Can The Brain Merge With Artificial Intelligence? — With Max Hodak
Channel: Alex Kantrowitz
Published at: 2025-12-12
YouTube video id: -1i4J0eubEk
Source: https://www.youtube.com/watch?v=-1i4J0eubEk
Hey everyone, we have a great crowd with us here today and we're live streaming it. So let's hear you make some noise. Come on. My name is Alex Kantrowitz. I'm the host of Big Technology Podcast and I'm thrilled to be here with Max Hodak, the CEO and founder of Science for a conversation about hacking the brain and the brain computer interface revolution. Max, great to see you. Thanks for having me. Today we have seen that there is research that the brain computer interface or the brain itself and the way that artificial intelligence works are actually a little bit more similar than we initially believed. Can you tell us a little bit about that? Yeah, so there's this there's this been this really interesting unification that's happening between uh what's happening in the artificial intelligence community and what's happening in neuroscience. I think when you look inside these AI models, there's like if you think about what people are talking about a year or two ago when they were they were much less sophisticated, they're thinking, "Oh, these things are glorified auto completes or they're kind of they call them stochastic parrots." But that that really does not seem to be the case. When you look inside these AI models, you see these mathematical objects that look a lot like what you see in what you see in the brain in neuroscience. Like the representations are converging. And this is an empirical finding. Like there's this uh there's an area of research uh there's a paper called the Platonic representation hypothesis where it there's this empirical finding where different models train different architectures on different data sets converge to the same underlying representations. And so this gives you this sense that these models are are learning something kind of deep about the universe. That there's this underlying thing that's like a physical fact that if you you train these big computational models, you get you get these representations that produce intelligence and they're the same things that the brain has figured out. And so there's this really interesting unification um that I don't know. I I mean I think it gives us a lot of reason to believe that the work in AI is on the right track. So it's important to digest what you just said, which is that there are similarities between the human brain and artificial intelligence, the way the models work today. Do you think one day we'll be able to merge with AI? I I think you have to kind of be specific about what that means. But yeah, I mean one of the things that I'm preoccupied with is is um the the binding problem with the So all of your experience, I mean it's important to realize that kind of all of this, like everything you see and hear and feel and think, like all of this is brain activity. Like that's at the end of the day that's all it really is. Like the brain is the thing that is you. It's the source of of uh your entire experience of the universe. But within that, it's composed of billions of neurons that are distributed over a large area and spread out in time. But you don't experience the activity of these neurons independently. You experience a unified moment. And even with within your brain, you've got two hemispheres, each of which is processing one half of your experience. So the left hemisphere is processing the right hemi field and vice versa. But you don't experience two hemi fields. You experience a a single visual field. And so how does that happen? There's some physics that we don't understand yet. And if you could understand that, then you could imagine adding you could imagine building say conscious machine. You could add a third hemisphere. You could connect this over the network. This would in some very fundamental sense allow you to redraw the border around a brain and change the like where your brain ends and begins and in that sense where where you begin and you end. Wait, how how far are we from being able to do that? I mean I think >> I probably not next year, but I think within the next decade. Wow. >> It's very possible that this stuff is going to get figured out in a meaningful way. Um I think like 2035 will be a very very different world than 2025. Um I think that the progress that's happening in AI is is much better known than any of this, but it's still not priced in. And yeah, I mean I think that there's a big change that's coming towards us over the next decade. So what happens when we're able to either connect with AI or even manipulate the brain using computing? Well, when I think about the trajectory of BCIs, so there's on the one hand there's there's very pragmatic near-term medical innovations that are coming to market now. So our our lead program at Science is a retinal prosthesis. It's a chip that's implanted in the back of the eye. Um we just we finished a major clinical trial last summer in age-related macular degeneration where patients that are unable to recognize faces have been able to read again. Uh there was a major publication in the New England Journal of Medicine about a month ago. This was actually on the cover of Time last week. And that is happening That that work is happening now. But then when you look at the trajectory of the next decade, I think it takes you to places that are kind of tough to talk about without kind of sounding like a lunatic. But let's do that. And yeah, I mean the Look, the the only organ that I really care about is the brain. The Like you can get a heart transplant or a liver transplant, but you can't in concept discuss talk about getting a brain transplant. And so I think that the way that like medicine might go is like if we're used to on on the one hand we have this this toolbox of drug discovery in conventional medicine. And on the other hand you've got this toolbox of neural engineering, which I think is just empirically much more powerful. If you look at the work I work in the retina, for example, people have been trying to restore vision to the blind for a long time, arguably for thousands of years. And there's a bunch of drugs that have been developed. There's a gene therapy that's on market that costs almost a million dollars a patient. And it doesn't First of all, it's only applicable for like 3% of of patients that have the type of blindness. And it also just doesn't really work. Like it very marginally slows the rate of descent for a small number of patients. And the because like dealing with the the dealing with biology in in terms of this these molecular details is very challenging. Humans just aren't good at this. But the brain is an information processing organ. And if you can think of it from the information perspective, like the power of of our retinal prosthesis is we don't really care why the rods and cones have died. We just care that we can get the visual information back into the brain and then that can be reflected in the mind's eye as a thing that you see. And so this this is just an I mean you you put this in and then these patients can read an entire eye chart going from not being able to read any of it. And so one of the reasons that I like the neural engineering approach is it produces these very large effect sizes. Like cochlear implants really if you are if you are deaf due to sensorineural hearing loss, they really work. Or if you've ever seen a a video of a a Parkinson's patient getting a deep a deep brain stimulator being turned on, they go from like unable to hold a pencil to like being very still. Um or like what Neuralink has developed where you're able to put a like put electrodes in motor cortex. I mean in an hour or so after that patient starts to use that, they can play video games. This just this produces effect sizes that you don't see in medicine. And um the uh Yeah, so I think there's a direction that this can go where you kind of you've got the brain and if you can interface with the brain and keep that healthy, then you can get like kind of everything else you can do and the rest become swappable parts. Like I'm going to be pretty disappointed if I'm murdered by my heart or my pancreas, which as far as I'm concerned is basically a support function to keep keep the brain going. Right. So if I'm hearing you correctly, what you're saying is this could be brain computer interfaces could be a science that develops in parallel with medicine or maybe even exceeds medicine and ends up playing into longevity because once you can manipulate, control, enhance the brain, then we can maybe live longer lives. >> Yeah, I mean like if So the number one killer is cardiovascular disease. Number two is cancer. We've made [snorts] great progress on each of these and there's you can find graphs online of like the mortality rate per 100,000 people falls over time, but still the the the loss rate here is remains 100%. And the curing cardiovascular disease, curing cancer, we don't know how to do this. But it is possible that um by connecting to the brain directly, you can just avoid the need to solve these problems in the first place. You don't need to solve these hard problems if you can avoid them. And so you talked about this a couple times and I'm going to give you a a moment here to expand upon it. Um because it's actually remarkable what you just spoke about uh a couple minutes ago, that you've built a device called Prima that enables people who have had loss of sight to regain that sight. By what happens? Do you stimulate the visual cortex? >> the first thing I'll say is we uh we're excited to bring this to market and we've been developing it, but it was originally invented at at Stanford and then we licensed it from um an inventor at a professor at Stanford who I think really deserves a lot of the credit for the central idea. And but I mean we're I mean it the results are amazing and we're really excited to bring this to market to patients. So it doesn't stimulate cortex directly. So if you want to restore vision to the brain, there's So first of all, there's many reasons that people go blind. You can there's cataracts, which are easy to fix through surgery. There's retinal degenerations like macular degeneration or retinitis pigmentosa or Stargardt's disease. Um where the the light sensitive cells in the eye, the rods and cones have died, but the brain knows how to see and the optic nerve is intact. And then there's diseases where the optic nerve has been lost. Um the most common of these is glaucoma where high pressure in the eye can cause the optic nerve to degenerate. Sometimes this can also happen through traumatic injuries. And then there's other kind of more like rare diseases where the the brain really never learned to see in the first place. So Prima is a chip. It's a It's a tiny little 2 mm by 2 mm chip that is implanted under the retina in the back of the eye that if you look at it under a microscope, it has all these little hex cells. And each one of these little pixels is essentially a solar panel. So it works in conjunction with glasses that are worn by the patient that have a camera that looks out of the world and a laser that projects into the eye to strike hit the implant in the infrared which you can't see, so it doesn't So if you have any These patients sometimes have residual peripheral vision that they can use to walk around, but they can't read or recognize faces. The the laser hits the implant and wherever the energy is absorbed, it stimulates. It creates an electric field. And so in this sense it acts essentially as an electronic photoreceptor. Like it it bypasses the rods and cones to stimulate the next layer of cells in the retina um called the bipolar cells. So in the retina there's really three layers of cells. There's 150 million rods and cones. These connect to 100 million bipolar cells. Bipolar because they have two ends. And then this compresses down to 1.5 million optic nerve cells which is this big cable that goes into the brain. And we So if you've lost the rods and cones, Prima will electrically stimulate the bipolar cells. And this empirically as a This was a finding of the clinical trial that if you do this in humans, they get an image in the mind's eye. They can see it. We We think that this is the right place to restore vision. Now if you've lost the optic nerve, then you have a more challenging problem, but most I mean these patients haven't. And it's the bipolar cells are easier to work with because it is before this 100x compression. So if you stimulate the optic nerve at that point, you've already you've already compressed a lot of the visual information like edges, color, relative motion. You'd have You have to figure out how to encode that for the brain, and nobody knows how to do that right now. But we know that if you stimulate an image onto the bipolar cells, you get an image in the mind's eye. And that that works. Now from the 1.5 million optic nerve cells, that eventually connects up to 1/4 billion to 1/2 a billion cells in visual cortex. And that that in humans is extremely difficult to stimulate. No one is I mean you get some flashes of light if you do that, but you don't get structured form vision. So about a year ago, I got a chance to sit with Nolan Arbaugh. He's the first patient of Neuralink uh company that you co-founded. He's quadriplegic. And since his accident 8 years ago or 9 years ago now, he hadn't been able to use a computer. And he got the implant like you talked about. And as you suggested, he started to be able to play video games. In fact, I played a video game against Nolan, and he beat me. Yeah, it's pretty cool. It's amazing. >> Yeah. And one of the things that he told me was that he was able to move faster than a typical human game player because often times we have the intent to move and then we move. We think about moving somewhere and then maybe we press something on a joystick, and then it moves. But all he has to do is think about moving let's say a cursor somewhere, and it's there already. And so he told me that it's given him superpowers, and he could probably beat world-class video game players at some point because his ability to move is much faster. So as we continue to plug computing into the brain, where do you think that's going to lead us? Do you think we're going to create a new class of humans that have superpowers over the rest of us that don't have brain computer interfaces? I mean I think it's important I mean So first of all, that is extremely cool. And yes, it cut a couple like tens of milliseconds off the connection from motor cortex out to the muscle. But I mean it's possible certainly. Um it's These are very serious brain surgeries. I think that it's tough to imagine kind of healthy 30-year-olds getting these soon, but many many people as you age, the body fails, and many people eventually I think will become patients. And yeah, when I think about the next generation BCI technologies that are on the horizon that are getting developed, I think it's very possible that in the next four or five years, there'll be some some patients that maybe had a stroke, maybe had um some other some other injury who you kind of go from like this terrible thing happened that in retrospect kind of became like this amazing opportunity. Um it's I mean it's a little tough to imagine, but yeah, I mean these things are going to substantially change the world. We have about a minute left. Uh this is one of those technologies maybe similar to AI that has a lot of buzz, but then you ask, okay, well, where is the business plan going to be? And we sort of stop to think for a bit and say, how's that exactly going to work? And is it just going to drive insurance costs higher? So what's your perspective on how to make this a profitable business? >> Oh man, I don't know if we can answer this in a minute. But I mean yes, in the near term, the for all of the BCI labs, it's these are reim- medical devices that will be reimbursed by payers. And these are for serious medical diseases. But I think that this is I mean there's a deeper tension here in healthcare. Like as techno- as time goes on and there are more technologies that produce better outcomes for more people and allow them to have higher qualities of life for longer, there's just more stuff to spend money on. But healthcare is kind of fundamentally this fixed bucket of money that increases very slowly. Like 20 years ago uh phones and computers were like they'd fallen in price by almost 90% over this time, but we spend 10 times plus more on them. And this This is a great thing, but if we spend if we had new technologies that meant that we could spend 10 times as much on healthcare, this would be a catastrophe. And so I think reconciling that like as these technologies uh really improve and we are really able to extend life and improve it um for for many decades, this is going to be There's going to be some kind of reckoning in healthcare. Well, it's a fascinating new frontier. I've called 2025 the year of the brain computer interface. And as I watch the progress from companies like yours, I'm feeling really good about that assessment that came in before the year. So Max, thank you so much. Thanks. Thank you everybody. >> [applause and cheering]