Why I'm Leaving My Company Immediately (Stability AI) w/ Emad Mostaque | EP #93

Wouldn't it be great to manage your portfolio on one platform? Well, now you can trade all in one place on Robinhood. That means you can trade individual stocks and ETFs and also buy and sell crypto using seriously powerful and intuitive tools at one of the lowest costs on average without needing to manage multiple apps. Robinhood makes withdrawing and depositing crypto seamless. Send crypto to your Robinhood account or send crypto from your Robinhood account to other wallets without deposit or withdrawal fees from Robinhood. Trade all in one place. Get started now on Robinhood Trading. Crypto involves significant risk Crypto trading is offered through an account with Robinhood Crypto llc. Robinhood Crypto is licensed to engage in virtual currency business activity by the New York State Department of Financial Services. Crypto held through Robinhood Crypto is not FDIC insured or SIPC Protected network fees may apply to crypto transfers. Crypto transfers may not be available to all customers. Investing involves risk, including loss of principal securities. Trading is offered through an account with Robinhood Financial LLC member sipc, a registered broker dealer.

When did making plans get this complicated? It's time to streamline with WhatsApp, the secure messaging app that brings the whole group together. Use polls to settle dinner plans, send event invites and pin messages so no one forgets mom 60th and never miss a meme or milestone. All protected with end to end encryption. It's time for WhatsApp message privately with everyone. Learn more@WhatsApp.com these organizations are telling you that they're building something that could kill you and something that could remove all our freedom and liberty. And they're saying it's a good thing, you should back them because it's cool. They don't care about the revenue. They have the political power. People are scared of them. Power should not be invested in any one individual. If I can accelerate this over the next period, I don't have to make an impact. I should not have any power. Whereas again you see everyone else trying to get more and more power. The only way that you can beat it to create the standard that represents humanity is decentralized intelligence. It's collective intelligence. The data sets of norms from that will be ones that help children, that help people suffering, that reflect our moral, upstanding and the best of us and gathers the best of us to do it.

A week ago Imod Mistaq was on my stage at Abundance 360 talking about the future of open source AI. Democratized decentralized AI. The day after A360 he stepped down as CEO of Stability. Now, five days later, I've sat down with Imad to talk about why he's stepping down, what he's doing next, the future of AI. He takes the gloves off. He talks about the dangers of centralized AI and the potential for decentralized, democratized AI to be the only avenue that truly uplifts all of humanity. All right, if you like this episode, please subscribe. Let's jump in. If you're a moonshot entrepreneur, this is an episode you're not going to want to miss. All right, now on to imod. Good morning, imod. Good to see you.

My friend Ash is always pizza.

So you and I were on stage literally last week at the 2024 Abundance Summit talking about the whole open source AI movement. You were beginning to talk about decentralized AI. You were talking about where stability was, the speed of the development of the different products. And the day after the Abundance Summit was over, the news hit that you had stepped down as CEO of Stability and stepped off the board. So let's begin with the obvious question. Why? What happened? And I have huge respect for you and I know a lot of the issues in the past, but I'd like you to have a chance to share with entrepreneurs out there and folks interested in AI exactly your side of what's happening.

Yeah, thanks. I think that Elon Musk once characterized being a CEO as staring into the abyss and chewing glass. Because you are looking at a very uncertain future, having to make decisions and the chewing glass is all the problems that come to you all the time. And it's required to steer the ship when things are incredibly uncertain. And Stability is a pretty unique company at a unique time. We hired our first developer and researcher two years ago, and then in those two years we built the best models of almost every type, except for large language, image, audio, 3D, et cetera, and had over 300 million downloads of the various models we created and supported, which was a bit crazy. And then generative AI is crazy in terms of usually in a startup, you don't have to deal with global leaders and policy debates about the future of humanity and AGI and everything else at.

The same time of building code.

At the same time as building code? Yes, and especially building code at a fraction of the resources of our competitors. We had certain teams were offered triple their entire packages to move to other companies. I was grateful that only a couple of researchers before I kind of announced internally I was going to leave, left for other companies. And that was just Startups, no one left for another big company, which I think is testament to that kind of loyalty in the mission. But what we've seen over the last year is, or the last half year in particular, is the question of governance in AI is something that's incredibly important. And who manages, owns, controls this technology and how is it distributed? So we saw everything from kind of OpenAI to congressional testimonies to other things. And as you know, one of my things has always been how do we get this technology in the hands of people all around the world and then who governs it and how can we then take this technology to have an impact from education to healthcare to others? So stability is a company that build great base models. And we got to that point, the revenue is going up, which is always nice, you know, finding the business model. And again, I think it was always curious to see that as a deep tech company. Two years in, people were asking us, you know, why aren't you profitable? And I was like, it takes a bit of time and investment to get to profitability. I think OpenAI, they're not there yet, but they have eight years. I think all the comparisons are perhaps unfair. So reviewing everything and kind of looking at it, I was like, do I really want to be a CEO? I think the answer is no. I think there is a lot imbued in tech, but it's a very interesting.

Position just to dive into that a second because we've had this conversation because there was a lot of pressure asking whether for you to step down as CEO. And I think founders typically want to see themselves or feel they need to be the CEO. And I've heard you say recently that you view yourself more as a founder and strategist than a CEO. Is that a fair assessment?

Yeah, I think everyone's got their own skill sets, right? So I'm particularly great at taking creatives, developers, researchers, others and achieving their full potential and designing systems. But I should not be dealing with, you know, HR and operations and business development and other elements. There are better people than me to do that. So now, for example, our most popular thing, stable diffusion and comfy ui. The system around it is the most widely used image software models in the world. There are great media CEOs that can take that and amplify that to make hundreds of millions of revenue. So they should come in and lead on that.

So why now, pal?

What?

Is there anything that specifically tipped for you that has. I mean, because, you know, it has. You know, you have done an extraordinary job. This has been your Baby, I mean, you have to feel a whole slew of emotional elements. And I've had to step down as CEO on two occasions over the 27 companies. I've had to sell a company for pennies on the dollar and it takes an emotional hardship on you.

I've had calls for me to step down as CEO since the. Well, since 2022. But I always thought what's best for the company and the mission and where I look at the world right now, there's a few things. A the company has momentum, it has spread, it's turning into a business. Like last year, I said let's not enter into large revenue contracts because technology isn't mature yet and our processes aren't mature yet and you have to deliver. So we did a lot of experimental things we were setting up and again now it's ramping on a business side, on a technology side, the technology is maturing. Diffusion transformers such as Stable Diffusion 3 and Sora are going to be the next big thing. And again, stability has got a great place there. But I think there's also the macro on this. So if you look at the OpenAI CEO board thing, Sam Altman said the board can fire me anytime. This is the governance of OpenAI and then they fired him and then he is back on and he appoints himself back on the board. There's clearly no governance. OpenAI I respect the people on the board greatly. I think there's some great individuals but who should manage the technology that drives humanity and teaches every child and manages our government, who's really leading on that, that can build these models and do those things. As you know, I've always wanted to build the science models and the health team's done that. I've been doing the education work and then my concept of a national model for every country owned by the people of the country, all tied together. I think it needs to be by a web three, not crypto or necessary token framework. That's something that's a brand new kind of challenge and one that I think there's only a window of a year or two to do if you have highly capable models. Let's put aside AGI for now, which you can discuss later. Really accelerating then no one will be able to keep up with that unless you build in a decentralized manner and distributed manner for data talent, distribution, standards and more. So there's only a small window of time here to do that. And realistically, again, successful companies and these things are all great, but generative AI is a bit Bigger than the classical norms. Just like the whole life cycle of the company was a lot faster than the classical norms. So that's why I felt now is the right time to make that change and hopefully play my part in making sure this technology is distributed as widely as possible and governed properly. As pretty much I think I'm the only real independent agent that has built state of the art models in the world right now.

Yeah, we've seen a lot of turbulence with OpenAI. We just saw Mustafa from Inflection become part of Microsoft. And I am curious, I mean you had a now famous conversation with Satya a couple of days after stepping down. Was that investigatory on your part or was that just a touch base with an old friend?

That's just pure trolling actually, you know, like to let off some steam. That picture was I think from a year or two year or so ago. But you know, I think Sachi is an amazing CEO and you know, he responds again like the top CEOs incredibly quickly when you message and he's got a great vision. But there is again this concern about consolidation in tech. We didn't take money from any trillion dollar companies at stability, you know, we remained and retained full independence, you know, and you know, to the detriment of some of the elements that we could have taken very big checks and other things. And even though you have good intentions, we have to remember that companies are slight, slow, dumb AIs that over optimize for various things and that's not certainly in the best interest of humanity. When you have infrastructure, it's like this is the airports, the railways, the roads of the future. AI isn't infrastructure, gently isn't infrastructure, but it should be. And should it be consolidated under the control of a few private companies with unclear objective functions? Again, the people in the companies may be great. I don't think so. And this is a key concern. And part of that was that commentary. Again, he's doing an amazing job. He's consolidating lots of power for the good of the company. And also he has a, I think genuinely good heart and mission to bring technology to the world. But it is a bit concerning. Right? Especially with the new types of structure.

You'Re speaking of Sam, in this case Satya here.

Oh, Satya again. I think the commentary was always interesting, like Satya playing 4D chess, assembling the AI Avengers. I mean he is, he's building an amazing mass of talent covering the bases and Microsoft is doing incredibly well here. Right. If you asked who's doing Best in generative AI, people would say Microsoft. But there is has to be concerns about consolidation of talent and power and.

Reach before I get to your vision going forward because it's so important and what you're doing next. I just again, as a founder, as a CEO of a moonshot company, we have a lot of those listening here. Can I ask, how are you feeling right now? Because the decision to step down has to have huge emotional. Are you feeling relief? Are you feeling anxiety? What's. What's the feeling after making that momentous decision?

It was a big feeling of relief, you know, because there's a Japanese concept of ikigai.

I know ikigai, and I love it.

Yes. Yeah. Do what you get at, do what you like and do where you believe you're adding value and other people do too. Realistically, again, I think I was an excellent research leader, strategist, other things, but I didn't communicate properly or hire the right other leaders in certain other areas of the company and they're better people to do that. And so I wasn't doing what I was best at a lot or I could have the most measurable value. And it was tying down, you know, there's a lot of legacy, technical, organizational, other debt, especially when you grow so, so, so fast. And, you know, we were lucky that we had high retention in the important areas and we could execute in spite of all of that, in spite of going the big company route. I think, you know, moonshot founders, you have to do because you don't have the resources at the start and you have to guide the ship, you know, as it goes out from port. But there does come that transition point there. And there is a competing thing where you typically take on VC money, which has its own objective function versus your overall mission. So again, if you look at the generative AI world right now, how many credible, intelligent, independent voices are there that have had the ability to build models and design things and make an impact? There's not many. So I was like, that's where I can add my most leverage. And also the design space is again, is unprecedentedly huge because an entire market has just been created. Like, where does gentrify not fit and where does it not touch? And what needs to be built there? We need to actually have the agency to go and build that. And so I felt tired, relieved. I felt that now there's a million options. I want it, rather than taking a long break, get on with things, I've just done the first thing in kind of web3 and I've got a whole bunch of other things that we're going to discuss kind of coming and catalyze stuff that can make exponential benefit because you know, like massively transformative purpose here is I want every kid to achieve their potential and give them the tools to do that.

And I love you for that because you've been true to that vision. And I know on the heels of your announcement you've been reached out to by national leaders, by corporate, you know, by CEOs and major investment groups. And you have a lot of opportunity ahead of you. So let's talk about where you want to go next. You mentioned publicly and you discussed on our abundance stage the idea of democratized and decentralized AI. Let's define that first. What is that, why is it important and what do you want to do there?

Yeah, I think that when I said I'm going to move to do my part in decentralizing AI, people were like, isn't that just open source? You give the technology right and then anyone can use it. But it isn't. Decentralizing AI has a few important components. One is availability and accessibility. Everyone should be able to access this technology. The fruits of labor and there's some very interesting political and other elements around that. Number two is the governance of this technology. Do you have centralized governance? Because the models are the data. There's a recent databricks thing model where they show that you have massive improvements from data. We all know that. Who governs the data that teaches your child or manages your health or runs your government? That's an important question I think too few are asking. And we need data transparency and other things like that. So accessibility, you know, you've got the governance aspect of that and then finally you have how does it all come together? Is it a single package or is it a modularized infrastructure that people can build on and is available kind of everywhere? You know, does it require monoliths and central servers where if it goes down and you have an outage on GPT4, you're a bit messed up or someone can attack and co opt it? I think that those are kind of the key elements that I was looking at when I was talking about decentralizing AI. And you know, I've come up with an infrastructure to do that I hope as well.

So if you don't mind, let's double click on it even further. So you mentioned we don't have long to get there. If that's a true statement, why don't we have long to get there and Then what does getting there look like? If you had all the capital available and if the right national leaders were hearing about this because a lot of this is supporting the populace of a nation to have AI that serves them versus top down, what's it look like 2, 5, 10 years from now?

Yeah, I think you'll have both proprietary and open source AI and they'll work in combination. The practical example I give is that this AI is like graduates, right. Very talented, slightly over enthusiastic graduates. You've got those and consultants. But I believe, you know, on stage with Nat Friedman last week at a360 when we were there, he said it's like we've discovered this new concept, what do you call it? AI Atlantis. Atlantis, yes. Yes. With 100 billion graduates, that will work for free. Yes.

I love that analogy. It was a brilliant analogy.

Yeah, we need to figure out how to say Atlantis. But there's a few things here. First of all is the defaults. Once a government embraces centralized technology, it's very difficult to decentralize it and every country needs an AI strategy. A year ago, one year ago was GPT4.

Yeah, crazy.

How crazy is that? The AI safety summit in the UK the king of England came on stage or came via video call and he said that this is the biggest thing since fire, you know, and that was like what, six, seven months later? Where are we going to be in here?

Yeah, I think he took that from, from, from the. Found the CEO of Google. AI is as powerful as, as fire and electricity.

Yeah, yeah, yeah. I've heard the same from like Jeff Bezos and a bunch of others know, not Kindle, fire propofire, you know. But then if you think about it, norms are going to be set in this next period. Like you know, I'm in California, Louisiana at the moment. If you don't set norms on rights for actors and the movie industry then you could have a massive disruption just occurring. As full length Hollywood features come in a year or two generated, you know, if you don't have norms around open models and ownership and governance by the people, it'll be top down governance because governments can't allow that to be out of control if they don't have a reasonable alternative. And I think the window is only a year or two because every government must have a strategy by the end of the year. And so I think if you provide them a good solution that has this element of democratic governance and others that will be immensely beneficial. I think also it's urgent because we have the ability to make a Huge difference, as we kind of may probably discuss later. Having all the knowledge of cancer, longevity, autism at your fingertips. We have the technology for that right now. We have the technology that no one ever, ever alone again on those things or to give every child a superior education literally in a couple of years. Like, there is an urgency both from there's a small window, but also from we must do this now because it can scale and make that impact we have dreamed of for so long. The enabling technology is finally in. It's finally good enough, fast enough and cheap enough.

Everybody, I want to take a short break from our episode to talk about a company that's very important to me and could actually save your life or the life of someone that you love. The company is called Fountain Life, and it's a company I started years ago with Tony Robbins and a group of very talented physicians. You know, most of us don't actually know what's going on inside our body. We're all optimists until that day when you have a pain in your side. You go to the physician in the emergency room and they say, listen, I'm sorry, sorry to tell you this, but you have this stage three or four going on and you know, it didn't start that morning. It probably was a problem that's been going on for some time. But because we never look, we don't find out. So what we built at Fountain Life was the world's most advanced diagnostic centers. We have four across the US today, and we're building 20 around the world. These centers give you a full body mri, a brain, a brain vasculature, an AI enabled coronary CT looking for soft plaque, DEXA scan, a grail blood cancer test, a full executive blood workup. It's the most advanced workup you'll ever receive. 150 gigabytes of data that then go to our eyes and our physicians to find any disease at the very beginning when it's solvable. You're going to find out eventually. Might as well find out when you can take action. Fountain Life also has an entire side of therapeutics. We look around the world for the most advanced therapeutics that can add 10, 20 healthy years to your life. And we provide them to you at our centers. So if this is of interest to you, please go and check it out. Go to fountainlife.com Peter when Tony and I wrote our New York Times bestseller LifeForce, we had 33,000 people reached out to us for Fountain Life memberships. If you go to fountainlife.com Peter will put you to the top of the list. Really. It's something that is, for me, one of the most important things I offer my Entire family, the CEOs of my companies, my friends, it's a chance to really add decades onto our healthy lifespans. Go to fountainlife.com Peter it's one of the most important things I can offer to you as one of my listeners. All right, let's go back to our episode.

So the let's talk about the objective function of democratized and decentralized AI. Is it that the compute is resonant in countries around the world? Is it that the models are owned by the citizens of the world? Is it that data is owned? And how do you get there from here?

I think that you can think of the supercomputers like universities. You don't need many universities. Honestly, if someone's building good quality models, that's one of the things, as I said, stability. And we did the hard task. We could have just stuck with image. We said, no, we're going to have the best 3D image, audio, biomedical, all these models. And no one else managed that apart from OpenAI to a degree. In fact, I think we have more modalities than OpenAI. Again, what I describe this move is accessibility and governance and a few of these other factors. So I think what it means is that this technology is available to everyone. But you see now that you don't necessarily need giant supercomputers to even run it. We showed you a language model running on a laptop. Stable M2 will run on a gigabyte on a MacBook Air faster than you can read. We were writing some poems, various things. We see stable Diffusion now at 300 images a second on a consumer graphics card. Our video model was like 5 gigabytes of vram. This really changes the equation because in web 2 all the intelligence was centralized on these giant servers and big data. Now you have big supercomputers. And I think you'll need less with better data training. These graduates that can go out and customize to each country, but they must reflect the culture of that country. Like the Japanese stable diffusion model we had. If you typed in Salaryman, it gave you a very sad person versus the base model, giving you a very happy person. Right? So you must have graduates that reflect the local culture and then reflect the local knowledge. And then global models again, that reflect our global knowledge and can be accessed by anyone. But who decides what goes in that? These are some very important questions. And who vouches for the quality as well?

What's your advice to a national leader? Because we're now starting to see ministers of AI in different nation states. And what, what's your advice to them right now in this area?

I think my advice to them would be to start collecting the data sets that they would teach a graduate that was very smart through school and kind of other things. This is national broadcast data, this is the curriculum, this is their accounting, legal and others. And note that those data sets are infrastructure. They will enable the local populace and others to create these models. Because models are just data wrapped in algorithms with a bit of compute. That's the recipe, compute algorithms and data. And it's not going to be as hard as you think to train these models, but you have to build them to get standards. So by the end of next year, probably a year after, I would estimate that a llama 70B model or a stable diffusion model, so these are two leading models in image and language, will cost about under $10,000, probably even $1,000 to trade. And then it becomes all about the data, and then it becomes about the standards.

You know, it's, it's, it, it's interesting. There is so much knowledge in the world that will vaporize, sublimate over the decade ahead as people die. You know, cultural data locked up in people's minds and stories and so forth, that's never been recorded. It's an interesting time to actually capture that data and permanently store it into the national models.

Yeah, and again, I think people over focus on the models versus the data sets.

I mean it's datasets.

Yeah, yeah, yeah. With the exponential compute, you can recalibrate and improve the data as well. So right now a lot of the improvements in models are actually synthetically improving data and data quality. As you said, there's so much that can be lost. But now we can actually capture this and the concepts and the other guidance and have cross checks like you can deconstruct laws, you can translate between contexts, you can make expert information available to everyone. Because again, you have this new continent of AI Atlantis and all these graduates soon to be specialists that are on your phone. And that's incredibly democratizing, you know, because otherwise the knowledge is. Throughout history, knowledge has always been gatekeeped, always.

I want to get to health and education next. But before we go there, I know you were meeting with a mutual friend, Jules Urbach the other day, and Stability announced a deal with OTOY Endeavor and Render Network. Are you still an advisor to that venture?

Yeah, no, this is part of the whole thing. It's the first of many Web3 kind of elements there. I think Web3 is 95%, let's say 90%. I'll be generous, speculative and rubbish. But there is that 5, 10% of genuine people that have been thinking about questions of governance, coordination and others and have built things that are proper. So Otto is the bridge to the creative industry. That's why we're with Ari Emanuel and Eric Schmidt and others are on the board. And the Render network is a million GPUs largely from creative professionals that are available. And so the first thing I announced there was the initial 10 million gaps, 250 million of distributed compute to create the best 3D data sets like Stability. We funded and worked with Allen Institute and others on obvious Excel, which was 10 million high quality 3D assets. We're going to a billion distributed. You don't need giant supercomputers. But then that is a community good that is owned by the people of the network and accessible to non academic and others as well. Why? Because you need high quality assets to create better 3D models. We have a new 3D model TripoSL that can generate a 3D image from a 2D image in 0.5 seconds and that 3D model feeds into better 3D assets. And then what does that mean? It means we're heading towards the holodeck. Without the data, you're not going to get there.

Jules wants the holodeck for sure.

Yeah, so Jules and I are on the same page a lot. And you're not going to get there without again, a commons of data that can train the graduates that then become specialized with Star Trek or Star wars or any of these other IPs and then also setting standards around monetization, IP rights, all sorts of other things. And so a network like Render is really good for that. But I've been talking to a lot of people in Web3 about the different elements of the stack, because what I basically see is that we have the opportunity to build almost a human operating system models and data sets for every nation, every sector, coordinated through proper Web three principles. Again, not speculative tokens or anything like that. Making it so that every child in the world or adult can create anything they can imagine, they can be protected against the harms and they have access to the right information at the right time to thrive. And again, that's infrastructure for everyone. It's a common good.

Access to GPUs has been sort of the limited fuel. Do you think decentralized GPU structures like Render is part of that future is an important part of that future, I think.

So right now it's far more efficient to train models on these again, big supercomputers, the university. But the rate of exponential growth there again is insane. Last year to train llama 2 cost $10 million. In a year it'll cost $10,000. It's a thousand times improvement from algorithms, data, super compute speeds. And that's crazy if you think about it. Right? So I think this will be the limiting factor. I think the GPU overhang for language models probably lasts until the end of the year, but then there's plentiful supply because what you have is Nvidia makes amazing GPUs at an 83 or 87% margin. Right. But the actual calculations aren't complicated. Like we took Intel GPUs and we ran the stable diffusion three diffusion transformer training. So this is the same technology that's used in Sora. And stable diffusion 3 is multimodal, so it can train SORA models with enough compute. And I think us and them are the only people kind of doing this. I think maybe Pixar as well. And then it ran faster on the Intel GPUs than the Nvidia GPUs. But we know that it can run even faster because it's not optimized for either and it's still running fast. So what you'll see is a commoditization of the hardware once the architectures get stabilized. Because GPT4 is just a research artifact. Still, diffusion was just a research artifact. You're not this engineering phase yet. And you've got to the point whereby this runs on MacBooks, it runs on other things. So I think it's a short term phenomenon of the next year because people were taking a point in time and extrapolating it without taking into account efficiencies, optimizations, and the fact that models that work on the edge and can go to your private data will be more impactful than generalized intelligence. Everyone's over indexing on generalized intelligence and building AI God versus amplified human intelligence.

Shall we say, before we leave stability, it's now in the hands of the chair and your past cto. What do you imagine the future of stability is going to be going forward? I know that you're not involved anymore. It's under different leadership and there's. What's your advice to them or where do you think they're going to go?

You know, I can. The very basic advice. I didn't want any conflicts or anything because I'LL be setting up lots of new companies and you know, being a founder and a shareholder and against ability. I'm a founder shareholder.

In fact, you're still the majority shareholder I think as of right now.

Yeah, just about. Just about. Yeah. That will change I'm sure. New money will come in like we saw cohere yesterday On I think 10 or 20 million of revenue run rate they're raising at 5 billion. Money's plenty. Yes. Yeah. With the right leadership, I think that it can again have an amazing part to play in media and that's what I've suggested to it. And again there's a great team that continues to ship great models. So last week there was an amazing code model. Next week amazing language, audio and other models are coming out. So you know, you continue shipping and great products around that too. So that was kind of my advice to them. Let's focus on media and take that forward. But you know, I'm not the expert on the business side of things. I did the best I could. My expertise on setting this up and taking it naught to 10 and yeah.

00 to 1 is, is definitely a role that, that you've played here and allow it, allow someone else to take it the rest of the way. But the area that I know if I could.

Actually there's something I wanted to kind of discuss here that I think is quite important. Please. For again the founders listening and the moonshot companies, there is an imbalance of power when you have very visionary, highly competent leaders there. Like what I found at Stability is that everyone would be waiting for me no matter how competent because I was the one that could see around the corners and I was a bit good at everything. Even if I hired people that built billion dollar startups or leaders in research at Google or kind of whatever because you have the outsized things. What kind of bezel says you have to speak last in some cases and some people in meetings. Otherwise everyone just does everything you say and they also wait on you. Now what I find and what I told the team is that you're flat as a power dynamic. You're all on the same page. You're all kind of relatively equal owners and it'll be interesting to see how it evolves from that given that there's actually a business. And again I think this is something that you probably had a challenge with and other founders here whereby they put more on your plate because you are so visionary and because you're up there in the future and they're always waiting on you. So you're always like, well, my schedule is completely packed. My schedule now is actually quite free, which is also quite nice.

I've had a chance to speak with you every day for the last few days. So that's been a pleasure to have extra time on your schedule. So we do have a world of visionary founder led CEO companies, right? So you've got Musk and you've got Bezos, historically and you had Steve Jobs and that's both powerful and dangerous. The power is the ability for that because we don't ever have a company that is preexisting. A new CEO comes in and has the same both chutzpah and also the power of their vision. The danger there, you're saying is not allowing your team to step up with their own vision or being overly indexed on your vision.

Yeah, I think that can be the issue. And that's why I wanted stability to again reach the point of spread and revenue rate increase and other things before I did any anything. And I felt again this external pressure in that if nobody or there's very few people in the world actually thinking properly about governance and spread and others, and a very small window given the pace of this, to make a difference in the dent, I believed I had a reasonable approach to that, but I couldn't while remaining CEO of this company. And again, it's a pretty unique scenario because you've never seen a sector move this fast that has such wide reaching human implications. And regrettably there's too few people I think with the right alignment and approach in this area. I've been very disappointed. Usually what happens is you have power maximization equations and this is what we're seeing from the industry consolidation. How many people want to genuinely bring this technology to kids in Nigeria or to the global south or to help those leaders build their own models and believe also in a positive sum game. That was actually my biggest surprise from the discussion at Silicon Valley almost entirely. They all believe in flat or negative sum, zero sum or negative sum things where there has to be a winner. Everyone's a winner in this. And again, I was just very disappointed.

Seeing that I've been asking you for a while to write and distribute your vision white paper because I've heard you describe it in detail and it's brilliant and I still hope that the world will see it soon enough.

Everybody, I want to take a break from our episode to tell you about an amazing company on a mission to prevent and reverse chronic disease by decoding your biology. The company is called Viome and They offer cutting edge tests and personalized products that help you optimize your gut microbiome, your oral microbiome, and your cellular health. As you probably know, your microbiome is a collection of trillions of microbes that live in your gut and mouth. These microbiomes influence everything, your digestion, immunity, mood, weight, and many other aspects of your health. But not all microbes are good for you. Some can cause inflammation, toxins, and actually lead to chronic diseases like diabetes, heart disease, obesity, and even cancer. ViovioMe uses advanced MRNA technology and AI to analyze your microbes and your cells and give you personalized nutrition recommendations and products designed specifically for your genetics, specifically for your biology. You can choose from different tests depending on your goals and needs, ranging from improving your gut health, your oral health, cellular function, or all of them. I've been using VIOME for the past three years. I can tell you that it has made a huge difference in my health. And because the data they collect and the AI engine they've built, it gets better every single day. I love getting health scores and seeing how my diet and lifestyle affects my microbiome and my cells. And I love getting precision supplements and probiotics tailored for my specific needs. If you want to join me on this journey, journey of discovery and improve your health from the inside out, VIOME has a special offer for you. For a limited time, you can get up to 40% off any viome test using the code Moonshots. Just go to viome.com moonshots and order your test today. Trust me, you won't regret it. All right, let's go back to our episode.

Before we jump into into health and education. Let's talk about governance a second because we've seen governance complicate this. What is the right governance structure for this super powerful technology?

We have representative democracy that I think can be improved by this. I don't think democracy survives this technology in its current form. It will either improve or it will end. I don't see anything else like yesterday.

What does end mean here? A benign dictatorship driven by an AI overlord.

Yeah. Like yesterday there was announcement of an app called Hume which had emotionally intelligent speech and they can understand your emotions and talk with emotion. You and I have discussed this.

Yes.

You know where that's going, right?

It's very powerful.

It's incredibly powerful. And governments have a tendency. I mean officially, but say it, say it here.

It's important for you to state what it means because we've discussed it, but help people here be ready for this.

Democracy is all about representation. And you see the questions of deepfakes and things. Speech is one of the most impactful elements there. But now you can't believe anything you see, hear everything. So one path that we have is a 1984 on steroids panoptica where life is gamified and you listen to whatever the government says and they're incredibly convincing and you're happy and you've always been happy and you've always been at war with Eurasia. You know, propaganda on steroids. The other part that you have is things like citizen assemblies, consultative democracy. The ability to take. Right now you can take any of the bills in Congress and completely deconstruct them and find what the motivations are. You know, you can check laws against the Constitution in a seconds. This is incredibly powerful. Empowering technology from a democratic perspective. So I see two routes, unfortunately, because I think that once the thing goes, it goes really fast. Centralized government control increasing because the governments want to protect themselves as an organization. And every party says the other party is crap. We've seen the increasing polarization in America already. And fundamentally, come on, you can do better than those two leaders that are currently competing. I'm saying this is clearly our system is sclerotic across this. Democracy is the worst of all systems, except for everyone else. We can have a better democracy where it's actually representative and empowers the people, or we will have the end of democracy where it is a 1984 Panopticon, in my opinion, because the momentum will go. Then of course, you will start using this technology. You're already seeing it being used, but not at scale and not intelligently yet, which is scary.

I think we finally have the technology for a direct democracy versus a representative democracy, right? Where I can have my ex. My desires directly represented on any specific law or. But I think the point you've made before is that speech, if you look back to everybody from Hitler to some of the most persuasive politicians, is a powerful tool. And AI can become the most persuasive speaker out there.

It can take anyone's speech and make it far more persuasive. Like, I think my voice is a bit whiny. I can remove the wine. You know, I can go in a very polished British accent and other things like that. Right. You know, let's fight them on the hills and the harriers and whatever public speaking makes a big difference. Someone took Hitler's speeches and put them through an AI and took them into English. Because when we're not in the German context, and we listen to them, it sounds like he's shouting, like, what are this crazy thing? You hear him in English, it is very different in his own voice. Just like someone took Xavier Millay's one in the United nations and put him into English again. He sounds a bit shouty, but then he sounds very reasonable when it's in English. And you can take the phenomes of Obama's best speech and a bit of Trumpianism and a bit of Churchill and you'll have full modulation wave control over all of this. People are already using this technology that everyone should have a passcode with their loved ones because people are getting calls from their mother saying, help, I'm in an emergency, I need to send money right now. And you cannot tell it. And it pulls at the emotional strings. And if you look at something like US Radio and one side of the political divide is taking that over, imagine if you're hearing optimized speech every single day that will have a huge impact. And then they control the visuals and they control the other things. We're not set up for defenses.

Yeah, if it's optimized speech for you, specifically for you. Right. For the kids, you have the age group, they have where you live, your historical background and so forth. And n of one persuasive speech coming at you. The brain is not set up for defenses.

We're not. And we take this as an example of the YouTube algorithm. Like YouTube as an organization is not an evil organization, but it's an organization optimized for engagement, which optimized for more extreme content. So there's some dark places in YouTube which then optimized for ISIS. So ISIS videos spread viral. I don't know. Sometimes viral is good, sometimes viral is bad. That one was bad because it was extreme and they didn't understand why. And if you look at it, two of our largest generative AI companies are Google and Meta. And their business is advertising. Their business is manipulation. And they are both amoral companies because why would you expect a company to have morality? Our governments are also amoral. And again, you can view these things as slow, dumb AIs. So you can see the way they will optimize unless we do something about it and they will have full control. Like again, you put on your Vision Pro headset with your spatial audio that is full sensory control. Not full, but you know what I mean?

Full immersion.

Full immersion. And so we have to be aware of this. And there's obviously other tools that can be used, like in the wake of the AB spring government's targeted everyone that was on social media. We can do that on again, hyper personalized basis. We need to set some defaults and standards here to protect democracy. But again, why democracy? We're not really trying to protect democracy. Again, people have different definitions. What we're trying to protect is individual liberty, freedom and agency. Education should be about enhancing the education of every child. It's not. Health care is sick here. Our government should uplift us. But how many people believe our governments do that rather than put us down because they couldn't encapsulate and cater to the brilliance of each individual because they didn't have the tools until now. So that's why I said which way? One man?

Yes, which way?

Infinite agency or massive control? These are the two ways. Do we control the technology or do these organizations control the technology that controls us?

You know, when we were on the stage at the Abundance Summit, we talked about a future of digital superintelligence. Right. And a future in which we've got AI a billion times more capable than a human, which looking at it just from a ratio of neurons, is the ratio of a hamster to a human.

Yep.

Do you believe that someday we could have a benign super intelligence that is supporting humanity?

Yes, and I believe that it should be a collective intelligence that is made up of amplified human intelligence is amplifying all of us pilots that contain our collective knowledge and culture and the best of us and data sets that are built from helping and augmenting us versus a collected intelligence and AGI that is top down and designed to effectively control us. Again, if you look at OpenAI statements on the road to AGI, they say this technology will end democracy, end capitalism, and maybe kill us all. I don't like that.

I remember seeing that you texted me, you said, read this. Does this sound the same as it does to me?

Yeah. So what I'd prefer instead is for this to be distributed. If you have data sets, formations that are built on enhancing the capability of the nation that reflect the local cultures, and you push for data transparency on models which I believe we must have, you know, especially language models, then you're more likely to have a positive thing. And again, the human collective can achieve anything from splitting the atom to go into space if we put our minds to it. But we have lacked in coordination mechanisms, They've not been good enough. So if you create the human colossus and every single person has an AI that's just looking out for them to enhance their potential and then coordination AIs that is a far more positive view of the future. And that is the AGI. That is a general intelligence. That's the hive mind. General intelligence. Not a Borg style hive mind, but one that's really thinking. Again, every child should achieve their potential versus this embodied concept of an AGI. That's a very Western concept. And you see that as well. We look at the Japanese concept of a robot. The robot is your equal and your helper. You look at the Western concept of the robot, it's Terminator and Skynet and all of that. And again, I think this is again where cultural norms become very interesting. And what do we want to build? Do we want to build AI God, or do we want to build that AI helper that helps us and we help it? You know.

Those listening now, I mean, you can see IMOD's brilliance and why I'm so enamored with the way you think about this, because there are very few individuals who are looking at this from an objective function of what's best for humanity, what's best for every nation state out there. Let's talk about your going forward future. Are you going to build something in the decentralized side of AI, the democracy size of AI? Is there a company there or a fund in your future for that?

Yeah. So doing the white paper finally, it's getting there with a bit of help from AI.

I can't wait to help broadcast that white paper.

But look, I think the basic thing is this. What I want to do is set up a AI champion in every nation with the brightest people of each nation, working with the organizations of each nation to help guide them through this next period. Because there will be massive job displacement from the graduates going, massive uplifts in productivity from the technology being implemented. And again, that organization can help govern and create these data sets and these models that are so important. And every nation should have that. But then I also believe that every sector should have a generative AI first infrastructure company that builds this and helps the healthcare companies, finance companies and others through that. And to coordinate all of that, you need to have a Web three type protocol. What is the protocol for intelligence?

So what is a Web three type protocol? Define that for folks listening.

Again, people talk about web3. It's not about the tokens or the meme coins or anything like that. What a Web3 protocol is is that everyone should have like AIs first of all aren't going to have bank accounts. They're going to need some way to pay each other or exchange Value and again, Web3 has done a lot of work in that there needs to be some sort of identity, attribution and other format because you'll have this mass influx of information. And so again, web3 concepts are very useful there. There needs to be an identity concept because you'll have real and digital people. Web3 concepts are very useful there. So data attestation, all these other things, verifiability. So when I look at it, if you've got sectorally, my plan is to launch almost a company for every major sector and we can talk about health and education and bring the smartest people in the world to solve that challenge of the infrastructure for the future, every nation. But you need to have some sort of coordinating protocol for all of that that becomes a standard and that's the substrate for this amplified human collective intelligence.

And is that where you want to play and focus your energy next?

Yeah, it's setting up these organizations and bringing the brightest, smartest people that really want to make a difference there because there's massive network effects in doing this. But again, I just need to be the founder and architect. I don't want to run the day to day of any of these things. And then because the most scarce talent. There are three types of capital as I view it. There's financial capital, human capital and political capital. In order to effect a change in the world, you actually need all three. But the financial capital actually comes with the people capital and the political capital. And the smartest people in the world in every sector, from healthcare to education to finance to agriculture, almost all believe that James Fi is the biggest thing they've ever seen. In the last year, everyone's asking you, all the smartest people, Peter, what's next? Right? And you know, many of the smartest people in the world. So I want to create organizations that they can come, the chefs and the cooks, the thinkers and the doers, and think, what is the future of finance? What's the future of education? And then the national champions that should be owned by the people of each country become the distribution for the amazing infrastructure that they built. And there's a nice kind of vice versa. But then again, you need the coordination function. So I'm trying to bring together people in each of these and there'll be public calls and things like that to build that infrastructure, the future. Because as mentioned, AI isn't infrastructure, but it should be. Maybe it's the rocket ship of the mind, right?

I love that analogy, my friend. It is the most important infrastructure that humanity will have going forward. Across everything it does. And I look forward to helping you build it.

Exciting, right? Like, again, I think one of the things I got, I got lots of messages that like, I'm so sorry for your loss. It was like my dog had died after I left to see you. I was like, what is that? You know, they're like, I mean, it's nice. It's nice that people care, right? But I'm generally excited kind of about what's next. Like, you know, again, it was like staring into the abyss and chewing glass every single day. And that's not what I'm best at or where I could have the most impact. But I want it to be a point whereby if I can accelerate this over the next period, I don't have to make an impact. I should not have any power on this. Whereas again, you see everyone else trying to get more and more power. I want to make sure it's set up properly, but I want to give it all away. Because power is obligation. It's dragging. And again, it should not be invested in any one individual. We shouldn't have to rely on anyone being nice or good for this technology.

I was talking to Michael Saylor during the Abundance Summit that evening and talking about the fact that because Satoshi, when he set it up, did not retain any power and did not trade on the founding blocks and so forth, that that's the reason it's been able to succeed because there wasn't that centralized power. And you know, bitcoin had been tried. He said bitcoin had been tried many times before, but because it didn't have that initial anonymity and the dissolution of founding power, that that's the reason it didn't succeed.

Yeah. I mean, again, I think you need to have it accelerate. And you see this with movements, right? The movement starts, but then it goes. Once you've got the DNA and the story there, right, you see the profits, you see the leaders, you see the others. But then it's about setting the framework correctly and reframing the concept. This technology is not beyond. Look, Stability is a company that started two years ago above a chicken shop in London. Right. You know, my first 20 employees, I went to the job center and I said, bring me people that have overcome adversity and I will train them. Young graduates and six of them are still at Stability, you know, like, because it was a program and they're doing things from cyber security to run the supercomputers. We only had like 16, 17 PhDs yet we built the state of the art Models in every modality. We built mind reading models like Mind's Eye. You know, I remember that we contributed to all these things, yet you're told it's impossible to compete. We have shown it's not impossible to compete. That's a reframing. The reframing is data versus models. You don't need giant supercomputers for everyone. You just need to have a trusted entity to build it. Right. And so I hope to kind of convey this and then figure out this organizational structure that can proliferate so I can take the holiday.

So before we go further, let's talk about one area of your next chapter in life that we both have as a passion, which is the use of generative AI in health. It's an area that you've given a huge amount of thought to and I think you're excited about. Can you share what your vision is there?

Yeah. So I got into AI 13 years ago. Gosh. I was a programmer before for 23 years building large scale systems as a hedge fund manager and other things when my son was diagnosed with autism. And then I built an NLP team to analyze all the clinical literature and then looked at biomolecular pathway analysis of neurotransmitters GABA and glutamate in the brain to repurpose drugs for him. And he went to mainstream school, which was great. N equals 1. And then I was lead architect on one of the COVID AI projects for the United Nations, Lord Stanford and others. And then because I didn't get the technology, I was like, oh, we've got to build it ourselves. But what is health again? I think we have this discussion a lot. Healthcare is sick care. We don't have all the information that we should have at our fingertips. Health assumes ergodicity, a thousand tosses of the coin, the same as a coin tossed a thousand times. But we are all individual. Across the world there are amazing data sets that could be better because when you write down a clinical trial or your own kind of experiences, you lose so much information. At the same time, you don't have all the information on cancer, autism, multiple sclerosis at your fingertips in the comprehensive, authoritative and up to date way. So when I look at the health operating system, we're going to build a GPT4 open for cancer. And it's going to mean that nobody is alone again on that journey and loses that agency because they know comprehensive, hard to upstate all their knowledge. But AI models today already outperform human doctors and empathy. So they're not going to be alone on that anymore.

Can I just double click on what you just said? Because it's really important. I have so many people because of my role as chairman of Fountain Life who reach out and say I just got diagnosed with this cancer or my brother or my sister or my wife and they're left with this decimating news and they're left googling. But a model that's able to have the most cutting edge information and then incorporate all their medical data and give them advice in empathic fashion. How far is that?

A couple of years if we focus, maybe even next year. And that's amazing because all of these topics that again we will have diagnosis that is superior, we will have research augmentation because again, even researchers don't have all that knowledge at their fingertips. And again, this is public infrastructure and a public good from primary care all the way through to that one is the open infrastructure of the future where this technology can come again to your own data as well. You have things like Melody and other things around, homomorphic encryption, federated learning that they were trying to figure out how to preserve privacy. We can run a language model on a smartphone right now that can analyze all your data and then just feedback stuff to a global collector. But people are people. So when I look at healthcare I see amazing data sets that we can activate by taking the models to the data and infrastructure that we can build. Like we had Chex agent with Stanford, the top X ray radiology model to build good standard things across the entire gamut of healthcare so we can actually get to healthcare versus sick care. So we can make it so that everyone is empowered to make the best decisions either as experts or individuals and make it so nobody is alone again. As well as increasing the data quality that will then feed better models that will then save lives, save suffering and again increase our potential. Like you've got a long behind you, right? Why don't you have all the latest knowledge of longevity at your fingertips at a GPT4 level right now? That will happen over the next year. We will launch stable health or whatever decide to call it and there will be the smartest people in each of these areas working on that. So again, you're never alone. It doesn't matter if you're with a hundred billion dollars and your kid has autism, asd, there's no cure, there's no treatment, there's nothing. Doesn't matter how rich you are. Yet with just a little bit of effort right now, we can build it as an open infrastructure for the 5% of people in the world that know someone with autism, the 50% of people in the world that receive a cancer diagnosis of them, of someone they love, and they feel that loss of agency. So we're going to return agency to humanity that way. And again, it needs to be an open infrastructure that they can then access private data sets and compensate them appropriately so everyone is incentivized. We need that fast.

Yeah, and that's a beautiful vision. It is, again, infrastructure. And one of the things that's so beautiful about it is guess what? All 8 billion people were all human or all running the same software. And the breakthroughs and the knowledge accumulated in Kazakhstan is going to be as useful in Kansas.

Yeah, but this is the thing. Operating system. This is the biggest upgrade to the human operating system we can imagine because we're going from analog to digital. Text is black and white, whereas these models only understand context. Daniel Kahneman just passed, amazing kind of guy. But he did have this concept of type one and Type two thinking we had one, which is these big data things that can only extrapolate. But now we have these models that understand context. We have the missing part of the brain, and that will allow us to extrapolate, allow us to have more rainbows, have the context of each individual push intelligence to the edge. That's why again, there is this imperative to do this now, because there's a window on the freedom, agency, democracy side. But the other imperative is no one should have to suffer as they're suffering now.

Amazing.

And how much does it actually need? Doesn't need that much, which is the really amazing stuff. The total amount spent in generative AI, I think I said at the conference, is less than the total amount spent on the Los Angeles San Francisco Railway, which hasn't even started yet.

And in building stable health again, if that's what it's called, the amount of capital required to build that is de minimis compared to what's spent on a single human trial of any, any drug.

Yeah, it is. But then, you know, you build it and you get to that 8020 incredibly quickly. That will change hundreds of millions of lives and that will attract the smartest people in each of these areas. Thinking about what is the open infrastructure of multiple sclerosis, of longevity, of cancer and more. But then you can amp that because the value is so, so huge. And I hope to build a trusted organization as part of this whole human operating system upgrade. That's what I want to build. I want to build human os. Well, at least catalyze it again. I don't want to run or control or own anything. I want to figure out how to give back that control. Because who should decide what cancer knowledge goes in there, who should decide what education, et cetera.

Let's talk about the second half of your vision, which is how we originally met when you were one of the winners of the global Learning X prize that Elon and Tony Robbins had co funded. Your vision around education speak to us about that.

Yeah, you know, we're deploying it kind of. The winners are kind of separate, but every child, my entire operating system is like. If you think about things in terms of the rights of child today, they have no agency and so we must respect their rights climate. Everything becomes a lot simpler now that we have language models on a laptop. Like I said, you can go to LM Studio AI, download stable LM and it will run on your MacBook faster than you can read. It's crazy. We can have a GPT4 level AI from us or someone else on a smartphone or a tablet by next year. One laptop per child was too early, you know. Now we have this transformative technology. You have an AI that teaches a child, learns from a child. Are you visual, auditory, dyslexic. That's the best data in the world for a national model. But also to teach these models how to be optimistic, how to be.

This really is the young ladies illustrated primer. This really is Neal Stephenson's vision in that regard.

Yeah, but Nell shouldn't have had to find the primer. She should have had it from day one. As a human right.

As a human right, yes.

Our school's education system are childcare mixed with the social status game mixed with petri dish. You know, they teach our kids not to have agency.

Yes.

Whereas they should be telling the kids. Yeah, yeah, they should be teaching. It's a relic of the industrial age where everyone had to be counted. And you can't measure what you can't manage. So you manage the creativity and belief out of people. Everyone in the world can do anything. Why? Because even if you don't have that talent, you can convince someone else who does have that talent. But they don't believe it, so they can't do it. So what happens if we have an entire nation of children that have this helper that brings in the right information, the right time and tells them they can always believe that supports them entire world. What can't you do? Then they have all of the cancer knowledge at their fingertips and all of the engineering knowledge at their fingertips. And it's a constantly learning, adaptive and improving system. Again, right now almost the entire AGI and AI debate is about these machine gods trained on giant supercomputers that bestow their beneficence down or may kill us or whatever. What about that human operating system upgrade? That is a decentralized intelligence where that kid in Mongolia or Malawi or wherever can make a real difference to humanity. Some of the contributors to our open code bases for our models are 15 years old. They just taught themselves and just happened to be their wizards. You don't know in this new age, right? And again, they should contribute to the whole because once something goes into this model or this system and again it needs the verification and other things that can be dynamic. They can proliferate to everyone using that system.

Do you think that once this capability is built, it will run into blocks in different nations? Or do you imagine that this will become a, again a human right and not all nations? I mean, listen, there's no greater gift and no greater asset you can give to a nation's populace than intelligence and education. But I'm not sure every national leader wants to see that.

That's why I think again, there is a gap here. There is a year maybe where you can go to any national leader and say, I'll bring this technology to your people and I will empower the smartesty with people. I want it to be owned by the people. And what option do they have this is positive for them? What happens is that a lot of the corruption in the world is because of local maxima. Actually it's weird because unpredictable corruption is the worst. Predictable corruption is a bit like tax. There's a good book by Fosso Jomo at Harvard about this. And then you have taxation kicking at 14%. If you can show them something bigger, and this is clearly big, they will embrace this technology and set new norms. And if you create the same across all these countries with talented individuals in each of those groups and talented individuals in each of those sectors with a shared mission, even though they're separate organizations, that's how you set amazing standards. That's how you build a network effect. And if you tie them all together with an intelligent protocol, when I talk about tokens or speculation or ramps or anything like that, but taking the best of thinking around coordination, that can work, that can break this open, but it's not going to be everywhere. And also when you look at the current debate, the current debate is, for example, we can't let China have this technology and you're like, what about the kids in China? Well, it's dangerous. They can have AGI. So under what circumstance would China ever have this technology? Never. You know, Pakistan, when should they have the technology? Never. That's really what they're kind of saying. It's also self defeating because China has 100 million people they can use to create data sets and two Exaflop supercomputers. Let's put that to the side. Again, it's a very western oriented debate. Whereas actually if you go to these countries and you talk to the leaders and the family offices that have power and the people, they will leapfrog in the global south to intelligence augmentation like they leapfrog to mobile. They want to embrace this technology. And again, you can set norms now versus what's going to happen is they will get a centralized solution. They'll adopt that instead. If you don't right now for hundreds of millions, billions of people. That's why I think again, it's a crossroads.

Is there anybody else working towards this vision that you know of?

No.

In a large AI?

Yeah, no. Certainly no credibility. And again, that's why I had to build these models, you know, and I had to kind of do this. Everyone's working on tiny parts of this, but they're expecting emergence. Build it and somehow it will spread. And again, this is why I find it fascinating the Web3 community, there are good people in there and I hope to be able to unite them. Just like hope to unite the people in health and others. And Peter, you've seen people working on tiny parts of this, but this isn't a Manhattan Project where we're facing an enemy unless the enemy is ourselves, you know. But this does require this big global coordinated push. And that's why I've tried to design this system that I believe will work because it's all about the talent and it is multiplicative.

Is the race against overly powerful centralized AI systems that achieve some version of AGI. Is that what we're racing against?

Yeah, again, we're racing against ourselves. Humans could scale through stories. You have organizations come and join abundance, come and go to Oxford, come and do this. But then when we scaled through text, text was a lossy information format. And there's this poem by Ginsburg. How about this Carthaginian demon of disorder, Moloch that comes in, Moloch comes in through the data loss. Our organizations are slow, dumb aisle. But now what's happening is they're configuring to achieve their thing of Getting more and more power. Again, corporations are technically a people under law, but they're not fully formed people. They eat our hopes and dreams. So I believe the competition here is against those organizations consolidating too much power and creating norms that are almost impossible to break. So we're almost competing against ourselves. And again, the question is this. Do you believe in amplified human intelligence or do you believe in artificial general intelligence? Do you believe in collective intelligence or do you believe in collected intelligence? Who decides is this infrastructure or is this a product like so it's not like a Manhattan Project against the Soviets or anything like that. But this is require us all to come together or at least the smartest people in each of these areas, from coordination to governance systems to healthcare to education, with a blank slate of how do we upgrade the human operating system. The time is now. It's our last chance to do it.

And I love you for it because I think you're right. You were there when Elon beamed in on X video over Starlink and from his airplane, which was a fun moment. And we were talking about the rate of growth and his, his statement, because Rick Kurzweil was there talking about still his prediction of AGI by 2029 and Elon saying we'll have AGI, whatever that means by next year, and the intelligence of the entire human race by 2029. So I am curious just to close out what you think about those timelines and that potential for a super intelligent AI system that is centralized because that's the people who are building that level of power are building centralized systems.

They're building centralized single systems that again take our collective intelligence, like all of YouTube in the case of OpenAI clearly, and other things, and they package it up, sell it back to us, but they don't give it, they don't care. You know, these organizations are trying to build a system that will take away our freedom, liberty and potentially kill us all.

Let's be credit fair about that and sell it to us on an incremental basis.

The selling to us is a complete canard. They don't care about the revenue of this. Again, let's kind of call a spade a spade. They are telling you that they're building something that could kill you and something that could remove all our freedom and liberty. And they're saying it's a good thing, you should back them because it's cool. It's not, it's actually shameful if you think about it. And we should not stand for it anymore. And again, this is Another reason I want to step aside a CEO, because you can't say things like that. I got canceled in Silicon Valley so many times. But realistically, it's ridiculous and it should not be stood for, but they're going to do it anyway because they have the political power, people are scared of them. So there has to be an alternative, and the alternative has to be distributed intelligence. When I resigned, I said, you can't beat centralized intelligence. With centralized intelligence, you're not going to beat it with the stability. So this is a great organization, it's going to do well. The only way that you can beat it, to create the standard that represents humanity is decentralized intelligence. It's collective intelligence. And the data sets and norms from that will be ones that help children, that help people suffering, that reflect our moral, upstanding and the best of us, and gathers the best of us to do it. Because if you work in health care, if you work in education, if you work in finance, if you work in any of these things, there's no organization for you to come and join or partner with on this. There's no kind of centralized mission. I have looked, I've wanted to help other people. I didn't want to do this myself. And I don't want it to be about me very, very quickly, which is why I'm kind of getting it out there now. I hope that I can capitalize something that then people will take forward. And time is now for that. Because AGI, when it comes, if it comes again, there's various definitions of this. Why on earth do you need any knowledge workers? Anything that can be done via a laptop doesn't need humans. And so you have concepts of UBI here. You have concept. I think when AGI comes, you don't need money. Money is a common story as a common good.

We head towards a post capitalist society.

I think the example I think you said was Star Trek versus Mad Max. You know, I'm like, Star Trek versus Star Wars I think is a bit of a better one, you know, and so, you know, you got the Sith Lords and all of that. But again, if you kind of look at this, I don't think we need money. Like it's cross contextual, like bartering with our AI systems representing us. Or it's you don't need money because you're told what to do. Again, our governments, the definition of a government is the entity with a monopoly on political violence. And an AGI can overtake any government that can then control the people. Because again, listen to it whispering. Look at the kind of Hume thing. So we have this opportunity to set norms right now. The way that the big labs are going to AGI is likely to kill us all. Elon and I signed that six month pause letter because even though people like Ahmed, you're an accelerationist, you put all this open source AI up. You have to think about the other side and who's involved in that discussion. And again, if we build an AGI as a centralized thing, is Windows or Linux safer as infrastructure? Our entire Internet infrastructure is built on open. Open can be challenged, open can be augmented. A monolith is like to be crazy. And the way that I put this is you and I both know so many geniuses. You know a side effect of genius is insanity. Honestly, we're not meant. Geniuses are not mentally stable. Why would you expect an AGI to be so? And you're putting all your eggs in one basket versus creating a complex hierarchical system that is a hive mind. That's intelligence, that represents us all. We should be working towards building that because it's safer, it's better, it achieves all the benefits that people are talking about and it's possible today.

Do you think Elon shares in this vision of a decentralized AI? Do you think he would play in that area? And do you think any of the national leaders that you've been speaking to would support that kind of a vision as well?

Yeah, I can't speak for Elon. I'll speak to him and see what he thinks and then I'll get back to you. He always says what he thinks, but you know, he's immensely concerned. He was one of the leaders in this area saying originally why Google, you know, now why Microsoft, OpenAI? It can't be centralized, but it's difficult. This is a difficult question. How many people have a feasible solution? Or have you even thought about this properly? You and I both know, just not many. And that's very sad. Should be everyone thinking about this on the leader side? All the leaders I've met are super happy because again, leaders want power, they want control and all of this, but genuinely they want to see abundance. They're not happy with where their countries are and embracing this technology. They know that they can leap ahead and you know, they will still have a say in all of this. It's not like it's kicking them out or removing them. There are still various kind of mechanisms there. And ultimately improving the health, education and capability of your people is not a bad thing. I mean like obviously I haven't talked to the completely oppressive leaders. You know, maybe that'll be an interesting thing, but honestly, I don't want to even be talking to leaders. I want to create again a system where the people of the country coming together with the franchise system can then build this technology for the good of their people in the open and not be reliant on anyone politically or any other kind of thing like that. Because we don't need giant supercomputers for where we're going. We need coordination. Need a few giant supercomputers. Yeah.

What's your timeline for putting out this white paper?

I'm working as hard as I can.

I've held you to this a number of times. I've said get the vision out there.

It's getting there. We're about to go off this call to a four hour session to dictate all the various bits and pieces. And again, it was impossible. When I was a CEO of stability, there was always another fire, there's always another thing. I didn't have time to think, you know, and I hope people can take that white paper and make it better. I don't have all the answers. I'm just trying to catalyze something, man.

I think after I heard you step down, I wrote you a text saying congratulations.

Yeah, exactly. Not commiserations.

Time to feel unleashed. Yeah, yeah. Imad, thank you, my friend. Thank you for sharing where you are, what led up to this, where you're going next and really pulling the gloves off on discussing the idea of centralized closed AI systems and their dangers and the importance of the vision that you portrayed. Because I'm fully supportive and full believe that what you've laid out is probably one of the most sane visions of AI in the future that I've heard.

I hope other people agree, you know, and they can take it forward. They're the real heroes. Thank you.

Thank you, pal.