David Eagleman, Neuroscientist and Author, on the Brain as a Neural Parliament, AI Hallucinations, and Rethinking Democracy with AI

00:09:26 · David EaglemanPlay from here

[00:09:26] S0: Exactly right. I mean, there are real differences between the way it's working and we're working. But to to my mind, it's, you know, like, clearly the proof is on the table that we're close, that we're figuring stuff out. But this issue about the LLMs not changing their weights, this is the issue about continuous learning. This is why they can't do the continuous learning that we can. Every moment that we are walking through the world from cradle to grave, our weights, our connection weights are changing as we take in new information. So there are these big differences, but boy, I am really optimistic that we now have good reason to think, Hey, you know what? This is computationally capturable, and we can surpass the human brain. I mean, it's already the case that, you know, just if if you had a friend that read every single thing on the planet, you could ask him about anything, you'd be really impressed with that friend. So even before we hit AGI, it's it's already already mind mind blowingly blowingly good. Good. Is recall equivalent to intelligence? You know, I don't think it is, but LLMs are doing so much more than recall. I think a big part of intelligence, or let me say for sure, human creativity, is the remixing of ideas. So what we do is we absorb things our whole lives, and we remix them, we bend ideas, we break them, we blend them, and we're coming out with new versions of things. That basic cognitive operation is what has led to our civilization. We build skyscrapers, we compose symphonies, we get off the planet, we do all kinds of stuff because we can take in the data and do something new, and LLMs do that too. They're really good at remixing information. And and, of course, you can prompt it anyway you want. You know, give me a poem about Godzilla standing on Mars and petting a cat, whatever, and it does it effortlessly. It's it's really good at this, and I think that is, one of its greatest strengths. In fact, just as a side note, a lot of people complain about AI hallucinations. I think often hallucination is the same thing as hypothesizing, which is what we do all the time. As scientists, we say, hey. What if this crazy thing were true or something? And then, you know, the only difference is that as trained scientists, what we do is we make a hypothesis, and then we follow it out. We say, Okay, what if that were true?

00:11:49 · David Eagleman, Dan TurchinPlay from here

Okay, this is what would happen, and then you see if it explains the data better. When Einstein says, Hey, what if I were writing on top of a photon? What would the world look like? I mean, that's that's great. So, anyway, I... When when I hear about AI hallucinations, I I sort of feel like, okay. Well, that's a good example of of remixing things, and maybe once in a while, get something good out [00:12:13] S2: of that. Most of the community that feels like we're rapidly approaching AGI, artificial general intelligence, also feels that LMs won't get us there and that there's maybe a neurosymbolic architecture or something post LLM. You... You're pretty complimentary of LMs. Do you think that this is the... I shouldn't say end state. That's not the right right terminology, but how far can we go with LMs?

00:12:40 · David Eagleman, Dan TurchinPlay from here

[00:12:40] S0: Yeah. Great question. I think LLMs won't be the final thing, and already exactly as you said, lot of people are working on ways of mixing things. And I think the models currently by Anthropic and OpenAI and others are actually mixtures of things. A lot a lot of this proprietary and locked away exactly what they're doing, but it's not your grandmother's LLM. And I I do suspect that the transformer model, this particular architecture, has its limits, but boy, everybody is working on this now. I mean, every kid I teach at Stanford, that's all they're doing is, hey, what's the next thing that I can do here? How can I contribute? And so this is going to move very rapidly, but yes, I do think we'll need other pieces of architecture. I think when we look at the human brain, what we see is it ain't one thing. It's not like, Oh, here's the structure. Here's the architecture. Instead, it's this weird melange of all these different architectures smooshed together. This is how evolution always works. Evolution never says, Okay, I want to solve this problem, and then stops. Evolution just keeps going and often comes up with many, many solutions to the same problem. And this is what we see in the brain is that... What I described in my book Incognito is that the brain is a team of rivals. That's that's how I put it. I was basing this off, Doris Kearns Goodwin. Kearns Goodwin. Right. I [00:14:08] S2: almost Yeah. Did What's that? The reference to Abraham Lincoln? Exactly right. His cabinet [00:14:12] S0: was... Exact... Exactly right. He he I love that.

00:14:12 · David Eagleman, Dan TurchinPlay from here

[00:14:12] S0: was... Exact... Exactly right. He he I love that. [00:14:16] S2: Framework, by

00:14:16 · David EaglemanPlay from here

[00:14:16] S0: the way. Yeah. When he got elected president, he said, you know, look. I I wanna get all my enemies into my cabinet so I can have this this this diversity of viewpoints. So, anyway, I've suggested that that's really the way to think about the brain. If I put some, you know, chocolate bonbons in front of you, part of your brain wants to eat that, part of your brain wants to not eat it because it's all sugary, part of your brain wants... You know, says, I'll eat it, but I'll go to the gym tonight. You can argue with yourself and cuss at yourself and contract with yourself, and I think this is often overlooked, that we are made of lots of different drives. By the way, this is something I wrote about some years ago, but I just recently found a quotation from Friedrich Nietzsche, who casts this in his own language in the eighteen eighties. He said, essentially, that the brain is made up of lots of drives, and each drive philosophizes in its own spirit, by which he meant, you know, each one tries to stand on a soapbox and say, look, this is why you should eat the bonbons. Another one says, hey, look, this is a really good reason why you shouldn't eat the bonbons and so on. They're all philosophizing, trying to get what they want, and so the way I think this comes together is that the right way to think about the brain is like a neural parliament where you're made up of all these different political parties, all of which love their country but have different ways of steering the ship. And so they all vote, and at different times, by the way, the vote will tip in different directions, and that's the direction that the ship of state moves in, you being the ship of state. So coming back to your question, what we're gonna get with AI is if if it comes out like the human brain, we're probably gonna have lots of different kinds of architectures bolted together and some sort of voting, and and that's how it'll operate. And and by the way, that's, as I said, currently, how a lot of systems already are. For example, if you ask CHAT GPT a math question, it doesn't go through the LLM. It detects it's a math question and sends it off through the calculator. It saves a lot of, you know, resources that way.

00:16:33 · Dan TurchinPlay from here

Ad[00:16:33] S2: I'm proud to share details from today's sponsor. It is the AI Playbook, mastering the rare art of machine learning deployment, a wonderful book published by repeat guest on this podcast, doctor Eric Siegel, a prolific speaker, writer, professor, rapper, I dare say. It is a great read, very entertaining, very relatable. It presents the gold standard six step playbook for ushering predictive AI initiatives from concept to deployment. It is just coming out in paperback form. It's been a crowd pleaser since it launched a few years back, and you will so enjoy learning from the great doctor Eric Siegel. Go ahead and head out to bizml.com to get your copy of the AI playbook mastering the rare art of machine learning deployment from doctor Eric Siegel. Now back to today's amazing episode. We are gonna spend some... Not not Dave and Dan, but other people will spend somewhere around a trillion dollars on compute infrastructure, building up data centers and things in the next twelve months. And this conversation, and particularly, I'm so glad you referenced the brain running at 20 watts. It's just a reality check about, you I think you call it the three pound machine. Right? Yeah. The inner cosmos. Right? It's such a stark reminder about what's unique about the brain and why kinda synthetically generating them in data centers feels a little bit like a brute force approach, a little bit naive. What's your sense about the approach we're taking and kind of the... What I'll call brute force, you can choose to use a different term, but is that the best path to synthetically generating intelligence? I I I think I probably share the intuition that