the company is in Zurich called TI Solutions. So basically brain cells are nonlinear low-pass systems, which is a fancy way of saying they act like old-fashioned AM radios. And so if you deliver multiple high frequencies to the brain, the brain kind of ignores them the same way that broadcasting across the city, the radio waves go right over your head, you wouldn't even know they're happening, right? But where the different frequencies collide, what a nonlinear low-pass filter does, what an old-fashioned AM radio does, is basically subtract off that high-frequency carrier and then you get the voice or whatever. In the case of the AM radio, in the case of neuroactivity, we get whatever signal we want to deliver. But to make a long story short, basically it's a way of having multiple high-frequency fields delivered to the outside of the brain, and then where they collide, you get information delivery. And so that's one technique that we built that is capable of non-invasive but deep stimulation of the brain. So the point I want to make though is just that, yeah, there are these different tools. How do we use them? What are they best at? Right now, the only way to know is to try. There's no grand theory, so we must be empirical about it. But my hope is if we can make really good models of the brain, you can be like, wait a second, this region and these two other regions are actually talking to each other in this way we didn't know about. What if we stimulate the three regions with these three different patterns? Could we restore this function that can be quite complicated and maybe we don't even understand it at all currently? But now because of the precision of our knowledge, we can dial in something that helps somebody and and relieve their suffering.

Yeah. Well, I feel like I keep trying to get you to rank things and put dates and predictions on stuff. I mean, well, I mean, my, my, all the stuff that I cover, I just feel like we're heading to some strange new world where, where humans and machines really are merging in some dramatic way. When I hear you talk, I feel like my timeline in my head is probably too quick.

Well, it depends. I mean, if you want to have a general strategy for any kind of brain augmentation, I think that's somewhat far off. But if you record from a specific region and your goals are very specific, already several groups— I've been marveling at the videos and so forth— can read out neural activity and do things like generate language or control a computer cursor, right?

Yeah.

And that seems to be a very well-posed problem. Now, does it work in the real world when the environment is rapidly changing? I mean, if you look at even a different field like self-driving cars, where in the desert, the DARPA Grand Challenge, driving across the desert was solved a long time ago. Getting it to work in a busy city with kids and pets and so forth running around, you could argue that still is not completely solved, right? And so I think one of the big questions for brain-machine interfaces is, What do they look like when they're put into the busy, complicated world? And you could argue that nobody's really put a, you know, en masse, a large set of people with brain-machine interfaces out into the real world, at least not to my knowledge.

And we have ability to read out data with optogenetics. You can control the brain as far as inputting data. As far as I'm aware, I mean, we don't seem terribly far along on something like that, putting information in.

Well, I mean, so optogenetics in the eye is in several dozen patients. Okay. So that's entering information into the eye. I think if any of that gets approved, that people might start wondering about putting it into the brain. But I think a lot of people are kind of waiting to see what happens. Yeah. Because if a lot of people are doing the eye, and the eye, of course, is much better understood than than the brain. Let's see how it goes. But yeah, if we could imagine a holographic projector that would aim light in 3D patterns and you can then make brain cells sensitive to light. And also, you know enough neuroscience that if I stimulate these cells, I know what's going to happen because remember, the brain is this complex network. I can be locally stimulating but still influencing things all over the brain. So it's important to be able to predict that, I think.

But that would be generating what if? Because I'm sort of talking about putting almost like an idea, a thought in. Aren't you describing more like a reaction in the brain?

Well, that's why I brought up the idea of holographic projection. So if you built a very good holographic projector, and so one of our collaborators, Valentina Emiliani in Paris, was one of the key pioneers of this. A hologram is basically a 3D sculpture of light, right? So you should be able to give every cell its own recipe, its own information, right? So if you do that with enough fidelity, you should be able to enter arbitrarily complicated information into the brain, is my hypothesis. Of course, you'd have to build such a machine, but optics is a field where with enough engineering, one could imagine anything eventually coming within grasp. And so, yeah, could you enter an idea or a concept with a holographic projector and optogenetics? Maybe.

Have you looked at the biohybrid stuff Max Hodak is doing?

Uh, tell me more. I don't know.

He co-founded Neuralink and then went off. That's right, he's a new science corporation, and they have— they used iPSCs to create lab-made neurons that live in—

oh, this is kind of gel.

Yeah, they cut a hole out of a skull, put the gel next to the hole, and, and, uh, the human-made neurons start to form axons and dendrites down into the actual brain.

Yeah, very intriguing.

Yeah. I mean, what do you make of that?

Well, again, I think it really boils down to what does the data show? If you can get consistent connections, if they're stable, if you're not getting side effects where neurons are getting wired up in the wrong way, then it's intriguing. But yeah, it all boils down to the data, right? Because you don't have a theory of neural development yet, but it is a very intriguing idea.

I know that you approach some of this stuff philosophically. What do you think humans and machines are like in 30 years?

Yeah. Well, if we can really build good models of small things the next 3 to 5 years, like we were saying earlier, and now in the same way that with earlier tools we spread it freely so that everybody's modeling biology in this way, Maybe we do understand the brain so deeply, and I'm just going to make this number up, but what if within 10 to 15 years we have good models of mammalian brains and maybe human brains? But if we understand what thoughts really are, if we understand what emotions really are, maybe we could try to augment them in certain ways. The thing about augmentation is that you really want to know what you're doing, right? Because if you augment something and then but you've introduced a side effect that causes something negative later, then that could be cause for regret, right? So I think we want to find ways to deeply understand what we're doing to the brain if we're going to go beyond the most pressing medical needs for sure.

But like in the thought of humans trying to keep up with AIs or merging with AIs or, evolving the species as this flesh-computing hybrid. Are these things you think about? Are these things you think should happen, are inevitable?

Well, I think at some point it's inevitable. The question is when. Is it now, 50 years, 500 years, or some other time point? My guess is it probably will happen in decades. But here's the thing. I think we want to understand things very deeply before that point. There's what the old Silicon Valley saying of move fast and break things. You might not want that if it's your brain. And so my hope is that if we can understand the brain rapidly enough, let's say models of small brains in 3 to 5 years, understand the human brain deeply in 10 to 15. Again, these are just made up numbers, but they're not entirely unreasonable. Then you can imagine putting the field on a very firm foundation, right? Okay, here's what you want to do if you want to help somebody with this, but not mess up these other 17 things. And so I think if we could do it right, there could be a lot of benefit of technologies that help us understand ourselves or improve ourselves in some way. That said, if you look at the history of biotechnology and maybe neuroengineering in particular, When we don't understand something, there often are side effects that result in something not working out well. I mean, there's so many examples, right? Like, I don't know, Merck made Vioxx, the pain drug. It had side effects in a small percentage of people and they pulled it off the market, right? I do some work in the Alzheimer's field and there's lots of examples of Alzheimer's drugs where it's just been a long, hard slog, right? And if they have a side effect, then You have to take this drug for years, right? It's a chronic disease. You know, will people do that? Right. So I think there's a lot of questions that are about like, will we really do it? Will it really work? Yeah. And so that I think it does make sense to put it on a well-posted footing and make sure we really understand what we're doing.

Yeah. Okay. You and George Church, I both see you involved with lots of startups. But then still heavily affiliated with your respective research institutions. Yeah, I'm always, I'm always interested how you make the decision of staying at a place like MIT versus just going all in on a startup and chasing some giant idea full time, all the time.

Sure. Everything. Yeah. Well, I feel every problem has a natural home. Some things go well in academia where you have open collaboration and you can publish and you can look at certain time horizons. Your startup will have to have a shorter time horizon. You'll be much more focused, less serendipity perhaps. And then a couple of my group members developed this FRO, Focused Research Organization concept and Adam Marblestone and Sam Rodriguez and so forth. And so maybe that also fills in a gap where you need something more scalable than academia, but not as profit-making as a startup. So that's great. For me, because I am just, you know, the risk in biology is so high, right? You can work for a decade and you can spend billions of dollars and then your therapy can fail with a side effect that, you know, in some part of the body you've never thought about, right? And that happens all the time. And so the way I think about it is we need to remove the risk from biology. And I mentioned this idea earlier of the microchip moment, you know, this moment where You know, after the microchip, you can be a dropout and start Apple, Facebook, and Microsoft, right? The risk of science that is represented by the microchip, as far as physics goes, is a huge reduction of science risk, right? And in biology, can we get to that microchip moment? Can we get to the point where, you know, a college dropout could start, you know, a company that cures a disease? We're not there yet, to my knowledge. But my hope is that if we can make simulations of living things, and now any kid in high school or college could use a simulation to try to go after a problem. Maybe that is what causes the true democratization of biology.

You have this reputation, your lab has this reputation for people having quite a bit of intellectual freedom, the ability to pursue their curiosity, not terribly sort of rigid and self-serving to your particular interests. Like, is this a— was this a conscious decision that you made when you started this? Is it just a reflection of your personality? I didn't know if you had some management genius wisdom to pass down.

I mean, the things we're doing in the lab are things I'm interested in. I guess I feel that my role, you know, is not just to let people do random things, but neither to micromanage. I'm here to persuade. Okay, here's this idea. If it's great, let's talk about it. Let's deconstruct it. Is it going to solve the problem? Is there a path we can realize? And so it's more of a process than a rigid set of rules. But if we have an idea, whether it's from me or from a student or so forth, and we can show to ourselves that it is likely to work, we show to ourselves that it will give us what we want in terms of solving the problem, you know, and there are some heuristics that I like. I love to build technology that's very inexpensive to use. So with optogenetics and expansion microscopy, anybody can do it because I really do believe in democratizing science. You know, then we can do it. So in fact, I'm starting to write a blog where I want to talk more about like our problem-solving methods. And so we posted a couple articles about how to generate ideas or how do you think about what to learn in order to solve a problem and so forth. And so I would like to just write down a lot of our problem-solving methods and get them out there. But I think if we apply those methods, then it's very easy for a group member and myself to be aligned because we will converge upon something that ideally is extremely impactful, but also is very practical.

This is a big question, I guess. Like, what do you make of the state of US science at this particular moment in time with all kinds of interesting political and sort of nation-state-level competitive things going on?

Yeah, yeah. No, I mean, science is being very disrupted with, you know, mass cancellations and reductions in National Institutes of Health, the National Science Foundation, and so forth, priorities. Yeah, I do think it has a real risk. I mean, it's already— the real risk has already happened of hurting national competitiveness in science. I've had people, you know, in the group who have gone to other countries or been unable to continue on. You know, funding, of course, is always an issue. Yeah, it is quite worrisome. And especially at this moment where it felt like we were poised to, you know, like with advances in AI, like we said earlier, with advances in ground-truthing of biology, you know, it felt like many fields are poised to really go to the next level. So in terms of timing, as well as, you know, both the state of funding as well as the sort of the general freedom of people to be and do things, it does seem like quite a disruptive state.

Where do the students tend to go, ones that leave?

Oh, let's see. Well, my group, like many, is very international. And so, yeah, I mean, many people, you know, if they have a visa issue or something, will try to go to their home country or other countries and so forth. But it's pretty chaotic, to be honest. I don't know if there's like a And all we can do is try to support our group and help people, you know, just as people. But one of my hopes is, well, is there maybe a new model that emerges of international labs working together? I don't know. I didn't really prepare a detailed thought on that topic. But it's got me thinking a lot more about like, you know, in an earlier time of crisis, right, with like the Cold War and mathematicians and physicists started talking across, you know, lines of political distance. And, you know, maybe that helped with some greasing of the wheels and communication and so forth. Is, you know, would there be the pendulum swinging the other way? And now maybe there will be more international collaboration in science. I don't know.

I don't know. Right. Because it, well, I mean, it would be nice if like the China and the US could, collaborate on all these fields and push the world forward together. That doesn't seem to be the way things are trending.

I guess this is one reason also I wanted to start writing this blog and writing more for the public and so forth, which is it seems that, of course, if somebody has Alzheimer's disease or terminal cancer or something like that, the disease does not care what nation you're from or what political party you are. It's just going to try to kill you. And so Is there a common enemy here? Is there a way to tell its story in a way that motivates us to go after the things that are causing so much suffering? And I feel like that story, maybe if fresh voices can tackle it, maybe there's a way of telling that story that makes people realize, wow, there's a lot of suffering here that we are just giving up the ground on.

You were You've had this prolific career and you started so young chasing big ideas. Is it harder as you get older to find the big ideas?

Well, I mean, I feel like the big idea that I had was sort of near the beginning, which is the idea of could we convert biology into physics? Could we solve the ground truth? And the basic thinking from day one was, If you could understand a complex biological system in terms of its parts and how they work together, that is the answer. You should be able to see all the parts, control them, and make a model of them. And my very first biology experience when I was still an undergrad here was trying to make a little model of a circuit in the songbird brain. How does the songbird clock out the syllables of a song? And I made a model that kind of explained the data, But how can you test it, right? And so that motivated me to go down this direction of building tools, of really getting into the molecular nitty-gritty, I guess. But, and then, I don't know, it seems like often the idea that solves the problem almost has this sort of feeling of inevitability about it. But sometimes we have to go through what I call constructive failure to get there. I'll give you an example. So we had the idea, the first idea for expansion in 2007, a long time ago. But we didn't know that it was an important idea at the time. And then in 2012, 2011, 2012 or so, two fantastic grad students in the group were trying to do nanoimaging with super-resolution microscopy. And we built our own, and also electron microscopy, classical techniques. And it was hard. And then even if it did work, it's slow and expensive. How would you ever scan a whole brain? And that's when we kind of learned a secret about the field, which is, wow, nanoimaging is really hard if you want to image anything that's like a big biological structure. And so then we mothballed those projects and started doing expansion. So I think in some ways it's almost like a process of elimination. Let's carve away— what was it? Michelangelo was like, carve away all the parts of the marble and the sculpture is simply what's left behind. Very often we'll fail as we try things. And then as we eliminate the different possibilities. There's kind of a nucleus defined by the failure of us in some ways, but in a constructive way to tell us what to do.

Whenever you talk about these models and simulating things, it just makes me think, I don't know, are you one of these people that thinks we live in a simulation? Hmm, it's a good question.

I'm not sure. I mean, the laws of physics and the laws of science are very intriguingly concise, and there's so many interesting coincidences. You know, like if this parameter was a little bit bigger, a little bit smaller, then the universe would be totally unrecognizable. That said, I guess again, I'm very data-driven. I would love just to analyze reality until we have some clue about what it is, but I don't feel like I have a strong opinion about whether, you know, It is one extreme of total atheistic nature of total chaos versus a contrived simulation or something in between. And I don't know, I feel like I'm very data-driven. I would like to find some evidence that one or more of these possibilities is the case. I don't feel a strong belief or opinion.

You said, hmm, like you hadn't thought about that before, but I'm sure you must have.

Well, I have wondered about it a bit. Yeah. Because And when I was a physics student, I was very intrigued by just how much structure the universe has, right? Like, it's kind of weird if you think about it that, you know, the laws of physics, there's only a couple of them, right? You know, and then there's only a couple things like electrons, protons, neutrons, that's it, right? And then a couple forces. And if you're big enough, then you wear gravity. But for most things, you know, you can kind of neglect gravity for many chemical and physical things. But then if you do the equations, you have to the periodic table of the elements, right? Now you got carbon and nitrogen and all that. And then take carbon. Carbon is this interesting atom, forms these long chains. You can get DNA, you can get proteins. No other atom seems to do that, right? Silicon is right under carbon. Silicon doesn't do those things. Is there silicon life somewhere in the universe? So anyway, there are all these interesting coincidences there. Is it evidence for simulation? Is it evidence for religion? Is it evidence of just Some people call it the anthropic principle, that there's lots of chaos and we just happen to be here and we notice it because it's the only universe where we can exist to notice things. But yeah, I don't know. I would love to see if we can discover evidence for one or another. And I have a sneaking suspicion that this is why investigating consciousness is so interesting, right? We don't have any measurement device or theory which helps us understand why this bit of matter feels its feelings, has subjective experiences, and this other bit of matter, presumably, of course, we can't prove it, does not. And yet it's so essential to everything, right? If you have a robot vacuum cleaner, nobody thinks twice about throwing it into the trash or recycling it when it breaks. But beings with subjective experiences, pets and humans and so forth, deserve justice and also are held accountable. And so it's very important, even, even though we don't have a scientific grasp over it.

Yeah. Yeah. Well, I got to ask most of my big questions that I wanted to ask you, and I'm just being cognizant of time. It was just an absolute treat to actually get to sit with you for a bit and chat.

Thank you so much. That was a lot of fun. Yeah. Let's stay in touch. The next couple of years are going to be hopefully Really exciting. Yeah.

I mean, is there— you went over so many things and you talked about your research, but is there a particular thing we should keep our eye on that you're excited about that you're working on at the moment?

Well, I'm very intrigued by this idea that you could map all the building blocks of life to the point where you could simulate it. I think that would be a real game changer.

Yeah.

And just in the last year, some papers have been coming out from us and collaborators and others, which suggests that you might be able to map the building blocks of life. So that's very, very intriguing. And if we can do that, this idea of converting biology into basically physics with computer science running on top of it, that'd be a big game changer.

Yeah, absolutely. And what happens in these? I was just walking down the hallways. I mean, is this where you would work on a day-to-day basis?

Yeah, yeah, yeah. Our group has much of the second floor of the McGovern wing of this building. And yeah, here we do things like expand brains. We build microscopes to scan things. We work on optogenetic molecules and are building optics to do the holography that I mentioned earlier.

Okay. So just pushing all those different fronts forward all the time.

Yeah. I think of it as 3 legs of a tripod, right? We want to image lots of things, perturb lots of things, and then make a map of the structure of the things and then integrate that to a model. So most of the projects in the lab are directly on the lines of one of those three things or opportunistically related to one.

Okay. Okay. Well, Ed, thank you so much. We really appreciate it. Okay.

So it was a lot of fun chatting.

Okay. Awesome.

Okay.

The Core Memory Podcast is hosted by me, Ashlee Vance, and/or Kylee Robison, or both of us together. It is produced by me and David Nicholson. Our theme song is by James Mercer and Jon Sortland, and the Show is edited always by the John Sortland. Thank you so much to Brex and Ewan Ventures for all your support, and thank you most of all to everybody for listening or watching. We love you. Please leave us a like, a review, a subscribe, all those tremendous things. Thank you, and we'll see you again.