2 Ex-AI CEOs Debate the Future of AI w/ Emad Mostaque & Nat Friedman | EP #98

What I think happened in the last year really is the starting gun was fired.

We spent $1 trillion on 5G. Is AI more impactful than 5G? Of course it is.

Do we actually know what's going on inside the models?

We can measure everything and still we don't really quite understand mechanically what's happening inside them. They're not engineered and designed. They're kind of grown.

AI is clearly more capable than humans now.

We know this is going to get better. When does it slow down?

Right now, it's a few companies doing this, but we need standards around it and we need the expansion. And again, every country will invest in this. Every company will invest in this.

We're here to talk about what happened. I almost say WTF just happened in the past year because a lot has happened. Imad, you were on stage with me a year ago and things were amazing then, but a lot has happened. So I want to split this into a couple of parts, like What just happened in the last year? And then I wanna talk about how far forward can you see? What do you expect is gonna happen? You know, I'd love to talk about the companies, the models, the dramas. What did Ilya see? I think that's gonna become a meme.

Yeah.

Nat, do you wanna kick us off? What's the big things that happened in the last calendar year in your mind?

Well, yeah, it's, it's been kind of an amazing year. I do think for people who just started paying attention to this sort of new AI and deep learning revolution in November of '22 when ChatGPT came out, things probably seemed very fast because we had Stable Diffusion come out over the summer previously in '22. We had ChatGPT in November, and then just a few months later, GPT-4 came out, sort of demonstrated a new level of capabilities that, you know, were shockingly improved over what was there previously. And so I do think there's a set of people who sort of extrapolated from those 3 data points in terms of the progress that was going to happen from there. And it's incredible, actually, how quickly people can adapt to these new things. And there was a point at which kind of ChatGPT blew everyone's minds. And then a few months later, It was sort of just something we accepted as part of the world, as part of reality. We can very quickly enter this slump. Where's the new model? GPT-4 is boring now. When do we get GPT-5? Come on, Sam, put it out already. What are you waiting for? This kind of thing. I think we had a little bit of that this year. Then over the course of the year though, what we started to see, and I'd been waiting for this for a while, but we started to see real adoption. taking off. Back in 2020, GPT-3 had come out. In '21, when I was running GitHub, we put out GitHub Copilot. It was, I guess, one of the first LLM commercial products. I thought that immediately after that, there would be this rapid wave of commercialization of large language models because developers would have seen how capable they were, what they could do, and start building products with them, but it really took the this kind of ChatGPT moment before entrepreneurs, developers, and product people started to do this. And so what I think happened in the last year really is the starting gun was fired on the actual exploitation of these raw capabilities, the building of products, the working of them into companies and organizations. And now that's what we see. The adoption's been incredible. ChatGPT is rumored to have gone from basically $0 to $2 billion in revenue in just about a year. That's amazing. There's, you know, organizations like Stability that put out models that have, you know, hundreds of millions of users. You have, you have Midjourney, you have now Google in the game finally with Gemini. I thought it was exciting to see them not only release Gemini but very quickly follow up with Gemini 1.5. And so there's a way in which Google has clearly internalized the pace here and the need to iterate and ship. And so that—

I don't think a lot of people realize how far ahead Google was for a decade. Yeah, right. So Google had really developed a lot of the early tech here and said it's not ready for release. Right. They were being responsible to a great degree. And then, and then how do you not release after ChatGPT goes— gets released?

They're shipping tilt now. They're going to ship.

They're going to ship. Yes. Emad, the open model story that you told us last year has really blossomed. And in fact, it's become sort of an ethos in the organization, right? It's like Elon is like really twisting the knife in with Sam and then says, but Grok's going to be open now. Can you talk about what's happened in the last year in open? Will open win? Do you think— and I'm curious for both you guys— and describe what open is versus closed for the audience here.

Yeah, so proprietary AI is you don't see the code, the weights, anything, and it's provided as a service, whereas open models and code are ones that you can adapt yourself, you can take, bring to your own data, and you own. And that's important because these models are something a bit different. They're like extensions of our mind. And so the best analogy I've said is that these are actually like graduates. So they can code, they can write, they can sing. They sometimes go a bit crazy when they try a bit too hard. You know, we'll give them better education. And so open models are like graduates that you hire, and then proprietary models are like consultants you bring in.

Interesting.

And in the early stages, we all needed consultants. Now people are saying, well, I want this for my own data, I want this for that. And open allows for innovation to happen. So we've had now 330 million downloads of our models on Hugging Face.

Wow.

So we just—

and what is Hugging Face for folks?

It's GitHub but for AI models. So it's where you go as a developer to download the models. And so people, millions and millions of developers are using this technology. But then you look at the language model side, people are taking it, adapting it, and optimizing it to have innovations that are now catching up with even proprietary guys. But they're complementary. You will have your own team and then you'll bring in the specialists. And I think that's the best way to think about this because you can't outsource your internal intelligence from a personal company or even country level. to models that you don't know what the provenance are. This is one of the debates that we saw over the last year. The first step was like, oh my God, exponential extrapolation. As Nat said, actually, the foundations for this were dug maybe a decade ago. Now we've been filling in the cement and now we're all building houses and we're like, well, the houses have become skyscrapers. It was like, well, they could kill us all and those are valid things to discuss. Then it became about sovereignty and, well, GPT-4 is amazing, but I want my own version as well for my own private data. And so I think as we advance, proprietary models will always beat open models.

Proprietary will always beat what?

Open models. Because you can take the open model and bring your private data to it and optimize it, because open models are like generalized graduates, as it were. But both of them need to exist. I think we'll see both of them continue to take off.

Do we actually know what's going on inside the models? I was listening to a conversation with the chief scientist at Anthropic, who will be on stage with us next year, and he's saying we actually don't understand really how these models work. Is that a fair statement?

Yeah, we have.

I mean, just so you understand, all right, it's like they're amazing, but we actually don't understand how the weights and connections and all are really working.

I mean, we understand it at this like very micro level, We can see the multiplications happening, and we can see the signal moving to the next layer in the neural network. But it's actually sort of similar, although not the same mechanisms, to the way we don't have a perfect understanding of the way thinking happens in our brains. We don't— even though in the case of our brains, when we try to do the neuroscience and understand what's happening at the neuron level and the organelle level, we're limited by our measurements, right? We can't measure the state of every single neuron in your brain while you're thinking. That's not something we have the sensing technology to do now. That's not a problem we have with these AI minds that we're building. We can measure everything, and still we don't really quite understand mechanically what's happening inside them. They're not engineered and designed, they're kind of grown. You know, they're the products of us, actually, because we have the internet, which digitized the world and put the sort of all the data we've produced online. And there's a way in which that process of building the internet was like a bootstrapping process for building AI because it was a precondition to making AI. We digitized the world, we put all the contents online, and then we could use all of that to sort of grow and train these AI models. But we don't quite know how they work. Now, there's a new field— new, not in time, but, but relatively small field called interpretability, which is about trying to understand what's going on in these things. What are they doing? How do they make the decisions that they make? Obviously, there's benefits if you can do that to making them better in all sorts of ways. You can make them smarter maybe and make, make better decisions. And you can also hope to make sure they do the thing you want them to do and not, not something else. But it's new.

Elon put a tweet out, which I read during my opening remarks, that said by 20— yeah, we're going to have human-level AI next year and by 2029. AI will be equivalent to all 8 billion people. Do you agree with that, Emad? Do you think it's going to move that fast?

I think we had a big discussion about generalized AI that can do everything, and that was the focus of DeepMind and OpenAI. But for specific tasks, AI is clearly more capable than humans now. So, for example, Google's Gemini Ultra model has a million, 10 million token context window. What does that mean? It means that someone can upload themselves debugging a piece of code in the codebase and it'll correct it. And no human could ever do that. You know, our RAM isn't big enough. You know, on image now we can generate images faster than anything, songs in seconds and other things.

So let's talk, let's talk about that a second. Um, Sora was, I mean, pretty like holy shit moment, right? Um, and someone is doing it right at OpenAI with ChatGPT and then Sora. Uh, but I remember last year You told us it used to take like 30 seconds to generate on Stability a single image, and then it was down to a second. And you've advanced the technology orders of magnitude since then.

Yeah. So, I think if we can put the thing up.

But yeah, put up the slide, Stable Image.

Yeah. And then if we can actually go to the next slide, talk about speed.

Here. Oh, should I click this? Let's see.

Oh, we can click this to give an idea of where it's gone. So image was kind of one of the things that kicked this off in 2022. All of these images can be generated on a MacBook.

On a MacBook?

Just from description. And now we've perfected text and other things, but the next step after you can generate all these beautiful images is that you want to move to control. So the models are just the first step. You have to have ChatGPT and other things to make them useful, but then you want to be able to take that guy and say upscale. That is all we said, and it upscaled him. You want us to replace a lion with a cat, and we can now do that real-time. You know, or tiger with a unicorn. Replace background with a forest. Now you have a lovely forest. All of this you can do pretty much real-time, because if you look at the next slide from this—

Okay.

I said it was 20 seconds to 1 second. This is live real-time. You can just type, and it automatically adjusts the cat. It gives him a hat. But then if you look at the optimization of that with the next slide, Back one. We missed it.

Oh, this one.

Okay, well, we missed it. Go back one slide. Sorry. With this process here, we're just releasing a new distillation model today. We've got it to 200 cats with hats per second.

200 cats with hats per second. So that's your speed of image generation. That's a new unit of image generation.

50 milliseconds per image.

Yeah, exactly. It's because there's not enough cats on the internet, right? So we're going to add more cats Oh, 5 milliseconds.

5 milliseconds.

5 milliseconds.

Yeah.

But I think with the new chips that are coming, NVIDIA is going to announce another one. We'll get up towards 1,000 images per second.

So that's real life. That's live video. It's times, times, times 30.

Yeah. And so OpenAI's innovation, there were 2 major things. There was the transformer architecture of the language models and diffusion that drive the media models. They combine the 2 together. So our new image model, Stable Diffusion 3, which is the best performing image model, combines those together as well. So if you go forward a few slides from here, Next one. Skip. Another one. Another one. That's the upscaling.

So you basically said upscale this image on the left and you get the image on the right.

We'll have that real time in a couple of years. So your boxy video games will look a lot more realistic.

So just literally as you're— you can take an old game and play it in surreal life.

Yes. But if you go to the next slide, these videos with our video model were all generated on a consumer graphics card with 5 gigabytes of VRAM.

So, I mean, the point is, it's— this is in everybody's hands.

It's in everyone's hands. And again, we haven't even optimized the data. So in an hour, next slide, we're releasing the world's most advanced 3D model. So all of these are just generated just from descriptions. And so the fastest version of this does it in 1 second. So you can generate that dragon. But then the next step is you'll be able to control every element, make its horns bigger, make these adjustments. So how long would that take for a normal creative? person in kind of industry a huge amount of time. But now it works on the edge and it works even faster in the cloud.

So a future here where you can be describing the video game you want created and it's generating all the characters and generating the play.

Yeah. And so if you go back a few slides to that thing with all the nodes, I think this is an important thing. Sorry, I didn't do the slides properly. I think one of the things that we're having right now is this is a system we built called ComfyUI that's used by just about everyone now. So you take the face, the pose, the dress, and then you have that output. But if I share that image with you, it reconstructs the entire flow. Because last year was the year of creation, models that create. Then ChatGPT ComfyUI allowed you to control and compose them. And the next bit is bringing that all together because ChatGPT, all the knowledge that you build from writing your speech, you don't have files anymore. You have flows with these models and assets there. I think that's the next step because, you know, when you go beyond just spitting out ideas to be able to control them like that, that's a huge deal.

Amazing. And you're announcing this in an hour?

The 3D model's releasing in an hour, open source to everyone.

Give it up for Emad here. You really have— and I don't know if you're able to say this, I mean, You're driving revenue and soon net profitability. What's your financial?

I can't say that publicly.

You can't. Okay.

But it's going well. We're ahead of forecasts. That's all I can say.

All right.

He told me backstage it was good.

Yeah. Okay. Go on.

We're doing all right. Yeah.

Everybody, I want to take a short break from our episode to talk about a company that's very important to me and could actually save your life or the life of someone that you love. Company is called Fountain Life. And it's a company I started years ago with Tony Robbins and a group of very talented physicians. You know, most of us don't actually know what's going on inside our body. We're all optimists until that day when you have a pain in your side, you go to the physician or the emergency room, and they say, listen, I'm sorry to tell you this, but you have this stage 3 or 4 going on. And, you know, it didn't start that morning. It probably was a problem that's been going on for some time. But because we never look, We don't find out. So what we built at Fountain Life was the world's most advanced diagnostic centers. We have 4 across the US today, and we're building 20 around the world. These centers give you a full-body MRI, a brain, a brain vasculature, an AI-enabled coronary CT looking for soft plaque, a DEXA scan, a GRAIL blood cancer test, a full executive blood workup. It's the most advanced workup you'll ever receive. 150 gigabytes of data that then go to our AIs and our physicians to find any disease at the very beginning when it's solvable. You're going to find out eventually. You might as well find out when you can take action. Fountain Life also has an entire side of therapeutics. We look around the world for the most advanced therapeutics that can add 10, 20 healthy years to your life, and we provide them to you. At our centers. So if this is of interest to you, please go and check it out. Go to fountainlife.com/peter. When Tony and I wrote our New York Times bestseller Life Force, we had 30,000 people reached out to us for Fountain Life memberships. If you go to fountainlife.com/peter, we'll put you to the top of the list. Really, it's something that is For me, one of the most important things I offer my entire family, the CEOs of my companies, my friends, it's a chance to really add decades onto our healthy lifespans. Go to fountainlife.com/peter. It's one of the most important things I can offer to you as one of my listeners. All right, let's go back to our episode. I am curious about the idea of how far are we from having an AI that I can have a conversation with and say, I'd really like to create a new business. That does this, this, and this, and just have that like a brainstorm partner AI, and it will do the incorporation, it will write the code, it will generate the marketing materials, and it will be able to— you know, we're all entrepreneurs here. That's all we, you know, find a problem, build a business, find a problem, build a business. How far are we from that reality of an AI created— well, I'll come back to the second part in a minute, but an AI that can be your thought partner and really staff up a company?

I think it's already happening gradually, and then maybe suddenly.

All at once.

We have— I think companies will— these models are neural, they're neural networks. And so I think the right way to think about it is that companies in the future will just be increasingly neural, and there'll be more and more of the company, and what used to be the departments of the company that are single models or swarms of models that are doing work. I think it'll, you know, at some point, probably very soon, we'll look back on 2024 and say, God, do you remember when companies had like hundreds of people in the finance department doing accounts payable and just like transforming information from one form of text to another, essentially, and doing this coordination? And so I think you will have some form of both existing companies that just adopt more and more AI because it gives them advantages, it makes them more efficient, Maybe it gives them better customer service. People enjoy the responsiveness, intelligence, politeness, clarity, etc., of the AI counterparty. And you'll also have new companies that will be AI native, started from scratch, neural from the beginning. So we— some people may remember when Instagram was acquired by Facebook for $1 billion, everyone was just marveling at the fact that a small company—

13 employees.

13 people could be worth $1 billion. How is that possible? And the joke was, you know, gosh, could you ever get down to a 1-person company that's worth $1 billion? And well, like, this is clearly going to happen. Maybe eventually a 0-person company.

So we talked about this getting ready, and I said, when are we going to have a scenario where a government— and we have some governments in the room here— says we're going to create a new regulation that allows an AI to incorporate on its own in our jurisdiction. I think it's— I think it's a winning scenario because all of a sudden the AI incorporates, you get the tax base, and every AI-incorporated company will move there.

I think Wyoming's done that.

What's that?

Wyoming has done that. They have a new structure called a DUNA for a decentralized autonomous organization.

I'm sorry, who has?

Wyoming.

Wyoming.

Of all places.

Yeah. So we have some Wyoming fans here in the audience.

Yeah. So technically that could happen today.

Amazing. I mean, I do think that that is a— so the speed of wealth creation, how should folks— so let's talk about what are the wow moments that might be possible in the next year? What are some crazy wow moments that might— you might see?

I think one of the most important things is the accuracy and then these long context windows.

So explain what the long context window is again.

So you're absorbing information right now, and it's quite high definition because you can see everything here and other stuff, but you're still writing it down. And the reason our organizations get big is because text is a lossy transmission format. We lose so much context. The final PDF loses all of that stuff. Now with the new Google models, Nat has some amazing companies in this space as well. You can upload hundreds of thousands of words thousands of documents and the AI can interpret them all at once. That doesn't need to be trained on it. So you can upload all of your ideas and say, build a business based on this and it will do that. Or you can upload like a whole bunch of movies and then tell it to write a script that incorporates all of that and it'll do that inference time. That again is something superhuman. But we all have these massive like repositories of all these ideas we've had. Being able to dump that now and then the AI without having to be trained spit back answers, ideas, and things like that, I think is a really huge step along with that composition step that kind of discussed before.

Yeah, I totally agree with that. I think there will be a couple of things coming probably pretty soon. It's hard to predict exact timelines on these things. Sometimes things happen faster than you expect and sometimes a little slower. But one of the clearly amazing, clearly possible now products to build is a voice-to-voice model that's indistinguishable from talking to a human, maybe for a conversation of up to a couple of minutes. Where—

what happens after a couple of minutes?

Well, maybe you can kind of just tell somehow that it's not quite human after a couple of minutes. I'm setting a milestone that I think is achievable this year. Maybe if you can do 2 minutes, you can do 12 minutes. I don't know. But yeah, you know, you would— it's actually about all the technologies there. It all just has to be integrated. And so you need the sort of the ability to recognize speech is there, the ability to interpret it with a language model and generate responses is there, and then the ability to turn that text into incredibly realistic voices there and kind of putting that all together into a package that has very low latency, that's talking the way we talk, where you can kind of interrupt me and maybe there's an avatar that's giving you this human-like— I mean, Aristotle was very impressive, but I knew that that was not a real person.

Right.

And so I think we could Yeah, he was a real person. Yeah, that's right. And then I think the other thing that's a very big deal is this, this idea of autonomy and agents. There's been a lot of talk of it with AI over the last year. Today, these things are not agents. They're tools. They're call and response. You go to ChatGPT, you type something, you hit enter, you watch the response kind of stream back. And I think what people don't necessarily understand is that when these language models are responding to you, that it's almost like a rap battle. They have a fixed amount of time to generate each word. And so they can't sort of sit there and ponder for a minute, you know, what they're going to say. They have to talk to a metronome. And so that's why when you see the words that they're writing out, that's not like they've thought about it a lot and then you see the words. It's actually the thinking is happening during the output.

Would you say that's how a human does it too? I often do. I mean, a lot of times, sometimes I'm really pissed off at what came out of my mouth because I didn't think about it in advance.

Yeah, yeah, I think so. I mean, Sometimes you have a conversation with a smart friend and you're just forming the words to respond, you have a better idea. I think we definitely do that too. The agent, the autonomy is about increasing the unit of work you can trust the AI to do without making a mistake. Right now, you can ask it one question, get a response, you interpret it, you figure out the next thing. What if it could go do 10 or 100 steps? You were talking about an AI business. Do you trust it to come up with the title of the blog post? that you're going to post, or do you, do you trust it to actually come up with the whole idea of the marketing campaign? Right. Come up with the strategy for how to execute it, all of the content it's going to generate, the partners it will reach out to and negotiate advertising or whatever it's going to do, have those conversations, multi-step successfully all the way to like measuring its results without supervision. And I think that agency thing, we're starting to see it in programming. There was There was a very impressive demo a week or two ago of a company called Cognition that I think it was maybe the— it was arguably the first really impressive demo of working agents in AI. And they did it with GPT-4. So not with a new brand new model. They did it by being very clever about the way they squeeze and distill the intelligence out of GPT-4 by repeatedly calling it. and analyzing and evaluating its results and choosing the best ones. And so sort of going from rap battle to like draft and redraft and sort of think and ponder about it a little bit harder mode. And it can do hundreds of steps in programming successfully.

So I think you might—

Okay.

I was going to ask you one quick question, then please jump into that. How far out can you imagine right now?

About 3 years.

3 years is a max. Yeah. How about you, Nate?

Well, you can dream. You can dream. Yeah. So I actually— the thing I find easier is to think about the long term because the—

yeah, where the asymptote of where we're—

yeah. And that's the key question of our time is the asymptote. It's like we know this is going to get better. When does it slow down and what's the shape of the curve to get there?

You might— I cut you off. Apologies.

No, it's fine. I think probably the thing to look forward to is the last couple of years— the first year was about the technology and the breakthroughs in the research. Last year was about the models. Next year going forward, no one will give a damn about the models. It's all about what you can do with the models, bringing them together, because the models have satisfied. They've got good enough, fast enough, and cheap enough. They will get even better, and there's probably 2 orders of magnitude improvement still. in the speed and reliability of the model. But this will really become about— and the stories of the next year will be about— we use this model to do this and it was amazing, you know. And so I think that's one of the things to look at. And that's what Nat said about tying them all together. You've got the ingredients now. What are the recipes we're going to make?

Very quickly, capital and regulation. I was in a conversation with the CEO of one of the major AI companies, and just, he was saying he's raising $3 billion. And I said, you know, I have a venture fund. I said, great. And I said, what's your minimum? And he said, probably $100 million. I said, okay, well, that's out of my ballpark. But, and how quick do you think you're going to raise it? And he goes, next month. I mean, there is a lot of capital flowing in.

Yeah.

I mean, have you ever seen a capital rush faster than this?

No, no. I mean, I think there are a couple of comparisons you can look at. I think the railways— when railway infrastructure was being laid in the UK, I think it was some double-digit percent of GDP investment. I think the solar explosion that's happening right now is actually pretty amazing. It's something like half a point of global GDP is being invested in solar. So those are pretty big. We're nowhere near that yet with AI. And so I actually think if AI keeps working, which it seems like it will, you should expect the future to look more like that railway or solar situation where there's— you're measuring the investment in intelligence because it's so valuable. Intelligence— AI is intelligence. Intelligence is power. Power is valuable. It's power over nature. It's power over others. And so you'll probably measure the amount of investment in it in points of global GDP. And whether that happens in 2 years or 10 years, I'm not sure, but it seems likely.

I mean, to put this in context, less money has been spent on private AI companies than the Los Angeles-San Francisco Railway that hasn't started yet.

There you go.

Wow. Right.

But that's almost done, I heard.

So really? Fantastic. They haven't even started.

We wrote down on it.

The AI right now isn't infrastructure, but it should be.

Which is why—

talk about your vision there, because I find it very powerful. Your vision there in education, in health. Talk to me.

This is a next-generation human operating system because these AIs extend our capabilities. And again, you'll need the AIs that are open and the ones that are proprietary to have the best outcome for everyone, because all of our companies here are all information systems, and we've seen how bad it can be. So, the total amount of spending in this will be trillions of dollars. We spent a trillion dollars on 5G. Is AI more impactful than 5G? Of course it is. Should it be infrastructure? Of course it should be. Should the data be transparent? Of course it should be, because we need to know how our railways are made, the information knowledge superhighway of the future. Right now, it's a few companies doing this, but we need standards around it, and we need And again, every country will invest in this. Every company will invest in this. I mean, is anyone here who runs a company not investing in this in some way?

At least your time, right?

Yeah. Multiplied by every company in the world.

So where are we percentage-wise at the investment in AI? Are we at a fraction of 1% still?

I would say so, yes.

How about you, Nate?

Seems likely.

So a lot of upside still. Opportunity.

It is crazy, though. I saw a tweet the other day. I mean, I lived through dot-com and All these, you know, little tech bubbles. And I saw a tweet the other day from someone that said, if you don't secure equity in an AI startup now, then your children will be chattel slaves for the machine god for all eternity. And I remember thinking, I don't remember anyone saying that during dot-com.

Yeah, like, and then you switched around and bought some more GPUs, right?

This is somehow a little bit more extreme. than the prior bubbles.

So, I mean, look, I think Web3 kind of received a lot without any results, whereas now you're seeing impact here.

Yes.

But again, you multiply this by the number of people it affects, the value created, it's insane because it will go orders of abstraction. You've had the base, the foundations, now the base. Now you're building the houses, you're building a whole ecosystem around this. And the transformation that that has and the amount of capital is bigger than anything we've seen.

Let me ask the question we're going to be debating today on The Asymptote. How concerned are you about digital superintelligence? I'm defining this for a purpose of conversation as AI a billionfold more advanced than the human being. How do you think about that? What's your position in that debate? Pro, con, any way you want to kick off.

My belief is that humans can break the atom or we can go to Mars. And so my vision is every single nation, person, company, country, culture has their own AI datasets and self-sovereign needs to figure out the governance of this because it's important. And then that AI is our collective intelligence, is the human colossus. So it isn't controlled by any one individual, isn't embodied like that. But again, it's amplifying all of us and it's solving every single problem that we can have. I think that's a positive version of the future.

I think it seems really likely that— I mean, it seems— Undoubtable to me, actually, that intelligence is just a material process like muscle strength. We have organs called muscles, we use them to move. We have an organ called the brain, and we use it to think. And so, if you look at— if you just sort of ask what that means, well, we've managed to exceed muscle strength with artificial artificial machines, with machines hundreds of years ago, and, you know, we're gonna do that with brains too. We're gonna have artificial minds that are much more powerful than ours. It's almost like you just have to be an AI doubter not to believe that, or you have to not believe in human ingenuity. All the most brilliant people in the world are now working on this. And there's a huge amount of capital going into it, and in a way we've just started. And so the idea that it wouldn't improve seems hard to believe.

Did you know that your microbiome is composed of trillions of bacteria, viruses, and microbes, and that they play a critical role in your health? You know, research has increasingly shown that microbiomes impact not just digestion but a wide range of health conditions, including digestive disorders from IBS to Crohn's disease, metabolic disorders from obesity to type 2 diabetes, autoimmune disease like rheumatoid arthritis and multiple sclerosis, mental health conditions like depression and anxiety, and cardiovascular disease. You know, Viome has a product I've been using for years called Full Body Intelligence, which collects just a few drops of your blood, saliva, and stool and can tell you so much about your health. They've tested over 700,000 individuals and used their AI models to deliver key critical guidelines and insights about their members' health, like what foods you should eat, what foods you shouldn't eat, what supplements or probiotics to take, as well as your biological age and other deep health insights. And as a result of the recommendations that Viome has made to their members, the results have been stellar. As reported in the American Journal of Lifestyle Medicine, after just 6 months, members reported the following: a 36% reduction in depression, a 40% reduction in anxiety, a 30% reduction in diabetes, and a 48% reduction in IBS. Listen, I've been using Viome for 3 years. I know that my oral and gut health is absolutely critical to me. It's one of my personal top areas of focus. Best of all, Viome is affordable, which is part of my mission to democratize healthcare. If you want to join me on this journey and get 20% off the full body intelligence test, go to viome.com/peter. When it comes to your health, knowledge is power. Again, that's viome.com/peter. Here's the challenge. It feels like potentially winner-take-all scenarios. Where if all of a sudden my company is able to utilize the most advanced AI and build out the next generation of systems, and I'm doing it with, you know, me and my versions of Hailey and Aristotle, that there's, you know, they're working 24/7. I'm feeding them as many GPUs as possible, and I have a chance to really outrun— outrun the competition. And it used to be that in the early days of the mechanical world, if you were using mechanical systems to outrun the horse, that was a local phenomenon, right? But now this is a global phenomenon because my bits are reaching around the world. And, you know, when you say, yeah, it can do my marketing, do my marketing campaigns, it can do my scientific analysis, it can do everything, And, you know, I keep on hearing that is like not 10 years from now, not 5 years from now, but that's like a 3-year scenario. I don't know how to think about that.

I think the, the pace of change, you know, is going to be very high. And global GDP growth has been kind of in the 2% range. And we have a society and civilization that's able to adapt to that amount of change per year, mostly, not entirely. It's been a little bit higher before. Maybe it's going to be much higher very soon. And whether there's a huge spread of outcomes that come from that, I mean, even if you don't have the extermination scenarios, the extinction scenarios in mind, I think you actually should have other scenarios in mind What does it mean to be human? Is the economy human-dominated, you know, 10 years from now or 15? Where are decisions made and who makes them? And then just, you know, this vision of sovereign AIs, you know, it's a new level of potentially cooperation and competition between countries and sources of this kind of power. And so a lot of change is coming. I think you should I think, you know, we went through this before. We had this sort of 5 big inventions of like the 1880s through the 1920s, and that was a wild time of transformation. It's like almost unimaginable. People who grew up with horse carts, you know, ended up riding on jet planes. And we're going to have at least probably much more than that amount of change happen this time.

Let me go to you, Imad, next. But when you're— you were about to go after this to part 2 of that tool. And I want you to be thinking about how can AI help you solve the challenges you listed? How will AI threaten the dominant positions or the strengths that you have? And how will AI this year help you meet your 2024 goals? That's the goal to putting it to use. Imaad, do you want to comment on that? And I want to also talk about medicine and You're dedicated to using AI models to solve autism and cancer and death, or let's just say sickness.

Yeah, sickness. Yeah, I mean, I just had a thought actually. We had Steve Jobs a little bit earlier. Maybe the AI is jet planes for the mind, right? There you go.

Here we go.

But the impact here, there's a sociological impact we can extrapolate way forward, but an ASI is just something we can't think. If you've got—

Advanced superintelligence. Yeah.

If you've got too many infinite supply of graduates, what do the existing graduates do? If the floor is raised, you don't need to hire as much, you become more efficient. What are the new jobs of the future in a knowledge-based economy? We have to think about that because as Nat alluded to, it's like slowly, slowly, then all at once like a turkey, you know, at Thanksgiving. I think if you look at this though, there's There's the negative side, but then there's a positive side. So I think mentioned kind of last year, Google's Med-PaLM model outperforms human doctors in medical diagnosis accuracy, but also empathy. And we have medical models like that. Anyone here who's had someone who's had multiple sclerosis, Alzheimer's, autism, something like that, you don't have comprehensive authority to update knowledge and nobody to guide you through that process, and you lose agency. Well, guess what? We will make open-source models and they'll be available to everyone and you You'll never be alone again through that. And then they'll be used for diagnosis and organizing on a larger scale.

Describe that a little bit more so people can feel what that feels like.

It means, again, anyone's experienced that, someone comes to you and says you have a diagnosis of Alzheimer's for your father or autism for your child. You lose agency because you're like, what now? There's no cure. There's no treatment. Where do I even go for information? A lot of our stuff is a coordination issue, whereas if we have a specialized medical model for that topic and retrieval augmentation and tie it into all these systems, you can have all that knowledge at your fingertips, and it can help guide you empathetically through that. It can literally talk to you. It can connect you with other people going through the same process. And then the next step is it can organize all the knowledge in this area because there's so many promising treatments. But what— how does anyone here find out all the treatments about autism or cancer?

There used to be patients like me, but it hasn't survived. But this It supercharges that.

And then you do that once, and it's infrastructure for the 50% of people around the world that will have a diagnosis of cancer. And again, you've transformed it because there will always be someone with every single person that can connect them to the right information at the right time. So this is the positive view of the future, and that creates boundless new potential in terms of both addressing our health, but science. And again, most of our science is based on files, a PDF, and we throw away all the stuff that doesn't work. If you look at that knowledge flow of that image with the different things, and again, if I drag and send you that image, it'll reconstruct the whole flow. I think most of our things will go from files to flows so we can remix our knowledge so that we can find out what works and what doesn't work. And that's how we get breakthroughs.

And is it true that most of the large language models were built on top of all the social media, Facebook websites, and so forth, Not crawling science magazines and most of the scientific databases out there.

Science was a part of it, but again, what we did is big compute was a substitute for bad data. It was like cooking a steak for too long. Now we see that high-quality data is even better, but we're only understanding what high quality means. And this is why we need specialized models which are transparent, especially for things that affect things like health, education, and others. So, I think dataset transparency will be a big deal. And this goes to Nat's point about interpretability as well. Like, you were giving that example of Arqiv earlier, right?

Science. You know, one of the things that I think is amazing is the potential for these AIs to help us discover new science beyond interpolation or extrapolation, new laws of physics, new new understandings of fundamental biology and chemistry. Do you think that is possible? When are we going to see that? And speak one second to the Vesuvius Challenge that you—

Oh, sure. Yeah.

I think definitely they will, which, by the way, for me is like the most exciting thing. If you unleash these AI models and say, go and create room-temperature superconductors, go and create, you know, Well, life extension processes.

So there's, in biology especially, biology, you know, there's this sort of analogy that's been put out there that the language of, of, you know, physics was mathematics. The language of biology might be machine learning because you're dealing with enormously complex systems. Machine learning is great at sort of understanding those. And so the potential for AI to transform biology is enormous. I think it's all in the very earliest stages right now. And we have not yet had the kind of GPT-3, let alone the ChatGPT moment for biology and AI. I think it'll come soon.

Since Google is our sponsor, I'd say the Gemini moment.

There you go.

The Gemini moment hasn't occurred yet. The Bard— what is it, the Bard moment? Maybe. I don't know, but it's coming. And, you know, I'm involved with some companies that are training enormous models to help synthesize and design proteins. There are a few great efforts out there to do this. And The capabilities that are popping out of these things are incredible. The ability to describe a target, describe a structure, and just have it produce a sequence of amino acids that you can then synthesize and test in the lab for safety and efficacy. It can short-circuit a huge part of this sort of cognitive work and experimental work that's been necessary historically for drug design. And so I think that's one thing. I think Another thing is there will be a surge of new discoveries that happen as AI digests all the existing scientific literature. And so there's a whole area of study, which is meta-analysis, where people will study across papers to find connections and correlations across existing research that haven't been noticed.

I'm bubbling with excitement on that notion and that idea.

Yeah.

Yeah.

So imagine, you know, I have a friend, Shauna Swan. She's a scientist at Berkeley and does a bunch of research on environmental toxins. She spent 2 years running one meta-analysis that, you know, I'm now working with her and trying to support her in the effort to use AI to automate this. And it'll— I mean, in theory, with these long context windows, this could potentially happen in minutes. And so I think new discoveries will pop out of that immediately.

And just because we're short on time, I'm gonna say you found Mount Vesuvius after the eruption buried a number of parchment scrolls that were, if you imagine, buried under all of the ashes and soot and so forth. You took some of those scrolls, you X-rayed them, gathered the X-ray data from those scrolls, and then you ran an order of a million-dollar Vesuvius Challenge, like an X Prize.

Yeah.

And you're— the winner solved it.

It worked. Yeah. No, it's true. So, yeah, 2,000 years ago, Mount Vesuvius erupts. It buried the town of Herculaneum under 65 feet of ash and mud. And then in the 1700s, some farmers digging a well found the villa 65 feet down. And then later, during these tunneling excursions, people kept running into these little chunks of charcoal. They didn't know what they were. They turned out to be carbonized papyrus scrolls that were not openable physically. They just sort of turned to dust in your hands when you opened them. They'd been stored in a library in Naples for years, and we used a particle accelerator to scan them at super high resolution. But then we needed to use AI and machine learning to unroll them. They're so badly distorted by—

How cool is that? Hanging up, you know.

Yeah.

Why don't you close us out here with what you're most excited about going forward? What's a vision of the world in the next 1 to 3 years that you want people to take away from here?

I think this AI is augmenting, not replacing. And so you see what kind of Nat said, and I look how creatives use it. AI can't do art, it can do content right now, but you can use it to riff and jam with so you're in flow more often. So the thing I'm most excited about is its impact upon education, science over the coming years, because everyone will have access to all the knowledge that amplifies them. I think things like creativity are also great because it's great to create. But realistically, again, every single person in this room will just have access. This is the worst it will ever be.

That's so important. This is the worst AI will ever be.

And view it—

And the lowest Bitcoin will ever be too.

Yeah.

View it amplifying yourself. And there's no— there's nothing you can't achieve with this.

I think people are sitting down here with goals for the year, objectives they want to achieve. What is your top piece of advice for the CEO, entrepreneur, philanthropist here that has like, oh man, I'm trying to do more and more. What should they What should they think about?

Well, I think what's happening or what's about to happen is sort of like we've just discovered a new continent with 100 billion people on it, and they're willing to work for free for us. And you should probably factor that into your plans over the next few years because your competitors will, and because it'll benefit you enormously at home, at work, in your We just discovered a new continent with 100 million brilliant workers willing to work. 100 billion.

100— what's that?

100 billion. They're going to outnumber us.

Yeah.

Yeah.

100 billion.

Okay. And they'll work for a few watts of power. And so this is the good scenario, and I think it's very, very likely. And so I would— and I think, you know, this is sort of like being an internet native. You want to be an AI native, and you want to spend time using this stuff and not looking for the problems. But looking for the— looking for the value and how to— what's it good at? What's it not good at? How do you interact with it? It's amazing to see people use Stable Diffusion, for example. You've improved it so much over time. But there are— there's a skill to being good at interacting with these models and partnering with them. And, and so I think we all have an advantage just to get that hands-on ourselves. You know, imagine being a CEO of a company when electrification was happening. And, you know, obviously every company should electrify and then not doing it in your own home, like that would be strange.

Amazing. Imad, you're going to be with us for the next few days. So thank you so much for that. Nate, it's a pleasure to get to know you. Thank you for joining us this morning. Let's give it up for Nate and Imad.