Wetwear

0:13We'll see you next time.

0:30Hey, this is Andy Mills, and for today... Just what are digital computers? The computer. Are they man-made monsters that perform mathematical miracles in millions of a second? From where we sit today, it's actually kind of funny to look at how we were thinking and talking about the computer just decades ago. Are they superhuman machines that can solve any kind of problem? How mysterious they seemed to the general public. No, nothing miraculous at all. nor monstrous. The working parts are transistors, vacuum tubes, magnetic devices, and other electronic components. But we've got to remember that the computer is still a relatively new normal in the world.

1:13For example, in 1980, only 1 % of Americans owned a computer. Let those think in, 1%. When the personal computer industry first got started, there was robust discourse for many years about why an individual would even want a computer, let alone need one. And of course, now, it's almost impossible to think of what would happen to the global economy, to our personal connections, to the very infrastructure of our modern society without these technological marvels. But our story today is about how the computer itself may be about to take its most radical and bizarre step forward in its evolution. And how that might prove to be even more transformational to the future of the human race.

2:12Imagine opening up your computer, breaking it open, like unscrewing the back and looking about the guts inside. And instead of seeing, you know, wires and chips and hardware and things like that, you see living brain tissue, human brain tissue, powering the computer. So this is Gregory Warner, the host of our series, The Last Invention, and he's been reporting for the last year on this new kind of computer. and before we jump in with him i just want to say that this is kind of a special crossover piece for us here at longview if you're listening to this on the last invention feed great we highly recommend that you go and check out our other show reflector where this episode is also being hosted and if you're listening on reflector and you are curious about artificial intelligence about why there is so much excitement so much hype so much fear about this new technology we highly recommend that you go start back at episode one and listen to The Last Invention.

3:19And now, without any further ado, here we go. All right, Greg Warner. Yes. Thank you so much for returning to the studio to tell us... The weirdest story yet? Definitely the grossest. The most goopy, for sure. It's a goop-filled story. It's a little smelly, too. All right, so where do we begin? How do you want to start this off? Well, one place to start is with this guy, Han. Han Wang Chong, CEO and founder of Cortical Labs. Han Wang Chong, he's a computer engineer down in Australia. I got to say down in Australia. I don't know why you always do need to say that. But he was really interested in this question of how do we get a better way to teach computers to learn?

4:01And he stumbled across this essay. Written by Demis Hassadis from DeepMind, where he advocated for machine learning and AI researchers to go back to our origins, right? Which is in neuroscience. So he literally took that advice and went over to the neuroscience department at the University of Melbourne. And he asked for a tour. And so they're like, okay, no problem. They give him like, maybe they give him like a white coat and goggles and they take him around. And they show him at one point in the tour, this tiny glass chip. It's about the size of a fingernail marked with these metal dots. attached to wires.

4:40And on the metal dots are like a little cloud of tissue. And he's like, what is that? And they're like, well, that is mouse brain tissue on that chip. That captures electrical activity between neurons and provide a little bit of a stimulus back. But Han was a computer scientist. Like, he had never seen anything like this. And so I asked him, so how does that work? and they say, well, we get a mouse pregnant. They're like, oh, well, first we get a mouse pregnant. We abort the fetus. We decapitate the head. We smear the brain tissue or the stem cells of the mouse fetus onto this piece of glass. He's like, okay, thanks for letting me know.

5:25You guys are disgusting, but keep going. No, actually, he was like, this is awesome. And then we put them on the chip and they grow intricate neural networks and we get their activity. and we test drugs on it. Then the way that the lab was using it is that they would kind of douse these brain cells with drugs and then zap it and sort of see how the cells responded. Like shaking the drug onto the mouse brain cells? Just like sprinkling it? Well, the cells are living in a solution. They're living in a Petri dish that's feeding them glucose and whatever stuff the neurons like to eat. But it's also, you could stick drugs in there.

6:02Gotcha. So imagine you wanted to test a drug. You wanted to see, okay, if I pour this anti-seizure medication into the little Petri dish in which the neuron cells are living, will those neurons fire erratically? Will they fire more calmly? Well, I can even induce a seizure in those neurons, which is pretty amazing, but I can induce a seizure and then see if the drug will calm that seizure. So it's a really useful way. Instead of giving a mouse a seizure and then feeding it the drug, I could just see real time what's happening in the brain, or to the brain cells, rather. It's almost like, you know, going straight to the source.

6:39Yeah. So he looks at this, Han looks at this, and he's like, wait a minute, why hasn't anyone tried to get these neurons in a dish to try to do some sort of computer intelligence? What if we don't just zap it and see how it behaves? Like, what if we could send it electrical information and have it respond? What if we could get it to compute? And I'm sorry if this is an obvious answer to this, but like, why would he even think that's possible? Right. It doesn't sound like a thing from like a biological lab. That sounds like a thing from science fiction. Well, there have been biological computers in the past.

7:18Kind of one of the earliest experiments of this was in the late 90s. It was something called the leech-olator. It used leech neurons attached to wires to basically perform simple addition. so it could like send in a two and send in a four and the leech neurons would send out a six. We're talking like blood-sucking leech wormy things? Oh, yeah, yeah, yeah. And they chose those neurons because those leeches apparently have like these huge neurons so they're easy to work with. Okay, so some people had already made a calculator out of the neurons of leeches and he's seeing what's going on with this mouse brain and he's like, let's build off this.

7:58Yeah. The way Hunt puts it is you can think about a neuron as like a little mini computer. It takes in electrical information. It processes that information, kind of does something with it, and responds. And of course, when there's a whole bunch of them in our brain, they could do a lot of big time thinking. But the thing about the leecholator and some other later biological sort of type experiments, yes, they showed that there could be some kind of computing that happens in terms of input outputs, but nobody had figured out a way to actually teach a neuron to learn a skill. I mean, the leechulator could add numbers, but it didn't become better at adding numbers over time.

8:35It didn't suddenly learn multiplication. So Han was like, okay, I need to just figure out a way to not just talk to these neurons, but actually get them to learn. Amazing. Well, before we get into how all that works and how he goes about it, I just would love a mental picture here. What does he actually build? What should I have in my mind's eye? Is the contraption the device that he puts together here? I don't even know the right words for it. It's like, I mean, well, it sort of looks like an Xbox. I mean, it's the size of one. Okay. But it's got a glass cover where you can actually peer inside.

9:13And what you see inside, it doesn't look like any computer you've ever seen. There's no circuit boards. There's no sticks of RAM in there. Instead, it kind of looks like a little fish tank, honestly. I mean, it's got a tube for humidity, an oxygen pump. There's a heater. And then at the bottom of it, there's Petri dish. And inside the Petri dish is a little glass chip. And there's that smear of neurons on the glass chip. You wouldn't even know it's there. It's just a little filmy cloud. Just a little bit of living brain. Right. Although in his case, he did not use mouse neurons. He used human neurons.

9:56All right, so he's got these human brain tissue, these real human neurons alive, living inside of this thing. Yeah. And they're sitting on a chip? Right. And under that chip are all these wires coming out to the back of the computer. So basically, when you type into that computer, when you program into that computer, you are sending electricity to those neurons. And those neurons are talking back. and they're sending electricity out. And so you have a two-way electrical communication with human brain cells. Wow. And Greg, before we continue on with the story, I have to ask, does it smell? It kind of smells.

10:36Does it smell like a pet shop or something like that? It smells like, I think it smells like a hospital. Sort of like that faintly sweet, but also sterile. Right. Also like kind of plasticky metal smell. Interesting. So, okay, he builds this contraption with the living human brain tissue in it. What does he actually do with this device? So Han was like, okay, I need to just figure out a way to not just talk to these neurons, but actually get them to learn. And he decided to start, again, maybe inspired by his icon, Demis Asabis, with a game. And he starts with a very, very simple game. a game called Pong invented by Atari.

11:21Pong. Now at last you can play at home. You know Pong, right? Classic. It's a classic. It's actually still fun, even though it was the first ever commercially successful video game. And for those who maybe are not familiar with Pong, basically it's like table tennis. You have a paddle on each side and you have a little digital ball that's bouncing back and forth. The only job is to keep that ball alive. Follow that ball and just make sure it doesn't pass you. And that's what he's going to teach these brain cells to do. Okay, so how is he going to go about doing that? Walk me through this. What does one do?

11:57I don't even know where you begin in this situation. I mean, if you think about how you program a computer, actually, I don't even know how this works, but you basically send it binary code, ones and zeros, and it's able to decode that information and make all these kind of amazing outputs. But the human brain cells will not understand ones and zeros, And if they do, they will not respond to ones and zeros. So Hans says what he realizes early on is that to work with a biological computer, it's a lot more like training a puppy or like teaching, I don't know, a baby because you're using reward and punishment.

12:33So for instance, as it moves further to the target, it gets rewarded. If it moves further away, it gets punished and all that kind of stuff. And so then it's like, okay, well, that's fine. But what are the rewards and punishments that will work on a little neuron in a dish? Like you can't give it a dog treat. You can't pet it. You can't tell it, you know, good boy. Right. And if you think about like, well, how humans respond, okay, we have lots of rewards and punishments all the time in our brain, but we also have drives. We are seeking pleasure. We're seeking warmth, safety, sex. I mean, social acceptance, like all these reasons that we make the decisions we are making.

13:09Neurons smeared in a chip have none of that. But they do have one core thing that they want. And they want it really badly. What they want is predictability. Interesting. We're using the word want here, but it appears as if the cells... Seek. They seek to minimize... Well, the technical term is they seek to minimize prediction error. They seek to minimize surprise. And I mean, the closest analogy I could think of was being in like a club, a dark club that is crowded. And you listen to the music, you look how people are behaving, and you basically dance along with them. Like you anticipate where to slide, where to back up.

13:58Where to shimmy. Absolutely. But so imagine the DJ changed the music. every second. And you were constantly trying to figure out, is it house? Is it techno? Like, what are we doing? And you would get all confused until the DJ was like, okay, no. I'm going to give you the same consistent music. Then you're going to predict what's going to happen next. Then you'd be, then you'd know what's happening.

14:34So you're saying that he learns that on a cellular level, our neurons, they like predictability, or at least they seek predictable information, kind of like a predictable beat that they can dance to. Yes. And so with that baseline of knowledge about what the cells seek or what they want, how does he build off that to get them to do what he wants? Okay, so this is where it gets a little weird. If, for instance, the neurons are doing the right thing, we want to give them a very predictable piece of information. So what is the most predictable pattern you can give a pile of neurons? Turns out there is something.

15:18It is a very beautiful sine wave and it goes like...

15:28Just it goes like that forever. And that is like... It's bacon! Bacon treats for neurons. They love it because it is predictable. Now, what is the most unpredictable pattern you can give a pile of neurons? White noise. Just random. It's the equivalent of static. It's totally random. It's totally unpredictable. So that's kind of like the punishment? Yeah. So if you can imagine, here is Han in the lab. essentially watching in real time as these neurons are playing pong and he can see if they are playing the game well and if they're playing the game well oh get more sine waves oh you played you you hit the ball again great job sine wave sine wave oh man you missed the ball you didn't move the paddle blast him with static you know blast him with white noise and then okay now you're playing good again okay you're playing better all right we're gonna keep giving you that sine wave oh you messed up again sorry guys bam you know like blast them with a white noise and just with this reward and punishment really simple really familiar sort of method these neurons mastered pong in like five minutes wow that's amazing

16:56At the risk of, you know, over-anthropomorphizing these brain cells. Well. I mean, they are human cells, right? Right. These are human cells. Yeah. Is there any cruelty in this? Like, are these living cells experiencing something like suffering or pain? You mean not being in a brain, not doing what evolution told them to do, but instead getting zapped by Han? Yeah, and forced to play Pong, or else they get zapped. No, no, they love it. They absolutely love this. Who wouldn't? No, I mean, the truth is, I talked to Han about this, and he's just, really, he doesn't know. Like, he does not know for sure.

17:42The scientists really knew on this. Like, one of the things that I learned doing this, which I found quite interesting, is that the number of neurons actually matters. So, for example, in our brain, I think there's like 86 billion neurons, and that's enough there to have consciousness. 200 ,000 neurons, which is what Han is dealing with, that's about three poppy seeds worth of brain. So it's tiny. Yeah, and apparently that's just not enough to do any real higher level thinking like, am I happy right now? However, that said, there is a technique that Han uses to train these neurons that I think definitely starts to cross the line and make you think, maybe these things are feeling something.

18:30Basically, if he blasts them with white noise and if they don't listen, like if they're still, I guess, misbehaving or not playing Pong very well, he could give them... We call it the silent treatment. The silent treatment. There's no information at all. It's just darkness. Because the only thing that neurons hate more than white noise is no information at all, like zero.

18:59So essentially this is like a cellular form of solitary confinement for these neurons. Well, even worse, because it would be like if you were in a solitary confinement and there were no walls, no lights, no feeling, no bed that you were sitting on, no sensation at all. And you're saying on a cellular level. Like humans obviously hate that. But it seems as if these cells also hate that. Well, let's just say they react very negatively to this. And either they fall into line and start playing Pong a lot better, or, and this is the other odd thing, sometimes they just go into what is kind of like a coma.

19:42So a comatose brain is a very distinctive electrical firing signal, and these neurons start to fire in that coma way. And once they're locked into that state, he can't reach them. Like there's nothing he can do. He just has to toss the whole thing into the garbage and start again. Oh, this is crazy. It's, if you leave them in solitary too long, they can't recover. So it's into the trash and a fresh batch of brain into the box. Precisely. Okay, but this works. He pulls this off. He pulls this off, starts a company, and he starts building his computer. Like a computer he's going to sell to people.

20:23A computer he's going to sell. Like a real-life biological computer. I think the most recent video they just put out was that now it's not just playing Pong, it plays Doom. Which is a much harder game. Who is buying these computers? Why would anyone want this computer? I mean, especially in a world where you can buy a Mac computer that doesn't smell like a hospital room or need oxygen. So one hope of actually not just Han, but a lot of folks looking into alternative computing is that this could change the energy equation of AI, right? Because right now, ChatGPT, Claude, all these AIs, they consume an enormous amount of electricity, of resources.

21:09whereas the human brain cell i mean it is remarkably efficient i mean think about all the decisions you made this morning on like coffee in an oat bar i had a smoothie but i hear i mean that's efficiency that is like the computers can only dream of there's no way that any hardware can ever match the efficiency of your brain. So, like, if you could swap out some of that hardware for wetware, then it would be so much more efficient that essentially the data centers that are currently like four or five football fields in size would only need to be the size of a Hershey bar. Holy shit. That'd be amazing.

21:54So that's one of the dreams. Right now, though, this biological computer is for sale and people are actually buying it. And who exactly is buying this? What are they doing with it like right now? Right. So Hans says that there's three categories of buyers in his words. One are actually medical researchers. Remember the medical researchers were interested before in testing their drugs on these things. So now they're like, great, we have a new tool to test drugs and not only see how the cells behave, we can actually see how they compute under these different drugs. So like we could get epileptic human cells, put them on one of these slides, basically put epilepsy medication in the Petri dish, feed them epilepsy medication, and see if they learn Pong faster.

22:38And if they don't have a seizure, you can see the neurons having seizures. So it's super useful in developing new drugs. All right, so first people, medical researchers. Category two is what he calls the crypto gamers. I'm not sure what to say about this category other than something to do with quantum biology. I don't know. But nevertheless, that's another episode. But the third group, Hans, are the most interesting and honestly maybe far more consequential. I think it's definitely the robotics people that are most exciting to me. Roboticists who are trying to figure out a way to put these biological computers into humanoid robots.

23:21And so if you could imagine the future robots instead of having silicon brains or just silicon brains, they would have brain cells inside them. You know, you could feed in sensory information from the external world and have outputs from the neuron actually control some sort of wheel or an arm. Wow, so this would be that you essentially are a god in a new Garden of Eden creating an entirely new organic-based intelligent life form. Right. And some of them believe this is actually how we get to AGI. This is how we get the superintelligence. Superintelligence. And on that cliffhanger, after a short break, we will meet one of these roboticists.

24:15Stay with us.

24:32Hello, everyone. This is Matt, co-founder here at Longview, where we report stories that are grounded in curiosity and context, not political bias. As we say, it's not the left view, not the right view, but the long view. One way we sustain this business is by advertising, but we are also listener supported. And if you would like to go ad free and support us at any dollar amount that you'd like, you can do that by clicking on the link in our show notes or by going to longviewinvestigations.com. Until then, here is a brief message from our sponsors.

25:13The Last Invention is brought to you by Quince. Every spring, I feel the same urge to clean things out. Closets, habits, routines. And with clothes, I keep coming back to the same idea. Fewer of them, but better quality. That's what I like about quints. The materials feel elevated, and the cuts are clean, and the prices, they aren't insane. They make everyday staples using premium fabrics like 100 % European linen and this incredibly soft flow knit fabric for their active wear. The kind of gear you end up wearing way beyond the gym. What surprised me most is the pricing. It's about 50 to 60 % less than what you'd expect from comparable brands.

25:51And that's because Quince goes straight to the source, working directly with ethical factories and skipping the middlemen. So you're actually paying for the quality itself instead of the markup. My favorite piece these days is one of their blue chore jackets. It fits great, it's really comfortable, and I know it'll hold up even though I wear it all the time. Refresh your wardrobe with Quince. Go to quince.com slash last invention for free shipping and 365 day returns. Now available in Canada too. Again, go to quince.com slash last invention.

26:29Now more than ever, technology is a dominating force in our lives. Then there's the threat of AI everywhere. And yet, tech can be inspiring and help level playing fields. I mean, a YouTuber with a self-funded debut movie just dominated the box office. I thought, hey, if you interview me, it'd be good for your publication. And that's not ego. I just have a lot of followers. But it's that stigma. It's like YouTubers, they're not real. Join me, Lizzie O 'Leary, the host of What Next TBD, Slate's podcast focused on technology, power, and the future. Follow What Next TBD now, wherever you get your podcasts.

27:19All right, I'm Andy Mills. You are listening to both Reflector and The Last Invention. Very special episode, and we are back with Gregory Warner. So there's this classic conundrum in robotics, which says that things that are really hard for humans, like chess and math, things like that, are really easy for computers. But things that are super easy for humans, like catching a ball or walking upstairs, are really, really hard for computers or robots. I mean, we've, for years, decades, have expected robot workers around us to be doing the laundry and doing all the messy labor in our factories, but that's just not happened.