Moonshots with Peter Diamandis - EP #16 AI is Creating Massive Entrepreneurial Opportunity w/ Emad Mostaque

Episode Date: December 15, 2022

In this episode, Emad and Peter discuss everything from AI-generated content and property rights to ethical implications along with the upcoming hyper-disruption wave of technology in all industries. You will learn about: 11:58 | How AI is going to affect Hollywood 31:31 | When AI creates your ideas for you, who owns the product? 42:44 | Emad's advice for 20-year-olds of today 1:18:47 | Should AI have a moral compass? Emad Mostaque is the CEO and Co-Founder of Stability AI, a company funding the development of open-source music- and image-generating systems such as Dance Diffusion and Stable Diffusion. _____________ Resources Learn about Stability.AI Access Stable Diffusion Levels: Real-time feedback on how diet impacts your health. levels.link/peter  Consider a journey to optimize your body with LifeForce. Learn more about Abundance360. Read the Tech Blog. Learn more about Moonshots & Mindsets.  Learn more about your ad choices. Visit megaphone.fm/adchoices

What happens when 20 extremely athletic Canadians who thrive on competition and won't settle for less than number one find themselves on a team? Taking on jaw-dropping obstacles all across Canada is one thing. Working together on a team with some pretty big personalities is another. It's a new season of Canada's Ultimate Challenge and sparks are gonna fly.

New episode Sundays. Watch free on CBC Gem. Will you rise with the sun to help change mental health care forever? Join the Sunrise Challenge to raise funds for CAMH, the Centre for Addiction and Mental Health, to support life-saving progress in mental health care. From May 27th to 31st, people across Canada will rise together and show those living

with mental illness and addiction that they're not alone. Help CAMH build a future where no one is left behind. So, who will you rise for? Register today at sunrisechallenge.ca.

That's sunrisechallenge.ca. So again, if you're an entrepreneur, you'd be an entrepreneur in this. If you are someone who can communicate, you communicate this to other people and you get paid a million bucks a year as a consultant, right?

You organize information. If you're an artist, if you're a creative user tool. You become the most efficient artist in the world when you lean in on this. You know, like systems can be out-competed. It's like the example of the steel mill, right? There were big vertically integrated steel mills that were out-competed by lots of little steel mills, micro mills, right? The big corporations, the big programs, the big things will be outcompeted by just individuals

and small groups. Building on top of this technology can do anything. And a massive transform to purpose is what you're telling the world. It's like, this is who I am. This is what I'm going to do. This is the dent I'm going to make in the universe. Welcome, everybody. Welcome to Moonshots and Mindsets. I am here with a old friend and a new friend, Imad Mustaq. Imad, welcome. It's a pleasure to have you here. Thank you for having me. We're going to have such a fun, wide, wide ranging conversation across everything that is of, I think, importance to anyone listening to any

entrepreneurs, any CEOs, any government leaders, any kids, everything that you're doing is really transforming the world. Let me do a proper introduction. Imad Mustaq is the founder and CEO of stability.ai focused on amplifying humanity's potential through AI. You probably know Stability.ai because of its text-to-image model, Stable Diffusion, released in 2022, which rocked the developer world and I think broke the internet is a good way to describe it. Previously, Imad has been a hedge fund manager and autism researcher, now has led multiple technology initiatives across multilateral organizations and governments, and an XPRIZE competitor. Today,

Imad is pursuing an incredible moonshot, actually a series of moonshots that could transform many industries. You know, I started by saying, which industries are you looking to disrupt and your answer was all of them yeah could do better yeah and i think uh that's not hype i think that's actually from what i understand what we're going to explore uh true you know uh scientists and technologists have been talking about ai for decades um but today is different, isn't it? Yeah. I mean, we always, I'd say this time is different, but this time really is different. I mean, if you kind of look at it, AI is basically information classification and we had classic AI,

which was information goes in and then you extrapolate from a data set. You create really custom models and it goes out like the big AIs. Internet two was Google and Facebook taking all that big data and then targetingA with like ads for rockets, you know, and things like that. You know, like rockets. In 2017, we had a bit of a step change where there was a paper called Attention is All You Need about how to teach an AI to pay attention to the important things. Fascinating. And learn principles. Yeah, principle-based analysis. Principle, yeah. to pay attention to the important things fascinating and learn principles yeah principle based analysis principle yeah so it's not good old-fashioned ai or kind of logic causal based

ai but it's kind of that because you've you remember like um there's a book who was it by um type one type two thinking a con yeah so you know there's the very logical thing there's a freaking tiger in that bush over there yeah you. And we didn't have that second part. We didn't have the ability to just leap to conclusions, principle-based, heuristic-based thinking. This is kind of the mindset thing, right? Whereby you construct these things, and it allows you to go very fast to just amazing conclusions. That's part of what makes humans humans. It's here now, and it actually works.

Leaping to conclusion. So it's no longer extrapolation based upon data. It's actually learning to understand the meaning hidden in the data? Exactly. It's kind of semantic understanding, kind of. Literally, it's kind of called latent spaces, hidden layers of understanding. So, like, I'll give you an example. One of the ways that i kind of entered

ai was um working with ai to do drug repurposing for my son who has autism yes so he was two years old at the time and the doctor said nothing could be done and i was like of course there's something can be done like let's try and figure out information because all we're looking for is information i need to come in time it's kind of the claude shannon theory of information theory information is valuable in as much as it changes the state. So how do you find the valuable information amongst all those autism studies and figure out what's going on? Cause everyone, no one can really explain to me what caused it.

So just one second there, because typically people are looking for a very clear cut answer in the data that's obvious, but there are answers, but it's hidden. It's hidden. You know, this kind of hidden. And so you have to look at things from a first principles basis when you can't see things just from scanning everything. You have to dig down to the net and then lay another layer. And kind of when I dig down, I kind of used AI to do natural entity recognition, look

at all the different compounds, all the different trials that were tried, built a team, then built a biomolecular pathway analysis model of neurotransmitters in the brain so it's like just like a common cold it seems like something very similar is happening it turns out there's two neurotransmitters in the brain one of them is GABA you pop a valium and chill out it chills you out and another one is glusamate and glusamate kind of makes your brain accelerate and so you know when you're tapping your leg and you can't focus and concentrate that's what happens because humans are really a filtering entity as it were we have so much information in the world it's just too much so our brain creates a simulation that's why we've got the optic nerve for example filled in instantly we are kind of constrained a little

bit and that filtering allows us to focus we all know what it's like when you're tapping your leg because there's too much stuff going on so with kids with asd they couldn't filter for various reasons but it was different reasons not enough gaba too much glutam like when you're tapping your leg because there's too much stuff going on. So with kids with ASD, they couldn't filter for various reasons, but it was different reasons. Not enough GABA, too much glutamate. When you're born, GABA and glutamate are actually both excitatory, and oxytocin flips the switch on GABA. Interesting. So we started digging down into kind of the things, and we're like, ah, different compounds can affect this in different ways. But then what was the upshot of that, and how is this long story coming to a conclusion? was the upshot of that and how is this long story coming to a conclusion because there's too much noise my son couldn't speak because he couldn't formulate the hidden layers of meaning and

connectivity on concepts so kind of we've got cup here yes and you can cup your hands you have a world cup which england hopefully will win you know and you've got um various other meanings of the word cup you form this latent space of hidden connective meaning to that so it can be applied in different things the principle based analysis By reducing the amount of noise you could then filter and pay attention and build that out The AI can now do the same thing So you don't need to have huge petabytes exabytes of data anymore with structured data It can figure out interconnectivity and I'm going to come back to that because that's the revolution that that you're creating right now

Just you know, these personalized data sets that are good for you, your country, your company, yourself. Your context, exactly. Because humans are heuristic animals. We learn by principle-based analysis. And we are animals that are story-based. So you have multiple stories that make you up, you know, from your work on the XPRIZE to your humanitarian to Bold VC Fund and all the things. We all kind of form connectivity there.

But those stories are very hard to map initially. Now we can do it dynamically, and we can start building tools to really augment human potential by doing this, because we can finally have that assistance. We're going to have so much fun on these conversations, but I want to take us back to some of people around the world for the last year or so have heard of DALI and DALI 2. And here comes Stable Diffusion. That is an open source version of DALI and DALI 2. And as I said, broken internet. You sent me an image of the speed of growth on GitHub of people using stable diffusion.

And it was so funny because you first sent me this image, right? Here you see on GitHub the users who are on Ethereum over time, right? And then there's this line that goes straight up that I thought was one of the axes for the graph. Yeah. Like what happened there? So first of all, what is what is stable diffusion? It's one product of your company, right? Well, stable diffusion is a community effort to build an AI that allows anyone to create anything. So words go in and then anything you describe comes out, which is a bit insane.

Okay. So what we did is, you know, in a collaboration with various entities, which we played an important part in stable diffusion too, is kind of led by us. We took a hundred thousand gigabytes of image label pairs, 2 billion images and created a 1.6 gigabyte file that can run offline in your MacBook. 1.6 gigabyte file, which is relatively, you can transmit it over the phone network. You transmit it over the phone network, but you only need to transmit it once. Yes. Everyone having this tiny file that basically compresses the visual information of a snapshot of the internet can create anything from a high resolution render of a apartment to

robert de niro's gandalf or anything else and they can now do it in a second and in a week they can do one of those images in 30 seconds so you suddenly have what you mean one 30th of a second one 30th second we've just broken real time we had a 30 time speed up in the last week so so when you mentioned that just as we were getting ready for this i said so you're basically able to render a virtualized world in real time yes and that's one step away from being able to virtual to render a movie in real time yes so. So we are, are we on the verge of basically, and in, you know, we're sitting here in, in Los Angeles and you've been in the Hollywood world. I mean, are we talking about rendering motion pictures by just describing writing, reading a script and having some, you know, stable diffusion three by O create a film in real time? Well, who needs to write a script? I did that.

That's right. You just say, I want something to make me happy, right? And then it'll pull together various models and it can live generate a movie, yeah. So we're live generating. So Hollywood. So in the next five years, I think that Ready Player One Oasis world

minus the microtransactions of wanting teenagers will be here. Create anything you can imagine. Iter, you know, dynamic, new forms of kind of communication, new affordances, like all these clicks and things like that. You don't need that anymore. Like then this is one of the biggest evolutions in humanity ever, because the easiest way for us to communicate is what we're doing now. We're having a nice chat, right? Back and forth. Then the Gutenberg press came and suddenly you could communicate through written form.

It's harder. Yes. You know, but still you see now GPT three and other of these AIs have made it easier now for anyone to do anything like chat. GPT just came out. It's amazing. Yes.

I saw from opening. Everybody's making extraordinary claims about what it's going to do to Google. Exactly. Well, this generative search engines are a very interesting thing. Um, but then visual communication is the hardest, you know, like you toil to be an artist or all of a lot of people on this call. How much have we toiled with PowerPoint?

We're going to rid the world of the tyranny of PowerPoint. But that's my expertise. I'm a PowerPoint expert. You're not a PowerPoint expert. You're a communication expert. So we remove the barriers whereby any visual medium can be created instantly. And you just describe to it how it shifts. And because it's got these latents, it understands happier or sadder. Or you can say flamboyant. And that's not a word, but it figures out what that kind of means.

But the ability to rapidly iterate it and say, okay, I don't like that movie scene. Make it happier, make it it brighter make it more dramatic and then being able to render any form of communications like you just said you know a movie and have a thousand variants i mean what's going to happen to hollywood i imagine hollywood will be quite disintermediated yes hollywood and everyone's thing i I mean, what is Hollywood, right? Hollywood emerged because it was far enough away from the East coast that IP laws didn't apply. Right. And then kind of, it got constructed around here as an entity that extracted rents from performers ultimately and creatives.

Like how many creatives like the Hollywood or the music industry or others? Not many, because they tend to be treated quite badly. Some people achieve superstardom. But a lot of this technology now is just truly democratizing, in that things are going from centralized to the edge, and we had to create centralized organizations as a people, because we didn't have the information classification and communication tools to be more dynamic. It's called the representative democracy.

It's not a true democracy. It's not a true democracy. It's representative democracy, exactly. At the same time, you know, everyone speaking at the same time doesn't make sense. But what we need, like the most successful organizations are complex hierarchical systems. It's groups of people working together loosely bound for a bigger story. And this is what we can achieve as humanity, like the human colossus when we all come together to do one big thing. But it gets blocked by a lot of the stuff because we're not communicating properly.

Like one of the things I tell my team is that, you know, roadmaps are not about resources. They're about communication. If you communicate properly and something's good, you will always get resources for anything. It's also about creating a common vision that everybody's aiming towards in order to get there. So you're not diversified in a thousand directions. So like Google is full of amazingly smart people right they did an amazing study called product aristotle yes because it's like why is one smart team better than another what was the finding the finding was it came into two things common narrative and ideally something that kind of engages you and there's a bit of sacrifice you know like the so this is what this is what Salim and I and Salim Ismail, who, you know,

well who sends his regards talk about as a massive transformative purpose, having a unifying, really compelling, emotionally charged vision that you're heading towards. Something deep and something that you've put yourself into as well. So it becomes part of your story as it were. And the second part was psychological safety, which I think was very. Amongst the team. was amongst the team amongst the team the ability to actually express yourself and talk about something without fear of reproach

that's what makes the most successful teams because we're too scared often about kind of our status and worrying people and things like that that's what's on the research side of things primarily but i think some of these lessons kind of come because when you feel comfortable as a community and you're working towards a massive transformative purpose, you can do a lot more. Because when you're falling behind, you can communicate it and you're not scared of people judging you. And if you have an idea that's considered, you know, divergent from the center, you feel open to being able to share it. In fact, it may well be the right idea. And I talk about the day before something is truly a breakthrough, it's a crazy idea.

And if you're scared about actually putting forward a crazy idea, then you're stuck. Well, again, it comes back to information theory, Shannon style, right? Information is only valuable in as much as it changes the state. Yes. Everyone's on the same page with everything. Flat equals dead. Exactly. And this is a time where we basically have to have exponential progress. And now we actually have the tools to help us make exponential progress. Like, you know, Facebook released Galactica recently, which is their language model trained on 42 million science papers. They made some claims that were a bit hypey. And then people like, oh, you can use this to create racist science papers. And so they were forced to take it down. But I mean, it's an amazing piece of tech and we're going to help

re-release it. And this is an interesting thing. You can use it to do things like a null hypothesis creator. You can use this AI to do all the sorts of things that enable more divergent original thought or creativity or any of these other things that classical AI couldn't because it was out of mode. It was out of the data set. Yeah, of course. If it's not in the data set, then you're screwed. Whereas now you don't need it to be in the data set. You can create new data. Yeah, that's extraordinary. This episode is brought to you by Levels. One of the most important things that I do to try and maintain my peak vitality and longevity is to monitor my blood glucose. More importantly, the foods that I eat and how they peak the glucose

levels in my blood. Now glucose is the fuel that powers your brain. It's really important. High prolonged levels of glucose, what's called hyperglycemia, leads to everything from heart disease to Alzheimer's to sexual dysfunction to diabetes and it's not good. The challenge is all of us are different. All of us respond to different foods in different ways. Like for me, if I eat bananas, it spikes my blood glucose. If I eat grapes, it doesn't. If I eat bread by itself, I get this prolonged spike in my blood glucose levels. But if I dip that bread in olive oil, it blunts it. And these are things that I've learned from wearing a continuous glucose monitor

and using the Levels app. So Levels is a company that helps you in analyzing what's going on in your body. It's continuous monitoring 24-7. I wear it all the time. It really helps me to stay on top of the food I eat, remain conscious of the food that I eat, and to understand which foods affect me based upon my physiology and my genetics. You know, on this podcast, I only recommend products and services that I use, that I use not only for myself, but my friends and my family, that I think are high quality and safe and really impact a person's life. So check it out, levels.link slash Peter. We give you two additional months of membership and it's something that I think everyone should be doing. Eventually this stuff is going to be in your body, on your body, part of our future of medicine today. It's a product

that I think I'm going to be using for the years ahead and hope you'll consider as well. You know, when you go to your website, I love the notion of AI by the people for the people. And then you say stability AI is building open AI tools that will let us reach our potential. Let's talk about that. Cause I, you know, there's a lot of individuals that you and I both know that are fear mongers around AI, you know, it's the devil, it's going to destroy us. It's going to, superhuman AI is the end of humanity as we know it. And I mean, my position is AI is the single most important tool we're ever creating to solve the world's biggest problems. And we can't solve our problems from where we were before,

but these are the tools that are going to allow us. How do you address the Bill Gates, the Elon, to the world on that side, the fear side? I think fears are valid because it's the most powerful technology we've ever created, and it comes from us. But then who is us, right? Like, if you look at our current data sets, they're massively biased. They're fixed towards the internet, and they're fixed towards manipulation. The way I kind of look at AI is that organizations themselves are AI. They are slow, dumb AI that feeds on us and turns us into cogs. In fact, there's this concept of Moloch, right,

from the Ginsburg poem, you know, how talking about this Carthaginian demon that pervades our organizational structures and turns us into these cogs that feeds on us effectively. I think this is the first thing that can actually defeat that, this particular technology that we have today, because is the world happy? No. defeat that, this particular technology that we have today. Because is the world happy? No. How many people in organizations are happy? We all know, you know? We should talk about my idea for happiness XPRIZE,

but that's a different conversation. Yeah, exactly, happiness XPRIZE. But what is happiness? Happiness is agency. Happiness is achieving your potential. And it's kind of going out there, so we need some help. But I think a lot of the AI discourse has been focused on gigantic language models and freaking supercomputers, which we still have, to create something that will equal and surpass us. It's very religious in its way. And you see parallels to religion across this.

Sure. In that you have the kind of declaiming people who like for it, the people who call it heretical. And then you kind of have most of ethics in AI. It's not actually ethics. Instead, it's ultra orthodoxy as it were like in islamic terms everything is haram unless it's declared halal you know classic judaism ultra orthodox judaism etc it's the same thing people look at red teaming they don't look at green teaming right and they look at this technology being too powerful so just like

cryptography we should keep it from the people and we definitely shouldn't give it to like emerging markets and people who aren't smart enough to do it and that's very interesting because again if you think about the power of it and you believe it's powerful then the question should be what is this and i believe that this ai is infrastructure clayton christiansen uh kind of the departed mentor of mine had an amazing, which is that infrastructure is the most efficient means by which society stores and distributes value. Obviously, that's ports and things like that. Sure. But it's also information. Sure.

And I think that's valuable when you combine it with the Shannon theory. So this AI is infrastructure for the next generation of human thought. And what should that mean? It should mean that it should be a commons that is accessible to everyone. It's a leapfrog. And you made a very definitive Decision to make this an open source movement here. Yeah Your community is how large now? I think we've got about 120,000 amazing in the communities And we created communities across verticals.

Classical open source is product related where you have MongoDB or something like that and then you've got that. Whereas we said language, healthcare, bio, audio. Let's get all the people who are fantastic in the private sector, public sector, actively independent together and let's jam on

how do we build a next generation infrastructure. The way I put it is let's go to the future and bring back AI with us. And we can choose if it's a panopticon controlled by corporations like web 2 or we can choose if it's open an infrastructure for humanity and if it's infrastructure humanity and it's an important point because i'm just finalizing that a lot of the naysayers around agi and asi and others i would agree with them if it's controlled by corporations which are this type of weird entity that we've created, like YouTube optimized for extremism because it was engaging.

So ISIS co-opted those algorithms and they adjusted eventually, but it was slow. If it comes from us and it's biased and we've brought in the conversation, I think AI is more on file is likely to kill us all. If ever to become sentient, which is a big question, you know, especially if it's

representative so i want to come back to this in greater detail later but i think we share a common belief that humanity ultimately is good yes at its fundamental level because that's a very important distinction if you believe that that humans at their base are good and you're enabling humans with more and more powerful technology, they're going to be using that to make the world a better place and solve problems on the whole. On the whole. But it depends on what the perspective of humanity is. You know, you've kind of got, what's it called, the thing when you go to space and you look at Earth? Yeah, the overview effect. The overview effect.

Yeah. You know, like hopefully more and more people get to space and have that. Because people are very narrow and have that because people are very narrow and they view themselves like that. The rabbi Sacks, um, former chief rabbi of the UK had a very wonderful concept called altruistic evil. Those who actually do evil believe they're doing good. Interesting. And you see that like, you know, if you talk to, especially in the religious,

religious or anything like that, I like isaiah berlin's conceptualization uh british philosopher from like the 1940s he had a conceptualization of positive liberty versus negative liberty so negative liberty was the freedom from anyone telling you what to do to kind of laissez-faire capitalism and things like that positive liberty was ability to believe in an ism, something big like fascism, communism, capitalism, Islamism, etc. And people use that as excuses, believing they were doing good to actually do bad. But then how does this all relate to this? I believe people inherently want to do good. It's just that what good is can become misdefined and co-opted.

And so my take was if we start building infrastructure where people can see bigger perspectives because they get the information they need what does that look like what if we mapped all the religious texts in the world so that a child in egypt could see judaism from the perspective of a child in jerusalem we have the technology to do that what if you could automatically translate tea party tea party republicanism into libertarianism and have commonality there again we finally have the technology to do that. But we didn't before. So people remain in their huddles. They look at the other, and they're encouraged to do so. We've seen that increase in political polarization.

So society is at tipping point. And by the way, that's been what social media has effectively done so efficiently. Because it's beneficial for the algorithms and for the slow, dumb AIs of the corporations that drive it. Let's come back to the company and the products just to lay out the tapestry here. Stable Diffusion is one product vertical area. What are the other ones? So, for example, our Luther AI community has GPT-Neo and X.

It's the open source version of GPT-3 by OpenAI. The most popular open language model in the world. It's been downloaded 25 million times. Incredible. And so can you take it? You customize it for yourself? No permission needed, right? Harmony has Dance Diffusion, which is the most advanced audio model in the world. You'll be able to create your own music. So you put your own music in it. And then you have your own music model that can create more music of your style. You'll be able to create your own music.

So you put your own music in it, and then you have your own music model that can create more music of your style. And you can just basically produce your own concerts, extrapolate in any direction you want. Exactly. We have Open Bioml, where we kind of have OpenFold and LibraFold, protein folding, DNA diffusion for DNA protein matching, and Open Biolm for, again um some protein kind of stuff so and you know we've spent some time talking about a passion that i have and and you share which is the ability to truly transform the healthcare industry and in human longevity understanding why we age maybe why we don't have to i think the why is the important part right most of medicine so again when i my son was diagnosed and scratching a walter's fingernail spread you know eventually went to mainstream school thanks to the interventions and i want

to i want to come back to those for those who are listening who have a child with autism or know someone i want to come back afterwards what were your learnings and what's your advice? Yeah. Because a lot in this space that we kind of occupy, of course. Um, but like what I realized is that conventional medicine and a lot of things view humans as a good, a thousand tosses of a coin is the same as a thousand toys cost at once twins coins tossed at once. Um, but humans are individual so for example a good proportion of humans have a cytochrome p450 mutation in their liver which means they metabolize

things like fentanyl yes far quicker yeah it's a very simple smp test but we don't do it so we just prescribe everyone the same thing and then a bunch get addicted you know because we are individualistic kind of creatures this is why I went to the first principles thinking kind of approach on this. And the question is, this personalized medicine thing has always been kind of out there. We've not been able to reach it. Again, I think the technology we have right now enables personalized medicine. We've seen things like CRISPR, obviously, and others. But more than that, it's about data availability and viability.

We do need to get to a point where every child at birth is sequenced and that is plugged into their personal AI and they understand exactly how every food, every medicine and every aspect of our living affects our physiology. Again, you go to the future and you bring back the AI with you and you say, what should it look like? So you take a whole country and you say, how do we build an amazing healthcare system, education system, et cetera? Is there any doubt that healthcare and education will have AI at the core in 20 years? Zero. Well, I think it's not 20 years, it's 10 years.

I know, but let's just say 20 years. You know, you and me, we're like, now, now, now, let's say 20 years, right? But then is that AI open or closed? Yeah. And is it better for it to be open or closed? Yeah, there's no question. Open is fundamentally critical.

Open is fundamentally critical because then we build, it is infrastructure. It is valuable. Like the way that I actually orient my rights, Vinay Gupta, who's one of the Ethereum guys, I think you probably know him big thinker he's crazy he had a very great conceptualization of rights which I agree with

which is the rights of children and so like effective altruism and all of that looking at people a million years in advance is kind of difficult

and it comes down to utilitarianism all sorts of weird stuff but the rights of a child rights of a child today yes today

what does that child have the right to? Achieving their potential. Yes. What infrastructure do we need to give that child in a refugee camp or in Brooklyn or in Kensington to help them achieve their potential? And this goes back to what's on your website in terms of using AI to help, in this case, a child, and in the broader case, humanity, achieve its potential. Achieve its potential. I mean, that's everything.

Yeah. And so for me, it's about, I've oriented that on agency and happiness. And it sounds fuzzy, but it's literally when you have the tools to be able to do anything, you know you can do anything. People underestimate their agency. Sometimes you've got to go for your shot, as it were. Yeah. But again, it's information.

What information can I bring to that person? What tools can I give them so that they can be creative? Cause we lose that creative spark as we get older, right? Yeah. Now it's coming back of it. You know, what can I give them so they can access the information they need to be educated and what's the optimal way to teach linear algebra. And by the way, it's all about play. It's all about play. It is all about play, happiness, play, all the neurochemistry of your brain is maximized for learning, retention, experimentation around that.

It's flow. It's flow. So I used to be a video game investor. That was my big sector. It was one of the biggest in the world. And I used to judge video games by time to fun, flow, and frustration. And so if we're building systems for humanity, we have to look at fun, flow, and frustration. Because if we can get those, you know, when you're just learning something like, wow, this is amazing.

And how do you get in there? It absorbs amazingly quickly. But our education system is not set up for that. Our healthcare system definitely isn't. And you've got a messed up healthcare system. I'd say it's maximizing for frustration. Yeah, that's what I speak about it openly saying. My mission is to crumble and destroy the healthcare system and also education system, which is

really sad. So again, it's very interesting. You look at the US and inflation numbers, education and healthcare, massive inflation. Yes. Everything else, not really. Yeah. And it's the percentage of the person's income is end. It should be going to zero, right? The top healthcare and the top education should be all AI driven.

It's all basically the cost of electrons. Life expectancy is falling in the U S a little bit over the last couple of years. Yeah. Before COVID it was falling. I mean, this is the thing, whereas again, it's not complicated, but it does require coordination. So the question is, can you create shelling points? So for me, open source software, this next generation, this model-based one, whereby stable diffusion is basically a programming primitive, just like you have a library to do various things for the programmers out there.

So in the good old days, when we started programming, we used to have to code everything by hand. Now you've got GitHub and things like that. It's more like playing with Lego, right? Yes. And you're sticking it all together. It's a new type of thing. A 1.6 gigabyte file that is hashed,

so it can be common across every single computer in the world, that you can call something in and an image comes out. And you know predictably what that is, no matter where you are. And you can also take an image and put text on the other way. Yeah. That changes the paradigm. But then what if you have that for language, audio, all these different modalities?

Okay. I'm going to go here next then on that, which is what happens to intellectual property rights, IP rights? Who owns the IP of those images? Is it the person who puts the prompt in? Help me understand where that evolves to. Nobody knows. Nobody knows. Okay. It's a fair answer. Yeah. I mean, nobody knows. I personally think it should believe it should belong to the person that prompts it because again, kind of,

we put this out as a commons for humanity. Yeah. Like we did put an ethical use license around it for various reasons that will be replaced by a purely open source license. But again, I'm viewing this as building blocks, right? It's really the person that has the action because these models do not have agency. If you're going to breach copyright, you can type in, I want a picture of Mickey Mouse. And then if you sell that, done. But it's like photos.

It's like telephone. It's like internet. It's like Photoshop. This is a tool, but it's a tool of a very different type. Because something like Photoshop is a million lines of code. This is a binary file that can do anything that Photoshop does. But you still need to act on it. Last year, there was a patent awarded to an AI in South Africa, which was interesting.

I think more of a gimmick than anything else. which was interesting i think more of a gimmick anything else uh do you imagine these uh these models will lead towards significant uh intellectual property on the invention side of the equation oh 100 they already are if you look at google's new tpu chips they're partially built by ai that they built sure yes i know like we just released um open uh lm from our CarpaLab, which is an evolutionary algorithm for code for robots. So we're trying to optimize robotics via that. And then when we bring in video, we'll have even better robots. So are we going to get to a place where I'm saying, listen, I'm, please invent a device that does this for me. That's under this price point. That's made of commonly available materials and over, you know, constrain it left,

right and center and then design it. Now go and print it for me and deliver it for me the next day. We're going from mine to materialization in one sense. Yes. Yeah. I mean, you know, again,

it's kind of the, was it the Star Trek thing? Yeah. It's a little bit longer. Yeah. It's the, the, the,

oh my God. We've both forgotten it. Yeah. Anyway, you know, when he, the replicator. The replicator. it. Anyway, you know what he gets. The replicator.

The replicator, exactly. Earl Grey Tea Hot. Exactly. But then you can kind of see this already because there's an app on the App Store right now. You put all your Lego out and it scans all the Lego. Ah, nice. What can you build with it? Yes.

It comes up with this is how you build it as well. So it figures out you can build birds and cars and things like that. So you go, so this is just an extension of that. Again, they use a transformer based architecture for that. Well, I'm interested in when it'd be fun to say,

okay, create me new life forms. I mean, we're not far from that. Well, you know, I'm going to leave that to other bioethicists. I don't want to go step my foot in that one.

Wait, there's, there's ethics involved in that. Exactly. I mean, the life form thing is very interesting though, right? Because again, everything is happening all at once. And so it's not just biology or physics or sociology, all these models, all these technologies all seem to be converging at once. So you can create anything and all the barriers are dropping at once so and that's complicated yeah so the world is i don't do you think anybody

truly understands how fast the world is about to change no i mean like look at creative industry video games 180 billion dollars a year like disney spends 10 billion a year amazon 16 billion a year on content all of that's going to change in the next couple of years alone, just from one tiny little two gigabyte file. Extraordinary. As well as healthcare, as well as education, as well as you said, every single industry we're going to. So here's another question.

Go ahead, please. Well, I was going to say industry is about, again, information theory. Yeah. That's most industries are based on on especially service-based industries and so once you can basically take a system you can have human input in the loop to train a system that's a generalist to understand principles you disrupt just about everything yeah i'm going to pause on that because it's it's probably one of the most important things any entrepreneur any ceo any parent any kid anyone needs to understand we're about to enter a period of of hyper uh disruption and growth and hyper opportunity

creation yes like what happens is that in any area you create there are value spikes you can look at it's like a flat area and then companies and individuals occupy certain areas so they've got like a mix of skills right and that's how you earn your living and these are like spikes that's all going to get shaken up and it's going to be the value is elsewhere yeah we don't know where the value will be we've got some guesses so like in a time when everyone can make anything what becomes valuable something yes so if a model that can make anything that means disney should have their own models right to create mickey mouses and things that. But now they can use that not only internally to save money on creation, they can use externally. Why can't we have Mickey mouse having coffee with master chief at a Starbucks and microtransact pay that it's been the promise of the NFTs and things

like that. Now with these models, you can do that seamlessly. Yeah. On demand over and over again. So one of the realizations for me is that this open source technology made available to everyone effectively has the potential to make everyone the equivalent of millionaires, billionaires, trillionaires, when whatever you want can be manifested, right right you can have the world's best education for your child independent of where you live and what wealth you have you can have the best health care available you can have customized entertainment you can you know the it's we end up in a world in which the cost of anything is raw materials in ip which is going to be disrupted itself, and electricity. Yeah, pretty much.

Again, like you look at what Elon Musk did with SpaceX, right? He's like, let's break it down to what is the constituent cost of this? And basically went to first principle thinking on rockets. First principle thinking. So what I've been doing is first principles thinking on information flow, social theory, and our systems. It all comes down to just information being in the right place at the right time to make the maximum impact.

And if we can give that as an open architecture to the world, then we can augment and then replace our existing systems with better ones because they out-compete. You go to an African nation and you teach every child with an AI that teaches them and learns from them. Within a few years, they will be out-competing children in the top schools in New York. What if you give them the ability to code the system as well and improve it? It becomes very interesting. The kids are helping train the AI and the AI is helping train the kids. Yes.

But then also the kids can improve the actual code that they have there. Yeah. Like this is something that we've seen in a virtuous cycle. A virtuous cycle. And you make that open and then you make it transplantable because what you have then is you have models that are standard like a base so i like to call stable diffusion one was a precocious kindergartner then it was a precocious high school stable diffusion two stable diffusion three looks freaking amazing it's going to be like a university level student so stable diffusion three by the way is the real time rendering.

No, that's stable diffusion two. That's two. Okay. So again, like to give you an example of the thing, when we launched stable diffusion one on a top end, a 100,

which is like a super souped up supercomputer chip, right? We have a lot of those. We can talk about the supercomputers in a bit. It took 5.6 seconds on August the 23rd. When we launched it. To render a single image? To render a single image in 512 by 512.

Okay. Today, it takes 0.9 seconds in 768 by 768. We've just sped it up 30 times. Amazing. 30 times. I'm still blown away by the real-time rendering of the world. Once you get below 200 to 220 milliseconds of response time, it opens up entirely new UI UX.

And we didn't just release stable diffusion as an image creator. We also released an in-painting model. So you could take, you know, MADS hat, and you could turn it red just by describing it. We released a depth image model so you could do transformations. We released an upscaler so you can have a 64 by 64 image, and the AI fills in all the details to take it to 1024 by 1024 transforms the storage industry for media amazing and this is real time now as well it's and so listen i have to ask you the question but come back to it later

which is do you believe we're living in a simulation um yes or no yes yeah as do i i think we're living in an nth generation simulation, but that's a, that's a different story. Okay. We'll come back to that for folks who are interested, but okay. How far out are you able to imagine this world that, that you're creating the,

the disruption of industries, the transformations. How many years out? I don't know anymore. Like when i started so i funded the entire open source art space from january of last year when it started because you had a generator model and then opening i released this clip model which took images to text we bounce things back and forth for each other we're like wow that's the way to do things creators and discriminators as it was and it's advanced advanced i was like this is the next big thing

humans can now communicate visually right it's the next biggest thing since kundalini press and it kind of went as expected plus minus six months i thought we'd get to stable diffusion in q1 q2 of next year and that was a big massive bet i built a gigantic supercomputer and everything to get to real time i think it would take another year or two and instead it took like four weeks and like again this is what you mentioned earlier the number of github stars now for stable diffusion is more than ethereum bitcoin and just about everything else and that took them 10 years in three months yeah i mean again it's the notion that people have no idea how fast the world is changing and is accelerating and it's the it's what you came back to with the common mission and the energy.

When you sent your first Bitcoin, it was an amazing experience. They got overtaken by raccoons and stupidity, and they're trying to create an alternative system outside the existing system. The interfaces were all the robbery and all the profits were made. This is something different whereby you're talking, when we talk to developers and the contributors that are increasing in the ecosystem,

they're so energized. And this is what drives things forward. Like when you see teams that do the biggest things, they have the energy. It's almost palpable, right? Where it's like that driven thing. But we've got people from all over the world. Like one of our latest developers

was an Amazon warehouse worker at the start of this year who taught himself to code. And now he's building the most advanced models in the world we've got 16 year olds to 62 year olds and it's a team of how big so our team is 137 but the but the developers who are creating 500 well but they're teams of one right their team and their individuals are able to use this to create oh yeah so like if you want to create with this you can just do it by yourself so um there's a fantastic twitter you can look at levels.io and he's like i'm want to create with this, you can just do it by yourself. So there's a fantastic Twitter. You can look at levels.io. And he's like, I'm going to create businesses by myself that are making millions of dollars

of things just because you could took this primitive and he wrapped things around it. And you're just making money. Like avatami, you can put your own face into the model, 10 images. You can create a Peter Diamandis model and we can put you in space. In fact, we'll do that a bit later. So listen, you're, you're listening. You're a 20 year old we can put you in space. In fact, we'll do that a bit later. So listen, you're listening. You're a 20-year-old entrepreneur. You're in college.

You're finished. You've skipped college, whatever it is. What's your advice to that 20-year-old listening right now? You should drop everything. You should focus entirely on this. This is the biggest shift ever. So self-driving cars, $100 billion went into. Crypto, hundreds of billions of dollars crypto into 100 billion trillion dollars going to go into this sector

say that again how much 100 billion in the next few years and then a trillion dollars will go into this sector because it's so transformative and so few people understand it there's never been something a technological advance that will diffuse as fast as this so this is electricity this is the Gutenberg press times a billion this is a this is a Gutenberg press times a billion because it's not just writing it's image 3d it's everything it's creation protein folding it's it's create it's creativity and extrapolation uh so what does it look like to focus on this? So you're, again, you say drop everything in focus. And I get that. And, and I, you know, part of me is like, maybe I should do that too. But what does it look like to focus? What would a person do?

So again, like if you're an entrepreneur, you'd be an entrepreneur in this. If you are someone who can communicate, you communicate this to other people and you get paid a million bucks a year as a consultant, right? You organize information. If you're an artist if you're creative if you're an artist you become the most efficient artist in the world when you lean in on this you know like systems can be outcompeted it's like the example of the steel mill right there were big vertically integrated steel mills that were outcompeted by lots of little steel mills sure micro mills right the big corporations the big programs the big things will be outcompeted by just individuals and small groups. Building on top of this technology can do anything. And we've seen that over and over again in every converging exponential field. We've seen entrepreneurs disrupting and the rate of, you know, I think of this as the asteroid impact that the slow lumbering dinosaurs all die and the furry mammals are the ones that rapidly evolve.

Yeah. I mean, like, let's take a practical example. And this is what we said a bit earlier about chat GPT. Why does anyone need to use Google image search in a year? Yeah. When you can create any image just by describing it and then iterate it just with your words. Yes. Just attribution. That's it. But that's an easy lookup table. Yeah. True. Absolutely. So now let's take it. Okay. The 20 year old, drop it, start experimenting, start playing, start using. Now you're running a company, you're running a hundred million dollar company, worse off a billion or $10 billion company. And you see this how fearful should you be and what should you be doing again you should be leaning in to understand this the classical innovation process inside a company is very limited you should be having kind of a

crack team of people who've just given freedom to say how can this potentially change my entire company you know um but it's hard because you are fighting against the inertia of your company you're fighting people being used to certain ways of interacting or certain ways of distributing stuff yeah so this is why i think you have to think and think again from first principles if this technology is real time fast across modalities and it can look and understand stuff let's go forward five ten years and work back does my company still exist in this current format can i out compete that if not how can i fold these into my current processes and procedures etc because this is

the other thing as a true exponential like the ai research actually in this area 80 of all ai research has become in this area in the last few years. And it's exponential with a 23 month rate of doubling a true exponential. The adoption of this now is also exponential because everyone from around the world is using it. And then each of those introduces another two people that goes four and four and four. So right now it's like a wave that's under the surface.

It just hasn't come yet. Next year is when it cracks. And the year after is when it crashes. I hear you. And, um, my reaction is I'm excited as hell about that. Right. It's like,

because, because it's about transforming the inefficiency of the world today. It's about taking individuals and empowering them to do, to giving them agency, like you said. But this is very interesting. Inefficiency is where value is created in the current paradigm in some ways. Or where value is created. Because inefficiencies exist.

Value is captured by inefficiencies. So we have to have, we go to the gatekeepers and we give them, you pay your lawyers loads of money, right? You pay your accountants and other things like that because there are inefficiencies in the system. So you pay them to remove the inefficiencies as it were. Some of these inefficiencies will no longer remain in the system, but that means that that value capture will also disappear.

And again, this is why you know you have to start thinking how where does the value look like in the new reconfigured landscape if do you have an example in an industry that exists today and just to make this concrete for someone sure you don't need lawyers anymore uh for everything awesome i mean like you know you've got do not pay.com has done massive things. They automatically write your tickets for you and kind of, yes. Right.

For those who don't know, it's if you got a parking ticket and you go to do not pay, they will figure out the legal loopholes and arguments to get you off your ticket. Yeah. Um, but like you still need litigators, right?

At what point can you have a robot litigator? It's like Phoenix, right on steroids. I know. Right. Um, you know, if you've got the, again, movie creation thing, movie creation is just going to transform completely.

Video games, you'll be able to have your user-generated content. You know, like, again, people can right now use stable diffusion, create any image. They can go to chat GPT, it's an amazing system by OpenAI, and they can chat with and like oh okay that disrupts a lot of industries as well because it knows the different

things but it's just a model it's a blob it doesn't look on the internet what happens when you hook up these models that are principle based to the internet stable diffusion plus google image search is actually incredibly powerful but people are just looking at stable diffusion stable diffusion as part of a architecture is incredibly powerful people are making looking at stable diffusion. Stable diffusion as part of architecture is incredibly powerful. People are making movies. So if you look at our friends at The Corridor Crew, for example, they did a movie called Spider-Man Everyone's Home, where they created a custom model based on the Spider-Verse movie with Miles Morales.

A couple of days, they created a three-minute trailer that blows away anything big studios can do. Just a few people, a few thousand bucks. Let's go to a few general questions on AI. First of all, you come out of the hedge fund industry. I have to imagine that the day of the trader investing on their own without the use of any of these technologies is long, long gone.

Does anybody have an, you know, have any advantage on their own? I mean, I think that's the thing. What's, what's an edge in trading, right?

And again, it comes down to information theory. What can move the Apple stock? 50% very little information. What can move at 5% decent amount of information, 1% quite a lot of information, right? So we're just looking at the narrative and the incremental narrative of this because again as humans we're heuristic creatures so like i used to be a specialist in a number of markets oil market

was one an oil barrel is fungible going around the world because you can just shift it on a ship sure but the impact of a libyan barrel going offline was the third as impactful as an oklahoma barrel why because of most of the money is in the West as opposed to near Libya. So they feel it more and the market reacts more. The market is a counting mechanism and a voting mechanism. I think the tools that people will use for investment

are going to change now because you'll be able to actually visualize stories better and people will kind of introduce that. This is kind of why people invest on themes like ESG and you have these index trackers and all these other things. So I think that's going to be very interesting. But then which industries will be disrupted faster than others?

We don't know. Who will be at the edge and the forefront of this, embracing this technology? And then who can be left behind? We don't know. So I think the entire stock market is going to change and adjust. It could be a period of supernormal profits and then out-competition as well. But this is against a period of super normal profits and then out competition as well but this

is against a backdrop of inflation and recession and all sorts of crazy stuff all at the same time so will you be using this technology to assist a hedge fund manager out there on investing i can run my own hedge fund no look i mean again i think this technology would be pervasive because people again will see the power of this and ultimately like i said investing usually comes down to stories you know like for all influencing influencing human desires and intention exactly like you know if you're doing vc to fund management to whatever you just basically you've got a story you can say for all you want that you're trying to be like quantitative and this and that. But nobody knows the future.

So you construct a story. But, you know, you're saying what is the evolution of that story going to be such that someone will buy this from me at a higher price? Yeah. That's all that matters. And part of it, like, you know, for entrepreneurs, one way that I suggest to look at things is what's the terminal value? You get someone to agree to that. All you're doing when you're raising money is you're de-risking your path to that terminal value. And people don't realize that's all of investment.

Yeah, it's a very simplified view of it. So I'm excited to have you speaking at Abundance 360 this year. Thank you for joining us in March. Thank you for having me. And, you know, I've dedicated an entire day. We've added a fourth day of the program focused on AI because it's just, I think, people need to understand how powerful and disruptive and a view of what's coming. Another one of our rock star speakers there, someone who you know, Ray Kurzweil. And Ray's going to be joining us to speak as well.

And, you know, it's been Ray's long-held belief that we're going to reach human level A by 2029. However you want to define that. What's your thought about that? Do you think that that is the case? How do you think about it? I'm not sure. It's singularity in the array.

Again, however you define it, I think my ideal view is that a future that's an intelligent internet. Every person, country, culture, and company has their own models. And they're all interacting with each other in an optimal way for humanity so again i kind of go to this thing of you know tim urban and wait but why when he wrote the article about neural link had this option this discussion over the human colossus i would like agi to be something of all of us working for us to make us better and i think part of this is why I'm basically pushing open source and these open source models. And I've created an organization that goes and has all these verticals

that will then be spun off into independent organizations so we can standardize the architecture across that. And it can be an emergent global consciousness, as it were. I think it's important for us to talk about the idea of these models and these personalized models to understand what that means. So when you build a massive model that is trained on everything out there, it's not really necessarily useful, but you talk about creating models for nations, for cultures, for companies, for individuals. Can you just give us the 101 on that

so people understand the power of that and the value of that? Knowledge evolves at different paces, right? There's knowledge on how to use a toilet that's been with us for many, many years, you know, versus knowledge on foundation models, which is just very new. It kind of signifies fashions as like pace layering, as it were. I think the way that models will be built is that you should look at them as like

pizza bases. So we're trying to figure out what an optimal pizza base is for a generalized image model. Then you've got maybe an Indian model with a bit of culture in there and then Indian fashion in there. And then Indian fashion for a mad to try and be fashionable when he's in India. And so you can kind of look at those as a basis.

Cause what we do is again, we flip the paradigm. So to train stable diffusion, like we built a 4,000 A100 cluster with our buddies at Amazon. To put that in context, the fastest supercomputer... 4,000 A100 cluster. Yeah. That's huge.

Yeah, to put that in context, the fastest supercomputer in the UK, Cambridge One, is 640. We've got about 10 times the compute of NASA. Wow. Roughly speaking. It's one of the top 10 in the world. And next year, it'll go 10 times bigger. Yours will go 10 times bigger?

Yes. And we make that available to everyone. And it's on the Amazon cloud? It's on the Amazon cloud, yeah. Because it would have been a bit too much for us to build it by ourselves. So we got them to lean in on that and kind of build this. Facebook has 16,000, for example, because they're pushing big metaverse. That's the fastest in the world, if you want to have an idea.

But again, this is an exponential thing whereby we have more than just about every single country. And we use that to take 100,000 gigabytes of images and compress it into 100 terabytes. Yeah, 100 terabytes and compress it down. So into two, effectively. So 50,000 to one compression of knowledge. All right. into two effectively so fifty thousand to one compression of knowledge all right and so when you say that compression when you're talking about a neural net you're you're not all of the connections or the pathways in neural net are useful or valid yeah right and so what you're doing is you're choosing which ones are yeah because it pays attention to what the most

important part of the attention. And then it creates these latent spaces that you poke with the prompts, which are the words that you put in. Right. So like a open AI, for example, made GPT three was 175 billion parameter model. The next step that after that deep learning, cause that's called deep learning,

this thing like, because we spend all the energy and all that compute, you don't have to, you can then take that base, that piece of base, you can inject some Peter and then you have a Peter model. It just takes 10 can then take that base, that pizza base, you can inject some pita, and then you have a pita model. It just takes 10 images, and then you can put yourself in anything. Or you can send 100,000 images and do an even more refined model.

But it forms a base. It's what's called a foundation model in a way. But you can even make it even more efficient. So GPT-3 is 175 billion parameters, so it's like 80 times bigger. And they said GPT-4 was going to be like 100 trillion parameters. Could be. I think that chat GPT is actually training GPT-4 right now because what happens is that you have this foundation But then in the classical day age of big data

What mattered was who you were so they took your data? They built these big models and they targeted you what matters now let's say, big models is how you use the models. Because not all those neurons are needed. We don't need all 2 billion images, right? Right. So GPT-3 was 175 billion. And they saw how people used it, and they identified the neurons that lit up.

Yeah. And they compressed it down to 1.3 billion parameters. So now when you're using chat GPT, you're training one of those models. Fascinating. And they're looking at how people are using it to compress it down. So stable diffusion can compress down even more but then this is personalized model that's very interesting because you can have almost this standardized base and then you can inject your

own context into it and the context of your culture your company your community and others and so that base is manual bill in that you can extend out the latent spaces. So like in stable diffusion one, we didn't really filter it. So Mona Lisa is overfit. There's too many pictures of Mona Lisa. So it's very hard to get her out of the picture frame. Now we adjusted it so you can make her swim very easily because it's not overfit anymore and it'll continue improving and adapting.

So we'll have better and better basis. And in a year's time, we'll have a very mature base that people can take and extend So like in Japan They took stable diffusion and they adjusted the text encoder for Japanese culture. Mm-hmm And then it meant that if you use normal stable diffusion because very Western oriented Salary man means a happy man. Yes in Japan diffusion is very sad man. Yeah, very sad man Indeed because I understand that context fascinating so in the future if I wanted to you know I coach a lot of entrepreneurs and a lot

of CEOs through abundance 360 and so forth if I wanted to create a virtualized version of myself that in in certain circumstances would react in a certain way that's pretty easy it It's coming. So pocket coach across, if I wanted to like have Tony Robbins in my pocket on in the right moment or the Dalai Lama, it's all possible. Yeah. And that'll be probably a four gigabyte file. And how far is that from now? So much straight now you do it now and you can do it fully realistic. Like,

um, you know, I think, I think the term is holy shit. That's amazing. Yeah. Pretty much do it now. You do it now. And you can do it fully realistic. I think the term is holy shit. That's amazing. Yeah, because you have the vulnerability. You can make it indistinguishable from a human. Stable Diffusion 2 is pretty much photorealistic now. Stable Diffusion 3 will break that barrier. And obviously we can animate now.

So again, that's kind of crazy. That's crazy. Metahumans, epic game style style or what nvidia doing you can also do human realistic voices with fully emotional range as well so like my sister-in-law ran runs a company called semantic um she reconstructed val kilmer's voice for val and top gun nice i was doing all the video games if you go to semantic.io you hear an ai tell you that it loves you and it's really creepy uh she just sold to spotify so i'm sure we'll have some really engaging podcasts and things like that um and like again across all modalities now in narrow you're achieving human levels and going on to human levels of performance and benchmarks from media to understanding to

output and the barriers to putting yourself in like I said you can train a model now in like less than an hour with ten a hundred images of yourself to put yourself in anything for a buck because we've done the heavy lifting of millions of bucks of pre-training the model as it were yeah whereas classically AI wasn't that. So if I wanted to create a virtualized host of myself, being able to on-screen play me, say what I want to say, that's here now? That's here now. The technology is here now.

The implementation needs to be there. One of the companies we work with is called Aletheia.ai. And so they use our language models. And so you can upload your scripts, and it'll learn how you speak. But then the voice technology, they haven't integrated the voice technology. We have our next generation audio technology. That'll come in a month or two and then it'll learn how you speak literally. I was using last year at Abundance360, I was using a company called Soul Machines and it had a virtual humans.

Yes. All right. And so I would love to create a virtual version of myself for anybody I want. And that's here. That's here again. Sol machines is upgrading now thanks to the technology that we're open sourcing. This is the other interesting part. Open source will always like closed source because closed source can always take

open and add in data and they can have very focused teams focusing on certain use cases. So we're literally upgrading the foundation of all of these companies as we release these models. Interesting. And then you mix and match and that's where the value is. Actually, who was it who said, was it my friend Jason or someone else? No, it was one of the other VCs. Most of the money in the world is made by aggregation and disaggregation. It just depends on which part of the cycle you're at. Interesting. is made by aggregation and disaggregation.

It just depends on which part of the cycle you're at. Interesting. I can see that. Yeah. I was having dinner with Reid Hoffman a month ago or so, and he said something which is interesting. He said every profession is going to have an AI co-pilot very soon. And I've been saying this for medicine. I think it's going to be malpractice to diagnose without having AI in the loop.

And we'll see what the time frame is. Five years is my guess. AI in the loop. We'll see what the timeframe is. Five years is my guess. But I can see an AI co-pilot as an architect, as a lawyer, as a chef, as everything. So how far is that? Well, I mean, it's here for code right now. So co-pilot literally was called from Microsoft, GitHub, and then CodeWhisperer now from Amazon. They help you write better code. It's about a 50% speed increase. Amazing. And that's what we've kind of measured so far, which is insane.

Like you type in, I want to have a piece of code that does this, and boom, it's there. And maybe you need to edit it, but it doesn't matter. It makes your life easier, right? Yeah. Like, say, well, Diffusion is a co-pilot for art, so artists use it to iterate rapidly on different concepts, and they take it, and there's a Photoshop integration,

and they use it as part of their Photoshop process process i used to say that the crowd was the interim step until ai right so github was the crowd and now you've got yeah yeah and you know there's questions around you know what was it trained on because it was trained on an entire snapshot of things but again it's like a different type of information flow to the web 2 economy we're going to skip over web 3 because it was a bit crap and we're going to go to web four or whatever it is now. And you know, maybe you can show it web four. I looks like AI.

You can do a cool logo for that. I should get that stable diffusion to make it. It's in fact, it's listening and it's made it. So I love Jarvis from Iron Man. Yeah. And I find Siri and Alexa and Google now kind of disappointing. How far are we from Jarvis?

So I think Jarvis is probably two to three years. Because right now you've got a MacBook M2 in front of you. 16.8% of that chipset is a neural engine that's optimized for these transformer-based architectures. Stable Diffusion is one of the first models to actually go down to that level. And so when will we see that on this machine? I think Apple is basically aiming for like 70, 80% of everyone to have this. And then you can go from Siri 1 to Siri 5. And this is why Apple has been talking about privacy and things like that, because the new paradigm of the internet, whereby the classical web 2 internet was

intelligence at the middle, coordinating us and feeding on our dreams and hopes and emotions to sell us ads now it's intelligence at the edge whereby you've got your own apple id you've got your own privacy layer and then you've got chips that can run ai at the edge that really understand you that's why the noggin's experience is seamless and google also realized that so they're building stuff into the Pixel phones. Yeah, I think people need to realize that the power of these future systems, call it Jarvis for lack of a better term, is when you give it open access to everything in your life.

You let it watch what you're eating, you let it read your emails, listen to your conversations, because it makes the world, the term I use is automagical in that regard yeah and you need to have the foundation models to do that because it needs generalized knowledge and then specific knowledge and contextual knowledge so it can adapt to your needs this human in the loop process is very important and so there'll be big ais in the cloud but then a lot of ais on the edge and they'll interact with and talk to each other and battle each other because what these models also do is they take structured data and turn it into unstructured

data. So again, stable diffusion, a few words, Robert De Niro's Gandalf, you get a photorealistic picture of Robert De Niro's Gandalf. Yep. That's structured to unstructured data. Yep. And then you can go both ways. A brief note from our sponsors.

Let's talk about sleep. Sleep has become one of my number one longevity priorities in life. Getting eight deep uninterrupted hours of sleep is one of the most important things you can do to increase your vitality and energy and increase the health span that you have here on earth. You know when I was in medical school years ago I used to pride myself on how little sleep I could get. You know it used to be five, five and a half hours. Today I pride myself on how much sleep I can get and I shoot for eight hours every single night. Now usually I'm great at going to sleep. If I'm exhausted, you know, I've worked a hard day, I'm right out. But if I'm having difficulty and it occurs, I'm having insomnia or my mind's

overactive and I need help to get that eight hours, I turn to a supplement product by Lifeforce called Peak Rest. Now Peak Rest has been formulated with an extraordinary scientific depth and background. Includes everything from long-lasting melatonin to magnesium to L-glycine to rosemary extract, just to name a few. This product is about creating a sense of rest and really giving you the depth and length of sleep that you need for recovery. It's a product I hope you'll try. It works for me and I'm sure it will work for you. If you're interested, go to mylifeforce.com backslash Peter to get a discount from Lifeforce on this product.

But you'll also see a whole set of other longevity and vitality related supplements that I use. We'll talk about them some other time, but in terms of sleep, Peak Rest is my go-to supplement. Hope you'll enjoy it. Go to mylifeforce.com backslash Peter for your discount. So let's talk about something which I have an opinion about, and I'm curious about yours. I think I know it, which is the idea of privacy, which fundamentally people all want privacy. I don't believe it really exists. I mean, when AI can read your lips, when there's data flowing everywhere, where encryption... What are your thoughts about

privacy and how do we deal with it? Do you have any ideas? I think for privacy, you should always look at kind of what the downside to not having privacy is, right? Actually, people are more than willing to give up their data. Too willing, in my opinion. Everybody clicks the yes, I accept this 15,000 pages of legal. Exactly. And then, you know, like, you have to think as well of different paradigms. Like, you know, in China, will there ever be privacy? Probably not.

And you have the social credit score, and it's an optical and being built by AI, etc. In the Western paradigm, what's the downside on the privacy thing? Like, what if your stuff isn't private? It's basically bad actors using you in certain ways, which can include AI algorithms trying to manipulate you. I think, again, what Apple's doing is building a paradigm for actual privacy because it's aligned with their business model. Even other companies now, Facebook and Google have enough information,

they don't need your data anymore. Like, who actually wants your data? I think is a question. We view ourselves as these snowflakes, like, you know, wonderful things. Who actually wants your data at this point? But the systems have adopted to do that and policies adopted to do that as well. With GDPR and all these other things. Some of them overreach, I think a bit.

But we are moving to this area whereby nobody needs your data anymore. And also the systems are now available to give you that privacy that you want and i think people want to opt in rather than opt out of a lot of different things to get more resources and other stuff finally the final element is that federated learning has matured now what does that mean so federated learning is when you take the model to the data so you used to have to ingest all this data and train the models. Now, if it's just like a freaking gigabyte, you send the model to the data, it can train. And then without saying it's PETA or EMAD, it can upstream the output. So we're seeing that in HDR UK, for example, health data.

We're seeing it with Melody, which is a thing with a lot of pharmaceutical companies coming through to open source analysis of patient data. You can finally get that where you don't have to sacrifice privacy to build AI and models. And that's going to be pretty amazing to, again, advance the field because you have access to so much more data to build better models. Amazing. Let's talk about the perceived downside. And, you know, I have to imagine that as much incredible compliments and the world should thank you for the work that you're doing because of the impact it's going to have. You're going to have to have detractors who are worried about technological unemployment or malicious use of AI or fake news and all of that. What concerns you? And I know you're, you're a principled man who thinks

about this deeply. What concerns you most? I don't have all the answers. Like, and that's, and that's a fair statement to me. I mean, like genuinely, like what I saw was that very few individuals had control of this most powerful technology. And then, you know, like it was very weird things. Like people like open source ai is like nukes i'm like so why should you control the nukes you know it was a very strange kind of thing they're like no only like it shouldn't be open source so why should big companies control it like again we live in largely a democracy we live in a society and so my take was like let's educate people get this technology out there and let's have a common conversation about it because I have my own viewpoints and they're there.

But again, I'm not representative of anyone. I'm just me running my own company, trying to catalyze this because I thought it was important. Given the fundamental change of society that will be caused by this technology now, because exponentials are a hell of a thing, for it to get out there. And so you need to make a splash. So, you know, we've got hate mail and kind of all sorts of things because it is disruptive. And we have to be aware of that. It is crazy and it will cause fear.

We have to be aware of that. And we have to decide together how to do that. So, like, for example, there are artists in the data set because it's a snapshot, right? Sure. It's less than 0.5%. And so is it ethical, legal, and moral to have them in there? So people can prompt an art style and then mash them together.

I think it is. But does that mean that we disregard artists who want to opt out of the data set? No. Yeah. Because they're part of the global community. So we've built an opt out and opt out mechanisms. And by the way, those artists are influencing other artists normally in the course of just them going to museum yeah exactly and you know what we have is now we've had like four or five thousand artists sign up for spawning

half of them have opted out half of them have opted in because they'd love to see their their work influence the world but how many people have really absorbed the parameters of the discussion? Very few. So like I said, my thing is that again, this is fundamental infrastructure. Access to this technology is a fundamental human right. Because otherwise what you're going to have, this is a discussion that,

you know, I know you've had many times superhumans and normal humans. Yeah. The ability to communicate and create makes you superhuman because just not only images, like it's presentations, it's being able to, like we have voice to voice technology that can allow you to speak more

confidently. It's interesting. I mean, people need to realize that today the poorest among us in society have more than the Kings and Queens had, you know, a couple of centuries ago. And this is about leveling the playing know a couple centuries ago and this is

about leveling the playing field this is about this is about this is the technology and this is what i care about deeply i don't know you do too uplifting humanity enabling every man woman and child to have access to food water health care education and have a voice and have a voice they are invisible and have dreams yeah and have dreams and have the tools to fulfill those dreams to have agency yes agency is the right word i had a bit of a flipping comment because again i can do what i want in my kind of role it was like humanity is creatively constipated we're going to make it so i can poop rainbows i think that's great it's a silly con but again it's the reality because people don't believe they can create they don't believe

they the mentality and mindset is wrong because people have more agents than they can do an individual can shake the world or the individual can make anything around them better but not if they don't believe they can and this is why art therapy is used in mental health settings to amazing things we've been conditioned to consume rather than create. We've been conditioned to be polarized rather than talk to each other and communicate with each other. And this can, again,

can change that. And again, that's why, like I said, this should be, in my opinion, a human right. It is infrastructure as important as 5g.

And what I'm trying to catalyze now is not that I build the company that makes the decisions for that, but that we put it out there and we're spitting off a Lutheran and other things. I just figure out governance structure. It's not the UN. It's something else. individuals are held back or restricted, feel they have no hope or a world where every mother knows her children have access to the best healthcare, the best education,

you know, the best ability to create. That's a more peaceful world in my mind. Yes. A hundred percent. I mean, look, all wars are based on lies.

Okay. Because for, for otherwise both sides can be couldn't believe because humans are humans like to kill another human is disgusting yeah right and so you have to tell the lie that that person is the other and you have to communicate it and control the means of communication you look at kind of again where conflicts are resolved when people realize they are humans and we're all part of a global society but our infrastructure has been set up to polarize. Literally, we can see it visually. This is how it happens.

The incentive structure is misaligned. So how can you fight polarization if not by communication? And how can you do that if you don't give people these tools and you create it so there is a base foundation for the world? So there are generalized models that are global. Then every country has its own AI policy using their variants of those models. And then because it's all standardized, we can hop between one and the other. That is a peaceful future.

Yes. And a future worth working towards creating. Working towards creating. But it's also now is the first time we can build that future because of this disruption in technology. Governments are, there's a definition of a government. It is the entity with a legitimate use of political violence. Right? The only one.

That's a sad definition. But it's the nature of it, right? Because you saw lots of political violence and then it was consolidated into one entity. They can imprison you. Yes. And they've got an army backing the currency and this and that and that right and governments rule on the basis of pure legitimacy to violence and again we see that kind of thing right so against this what typically changes a government or a

society it is an act of violence in some ways it's an act of disruption it can be a technology it can be a revolution or anything like that this This is a revolution that's happening just after COVID when everyone's thinking, holy crap, the system was rubbish. Let's do better. Yeah, the ability to respond. Yeah, I agree. And the challenge is that in the world today, you can't transform a government gradually. Yes.

And this is why as well, crypto, there were some amazing things in there and amazingly smart people work in there. It's rubbish because it tried to build a system outside of the existing system. And there was this system. And then there was the interface fortunes were made there and fraud happened there. Yeah.

Whereas this AI, because it can bridge structure and unstructured can actually go into our systems, out, compete them and make them more efficient and bring them forward. And it's the first technology that can do that dynamically and at scale. Or build virtualized systems that are de novo that we spend our time in and opt out of the existing system and into the new. Exactly. And it can out-compete. and opt out of the existing system and into the new. Exactly.

And it cannot compete. Or the final thing, of course, is that if you keep it as it is, whereby it's controlled by the few, they will ultimately use it as a system of control. It is the panopticon forever. And again, we're seeing this in China and other places with the social credit score that's about to be augmented with AI. Everyone's looking at everyone else, monitoring everyone else. What is freedom there? Maybe it's still happiness. Maybe it's AI. Everyone's looking at everyone else, monitoring everyone else. What is freedom there?

Maybe it's still happiness. Maybe it's control. Let's give people the option, right? Let's give people the infrastructure and the building blocks they need to be independent, happy. And maybe they don't choose independence. Maybe it's a bit more kind of Borg-like. That's fine.

At least you got the choice. Yeah. Because, I mean, this is, again, kind of if we tie this all back, the fact that you can now have AI that can write better than a human, Yeah. Because, I mean, this is, again, kind of, if we tie this all back, the fact that you can now have AI that can write better than a human, that can draw better than a human, that can emote and speak to emotional tones, means that, let's say, for instance, take one aspect of it.

Companies that are ad-driven, that sell ads, they can create the most manipulative ads in the world ever, and regulators will not regulate that in the world ever. And regulators will not regulate that. That's interesting. And they do know, but now it's the next step up because they have these latents that resonate. So like now, when you look at some AI art, artists can complain what they want. It's resonant. When you listen to the most advanced AI voices, it's emotional. You can feel it. It tugs at something. And again, this is a breakthrough that's literally a year old. Let's talk about the ethics and morals. Does AI have a moral compass? Should it have a moral compass? Well, I think the creators of AI,

technology is not neutral. Okay. The creators of technology do have a responsibility and they will never make it neutral because it embodies their perspective and it embodies the data set and other biases and things like that.

So I think that AI itself, this particular type of AI, again, if we just take the model, it is the action upon the model that then leads

to the output. So there's a responsibility there. But then, how do we adapt the model? Do we just have a one and done thing that's how they train in western values and norms and mores which is which is the way it's been going historically in the large corporate setting yeah because there's nothing you can do like you can't build a swahili version of it because you don't have access to technology whereas now with the pre-training and other things like that you can do that with one graphics card. That's great.

Right? Because again, we kind of flipped the paradigm of an AI needs to be going all the time to this pre-compressed foundation model. So I think that, you know, then there is the things of like, when you've got self-driving cars and other things like that,

what are the ethical norms, the trolley problem and everything that you can put on that? These are not easy questions because you're extending humanity, which then means you're also extending the ethics of humanity and that is not the same around the world education one of your moonshots uh we first got to know each other through the global learning x prize that uh elon musk and tony robbins funded uh you were one of the leaders of the one billion team. I wasn't the leader.

I just kind of helped implement it with Joe and Imagine Worldwide. Yes. And so what, so if you don't mind, what did your team do in that X Prize? And then where are you taking this vision? So my co-founder, Joe and I, Joe led this, made Imagine Worldwide to take the winners of the X Prize Kit Kat school and 1 billion, who are the real champions, and implement it around the world.

Into the Rohingya camps and Malawian camps and others. Like, I just support from the side and cheer him along as he kind of goes and does the really hard, valuable work. But now we've seen, I believe, the latest stats from the randomized controlled trials, because you need to actually implement it and see what happens. I think 76% of kids in refugee camps learning literacy and numeracy in 13 months without internet yeah it's immense it's like one hour of use per day is the equivalent of them being in school pretty much yeah because like it's one teacher for 300 students 400 students but then is it enough no it's not enough what we've done at the

moment what needs to happen is there needs to be a big grand challenge whereby, you know, Malawi and Malawi kind of has said that they want to roll this out at super scale and so have multiple other governments now that we have the RCTs. Let's make an open source ecosystem that has AI at the core, that teaches kids and learns from kids. So you take from what's happened in Malawi, move it to Ethiopia, Sierra Leone, ring accounts, what's happening in Malawi, move it to Ethiopia, Sierra Leone, ranking accounts, Brooklyn, everywhere. And so there is an actual superstar, amazingly well-created ecosystem for education.

And again, go to the future and bring it back with us. Love that. This year at Abundance360, Sal Khan is going to be joining us as well. Have you spent any time with Sal? No, not yet. Okay, so I'm excited to connect you guys because you know, he's built something pretty extraordinary, but his vision is to bring AI to it so that it's AI is generating the content, not him. And it's, and it's able to rapidly iterate for cultural appropriateness and so forth. This is why we need to build. So one of our things is that we're building national level models um from india to other ones where there's localized data sets and other things if you can get the education piece

going remember how i said that um this ai is like a kindergartner or a grade schooler sure it matters what you teach it so right now we're teaching everything do we need to teach everything no so if you have an ai that teaches the kids and learns from the kids that's the best data in the world to teach an ai for a kenyan model or a Nigerian model or others. And you know who should run that technology? Nigerians and Kenyans and others. And so one of the things we're doing is... It's almost like family-based learning and extrapolated from there.

Exactly, because we don't know best. We can give tools and we've reduced the barriers to create national-level, localized, cultural models. And then those models together form a constellation that not only have you got base learning of, like, what's the optimal way to teach linear algebra getting better, then you can go beyond that to that. And the plan is to have an integrated system where it's hardware, software, deployment, curriculum. Because then we can update that through mesh networking or the amazing work of Project project giga which is putting high-speed internet into every school in the world by the un and then you can put health care on that yes and then you have a self-adaptive improving health care system self-adaptive improving education system and then the world i mean for me that it is the that's the calling uh that i think both of us have and hopefully

many entrepreneurs here it's like what greater purpose could you have in life than uplifting humanity in that fashion? Exactly. And then as an ex-hedge fund manager, you can fund that at scale by bonds. Yeah. And the world's biggest problems are the world's biggest business opportunities, right? Exactly. Why to become a billionaire, help a billion people. So one of the ways that we kind of doing that is results oriented bonds, whereby you can pledge $20 million for million kids are provably educated on this standardized architecture. The invisible become visible and measurable, measurable infrastructure banks and the world

bank fund the remainder held by the pension funds. And you can divert billions and billions of dollars into this. It's kind of the promise of one laptop a child can finally be done. But rather than building a system for today, we build a system for tomorrow that constantly adapts and improves, is understandable and standardized because that is infrastructure. Yes. Again, the thing I really want to emphasize is this AI is infrastructure. Yeah.

It's more important than 5G or anything else. It's oxygen in the room. It's oxygen in the room. Yeah. So when I think about the future of education going out 10, more important than 5g or anything else oxygen in the room it's oxygen in the room yeah um so when i think about the future of education going out 10 20 years and bringing it back to today uh for me it's not a book and it's not a flat screen it's going into a virtual world if i want to learn about plato there's a guy sitting on a slab of marble over there and he says hey peter come on

over let me show you around introduce you my friends and it's experiential yes and that that npc of plato is trained up by all the knowledge about plato by every historian and it's accurate and the imagery and so forth and and what you just said earlier about real-time rendering um from diffusion enables that, right? And the ability to take every historian's work on Plato and train up a model on Plato enables all that. Exactly. And particularly when it is at the hardware level, because typically what software happens is that, again, you build layers and layers and layers of kind of compilers and translations. So you're far, far from the hardware.

This model is already efficient at the top level. What happens when we optimize them and push them down to the hardware level? Don't need internet. You don't need anything. You can form it. But then all of a sudden you have the young ladies illustrated primer. Yes. Neil Stevenson's incredible book and a vision for,

for the future of education. Exactly. Yeah. And we can make it, finally. But we can make it more. We can make it closer to the prime radian and foundation. So are you building hardware, Imad? I'm getting other people to build it for me.

My life is not complicated. But you're setting the specs. We're setting the specs. But this is the thing. Who is we? Okay. The way that we're going to do it is that it's similar to the grand challenges and the prize and that like let's get together we will drive the process because rule by committee you

know never works but let's invite everyone to participate from the kids using the tablets to code them yeah to the most advanced developers in the world and let's build something for humanity that's the way to do this amazing amazing so So you don't like the term Web3. You wanted to jump over to Web4. But this virtualized world, which is the convergence of AI and VR, AR, blockchain, and so forth, where do you see it in the near term going? It's just going to go insane. So we have technology.

I'm going to make an announcement in January about some of our technology. You've got Apple likely having AR glasses, Snap, Oculus, all of this. It'll be fully immersive worlds where you can engage. When I was a video game investor, I looked at time to flow fun and frustration. Yes. And I think this technology can adjust all of those and create worlds for us, but it'll be a year or two because again, now it's percolating and it's getting ready to then create brand new

experiences on the fly for everyone. And I think in a couple of years, it starts going in five years, creativity, imagination, engagement, entertainment is completely transformed within five years, it starts going. In five years, creativity, imagination, engagement, entertainment is completely transformed. Within five years? Within five years. It's going to be the biggest shift change that we've ever seen because the incumbents can't keep up with entities that are using this technology that can do everything.

Like, what does making a picture look like when you can change it live with just words? Right. And you can say, actually, make his hair a bit longer. Or, you know, get rid of those nostril hairs. And it just understands and does it without having to be coded. Intentionality, yes. Intentionality and action. Yes.

You know, it's kind of, you know, the whole military thing of observe, orientate, decide, and act, right? These systems enable that almost flawlessly. Like if you use OpenAI's Whisper technology, the translation of the podcast is just immense and all these other things because it's learned so many principles and they're tiny files. And then you can make it self-referential and say,

improve it the way you think I'd want it improved. You can. And like we're building technology, for example, that tells you how good a story is or how good code is. Yes. So then you have a creation and a discriminator. They bounce back and forth against each other. Yeah. And they learn from your personal context.

Amazing. You took a break from being a hedge fund manager to address your son's autism, which was, I mean, there are individuals like yourself, like Martine Rothblatt and others who are like, you know, yourself like Martine Rothblatt and others who were like, you know, I refuse to believe that I, that something can't be done. Yep. And you jumped in, um, for those who are dealing with autism, uh, uh, personally or in their families or in the community, what were your learnings and what would you advise?

There's always room for improvement and everyone struggles with something. How old is your today he's 14 14 he's super happy um so i've asked burges and adhd myself you know i think they balance each other out that's what they say you're doing pretty well it's all right um so i had a lot of kind of issues around those things um but people again everyone's different and then diversity is our strength as it were but sometimes kids and others need help because they can't achieve their potential because there's too much going on so like i said i'm buy into the gabaglutamate balance theory of it and so we had some drug repurposing and other stuff on an n of one equals case that's not science as it were it's first principles and i don't think it's coherent enough to kind of be brought forward but it's interesting instead i think there are

certain things that benefit everyone such such as applied behavioral analysis, whereby because you haven't built up the words, there are short trials to try and reconstruct what a cup means. It's quite an intensive process, but it's also what people do after strokes and other things that make them lose their connectivity in their brain. Most of this thing, though, is about noise filtering and reduction of that. But I think percolation of that means that you have to look into first principles analysis of some of the science of what can cause it.

And then you have to bring that forward to what is safe. So don't do kind of crackpot theories. But there's an emerging science and studies showing things like N-acetylcysteine, sulfur refrain, other things that calm you down are probably the best. And a large part of it's just connection and engagement there so one of the things that we'll be doing next year is that we're taking all the research that i've done and formalizing it properly because again it's not a case of well i can do it so anyone should have it i did an n equals one case yes but then that should be extended out and this is what i realized when covid came along so i designed and launched, with the help of loads of people,

a collective of augmented intelligence against COVID-19, launched at Stanford in July of 2020. Was that the actual origin of stability? That was the first origin of stability. So we didn't incorporate at that time, but kind of put it together. I mean, it's insane how far things have gone in two years. Yeah. I mean,

we probably actually kicked off in 13 months ago. So yeah, it's, it was insane because I thought I saw it coming and I saw it like autism as a multi-systemic inflammatory condition where even now, if anyone asks you, how does COVID actually work? If you ask a scientist,

they'll tell you, we don't actually sure. Like why are ferritin levels high and why is this and that the reality is our base foundation is not good enough we don't have enough shared knowledge so i created that to create a system that's comprehensive authoritative and up-to-date so there's a nice blog about an ocd type site and things like that a lot of private sector enterprises that promised a lot didn't deliver and so i realized again this technology was the future and we needed to create open infrastructure for that so when we do our autism releases next year all of the knowledge and everything like

that will be available there will be a semantic scholar on steroids which allows you to access the information relevant to the type of autism that your child might have because I basically figured out there were 16 different etiologies or driving conditions but 30 of kids would get worse and 26 of kids would get better with the same treatment so it doesn't work yeah this era of personalized medicine needs a foundation and again that foundation must be common but we can't wait around to do it and it's interesting right the more i learn about the fundamentals of human biology the more complicated you can dig layer after layer after layer. And I don't think it's possible for, I mean, we're discovering so much because the tools

we're using, but we're going to need this level of AI to cognitively understand the interactions of the systems. Again, it's a complex hierarchical system, you know, classical Herb Simon style. And so we have to build new tools to enable that. The question is, are these tools closed and the company's trying to, are they open? And so our take is open. And then, so our business model is just scale and service around that, which is how all the servers and databases are. But open is secure as well.

That's why Linux is used everywhere versus Microsoft Windows, et cetera. So I think the final part of this is that, you know, again, it's all interrelated. Like the body is such a wonderful, powerful thing. If you look at longevity and things like that, we need, again, this first principles thinking to both make us healthy and live longer and be happier. Yeah. Have you thought much about what's coming down the pike with quantum technologies and quantum computing?

So like, I think very sympathetic to various approaches there like i'm quite i quite like the quantum annealing on the kind of d wave um because i think that you know carl christian's theory around kind of free energy principle and having these low energy states makes a lot of sense in fact this is similar to what the ai models do in that when you look at the latent spaces you're going to the low energy states of what that could kind of mean right um so that's quite a lot of jargon i think for a lot of listeners but basically um quantum computing kind of is another part of the puzzle a lot of people try to say one ai model and say oh we're going to put something it can do everything you know they look at agi this is the thing that can do everything the human brain is made up of so many different

parts and we're just filling in some of the missing gaps right now of which quantum computing is one of them yeah and it's it i think it's adding fuel to the fire of how fast things are moving ah yeah it's gonna be in you know we can see the cubits increasing and we can see you getting there i mean again part of this is a supercomputer thing right like our supercomputer we've been the fastest in the world five years ago yeah which is insane for a private company right absolutely you know like think about bigger than everything but if you look at like nvidia is the one that kicked it off because they put ai in the core of their graphics cards before this even happened which is why it's now a 32 billion dollar part of their business the av100s four years ago with the first iteration the a100s were the next now the h 100s are like up to nine times faster.

It's literally going exponential. The compute is going exponential. The research, the technologies, we see exponential technologies everywhere. And we're like, it's so difficult to piece all these things together to see. I don't even know where things will be in a year,

let alone five years, let alone 10 years. I think that's an important part. And this is what we're, when Ray talks about the singularity, you know, it's the notion that we're, you know, the ability to predict what's going to happen next has become impossible.

And that timeframe, you know, people say, what's it going to be like in 20 years? I can barely think about 10 years from now or five years from now, let alone 20 years from now. And this affects the way that we act. So the way that our brains work is that our default is decision-making under risk. So we look at the upside downside of something and we make an expected utility calculation based on that because systems are stable. Is the world stable now? No.

Yeah. You know, so what do we do instead? We make decisions under uncertainty, which is we minimize for maximum regret, you know, or actually just minimize for regret. So with these powerful technologies, people like don't give them out because I don't know what can happen. It's the amygdala. It's a dystopian point of view. It's minimizing our downside, protecting what we have. It's a scarcity and fear-based animal brain, you know, reptile brain that we default to.

And this happens synchronicity. That's what we saw with COVID. It just happened. All of a sudden, everyone thought the same. It's what we're seeing with tech layoffs and things like that. Facebook makes loads of money. Why are they laying off? Things like that.

And so we have to be kind of aware of this because what's happening now as well is that the classical system has reached its end. We've borrowed too much money from the future. And now we are going to a negative sum game with inflation and other things. People are losing wealth rapidly. That leads to unstable Nash equilibrium from a game theoretic perspective. So we go from one steady state and lurch to another. Inflation to deflation to maximum employment to job losses to fiscal stuff. And it's just going to be a crazy few years.

And that's the reason I called it stability. That was one of my questions. And that's the reason I called it stability. That was one of my questions. Obviously, there's a terminology of stability in the models that you're building, but stability in order to help stabilize society? Yes. Build a human OS or catalyze a human OS because it can only be built by society.

We're just a catalyst. Yes. Help guide that in a way because this ai can be finally the thing that can stabilize a complex system that is humanity and then i'll just achieve our potential that's as a platform a infrastructure platform that uplifts all of humanity yes and again it should be run by the people for the people so like our subsidiaries in the countries we're putting aside 10 of the equity for the kids that use the tablets and i'm never going to take a penny out of any of them until IPO-ing them because they should be owned by the people. I love that. Okay. I want to close this out with two topics around the moonshots and mindsets. If I were to launch an Emod XPRIZE, fully funded, and push it out to the world.

What's a grand challenge in XPRIZE that you would love to see materialize out there? I think this education is the key one. I think it's the next step of global learning, which is a grand challenge just to build this system. And especially for low-income countries. Again, the delta on the impact can be so massive on this, because it's infrastructure for the next generation. A lot of these emerging markets leapfrogged straight to mobile phones and skipped computers. Now they can skip to the AI age.

How amazing will that be? It'll uplift everyone it and and there's no greater asset to a nation a company anybody than the intelligence of its uh of its citizens citizens are intelligent they don't have access to be able to take that intelligence and build yeah to build for themselves and extend that we can make that infrastructure now to do it for the first time ever yeah amazing amazing what are the mindsets that have allowed you to be successful? Do you think I talk about a abundance mindset, a longevity mindset, a moonshot mindset, exponential mindset, curiosity mindset. Do any of those resonate for you? Or are

there other mindsets? Cause I think mindsets are the most important differentiator that we have. So like I've always been very lucky and i've achieved very interesting things i was never really motivated the last few years when i finally applied myself and what i'm good at is first principles thinking on i suppose bits rather than atoms but i realized there's nothing i can't do there's nothing people can do anything yes if they put their mind to it but they have to think in a structured way and they have to let almost water flow as it were. So what I always try to do is I try to make it a win-win for everyone to participate, to help, to extend this. And at a time of absolute change, you can make that happen. Um, so this is why I go to the future and I go back to the past and I work back that way, which I think a lot of people don't

because they get stuck in the present. So it's the moonshot mindset, looking to the future. And again, the present. So it's the moonshot mindset, looking to the future. And again, recursively back propagating. And I'll close out with, uh, if you were going to list the moonshots that you're working on right now, uh, you're clearly working in education. We've talked about that, but as someone who's like disrupt all the industries, uh, uh, healthcare as a moonshot i am working on everything everything i've got about 18 different ones but education is core and creativity is core those two enable everything else if you want to fix climate if you want to fix hate if you want to fix a lot of different things get those two right and that's the foundation for the future it It's a pleasure, my friend. Uh, what a fun conversation, uh, excited to share your wisdom and vision with the world excited for what you're

creating as a, as a, uh, fundamental platform and infrastructure for humanity. Uh, I wish you all the luck and look forward to supporting you any way I can. Thank you very much. Again, it's just a little catalyst. It'll be everyone else that drives this forward. See you in March at A360. Cheers. Cheers.

Everyone, this is Peter again. Before you take off, I want to take a moment to just invite you to subscribe to my weekly tech blog. Today, over 200,000 people receive this email twice per week. In the tech blog, I share with you my insights on converging exponential technologies, what's going on in AI, how longevity is transforming, adding decades to our life. In the tech blog, I often look at the 20 meta trends that are going to transform the decade ahead and share the conversations I've had with incredible tech thought leaders on how they're transforming industries. If that sounds cool to you and you want to try it, join me. Go to dmandes.com backslash blog, enter your email, and let's start this weekly

conversation. Let me share with you the incredible progress we're seeing in the world of technology and the positive impact it's having on our lives. Again, that's dmandes.com backslash blog. Looking forward to sharing my insights and incredible breakthroughs I'm seeing with you every single week.