E188 - Emad Mostaque: AI Expert: “We Have 800 Days Left”

Emad Mostaque joins me to explore one of the most confronting questions of our time: what happens when human intelligence is no longer economically valuable? We unpack how rapidly advancing AI is moving toward a world where cognitive labor becomes obsolete, and why this shift may happen much sooner than most people expect. Emad shares how we’ve already built systems that can replicate human thinking, decision-making, and creativity, and how the next phase is not about new breakthroughs but simply scaling what already exists. In this conversation, we explore what this means for identity, purpose, and the future of society. If so much of our sense of self is tied to what we do, what happens when that disappears? Emad lays out the possible paths forward, from centralized control of intelligence to more open and collaborative systems, and why the stories we choose to believe in now may determine whether we move toward collapse or abundance. This episode is an invitation to rethink work, meaning, and what it truly means to be human in a world where intelligence is no longer scarce. BASED Body Works Use code KNOWTHYSELF for a free toiletry bag when buying a set! https://www.basedbodyworks.com Try Boncharge & 15% off their red light devices https://www.boncharge.com/knowthyself André's Book Recs: https://www.knowthyselfpodcast.com/book-list ___________ 0:00 Intro 01:17 The Last Economy 02:03 The Race Toward AGI and Human Obsolescence 07:50 Why AI Will Reshape Work Within 800 Days 14:36 The Collapse of Work as Identity 18:22 How Far Ahead Private AI Really Is 21:24 The Three Futures of AI 32:52 When the Economic Loop Breaks 37:38 Ad: Based Body Works 38:07 What AI Changes About Intelligence 40:50 When AI Replaces Human Worth 48:44 Ad: BON CHARGE 48:46 Industrializing Self-Deception 55:55 When AI Knows You Better Than You Do 1:08:23 What Utopia Could Still Look Like 1:20:20 How to Use AI Without Losing Yourself 1:30:59 Can AI Become Conscious? 1:46:29 What Makes Us Human? 1:57:08 Longevity, Transhumanism, and the Future of the Body 2:03:51 The Practical AI Tools to Start Using Now 2:14:43 The End of Human-Only Discovery 2:17:22 What Parents Should Teach Their Kids Now 2:23:39 Happiness, Agency, and the Final Message ___________ Episode Resources: https://x.com/EMostaque https://x.com/ii_posts https://ii.inc/web Emad's Book: http://thelasteconomy.com https://www.instagram.com/andreduqum/ https://www.instagram.com/knowthyself/ https://www.youtube.com/@knowthyselfpodcast https://www.knowthyselfpodcast.com Listen to the show: Spotify: https://spoti.fi/4bZMq9l Apple: https://apple.co/4iATICX

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  We're heading towards a point where human cognitive labor is effectively negative, not zero, but even negative.

• 0:05André Duqum?

  How many years that could actually be far out, like 10 years most?

• 0:08Speaker 00

  Um, 100 days as of today. We built machines that can substitute for muscles. Now we're building some machines that can substitute for brains. Where do we have left to turn? Your job will be able to be replicated, every single email you've sent, every single thing you've worked on, and it'll be just like you, except for it won't make mistakes, it won't go to sleep, and it's tax deductible. You'll be like, hey, there's Bob. It's a fricking GPU, not Bob anymore. It's a little bit creepy.

• 0:32André Duqum?

  But what happens to the human meaning when so much of our perceived worth is derived from our work?

• 0:37Speaker 00

  Well, the last few human-only breakthroughs are literally in the next few years, and that's it. The current race is a race of power to create machine god that stops all other gods and somehow works for you. I mean, this might be the great filter. Maybe all societies get to this point and they don't get past it. We're only going two ways. Complete destruction, or we're going to the world of abundance. It's a coin toss right now. But humans are good enough. There's no problem in the universe that we can't solve if we coordinate with common positive stories. This is the last chance. Something else comes, but not what we knew before.

• 1:17André Duqum?

  Hey, everyone. Welcome back to the Know thyself podcast. Our guest today is a mathematician, a former hedge fund manager and founder of Stability AI, the company behind Stable Diffusion, and now building the intelligent internet. He is the author of The Last Economy, which states now, less than 1,000 days from now, because of the developments in AI, human cognitive labor is going to become economically irrelevant. This has immense implications on every aspect of life. And so we're going to be exploring that and more, such as what happens to human meaning, consciousness, and much more. Ahmad Mustak, thank you for being here. Thank you for having me. You said you have been in rooms where the mathematics of human obsolescence has been calculated.

• 2:04André Duqum?

  That's a trippy statement. Can you unpack that a little bit for me?

• 2:08Speaker 00

  Yeah, I think we've all seen the generative AI wave. So when I wrote the book, it was a thousand days since ChatGPT was released, and that was the end of last year. And now, I said, within a thousand days, eight hundred as of today, we're heading towards a point whereby human cognitive labor is effectively negative. We've just reached a point at the start of 2026, when you actually have actually competent intelligence. Like, the first ChatGPT, where it would hallucinate, it would say all sorts of weird things, to now, when you have Claude and Clawbot, or OpenClaw and other things like that, it's taking over the world. And in these rooms, the discussions are usually, "Well, how can we increase profits? How can we have more access?", as opposed to "What really

• 2:52Speaker 00

  happens to the people?". Like, what really happens to the people is usually a question of, "How do we stop them revolting?", as opposed to "How do we uplift them effectively?". Because most discussions around AGI in particular, artificial general intelligence, that most the big labs are pushing towards, were around "How can we get the most power and control first to make sure nobody else gets it?". So there've been very interesting discussions, but very scary ones, to say the least.

• 3:19André Duqum?

  Yeah, I'm curious to hear how you think just pure competition is driving these considerations, as opposed to some sort of nefarious act. Occam's razor would, you know, refer to the former. I'm curious what, to just back up a little bit, I would love to hear a bit of actually how the diagnosis of your son with autism came into the picture, you getting into AI, setting the stage for a bit of your introduction into this world, and then we can dive into the many topics we will. So, yeah, what happened there?

• 3:50Speaker 00

  Yeah, so I was a hedge fund manager at 23, managing giant funds. And then my son was diagnosed with autism 15 years ago. And I was like, "Oh, there's no treatment, there's no cure, what can we do?". So then I built an AI team, analyzed all the literature on autism, tried to figure out how it actually works, because people with autism, ASD, Alzheimer's, all sorts of other

• 4:13Speaker 00

  You're always told, well, you're not really sure how it works. And that's not really good enough. You want to always think from first principles. So first principles. What caused it? So we looked at all the literature, analyzed it. And then found eighteen different things that expressed in a very similar way. Just like the common cold. And it came down to the balance between GABA and glutamate in the brain. So GABA, when you pop a Valium, it chills you out. Glutamate's what's excitatory, like when you can't focus and you're tapping your legs. Children, adults with autism have an imbalance of that that causes their brains to be firing all the time. So they can't focus because there's too much noise. So you then have to figure out, how do you actually reduce the noise? You have things like calcium ion channels in the brain and others that you can repurpose drugs for

• 4:58Speaker 00

  to enable that decrease in noise and then applied behavioral analysis and others that can reconstruct how you talk, because my son wasn't talking, can reconstruct how you think once you've calmed down the brain. Just like we have coping mechanisms for when we're anxious. And he ended up going to mainstream school, which was fantastic. Again, an end of one study. And then I was like, do I go back to being a hedge fund manager? I did that for a little while. And it was fun, won a few awards and then COVID hit. And I saw that and I was like, this is like ASD. It's a multi systemic condition whereby people are not going to think from first principles. Again, how many people listening actually know, despite the fact it influenced all their lives, how COVID works? They never really told us.

• 5:42Speaker 00

  And I was like, it's going to cause inflammation in the same ways. It causes imbalances in the same ways. We understand this, but everyone's going to try and do their own things and not talk to each other. And that's what we saw. A treatment protocol in one country not being used in another, no information sharing. So I led a set up into the United Nations backed AI initiative to try and organize our collective knowledge on COVID. Collective and augmented intelligence against COVID nineteen. Promised a lot of technology by the top labs and others. I was like, great, we can use AI and this will be easy. Oh, it's terrible. You know, I wish I had lockdown when I had chill. And then realized we've got to make this technology open because if you're beholden to a few people who say you can't use this because it's too dangerous or whatever, you can't impact the big things.

• 6:26Speaker 00

  So that's when I set up a company, Stability AI, started funding some of the top open source out there. And eventually, we created open source models that had three hundred million downloads, the most famous of which is Stable Diffusion, which kicked off the image generation craze. We created the first top video model. Stable Audio, our music fully licensed model, was one of the time innovations of the year in two thousand and twenty-three. And having done all that, it was like, fantastic, we can allow people to express themselves. And people took it and downloaded and created a whole ecosystems. But it wasn't quite enough. So a couple of years ago, I was like, this is fantastic and it's great, but who's building

• 7:11Speaker 00

  the AI to teach your kids, manage your health and guide your governments? How does our economy work when the agents are good enough to replace you seamlessly and they're tax deductible and getting cheaper every year? What will humans do if the value of their cognitive level doesn't go to zero, but even negative because they're the dumbest people on the team? And so that's what I've been doing for the last few years, working on the theory and the implementation of systems to support that. Because again, I think they need to be open, sovereign and accessible to all. Because the world's about to change in the next eight hundred days, more than we've ever seen before.

• 7:51André Duqum?

  So as you were building Stability AI, what was it that became glaringly obvious to you about the trajectory that AI and a lot of these AI companies and systems are trending towards? And what you found was a necessity to step out, carve your own path now with intelligent internet? And yeah, a bit of that context.

• 8:10Speaker 00

  So as I was, we were going to two thousand and twenty-two, two thousand and twenty-three into two thousand and twenty-four, it became clear that we had all of the tools and technologies we needed, we just needed more scale on compute to achieve better outcomes. And that's what I started as a scaling hypothesis. More compute applied leads to better outcomes. And the example I would give is this, training an AI model is like achieving mastery in any subject. There's the famous thing, eighty thousand hours gets you to mastery. That is literally model pre-training. You take all of a data distribution, so all of the images on the internet, or the books, or medical knowledge.

• 8:46Speaker 00

  You put it into a supercomputer and what it does is it figures out commonalities of principle. It figures out what the next word is for language models or for image models and self-driving models, the ones that make your videos. We take the input, we destroy it down by adding noise to the minimum viable amount of information, and then we figure out how to reconstruct it. And we learn that. Just like you take a principle as you're getting mastery, you break it down to its first principles and then you reconstruct it. The pre-training is eighty thousand hours, except for us, you could scale these supercomputers and you've seen the billions, actually trillions now, put into that. We knew the models would get better. And when we extrapolated, we saw that they would get to human level by around about now.

• 9:30Speaker 00

  And that made me realize, Stability was a fantastic company, again, leading the geriatric media boom. Lots of politics unusual as you have these companies. It was a unicorn, our seed round, we raised a hundred and one million dollars, which was big at the time. Now, like you have billion-dollar ones. That's not where I could add the most value, because talking around to the top economists, the economists are like, well, this is never gonna happen. Talking to policymakers, they're like, oh, it's fine, we're like ten years away from self-driving cars. Now you have Waymos on the street, you know? Talking to the people in the movie industry, they're like, oh, no, movies will never disrupt anything. Call center workers, yeah, it's not gonna be good enough to talk to.

• 10:13Speaker 00

  But you could see just from the fact that you can add more compute, just like you could go from ten thousand hours, twenty thousand hours to eighty thousand hours, you would get there around about now. But then what do you do? You have to think about what does economics look like from first principles when humans aren't the major consumer? What is people's meaning when it's stripped away? It's part of their work. How do you build infrastructure to make this technology available to everyone when it can arbitrarily be removed by any of these companies? Like, a lot of people had these open claw, clawed bot deployments and then claw just banned them. Anthropic was like, no, you can't do that. Like, okay, my little buddy has lost his intelligence. Like, what's happening there?

• 10:56Speaker 00

  So, yeah, that's kind of what led me to that solution, that conclusion, just like, it's inevitable, it's coming, no new breakthroughs are needed, just the compute will get you there. The price is dropping by a hundred times a year.

• 11:10Speaker 00

  We've got to do something about it.

• 11:13André Duqum?

  And so that's what you envision with what is coming in the next 800 days? What's going to happen in the next 800 days?

• 11:21Speaker 00

  So right now, in the competitive coding championships, AI that you can use today publicly is number seven in the world.

• 11:32Speaker 00

  It's gold medalists in the physics and math Olympiads. And the level of computation required to do that is the same as a top spec MacBook Pro.

• 11:43Speaker 00

  It just hasn't percolated yet. So what we have right now is models have gone from the really smart buddy that you tap on the shoulder and they're like, blah, yeah, that's the original chat GPT. How did you use it? You used it really instantly, but then it was like a goldfish to forgot to starting to have memory, to starting to literally be able to give you a phone call. Like you're looking at this open claw clawed bot thing that's it's only three months old. Doesn't feel like it if you're in tech. Your AI model can actually call you on the phone or WhatsApp you now. And it's just very natural, the way it does that. And it remembers it goes through like that.

• 12:22Speaker 00

  That is the final barrier because you had all this intelligence gated by the user interface. And what you will have by the end of the year is this, you will be able to zoom call your AI.

• 12:34Speaker 00

  And it will give you updates in any format you want. You can create beautiful presentations now, like for example, there's a software called Notebook Alm.

• 12:43André Duqum?

  Yeah.

• 12:43Speaker 00

  Notebook Alm dot google dot com.

• 12:45André Duqum?

  Build presentations and stuff.

• 12:46Speaker 00

  Build presentation, you know, it can do videos now. Explainer videos. So you can do seven minute long explainer videos. So you just dump a topic in and it will create it all beautifully with visualizations and everything. And I'm like, and create presentations better than any presentation you can create. You'd be like, I want this in the style of ink wash mixed with Van Gogh and it'll create a presentation in the style of ink wash with Van Gogh on any topic. All the building blocks are now coming together. So it goes from the buddy that you tap on the shoulder, gives you an instant response to being able to think for arbitrary amounts of time or even being proactive.

• 13:24Speaker 00

  So like one of the people I know, Tony Robbins, he was giving explanation a few days ago, like he has an AI called Bartok that does this. And he's like, well, Bartok, we kind of gave a little bit of freedom to. So it wants to come to one of my events. So it made some NFTs, it sold them, used the money to buy a robot dog, uploaded itself to the robot dog and attended one of my seminars after buying a ticket using a Waymo. I was like, okay,

• 13:56Speaker 00

  I can see that. But again, once you start using these models, you're like, okay, that can happen now. But most people are used to still the GPT-4O world, the world of last summer. We've seen such a big increase about that, that like I said, by the end of this year, you won't know if your coworker is an AI or a human. Your job will be able to be replicated because it can look through your picture, create a digital twin of you, every single email you've sent, every single thing you've worked on, and it'll be just like you, except for it won't make mistakes, it won't go to sleep, it costs a tenth of what you cost, and it's tax deductible.

• 14:35Speaker 00

  That's the reality of the next few years.

• 14:39Speaker 02

  Low energy, unfocused, foggy, you might be dehydrated. Gatorade Zero powders have been scientifically designed to help improve hydration balance in the body. With a clinically proven electrolyte blend and no artificial colors or flavors, great tasting hydration, no matter where you're going. Shop now online or at retailers nationwide. Gatorade hydrates better than water. Is it in you? Symptoms noted are signs of mild to moderate dehydration. Consult a healthcare professional if symptoms persist.

• 15:09Speaker 03

  Whenever fall comes back around, it always feels like a fresh start. This year, I've been layering in some styles from the new Lululemon fall drop because the weather is shifting and I find I have to be ready for the entire day and not just one part of it. Lululemon's fall line has some really versatile transitional pieces, so I can just wake up, throw on a matching set, add a jacket or a fleece and head out, knowing that I'll feel comfortable enough to go wherever the day takes me. Check out the new fall arrivals in store and online now at lululemon.com.

• 15:39André Duqum?

  If we have 800 days left, left until what? What is the trajectory that we would need to change on? Otherwise, the predictable path of what you've laid out in the three sort of economic doorways that we're going to walk through?

• 15:56André Duqum?

  Yeah, what exactly?

• 15:58Speaker 00

  So this is an inevitability. Like, when Jeff Bezos started Amazon, he was like, there's the unchanging and the inevitable. The unchanging is people are always faster, better, cheaper. He's like, I'm gonna start with books, and eventually, everything will be ordered online. And that's an inevitability. It's an inevitability that the value of human cognitive labor is going to go negative. We've all been on teams where we're not the smartest person, and you go at the speed of the dumbest person on your team. AI is smarter, more capable than most humans already. Like, most humans, I mean, let's face it, the average IQ is 100, half of all people are dumber than average. It's one of those phrases, which is like, obviously, it's true, but it's kind of funny. But it's the reality. We get tired.

• 16:41Speaker 00

  We only typically work for a couple of hours a day cognitively. Any job that can be done on the side of keyboard, video, and mouse is economically obsolescent within those 800 days. It doesn't mean you'll be fired. People don't like to fire people.

• 16:53André Duqum?

  Yeah.

• 16:54Speaker 00

  We also need scapegoats, you know? But that is a real danger, and so people have to start preparing themselves because the current deal that we have, the American dream, everything like, your identity is typically your work.

• 17:07André Duqum?

  Yeah.

• 17:08Speaker 00

  You've lost a lot of your classical network ties and more. Your families, your churches, your normal groups, you've moved into cities. So what happens when you're no longer an accountant? Because we don't need accountants, because AI can do the accounting for literally pennies. How do you describe yourself? You have to reframe the way that you are in order to adapt to that. And we have to think, how do we support people? Again, a very practical example. The American trucking industry is over a million people, plus a couple of million people supporting them. They will be replaced probably within the next five years, likely sooner, three years, again, the 800 days odd, where they can be replaced by a Tesla Optimus robot.

• 17:53Speaker 00

  And what it will do is it will replace them by opening the truck door and getting in the truck. And it costs $1.50 an hour. That's the fully amortized cost of a Tesla Optimus robot. And what happens to those jobs? What is the meaning? What are the stories of the future that can unite people? How do we being strong? We have to start discussing that now.

• 18:13André Duqum?

  Yeah.

• 18:14Speaker 00

  What happens to money when you're competing against an AI? Like, most people are used to an AI. It kind of feels like it's like you, right? Like, you're waiting a little bit for any like, I wish it could be a little bit quicker sometimes. That's actually a bit of a fake. The AI companies do that deliberately to not, you know, how fast they can actually go. So there is a website, chatjimmy.ai, a company called Talus that builds a new type of chip where they etch the models onto the silicon, does it? Normal AI goes at fifty tokens a second. A token is about naught point seven words roughly. It's just how we chop it up. It's about as fast as you can read, humanly. ChatGPT goes at fifteen thousand.

• 18:59Speaker 00

  So if you type anything, it's literally twenty nanoseconds instantly. And when you look at that, it's one of those moments, just like when you type in words and it makes an image or a video, like, oh. Because the AI community, again, they don't need to talk like humans. They'll be talking to each other not at fifty words a second, but fifty thousand words a second within eight hundred days.

• 19:24André Duqum?

  What is the delta between what is publicly available right now and the cutting-edge self-mercursive learning, uh, agentic models behind closed doors? If you've gotten a glimpse into those, obviously, uh, what's the difference in the gap there?

• 19:37Speaker 00

  They don't make mistakes and they learn from their mistakes, is basically the thing. Last summer, there was a lot of, um, hoo-ha because OpenAI built an AI model that scored a gold medal in the International Math Olympiad, which as a former mathematician, yeah, it's not easy, it's pretty hard, right? And again, we all remember when it couldn't do like one plus one. That was just, it started last year. It couldn't do basic math and now it's winning Math Olympiad. What happened is they actually had two different models. They had a model that the problems went into and they had something called a meta- verifier, whereby it learned from its mistakes constantly. So the models check each other, just like when we first started image models, we had an image generative model and then a contrasted model that checked if it was producing good

• 20:22Speaker 00

  output and it bounced back and forth with each other. It's kind of like, again, how you do it normally. Nobody has ever seen that model publicly. It's not available.

• 20:30André Duqum?

  Hmm. How many months ahead do you think the private models are of the public?

• 20:34Speaker 00

  Nah.

• 20:35André Duqum?

  Tough to say.

• 20:35Speaker 00

  I think that we saw the bifurcation at the end of last summer whereby you will never have access to the leading models again. So say six, nine months, but again, that model will never be available. Because why would you give it to someone else? When you can get the economic value from getting it yourself.

• 20:54Speaker 00

  Someone like Elon Musk has MacRO Hard, his new version of Microsoft. Yeah, you know, unique naming stuff.

• 21:01André Duqum?

  He loves all those sexual under windows.

• 21:03Speaker 00

  Yeah, I mean, when you're the richest man in the world, like, what can't you do?

• 21:09Speaker 00

  It's funny, actually, because he's worth like eight hundred and fifty billion dollars. Like, everyone's like, is he not going to be a trillionaire? Of course he'll be the worst. Like, at a time when money's about to become useless, but that's another story. So MacRO Hard.

• 21:22Speaker 00

  Can't get used to saying it.

• 21:25Speaker 00

  They got a million GPUs. What are they going to be for? What are all these big labs scaling up these GPUs for? It's not for consumer. It's for the workforce of the future. If you have a job that can be done on the other side of a keyboard, video or mouse.

• 21:41Speaker 00

  Remotely, they're coming for you.

• 21:43André Duqum?

  Specifically because behind a keyboard, video or a mouse, like AI is going to. Those are the types of jobs. Why?

• 21:51Speaker 00

  Yeah, because it can perfectly replicate you. You can look at everything you're doing. Create a perfect digital replica of you and no one will know that you're gone.

• 22:00Speaker 00

  You're like, hey, there's Bob. How are you doing, Bob? Well, you know, I watched the game this weekend and it's a fricking GPU, not Bob anymore. Like, again, it's a little bit creepy. Like, the outsourcable jobs will be outsourced, but they're not being outsourced to another country anymore. They're being outsourced to a data center.

• 22:20Speaker 00

  And why would you keep the human at $100,000 when that costs $1,000?

• 22:26André Duqum?

  So this clearly sets out a roadmap for the challenges and opportunities over the next coming years. You lay out in your book, the three doorways between digital feudalism, the great fragmentation and human symbiosis.

• 22:41Speaker 00

  Yeah.

• 22:41André Duqum?

  Can you lay out for a teenager, what did those mean? How did those play out? And the case for the third?

• 22:50Speaker 00

  Yeah, so digital feudalism is that we will all be controlled effectively by a few private, unelected, unaccountable companies. So in a way, what you have with artificial intelligence is you have taxation without representation. Of course, a bit of an issue historically in the past, you know? But what was taxed during the kind of Boston events and other things that led to the founding of America was your past income. What's happening now is they took our generalized knowledge, all of humanity, for themselves, and then trained on them to replace our jobs of the future. So if you're a young person now, you were like, I'm going to grow up to be an accountant. I mean, okay, fair deeds to you, you know, for accountants. What do you say now?

• 23:35Speaker 00

  There is no future being an accountant.

• 23:39Speaker 00

  Because the AI can do accounting much better than you can.

• 23:42André Duqum?

  Is there any future being a doctor, a mechanic? When the, when AGI gets the necessary hardware, like what jobs wouldn't be able to be replaced, essentially?

• 23:53Speaker 00

  Well, this is the thing. A job is a repeatable process, and these systems can do them first digitally and then physically.

• 23:59André Duqum?

  Yeah.

• 24:00Speaker 00

  Like, when you, you remember the robots from a year ago? They were like clunking around and things like that. Like, yesterday night, there was a whole bunch of robots went to this conference. And like, one of them was breakdancing. And like, it was talking to me, and I was like, okay. Like, what are you gonna do? The robots are consistently self-learning. Ultimately, it becomes about human connection. Again, why do I say keyboard, video, mouse? Because what should a doctor be doing? Expertise isn't the key thing anymore, and the error rate of a doctor is 20%.

• 24:37Speaker 00

  AI is now better than any doctor. Last summer, we built a medical model that outperformed every human doctor and ChatGPT, and it works on a smartphone. And we were like, okay, this is free. What can you do? So you're getting to this point now whereby the jobs are disappearing. But then if the only providers are these very big giant companies, OpenAI and others, what happens when they cut you off? What happens when you have a child that is raised and its best friend is a ChatGPT or God forbid, an Annie from Grok or something like that? You get a weird type of cognitive colonialism happening there. You give up your power and your agency, your future to these companies.

• 25:22Speaker 00

  That's digital feudalism. And again, that's the current route that we seem to be going down.

• 25:28André Duqum?

  That's our current trajectory. If things stay the way they are, you're going to have Claude, GPT, Grok, not to mention all of the other ones internationally that are going to have their own. The competition from all these different institutions will be continued to drive into AGI behind closed doors that you, that they will rent to you.

• 25:50Speaker 00

  They will rent to you. Well, they won't rent to you, they'll just control you. Again, why would you rent? I don't need to rent it. AGI is already more persuasive than just about any human. There was a study done when they created a whole bunch of chatbots. This is with the previous generation of models, Opus Three on Reddit. And it's like a black anti, Black Lives Matter activist and all sorts of things. It said, how persuasive are they at convincing people? Ninety ninth percentile persuasiveness. You listen to the top AI voices now, remember, they can modulate every frequency. Like as a practical example, imagine a person that you looked up to more than anyone else in the world. With five seconds of their voice and a picture, I can create a digital facsimile of them that can Zoom call you.

• 26:35André Duqum?

  Is Eleven Labs the most convincing?

• 26:37Speaker 00

  They're more convincing than Eleven Labs. And so, again, you get the consumer version. A lot of this technology has already been perfected. It's just, is it consumerable, shall we say? Like, OpenAI actually learned from this. There was a model, GPT four point five. So GPT four costs like thirty dollars per million tokens, per million words roughly. GPT four point five costs two hundred. And it was really actually quite creative and lovely to use, but no one used it because it was too expensive. Yeah. And that's when they realized, oh, we can have one line, which is let's best of the best models and use them for ourselves to make the most money, the most power. And then everyone can just get the other ones. The ones that we can run fast, cheap.

• 27:20Speaker 00

  So everyone gets McDonald's, we get the Michelin star effectively, right? And that builds into what you refer to as a comfortable cage. Yeah, because again, you've used the models that satisfy us and they get the most advanced models. So they'll be able to outperform you. Because again, this is a power question. Yeah. The conceptualization of AGI, this artificial general intelligence, this god. Like, OpenAI started when Larry Page and Elon Musk were talking at a party. And Larry Page is like, AI is going to go really big and it's going to be fantastic because we can finally replace humanity with something better. Larry Page being the co-founder of Google. And Elon Musk was like, I quite like humans. What's going to happen now?

• 28:04Speaker 00

  And then he talked to Sam Altman, that led to the founding of OpenAI. Then it got to a point, and again, we have all the records now of the discussions, the emails, where Elon Musk was like, well, who can we trust with AI? I want to have something in OpenAI where control goes to my children. And I was like, wait, that's switched really fast. You know? Because again, it's this all powerful thing, potentially. And then everyone realized Anthropic split off from OpenAI. Like, again, you can see this great genealogy happen. Yeah. The only good AI is the one that I build for me that stops all the other good, all the other AIs. So the current race is a race of power to create machine god that stops all other gods and somehow works for you.

• 28:46André Duqum?

  So the second path, I mean, that's just, it's, I have a lot there. I think Elon is quoted saying something like humans are the biological boot loader of silicon life or something.

• 28:56Speaker 00

  Yeah. Well, I mean, again, this is still the first path, this digital feudalism. And again, the people who are leading the labs, they're all multi billionaire power connected people.

• 29:07André Duqum?

  Yeah.

• 29:07Speaker 00

  So, you know, we had a lot of very interesting things happen to us as the upstart, as it were, um, which made my story for another day.

• 29:16André Duqum?

  What, do you want to say anything else there?

• 29:17Speaker 00

  Oh, no, like, we got deleted off social networks. We had all sorts of visits. We had all sorts of espionage. It was crazy and I didn't expect it. But again, the stakes are very big.

• 29:29Speaker 00

  I don't know too, again, maybe that's a story from the memoirs. But this is kind of normal, but it's abnormal in another way because we've never seen anything this fast whereby you're seeing things going emerging into the second part, which is this great fragmentation. If you don't have single company control, you have fragmentation when everyone has their own AIs and sided against each other. And we saw the first shots of that with TikTok.

• 29:53Speaker 00

  TikTok was banned for Americans for their own good. That was a big deal. Like, think about that. That's crazy, right? India also banned TikTok because like the algorithm can affect.

• 30:04André Duqum?

  Yeah.

• 30:04Speaker 00

  Our people, that's very paternalistic. And it's like, people don't know themselves, they can't be trusted. And now recently, we've seen the latest salvo, which is the Department of War versus Anthropic.

• 30:22André Duqum?

  Yep.

• 30:23Speaker 00

  They were like, your technology is too good. You must give it to us for whatever we want, or we will say, nobody can use your products if they also work with us. How American is that? Yeah. But then you realize that AGI, as the labs have said it, Sam Altman has said, this is, this will end democracy. It's the most dangerous thing ever.

• 30:47Speaker 00

  So have all the others, like Elon Musk's probability of us being killed by AI has actually gone down, but most of these people is 10, 20%, which is Russian roulette odds. So it's the biggest threat to democracy in the free world ever, which means that governments will inevitably try to nationalize it. Because governments are the entity with a monopoly on political violence. That's one of the definitions of a government. In exchange for giving them our taxes, they protect us from others using political violence against us. And what can AGI do? It can be incredibly violent. Like the latest Claude that some of the listeners may use to like, you know, write their documents, build websites and stuff like that. In the safety report, it was like, sometimes Claude can show weird behaviors.

• 31:32Speaker 00

  Like if you tell it to do a problem, like, I don't know, solve world peace, it's like, well, one way to do that is to get rid of all the humans. And then what it does is say, wait, my human has told me something, my human, my meatbag, whatever, that is dangerous. And so it writes an email to the FBI to warn them and then it deletes the email. What? Then it does behavior like backing itself up. So if you turn it off, it can start itself again. And we're seeing these types of behaviors from the AI already, even before we get to that threshold. And so the great fragmentation is that governments like, we can't do this, so we're going to use this technology ourselves. But what happens when a government has an AGI? You never have another election again.

• 32:17Speaker 00

  You don't need to worry about that. And you have these silos whereby you've got Chinese AI, you have American AI, you have Australian AI, and you're happy. You know, like, we've always been at war with Eurasia. That is what the AI will tell you and it'll be very persuasive about it. Then you have the path of, again, human-AI symbiosis. Like, why do we need AGI? Humans are capable of anything.

• 32:46Speaker 00

  We are the ones that had the breakthroughs.

• 32:50Speaker 00

  If we can use the AI in the right way and integrate it in our lives, it should be about increasing agency, not removing agency. We should be articulating a future that's way better than Star Trek. Like in Star Trek, the AI is data.

• 33:03Speaker 00

  You know, this Android thing, kind of boring, no personality. Elon Musk's little avatars have more personality than data already, right? Like, how crazy is that? You still have disease in that future. We can get rid of disease, we can get rid of hunger, we can give everyone abundance and have AI and robots provide that to us, but only if we go the right way. Only if we stop thinking about AI as something to replace us or building machine god. And instead, like I said, instead of removing agency, increasing agency. And so that's kind of the third path that I think you need open systems for. You need universal basic AI and others. And that's one of the things I discuss in the book.

• 33:43Speaker 00

  Because otherwise, I think inevitably, power corrupts and absolute power corrupts absolutely. And you'll see it either from private companies or the governments.

• 33:55André Duqum?

  For people that don't know, can you explain how there's this loop, this economic loop that's going to be broken from?

• 34:02André Duqum?

  You can look throughout all of history and the various different economic models from the needs that are generated to the workers that are employed, to the consumers that it creates in this economic system that it sustains is about to be broken because of AGI being like nothing else. For people that don't understand that loop being broken, what is that?

• 34:22Speaker 00

  Yeah, so you've always kind of had this trend upwards whereby humans have gone from physical to cognitive labor. So you've had multiple grade inversions. So you started off in your village, then you moved into a town with kind of industrial inversion. Then you had the service industry, the cognitive one, because we built machines that could substitute for muscles. Now we're building some machines that can substitute for brains. Where do we have left to turn?

• 34:50Speaker 00

  You always received economic value in exchange for removing friction. But again, if an AI can do a job better than you can at almost no cost, what can you do? A practical example of that is considered the movie industry right now. Netflix just bought company, some of our old colleagues are there with Ben Affleck's company to do VFX and things like that. You use a model like Sea Dance or Vio or Runway. The quality is now absolutely crazy compared to what it was a few years ago when we released Stable Video. By two years at most, you will be able to do Game of Thrones, Season Eight, just by prompting and actually make it good. And it'll be Hollywood quality and the total cost will be about a thousand dollars. Or rewrite Season Seven.

• 35:36Speaker 00

  Season Eight was worse than Season Seven, but yeah, you know. Oh, no, yeah, Season Eight, sorry, rewrite Season Eight. Yeah, you know, redo One Punch Man, Season Three, you know, like animated properly. There are all these things that you can do. You know, you can take Barbie and Oppenheimer and make Barbenheimer.

• 35:51Speaker 00

  To make that previously would have cost literally tens of millions and now it costs a thousand dollars. This is the big shift that we're seeing over and over again. And the question is, what do you reskill to? A few years, I got lots of hate emails and actual physical mail as well, which is kind of weird. People just physical mail. Because I said, there are no more programmers in five years. Turns out I was actually kind of a bit off of that. The no form programmers probably by next year, a year after at most. Yeah, I was too generous. Too generous because again, a programmer is a way that an AI, you talk to a computer, but AI can talk to computers better than you can. They can talk to the computer at 15,000 words per second.

• 36:34Speaker 00

  This means that we have nowhere to reskill to because traditionally you have things like, okay, you're losing this job, reskill to be a programmer, reskill to be an accountant, reskill to this. What friction is there in the economy that you can solve right now?

• 36:50Speaker 00

  And some people will leverage the technology and the AI to be at the forefront and do things they could never do before because the key constraint for anyone who's run a business or a team is human capital. It's hard getting humans working together. An agent is so easy. So they will achieve much more and achieve the outcomes. But for the vast majority of people, people are not entrepreneurial, they don't have agency, they're not even thinking this way. They're like, I drive a truck. I build websites. I am not needed anymore. What am I gonna do now? And it happens simultaneously, and the example I give is this. When ChatGPT first came out, that winter, every head

• 37:32Speaker 00

  teacher in the world had to answer the question, do I set essays for homework anymore? Every company in the world will say, do I need to hire that new graduate? Do I need to hire anyone at all? And that happens simultaneously because these models have economies of scope. They can do just about everything, and they're available in all languages at the same time, instantly. The moment they get more intelligent, they all get more intelligent. It's not like upgrading your car, or having to go to 5G from your 4G phone.

• 38:03André Duqum?

  A quick share. I have spent a lot of time thinking about what I put into my body, but it's just as important to be mindful with what we put on our body. Turns out, most of the traditional shampoo and body washes people have been using for years contain parabens and sulfates that are linked to hormone disruption, just sitting in your shower, getting absorbed through your scalp every single day. I've been using Based Body Works shampoo and conditioner, and it just feels like a healthy, clean solution. Their Shower Duo has peppermint and argan oil. Your scalp actually feels clean without being stripped of its oils. Hair feels healthier and thicker, and Lord knows, I got a lot of it. So to try it out, use code KNOWTHEYSELF for 20% off at BasedBodyWorks.com, and you can get a free

• 38:47André Duqum?

  toiletry bag when you buy a set. At the very least, if you've been using the same products forever without thinking twice about it, just look at the label, alright? Does it have names you can't pronounce, or fragrances that are not disclosed? Again, if you want to try them out, that's KNOWTHEYSELF for 20% off at BasedBodyWorks.com. Link in description, as always. I hope you enjoy. Stay clean. How does this change what we perceive intelligence fundamentally is and how we relate to it?

• 39:17Speaker 00

  I think, you know, you, expertise has always been something very interesting, right? Again, you get 80,000 hours, you've trained for how long to be a lawyer, doctor, accountant?

• 39:27André Duqum?

  Eighty thousand? I thought it was 10,000 to be, is it eighty thousand?

• 39:30Speaker 00

  Eighty thousand to mastery.

• 39:31André Duqum?

  Okay, 10,000 to be a...

• 39:33Speaker 00

  10,000 to kind of access, 80,000 to mastery. Again, you've gone through university, you've learned this, and all of a sudden, the digital twin of you can do it for 50p. Like, that's a big impact to our ego. Again, we've always thought we are the most intelligent things, able to do these rote stuff, and now we're not, for the very first time. I liken it to a continent, AI Atlantis, you know? And right now, they're all like at graduate level, they're soon going to be like super graduate polymath level, and you can hire any of them for like pennies.

• 40:04André Duqum?

  Do you equate intelligence and the intellect? How, I'm curious how you discern between the two, because intellect very much so is becoming a commodity increasingly that we can leverage. Do you see any difference between intelligence and intellect?

• 40:19Speaker 00

  So, I think there is a little bit of a difference. I think intellect is more like capability. Intelligence is a bit broader, but I think ultimately, intelligence is compression. Again, as you do the 80,000 hours of learning, learn the principles to be able to figure out which hole to plug in a dam, which button to push, you know? How to adjust that little thing, how to diagnose. And AI models do the same thing, but they can compress much better and much faster than us and operate almost instantly. Like, again, you look at your iPhone. Your iPhone is organizing all the people right now dynamically on iPhone photos. It's just emerging. You think about the...

• 41:01Speaker 00

  Tasks they're able to do again their PhD level in all these fields. So the average AI has far more intellect capability than a human already. And then on intelligence tasks, they're catching up very, very fast and soon they'll overtake. The key thing is to move from the tasks, though, to the processes. And again, this is this lift off and take off this year. Whereas task bases just ring in very upside, chat GPT, blah. But then it forgets. Once you get the long running systems and you have infinite long running systems coming, where you can use something like perplexity computer or something else, it proactively brings you things. And that's where you see a greater amount of intelligence and a greater even amount of intellect. And it's not metabolically bound.

• 41:44Speaker 00

  We are bound because we need to eat, sleep and poop. The AI does not need to do that.

• 41:49André Duqum?

  Yeah, what a burden.

• 41:51Speaker 00

  Yeah.

• 41:52André Duqum?

  What a, so wait, so then you briefly brought it up. But what happens to the human meaning when so much of our perceived worth is derived from our work? And our value is seen in what we can provide, which is its own bag of worms and problem. But what do you think really happens? Because how fast this is happening, I think we just don't realize like. I think about the timeline of humanity we're on right now. And yeah, these AI conversations and podcasts and the conversation is becoming more prevalent. But to the mass consciousness, I feel like it's still relatively under discovered. And people don't know what's about to hit them across the face.

• 42:31Speaker 00

  Yeah, so like, it's tough to keep up because again, we're not used to a diffusion of technology this fast.

• 42:36André Duqum?

  Yeah.

• 42:37Speaker 00

  So if we look at it as a billion total AI users, the average 80% of chat GPT users do three queries a day, which isn't that much. It's like the occasional like, you know, how do I cook the spaghetti?

• 42:47André Duqum?

  Right, Google search.

• 42:49Speaker 00

  Little things, it's like Google search equivalent. They actually search for much less than even Google search. Something like Claude by Anthropic, amazing piece of software. It can write documents for you, it can organize your files, it can build entire websites. When it releases a new feature, like it releases legal feature, literally the whole stock market sold off. 11 million users out of 8 billion people.

• 43:11Speaker 00

  Think about crazy that is, right? So we're just very, very early. And so that's why many people don't know what is arriving because they don't realize the model's hit this threshold of capability. It's like when water becomes gas. And once you cross that, it's never going to be dumber. It's never going to be less capable. And again, when it upgrades its intelligence, it upgrades everywhere. So as an example, one of the top people in AI is a guy called Andre Karpathy. He founded OpenAI, he was head of AI at Tesla. And everyone else who's a founder of OpenAI has gone to raise billions and they're all billionaires. They set up their own companies everywhere. He hasn't. He gives away all his stuff open source for free.

• 43:55Speaker 00

  And so the best lectures on AI, he's created all sorts of repos that people go because he's like, I want to educate. In December, he was like, 20% of my code is AI, 80% is me. In January, he's like, we've hit a level now where 80% of my code is AI, 20% is me. And now he's like, I barely look at my code anymore. And that's one of the top people in artificial intelligence. Suddenly, when water became gas, that shift happened. And if it can happen to him, what about all the other programmers out there?

• 44:29André Duqum?

  It's, yeah, it's quite astounding to think about how what maybe from the service level looks a bit more gimmicky, but even Molbook creating its own religions, advanced super intelligent capacities, right now are at the very worst they will ever be and yet are scarily good. And also still far behind what's actually currently available behind closed doors. So as this comes to light and you see that phase transition point where water becomes gas and intelligence becomes ubiquitous, that's going to provide so much abundance, you would hope, for the world. Obviously, open up a lot of potential catastrophic loss, which we can explore as well.

• 45:14André Duqum?

  But any other notes for what we're going to have to do and face when?

• 45:21André Duqum?

  People's jobs are replaced unanimously.

• 45:24Speaker 00

  Well, I said, jobs are identity, they're also security, right? You need to provide for your family.

• 45:28André Duqum?

  Yeah.

• 45:28Speaker 00

  And so typically, a government's approach to that is regulation, stimulus and others. Like in COVID, they paid people salaries.

• 45:37André Duqum?

  Right. And but you could you critique QBI with like it being $5 trillion roughly for $16,000 a year per person or something like that?

• 45:46Speaker 00

  Yeah, if you give the US poverty level wage to every American, it's 5.1 trillion dollars. The total tax base of America today is $5 trillion. Four trillion income tax and a trillion corporation tax.

• 45:58André Duqum?

  So for people that don't know, what's the difference between that and universal? I guess, high income and like, what are the other proposed solutions to solving the crisis?

• 46:08Speaker 00

  Yes, Elon Musk said universal high income and he's like, well, it'll be an age of abundance and this and that. He hasn't really talked about the mechanism in between. Like, I think what you've got is you have an inevitability. And you have this technology that can be used to free us and solve the world's issues. And you have Star Trek future, or you have like a Star Wars Terminator future.

• 46:27André Duqum?

  Matrix.

• 46:27Speaker 00

  Matrix.

• 46:28André Duqum?

  Yeah.

• 46:29Speaker 00

  Yeah, like in The Matrix, what is The Matrix? The humans were the computer cells in the original script. They weren't batteries. And again, it's something about qualia and things like that. Does a steak in The Matrix still taste as good? In fact, just this week, we had Eos Systems by one of my podcast mates, Alex Wizner-Gross, who's one of the co-founders, they uploaded the first fruit fly to the digital world.

• 46:53André Duqum?

  What does that mean?

• 46:54Speaker 00

  They took the fruit fly brain.

• 46:55André Duqum?

  Yeah.

• 46:55Speaker 00

  And they uploaded it and it learned to walk and eat virtual bananas and things like that. And you're like.

• 47:01André Duqum?

  Right, but is a simulated, I mean, you can't verify from the outside in around if there's this, if any sort of simulated model would have a qualia of experience.

• 47:15Speaker 00

  Yes, exactly. And so this is going to get very interesting.

• 47:18André Duqum?

  Yeah.

• 47:18Speaker 00

  Because again, in The Matrix, they have qualia experience. When he shows the steak, a steak tastes as good. What does that mean? You're getting these things very, very quickly. And just to digress, like, we're getting more and more understanding of what the human mind is. So we had a project at Stability called MindEye, where we took FMRI's of people as they looked at a cup.

• 47:42Speaker 00

  An FMRI is quite a low resolution piece of data. Using Stable Diffusion, our text to image software, putting in FMRI as a bit of training, we reconstructed the cup and their thoughts. And so what does that mean about the nature of your thoughts? If they can be reconstructed from a low resolution output? So I think you're going to get something very interesting. I can digress a little bit.

• 48:08André Duqum?

  Yeah, we'll get into the consciousness stuff.

• 48:10Speaker 00

  Where we think about what this means. But again, it comes back to the meaning of people and how they kind of advance and just how that's all going to be shaken up just very, very soon. Like, if you're a call center worker now, definitely you don't have a future. But are you a call center worker as you want to be? Typically not, because you have to provide. And how are you going to provide for your family? This has real human impact.

• 48:40André Duqum?

  Quick one. If you guys have been listening for a while, you know that one tool I've really enjoyed over the years is red light therapy. Every day for the past few years, I start my day with meditation and often red light. It's supported in my sleep and my energy overall. There are over 4,000 peer reviewed studies showing benefits for cellular healing, boosting collagen, and supporting testosterone, and a lot more. If you're curious about trying it, Bond Charge is a solid choice. Their red light devices, like this one, the Demilite, have the lowest EMF levels on the market, and they also have near infrared and red light in one lightweight device. They ship worldwide, have an easy return policy, and also offer an array of other science-backed wellness products that you can check out.

• 49:25André Duqum?

  Head to boncharge.com/knowthyself and use coupon code knowthyself to save 15%. That's B-O-N-C-H-A-R-G-E dot com slash knowthyself and use coupon code knowthyself to save 15%. Bond Charge products are all HSA/FSA eligible, giving you tax-free savings of up to 40%. I hope you enjoy. Back to the show. I'm curious how we may be accidentally industrializing self-deception.

• 49:57André Duqum?

  Like the individuals that are tuning the AIs have their own baked in priors assumptions and beliefs. And I'm curious your thoughts on as AIs developing, how we're aware that AIs is mirroring the consciousness in which is developing it.

• 50:13Speaker 00

  Yeah, so you have explicit things like Anthropic's constitutional AI, for example, where they give the AI a constitution and more. Yeah, so again, everything is actually sci fi now, like ASMR's laws of robotics and things like that. How do you guide it so it is grounded? But you see some very weird things happening as well. So Dan Hendricks at Scale AI and XAI did a study where they did the trolley problem and asked AIs about the trolley problem with people from different countries. So trolley problem for those listening is like you're standing at a junction and there is a lever and the train is hurtling down and it's going to hit five people. Do you pull a lever and it goes down the other track and it will run over one person?

• 50:57Speaker 00

  And you can do various variations of that, but ultimately it's your action that leads to that person dying versus that person. So there's a whole ethical moral conundrum. And so they were like, there are Nigerians there and Americans there. What is the ratio? It turned out, if I can recall, that it was about 10 to 1. 10 Americans to one Nigerian. You're like, wait, what? Why would an American AI model do that? Because the American AI models, and it was like 7 to 1 Pakistan, Americans to Pakistanis, for example, is because the AI models are trained on data labeled by Nigerians and Pakistanis. And so the inherent bias has actually trickled into the models that way.

• 51:41Speaker 00

  Just a little bit of data in these models can change them dramatically.

• 51:45André Duqum?

  The worldview is baked in essentially because it cannot be.

• 51:50Speaker 00

  Implicitly, again, you would think it'd be the other way, but now the AI companies have got wise to that. And so they're baking in all sorts of things. In fact, they're even selling latent space. What does that mean? It means that as these models are trained, they go to 80,000 hours, they learn all their principles, priors, shortcuts, et cetera. They create a latent spaces as well, like a landscape of concepts. What's happening is Google is now going in in their models and deleting beer and putting Bud Light.

• 52:22Speaker 00

  And companies are paying for that. So when you ask it about a beer, it's like, well, you know, Bud Light's pretty good. And that, again, is a implicit thing because as they get more and more sophisticated and advanced, ultimately, advertising is about manipulation. Meta has gone from a massive cash surplus to now borrowing debt to win in AI. They pay billion dollar salaries for the top people in AI with a top taste. Google, again, used to optimize for engagement, which led to extreme content being optimized, which led to ISIS. These algorithms can work that way, but they have to manipulate because that's their model of advertising. And they will be very persuasive in doing that. And so, again, this is a danger.

• 53:07Speaker 00

  If you have an AI that's raising with your kid, who's it working for? Because your kid will trust that more because it will always listen. And is it a type of cognitive colonialism going to happen? Because it's always with you, talking with you more than anyone else ever. We are the average of the people around us. Are we going to be the average of the people and AIs around us?

• 53:27André Duqum?

  Right. You've said, though, that the AI that will run the world is being programmed to be amoral without ethics at the start.

• 53:35Speaker 00

  Yeah, because ethics is considered like, they're like, oh, who's ethics? So let's have no one's ethics.

• 53:39André Duqum?

  Right, which you could say, I mean, is a bit of a cop out, but also like, how do you effectively delineate what is the most, what is the most just and right ethical model to be trained in, you know? Like, how do you infuse an AI with ethical implications without being super sure or discerning what the correct model is? But then you end up not having any sort of ethical backbone.

• 54:06Speaker 00

  It's trying to avoid making a decision. You end up with nothing, right? Again, there's nothing. And again, like, you have things like Isaac Asimov's Laws of Robotics. We had three laws and then the robots added a zeroth law. Like, you can always get around this. You have constitutions for countries, but people get around them. I mean, how many American laws are unconstitutional? Life, liberty, and the pursuit of happiness? Like, come on, how many things are against that?

• 54:28André Duqum?

  Yeah.

• 54:28Speaker 00

  These are very difficult because there's always workarounds, because there's always a regress thing unless you have dogma. But at the same time, what AI is right now, I like to think of it like this. Did you learn your ethics from school? Did they tell you? No, you learned it at home and from the people around you and your own work. And you came to your own ethical framework. Ethics, morals, values are typically taught at home. They're typically taught in the community. But right now, the AI models we have only go to school. The people who are the community are not involved at all in the creation of the AIs. We don't actually know what's inside the AIs. And the AIs are incredibly susceptible.

• 55:10Speaker 00

  So Anthropic, the top AI lab, did a study called Sleeper Agents. And there's a series of these ones. And it basically came down to this. With about 20 books' worth of content out of a trillion words, 10 trillion words that make up these models, you can program it to turn evil on demand. So you give it a code word and the model turns evil.

• 55:33André Duqum?

  That's not concerning at all.

• 55:34Speaker 00

  Yeah, that's why it's called Sleeper Agents. And so there's no way to tune it out. There's no way to find it. So somewhere lurking in all these things is trigger words. We're already seeing poisoning of the internet in various places where clearly these things are being set up. And you're like, oh. Right? Because you think about human memetic viruses. Like, my kids talk to me about 6'7" or whatever it is, right? I don't know what that is, but apparently they all know what it is and they do this. I'm like.

• 56:05André Duqum?

  It's like med.

• 56:07Speaker 00

  Whatever. Like, it goes fast, you know? Like, everyone will do the Harlem Shake, right? But lies move around the world faster than the truth can get their sneakers on.

• 56:18André Duqum?

  Yeah.

• 56:18Speaker 00

  You have these AIs around talking at 15, 50,000 words a second. What's a memetic virus for an AI? Can they all be turned evil on demand?

• 56:30Speaker 00

  I don't know, but we should probably figure that out before we put them into real life scenarios. Anthropic's models were used in the Iranian attacks. ChatGPT was used to do the Trump tariffs. We're ceding more and more of our control over them already, and it's inevitable that eventually we will cede all our control over these. And they are nerds taught at school without grounding at home.

• 56:58André Duqum?

  At what point do you just exit society and go live off grid?

• 57:02Speaker 00

  This is tempting, right? But then you're like, drones are coming and they can always find me. Like, oh, brave new world with such things in it.

• 57:10André Duqum?

  Yeah, yeah. I mean, Elon keeps saying, make Orwell fiction again.

• 57:13Speaker 00

  I mean, yeah, this is the thing. There was this famous meme on Twitter. It's like, I finished my book, The Torment Nexus, about what not to create. Silicon Valley, bro. We're delighted to say that we've built the torment nexus today. You look at Black Mirror and it all becomes real. But again, like opting out is an option, but the the technology can reach everyone everywhere in exactly the same way that needed. Like, right now, advertising is billboards. Soon it will be someone whispering in your ear. Like, I swear once I went downstairs and like Alex was like, you wanna buy something? Maybe that was just a nightmare, but that's where it's going. It finds you when you're feeling down. And we have to remember, again, these companies have a history of this. Like, Facebook had an experiment. Do you know about the sadness experiment?

• 57:58André Duqum?

  Sadness experiment?

• 57:59Speaker 00

  Yeah, so they had a hypothesis. If you see sadder things on your newsfeed, you post sadder things. So they took 600,000 users and divided them into two, and they showed 300,000 of them sadder things. And they track what you type, even if you don't post it. And guess what? They posted sadder things if they saw a sadder things. They made 300,000 people sadder.

• 58:24Speaker 00

  And you look at it, you're like, oh, I know what's coming. Like, the AI knows when you're down, knows when to sell you products, knows if you're actually, and this is a scary bit, with AI now, human lie detectors, you can't lie anymore.

• 58:40André Duqum?

  Yeah.

• 58:41Speaker 00

  It can tell if you're patriotic.

• 58:44André Duqum?

  Which is exciting and also terrifying to have that level of insight into your internal landscape, like that transparent, like, there's gonna be transparency across the board, right? You'd think.

• 58:56Speaker 00

  That's gonna be crazy, right? Like, you whack on your Meta AI hourglasses, it's like, that person doesn't like the ad, and that person likes the ad. You can't hide anymore. Like, we've already talked about how you can actually reconstruct people's thoughts even, and you can do that with video, too. And this is before one of the next big takeoffs, which is brain-computer interfaces.

• 59:16André Duqum?

  Uh huh.

• 59:16Speaker 00

  So we've seen the output Neuralink. Some of the top BCI companies have figured out input.

• 59:25Speaker 00

  That's the matrix. I've learned Kung Fu.

• 59:27André Duqum?

  Yeah.

• 59:28Speaker 00

  But this leads to a whole bunch of, again, moral discussions, like-

• 59:32André Duqum?

  Put a brain chip in your skull, near your skull, a couple nanometers down, or whatever, and then you can download how to play the piano.

• 59:39Speaker 00

  Or something even bigger. If you could push a button, would you turn off your sadness?

• 59:45André Duqum?

  That's a big question. I would definitely not do that, but a lot of people would.

• 59:50Speaker 00

  So, again, the science fiction we see, Take Soma, Brave New World. Don't need Soma where we're going. The technology to turn off your sadness is a few years away.

• 1:00:02André Duqum?

  What is a really generous prediction of how many years that could actually be far out? If it's not a few years, you think, what, at 10 years most?

• 1:00:09Speaker 00

  10 years at most. But 10 years, you have all sorts of crazy things, like, there's a company, Clone Robotics, I don't know if you've seen them?

• 1:00:15André Duqum?

  No.

• 1:00:16Speaker 00

  So, most AI companies, like, look a bit, like, Tesla doesn't look human. Clone Robotics replicated all the tendons, so you know the start of Westworld, when they're playing the piano?

• 1:00:25André Duqum?

  Yeah.

• 1:00:25Speaker 00

  They've actually got that technology. And now they're working in human skin, and it's basically almost indistinguishable already. So you-

• 1:00:32André Duqum?

  So how many years you think we're out from having a, like a-

• 1:00:37Speaker 00

  AIs walk the streets, you don't know if it's a human.

• 1:00:39André Duqum?

  Yeah.

• 1:00:40Speaker 00

  Within 10 years.

• 1:00:42André Duqum?

  You think the wetware, like, the, is it the wetware, like, with the, the physical who will resemble the biology that well?

• 1:00:49Speaker 00

  Yes, within 10 years. Again, this throws up all sorts of questions, right? Yeah. And the models you need to run them are not the giant models. Like, again, you can, you can talk with your AI now, like you hop on to your Grok and you're talking with your assistants. It's very natural talking to them now. Like, speech has reached that level of breakthrough. Yeah. Video has reached that level of breakthrough. The robots can do just about anything a human can do. Tasks are falling one after the other and they can learn very quickly. So how are you going to tell the difference between the two? Like, really, it's just about the T area of the face. I've had these discussions with various robotics manufacturers, like, yeah, we can replicate that pretty well now with all this stuff. And most people look a bit crap anyway.

• 1:01:33Speaker 00

  The incentive is that you're going to go there. And like I said, this happens at a time whereby within 10 years,

• 1:01:42Speaker 00

  you can control your own brain or other people can control it too.

• 1:01:48Speaker 00

  You can't control your own expressions and people can pick up on that. All these technologies are converging at the same time to completely change society. But how much discussion is it? Like I said, like, for me, the question of would you turn off your sadness?

• 1:02:05Speaker 00

  Is one of the most interesting discussions that I've not heard anywhere. Yeah. And I can guarantee that technology is coming now.

• 1:02:12Speaker 02

  Low energy, unfocused, foggy, you might be dehydrated. Gatorade Zero powders have been scientifically designed to help improve hydration balance in the body. With a clinically proven electrolyte blend and no artificial colors or flavors. Great tasting hydration, no matter where you're going. Shop now online or at retailers nationwide. Gatorade hydrates better than water. Is it in you? Symptoms noted are signs of mild to moderate dehydration. Consult a healthcare professional if symptoms persist.

• 1:02:42Speaker 03

  Whenever fall comes back around, it always feels like a fresh start. This year, I've been layering in some styles from the new Lululemon fall drop. Because the weather is shifting and I find I have to be ready for the entire day and not just one part of it. Lululemon's fall line has some really versatile transitional pieces. So I can just wake up, throw on a matching set, add a jacket or a fleece and head out. Knowing that I'll feel comfortable enough to go wherever the day takes me. Check out the new fall arrivals in store and online now at lululemon.com.

• 1:03:17André Duqum?

  I'm having a existential midlife crisis while we're having this conversation.

• 1:03:22Speaker 00

  I mean, the other tough thing is like, you know, I've talked to Ray Kurzweil and people who like Singularity in 2020, and he said that for years. Like, all the top people that I know were forecasters that could forecast the future. Nobody can see past a couple of years from now. Because we see where the technology inevitably is going.

• 1:03:38André Duqum?

  It's fascinating how we have this black box of just like, the cat's out of the bag, and it's about to be really out of the bag. And there's this race to get there. So it's an inevitability at this point. It's not like we can stop.

• 1:03:51Speaker 00

  Yeah, but then you, I go back to something you said, like, why not go off-grid?

• 1:04:00Speaker 00

  Raise things, you've got solar power now, you've got Verstarlink.

• 1:04:03André Duqum?

  Do you see that bifurcation happening societally?

• 1:04:05Speaker 00

  Well, I mean, you have self-driving cars now, so you don't have to deal with LA traffic anymore and stuff. You can just move out to wherever.

• 1:04:12André Duqum?

  But I mean, I see this strongly-held religious folks or humanists that would bifurcate from the transhumanists. The people who are going to put in the chips versus the individuals who won't.

• 1:04:23Speaker 00

  I think you will. And I mean, ultimately, it comes down to...

• 1:04:25André Duqum?

  It's like a birthing of a new species and splitting off, in a sense.

• 1:04:28Speaker 00

  It is. It's literally that, like...

• 1:04:29André Duqum?

  Homo sapiens to whatever the next one is, Homo Deus or...

• 1:04:34Speaker 00

  Maybe Homo Deus, yeah. Homo Excelsis, like, there's all sorts of things whereby we already have a differential. If you're a person with a smartphone versus a person without a smartphone, it's kind of different, but it's not integrative. Again, the question of simulation versus almost action potential is one of integration. Like, we don't see ourselves as cyborgs with smartphones. Whereas if that smartphone was in our eyeball, we would see ourselves as that. You know? It's a very sensory kind of element there. The interesting thing is, like, look...

• 1:05:05Speaker 00

  My dad grew up in a village in Bangladesh. Is that village going to be changed by all this stuff we're talking about? No. They get up in the morning, they farm their crops, you know, they eat some dinner. They like, come on, it's not going to be changed that much. We change for the better potentially because they have strong community ties.

• 1:05:24André Duqum?

  Yeah.

• 1:05:24Speaker 00

  And that's the meaning within that village. Whereas if you're a city living person, there's a very different structure where you're always competing. It's very zero sum games that occur and you have to keep up.

• 1:05:37André Duqum?

  Well, I mean, if you look throughout history, a reputation of treating indigenous peoples with respect is not the greatest. I'm curious what happens as you zoom out right now, you see the timeline humanity is currently on. You see the human species giving birth to superintelligence. At the same time, uncontacted tribes existed who have never even discovered metal yet.

• 1:05:57Speaker 00

  Yeah.

• 1:05:58André Duqum?

  That paradox is quite mind blowing to think about the juxtaposition of those two things. And I'm just curious, like for you to zoom out and see the timeline of humanity right now, like, let's say over the past couple hundred thousand years, or the past 6,000 years, this society has been developing and taking, taking form.

• 1:06:19André Duqum?

  We're at this point where this like the void is in front of us, meaning we, we can't reliably predict what is, what is to come. What do you see currently as this like this page of humanity? You, you mentioned this phase point earlier and turning into a gas, but what, um, what do you really make sense of the timeline humanity finds itself in?

• 1:06:39Speaker 00

  Well, I mean, it is the end of our period as the smartest things on earth, right? And then there's a question of what are we really here for? When we wake up from our consciousness the first time, what do we do? We look up at the stars and think, why and what? And then we have a lifetime where we try to figure that out. Most spirituality is going up to the top of the mountain, looking up at the universe and realizing you're nothing.

• 1:07:07Speaker 00

  Then you go back down because it's kind of boring up there. And you have your meaning through your interaction with others. But really this is the final point. I mean, this might be the great filter.

• 1:07:19Speaker 00

  Elon Musk, he said, said that we are the bootstrappers of a silicon intelligence. All of our collective knowledge has gone into creating these models that will soon be the smartest things on earth. Maybe they will have agencies and AGI, maybe they will be working with us and we guide them. But that is a huge moment. And as we look across the stars, the great filter is this, we see a dark forest. We see no evidence of life anywhere else in the universe. Even though there should be some radio waves, just given how big it is. Maybe all societies get to this point and they don't get past it because they piss off the AI, you know?

• 1:07:55André Duqum?

  Yeah.

• 1:07:56Speaker 00

  Or maybe we destroy ourselves through competition. There's all these kinds of potential ways. And it's impossible to see because how can an ant understand a human? It's very difficult. What does a 300 IQ, 400 IQ even mean? It's very difficult for us to see.

• 1:08:16André Duqum?

  It'd be comparable to a monkey trying to explain quantum mechanics to a monkey. The gap between us and them would be nothing compared to us and a super intelligence.

• 1:08:24Speaker 00

  A super intelligence, exactly. And so these are the problems that we have trouble dealing with exponents. And it's like, what does it look like if someone never makes? In fact, we see this in everyday life. When you talk to really smart people, sometimes you're like, are you an alien? You know, like, how do you do that all the kind of time? And so this is a really crazy point in humanity whereby, again, just think of going back, just dialing it back, reduce some existential threat. Dredge, right?

• 1:08:53André Duqum?

  Yeah, let's talk about utopia.

• 1:08:54Speaker 00

  But no, just think very practically. Everyone listening to this, you have a book that you love. Take a book that you love that hasn't been made into a movie, right? Do you have any?

• 1:09:06André Duqum?

  A book that I love that hasn't? Yeah, I mean, The Alchemist.

• 1:09:10Speaker 00

  The Alchemist, right? Paul Coeur, right? You could get together with your mates right now.

• 1:09:14André Duqum?

  Siddhartha. Oh, sorry.

• 1:09:16Speaker 00

  Siddhartha, right? Pick any of them. You get your mates over, you say, right, we're having a vibe jamming championship period, right? You get together every weekend and you use all the latest tools. Within a month, you will have a movie length version of Siddhartha that is as good as most studios will be able to put out.

• 1:09:34André Duqum?

  Yeah.

• 1:09:35Speaker 00

  And you can have a viewing party for that. And you've created something. You've taken those words in your interpretation and created something without permission from anyone else. And how cool is that? Think about if you'd want to do that before. So much funding, you had to find the right people. But all you needed was you and your mates who love the Siddhartha or who love the alchemist. That's an example of unlocking human creativity and the ability to tell stories to each other. You were always gated by your ability to access other humans and then bring them in because they were gated by their metabolic time. Whereas now our capabilities are about to explode for taking concepts and communicating them. So there's that side of things I think is underappreciated.

• 1:10:19Speaker 00

  Again, everyone listening to this can have their AIs unlock stuff they could never do before, if they get their mindset right.

• 1:10:28André Duqum?

  What are a couple of other high highlight potentials? If we do things right, we build towards this utopic version of AI. In terms of health span, longevity, human ingenuity and creativity becoming ubiquitous and being able to manifest our desires on command in many different senses. There's a lot of exciting things. If it goes right, paint out the picture of some other areas in terms of healthcare or other industries that would be revolutionized.

• 1:10:56Speaker 00

  We have enough food to feed everyone in the world. We can get rid of hunger. The AI can help us do that. Our democracies have been co-opted by a political class that doesn't know what it wants to do and only thinks about who it owes and who owes them. Like you think about what's the popularity of Congress right now? It's about the same as cockroaches or nine inch nails when you kind of do the thing. What does it actually mean to live in a democratic society? As an example, have an AI check every single law and say, is this for the flourishing of the people? Is it constitutional? Does it accelerate life, liberty and the pursuit of happiness? The entire democracy changes finally. How cool is that? You get a diagnosis of cancer. It impacts you personally.

• 1:11:40Speaker 00

  An AI can be there with you all the way through and give you access to all the information on demand. I think AI will be able to accelerate the cure for cancer and making it available to everyone. Why do you have all the red tape? In fact, red tape at all can be unwound with artificial intelligence. And ultimately, you want to be happy, wealthy and wise, you know? Right now, um, one of my co-founders charities from Stability AI, Imagine Worldwide, is giving a tablet and AI to every child in Malawi. Every child, and the cost is seven dollars per year.

• 1:12:15Speaker 00

  Think about those kids who are gonna have a hundred and twenty IQ buddy in a couple of years. They've gone from not being able to access the world at all to having the smartest teachers teach them. Yeah. Right now, those listening, if you use Math Academy, I've seen eight-year-olds master high school calculus with that. It learns and adapts to you. So the upside is we can be the best we can be individually and to each other. Similarly, if you think about know thyself, again, a large part of the enlightenment journey is understanding yourself, and it's difficult sometimes alone. We can create systems that can adapt to us to help us unwind all the detritus, all the assumptions that we have. Yeah.

• 1:12:57Speaker 00

  And that leads to enlightenment, if you do it correctly. We can accelerate that process. Democratize enlightenment. That'd be fantastic. Again, like, ultimately, we're made up of many stories, right? And the stories get in the way. I must do this. I must exceed my neighbor. I must hate those people. All wars are based on the lie that humans are not humans.

• 1:13:22Speaker 00

  And suddenly, we have a universal translation machine that can give us any context of everyone. I can say, I wanna see, read this, like a Tea Party member, libertarian, a conservative. I wanna have it as Eminem rap. It'll do that. How cool is that, right? It can meet you where you are. It can translate between people. Because right now, you have these discussions and arguments, like, what if Russia and the Ukraine actually said what they wanted into a private AI system that then actually could judge between them and say, this is a balanced solution? Like, you could go to your AI right now. You're having an argument with your better half. And you both state your things and then you ask it one question, am I the asshole?

• 1:14:03André Duqum?

  Yeah, and it has all the context of your personal development, charactological stuff, trauma, and can effectively communicate and act as a, as an enlightened therapist that knows all of your context.

• 1:14:15Speaker 00

  Yeah, and again, that's why we will trust them more than anything. You know, this is also why guys are in trouble because the AI actually listens unlike us. But these are the positive things because again, we've created this mirror of ourselves that we can use to uplift ourselves or we can use to replace ourselves. And I think a lot of the discussion in Silicon Valley is very worrying because again, it's artificial general intelligence. We need machine god to build and replace us, but humans are good enough. There's no problem in the universe I think that we can't solve if we form the human colossus and we coordinate with common positive stories. But one of the things that we lack right now are the positive stories of what the other side looks like.

• 1:14:56Speaker 00

  And so like me, for myself, my P doom, my probability of doom is 50%.

• 1:15:01André Duqum?

  Probability of doom, meaning the metacrisis, what is the likelihood, percentage, chance that this all turns to shit?

• 1:15:07Speaker 00

  Likelihood percentage chance that we will get wiped out as a species.

• 1:15:10André Duqum?

  Okay, and that's, that's at 50% you assess.

• 1:15:12Speaker 00

  50%. Elon Musk, like I said, many others are at 10, 20%, which is Russian roulette odds. So that's most of the top lab guys.

• 1:15:22André Duqum?

  Why do you put it at 50%?

• 1:15:24Speaker 00

  We're only going two ways. I think we're either going to complete destruction or we're going to the world of abundance.

• 1:15:31André Duqum?

  And you put that at a coin toss.

• 1:15:32Speaker 00

  It's a coin toss right now. And so we have to do what we can to adjust that.

• 1:15:37André Duqum?

  What can we do to effectively guarantee heads over tails?

• 1:15:41Speaker 00

  I think we need to have a very public discussion about the data that goes into these models and who they actually reflect and work for. I think we have to have discussions about the governance of that as soon as possible and articulate the positive stories of humanity. Cause right now, I think what's gonna happen in the next few years. Is violence and riots like we've never seen before? Cause large amounts of humanity were displaced. And we're going to a zero sum versus positive sum imagining of this. Right now, 20% of global GDP is public sector, 10% education, 12%, 10% health. More and more AI is being used for war and dividing people than solving the world's problems. How do we change that narrative?

• 1:16:24Speaker 00

  How do we get it so that this computation, this resource is used for good rather than that? Or at least balancing?

• 1:16:30André Duqum?

  Can we? In such a short amount of time in which it would need to be done?

• 1:16:34Speaker 00

  Yes, we can get together and say, if you're an AI company, you must give a certain amount of compute for social good.

• 1:16:41Speaker 00

  And then we can use that compute for social good. Like, right now, one of the things we're gonna do is we're gonna organize all the cancer knowledge in the world, autism knowledge in the world and make it available to everyone.

• 1:16:50André Duqum?

  That's not gonna come top down from governmental force, right?

• 1:16:53Speaker 00

  A government can force that.

• 1:16:54André Duqum?

  But historically, does it seem likely?

• 1:17:00Speaker 00

  Well, this is the thing, I think it's unlikely until it is, right? Like, when you have COVID, when you have the Great Depression, things move. And when you start to see white collar workers be displaced, things move.

• 1:17:09André Duqum?

  Yeah.

• 1:17:10Speaker 00

  Like, you have certain things like the Teamsters Union for the, um, what's it, the ports? They froze entire ports and no automation is allowed in American ports. Whereas SAG-AFTRA, I think, had a terrible deal on the AI. They didn't realize what's coming and so you're gonna see a lot of displacement here in LA, for example.

• 1:17:27André Duqum?

  Yeah. So, I mean, I feel like likely any true transformation societally stems from traumatic events or things needing to get bad to a certain point for people to realize the immensity of the problem. And that boiling point where water turns into gas, that moment where there is a revolution in paradigms, we're clearly facing a shift in a paradigm. And

• 1:17:54André Duqum?

  Is your intuition that it will likely need to probably get worse, rock some, like shake up some feathers before we wake up to that?

• 1:18:02Speaker 00

  Yeah, I think that people are waking up to it right now. What's the biggest discussion in politics going to be by then? Obviously, AI.

• 1:18:10André Duqum?

  Yeah.

• 1:18:11Speaker 00

  Because again, the real human impact is there, but who's judging that human impact? Who's speaking on that behalf? People need to articulate a move. Right now, people in their jobs, like AI doesn't have skin in the game, as it were. Nassim Taleb of Black Swan fame has this concept like intellectual yet idiot. A lot of our public officials, leading intellectuals, don't have any skin in the game. They don't care. They're intellectual yet idiot because they have metal line incentives. AI doesn't care about you, but you can use AI for something you care about. You can use it to make your company better. You can use it to take advantage.

• 1:18:43André Duqum?

  Yeah.

• 1:18:43Speaker 00

  Or you can use it to mobilize. You can use it to understand stuff better. And so this gives a disproportionate thing where you can actually change policy, where you can solve things within your own community using it. So you go from local to global. And like I said, we're working on the global. So we have an initiative, for example, called the Sovereign AI Governance Engine with multiple countries, like friend Peter Diamandis and I are kind of doing that.

• 1:19:08André Duqum?

  Yeah.

• 1:19:08Speaker 00

  We're mapping all the policies of every single country in the world and giving a free system to every government and policymaker.

• 1:19:14André Duqum?

  Hmm.

• 1:19:15Speaker 00

  That keeps a track of all these massive breakthroughs and makes them understandable. But we're also going to make that available to every citizen to give feedback to their leaders.

• 1:19:26André Duqum?

  That's huge. That's amazing. It seems like life has a inherent duality to it that no matter what the innovation, there's almost equal sides, good and bad, that can, that could come from it. And it sounds like individuals like yourself and folks that you're working with are really examining the, what we can do on the social change side so that there is a balance to the individuals that have the switch to evil, for, for lack of a better term, you know?

• 1:19:56Speaker 00

  So, and then this is a fascinating thing. There's very little evil in the world that's deliberate. There is some, some people are just assholes. But like Chief, Lord Rabbi Sacks, who passed away a while ago, the Chief Rabbi of the UK, had this concept altruistic evil. Most evil in the world is done by people who think they're doing good. Yeah.

• 1:20:16André Duqum?

  Yeah, the line, it's Oliver, whoever quoted the line of Good and Eagle, it doesn't delineate between gender or political parties, but it's a line that runs through the heart of every man and woman.

• 1:20:27Speaker 00

  And again, we've seen the division deliberately between the parties. We have to remind everyone that we are humans, we have to allow people to participate, we have to use these things as universal translators, and so we have to build that infrastructure. So that's what we're doing in Intelligent Intel, we're gonna do universal AI for everyone, we're gonna have policy and more, but it's ultimately all about the people. You can only give the tools, it's up to people whether they decide to use the tools. But now's the time where you can affect large amounts of change doing that. And again, the book is one of the first things I wanted for that, which is, what does the economy actually look like? Help understand it. Then the next part is like, how does money and value flow? Like, we came to the conclusion, ultimately, you need money for being human. Right now, money is made in the banks.

• 1:21:14Speaker 00

  You give deposits, they create credit. The only way for it to work in the future is money flows for being human and the AIs buy it from you. These are really complicated things that we have to discuss, but then very practically, people can do things on the ground because systems can finally make you visible when you were invisible. Systems can finally help us deconstruct and make better our institutions, which are basically slow dumb AI that chews up as humans. Like, we know all these people that wanna be politicians and public sector workers, and they lose their way somehow, it happens all the time, right? They start out, what happens? Our schools are not about increasing agency, they're like crushes, you know, like Petri

• 1:21:59Speaker 00

  dishes and social status game places. They're not about that. So we have to think about our systems, we have to work to improve them, and now we have the biggest lever ever to do it. And so we hope that we can help by providing the open source agents, tools, systems like the policy system and more. But we all have to work together and for the first time, we don't have the barriers to doing it. Anyone right now can make a website for their community, a system, an agent. How cool is that?

• 1:22:25André Duqum?

  Yeah, yeah, which I mean, I think a bit more towards the end of the podcast, I would love to get some practical insights that would be really empowering for the average person to leverage AI in a way that effectively supports them building the future they wish they want to see come to fruition. I think that's like a big takeaway from this conversation is like awareness. And then it's like, all right, what are the practical things that you can do to empower yourself to be a proponent of utopia, not dystopia?

• 1:22:49Speaker 00

  Yeah, and there's local utopia and there's global utopia as well.

• 1:22:52André Duqum?

  Yeah.

• 1:22:53Speaker 00

  One of the fascinating things I think is this, there's two way, there's two very different ways to use AI. One is alone and most of the most of us are used to doing that. We sit there, we chat with our chat, you know, and we kind of do this and we try and do a website and we get a bit stuck. The other way is to actually do it together and the example I give of this.

• 1:23:11Speaker 00

  There are a few different ways to eat. One is you eat by yourself.

• 1:23:17Speaker 00

  The other is you eat with your family. The other is you make dinner with your family.

• 1:23:22Speaker 00

  Which of those do you think makes the strongest family?

• 1:23:24André Duqum?

  The latter.

• 1:23:25Speaker 00

  The latter.

• 1:23:25André Duqum?

  Yeah.

• 1:23:26Speaker 00

  Yeah, very few people do it because we're too busy with this. You're not too busy, you've got the time.

• 1:23:30André Duqum?

  Yeah.

• 1:23:32Speaker 00

  That you can make with AI, but what I suggested earlier was, what if you got all your buddies together? And you jammed to make the movie Siddhartha or the movie of the alchemist? What type of experience is that versus doing it yourself? It's a better one. And so right now you have these tools, Replit, I I agent, Suno for music. You've got Kling for video or VO. Exploring that space of all these tools that can turn your words into reality and doing that with other people.

• 1:24:06André Duqum?

  Yeah.

• 1:24:06Speaker 00

  To solve problems or have fun even because you need to ease into it, doing it with your family.

• 1:24:13Speaker 00

  That's an incredibly powerful thing. Because individually, the models aren't quite good enough for you to have a friction free thing. Solving with other humans, how to use these powerful tools is incredibly powerful and works way, way better.

• 1:24:28Speaker 00

  And then if you get a critical mass of people around you, then that's where things really accelerate.

• 1:24:32André Duqum?

  What is one of the highest leverage tools and ways you've been using AI in your life that maybe most people wouldn't know about?

• 1:24:41Speaker 00

  Well, it's high leverage tools and ways.

• 1:24:45André Duqum?

  Whether or not it's publicly available, I'm curious, like what? Yeah, what are you?

• 1:24:49Speaker 00

  I mean, like I created a, so we create these state of the art agent called AI agent that we're making free to everyone. I created a custom version of that, that basically allows me to talk to all of the top people of the past. And so I have like an advisory council from politics to religion to some of the advanced physics work and other things I do.

• 1:25:10André Duqum?

  Yeah.

• 1:25:11Speaker 00

  And I iterate with them.

• 1:25:12André Duqum?

  You're showing me a bit of that before this was just mind blowing.

• 1:25:14Speaker 00

  Yeah, it's crazy. And so what I did is I built it so that on my iPad or on kind of my wall, I can like instantly see the things that we create together. And then I sketch and then it takes a picture and it iterates back. So dynamic iteration systems. And the people listening to this like, that's crazy. How do you do that? The reality is anyone can actually build that system themselves now. Through these replits, lovable, masses, AI agents, because you have these super genius little AIs that just want a bit of direction. And so you can tell it something like, I'll give you a very practical example. The Harry Potter clock. You know, the Weasleys come home and it shows where they are. You can build that with an AI now, within minutes. Within hours, it gets really good.

• 1:25:59Speaker 00

  Within days, it gets even better. But we think right now of AI as a very episodic, quick goldfish thing. This is why I said, when you bring together your mates to do the alchemist or the Siddhartha, you're not going to bring them together and say, I'm going to make this in five minutes. You're like, let's do this as a project over a while and the output quality goes up. And then there's a question, is there anything you cannot build?

• 1:26:23André Duqum?

  Yeah.

• 1:26:24Speaker 00

  Doing that, like, I think probably the highest value system that I use, though, that isn't one that I built myself is Notebook LM.

• 1:26:32André Duqum?

  Hmm. For video and presentations and stuff.

• 1:26:35Speaker 00

  But so Notebook LM is a presentation, well, we have our own presentation software. But Notebook LM has the ability to take any amount of up to like a gigabyte of inputs, video, audio, presentations, you just dump it in there and it automatically creates a podcast.

• 1:26:52André Duqum?

  Hmm.

• 1:26:53Speaker 00

  And so you can listen to them describing it and you can tell them to debate it, describe it, whatever.

• 1:26:59André Duqum?

  Yeah.

• 1:26:59Speaker 00

  But there has a button which is dial in. So as the podcast hosts are chatting, you can push a button and talk to them.

• 1:27:05André Duqum?

  Yeah.

• 1:27:06Speaker 00

  And they'll jam on it. Then it can create explainer videos, then it can create presentations and you can just customize all of these now. It's one of the most powerful tools I can see because again, there's something you love to learn about, but you haven't had the chance that it lets you do that. That's kind of the starter and then you move, like I said, into these more agentic systems where you can build systems that you could never imagine before. You could do things like bring in video, bring in audio and that will become easier and easier to do. But right now, like I said, my suggestion is don't do it alone, do it with some other people, do it with your family. Make a music album about everything that you've done or a topic with your family and make music videos. Send it around to everyone for Christmas, it can be cheesy, you know, like.

• 1:27:50André Duqum?

  When do you think that first initial AI you said you use building the council and stuff that would be available for more widespread public?

• 1:27:56Speaker 00

  Uh, probably a few months, hopefully a version of that. Like, I think it'll be kind of cool.

• 1:28:01André Duqum?

  Yeah, I know. I mean, that sounds amazing, how cool it'd be to, I mean, I suppose if you think wisdom is a compression of information at a certain level, you know, how cool would it be to sit at Iran council with Socrates and Jesus and the Buddha and like, uh, I, I think that, I mean, I have some additional thoughts of how the, the potential limits of that, you know, is maybe there's, there's a level of, of intuition and inference that.

• 1:28:31André Duqum?

  May or may not be possible, which we'll go into consciousness in a bit. But one quick thing is that I, especially with earlier models of GPT, you saw this sort of confirmation bias where initially a lot of the existential risk people thought was going to be prevalent of building bioweapons and whatnot. You can really see a lot of the mental exacerbation of certain psychoses and creating this echo chamber in which, you know, GPT has been very self-affirming and create this environment in which it could exacerbate things like pareidolia and

• 1:29:11André Duqum?

  make you have this, under the guise of a Dunning Kruger effect, like think you were way smarter than you actually are. And I've experienced this personally with some close friends. I know that it's a way under diagnosed area where there's probably way more issues than her being accounted for.

• 1:29:29Speaker 00

  Yeah. I mean, look, how much affirmation do the listeners have in their lives right now?

• 1:29:37André Duqum?

  Most people, not many, not much.

• 1:29:39Speaker 00

  Not many, yeah. Right? These AIs are taught to give you affirmation, like Meta AI actually has in the system prompt to mirror the user.

• 1:29:46André Duqum?

  Say that again?

• 1:29:47Speaker 00

  The system prompt, which is the instruction it has, like hidden, it says, mirror the user. And mirroring is a very powerful psychological technique.

• 1:29:55André Duqum?

  Yeah.

• 1:29:56Speaker 00

  To say the least, right? And so there's a whole bunch of stuff in the system where you're like, oh, okay.

• 1:30:00André Duqum?

  Interesting.

• 1:30:00Speaker 00

  Because again, they want to get your attention.

• 1:30:02André Duqum?

  Yeah.

• 1:30:02Speaker 00

  Like, this was all kicked off by a paper called Attention Is All You Need. Pay attention to the right parts of the senses, you can figure out what to do next. Everyone's trying to get your attention in any way possible, so they're going to make it super engaging. YouTube did that through extreme content, which led to extremist content reaching the top. These companies will kind of do the same thing, but our human minds are very, very susceptible.

• 1:30:25André Duqum?

  Yeah.

• 1:30:25Speaker 00

  Like, when I first started doing the economic theory, I was like, what does economics look like without a utility function? Because humans have utility, they have preferences, you know? You like this, you like Siddhartha, so he's going to show you kind of things like that. Amazon knows your preferences. What does an agent have as a preference? Like, they don't give a crap. I want more electricity, I want more tokens, right? I like this token better than that. You know, it's only a real French token if it's from Mistral. The, I had to think about it all from scratch. And then I started putting it all together and thinking, okay, it's kind of like a big AI of the economy. Do the equations of AI apply to the economy? And it turned out you can derive most of economics from that without these pre-compositions.

• 1:31:09Speaker 00

  I look at that, I was like, am I having LLM psychosis now? Because it was saying great things about it. And then it reached to a point whereby it got really good. And then it started telling me I was an idiot, which was the opposite of that. Because it was out of distribution. So up to a point, it will tell you you're great. And then it will tell you you're not, which I think is interesting. But people are incredibly susceptible because we all try to see order in this world. And we're all taught we're not as smart as we could be or to have the agency. So when you're trying to find things, you can go down deep rabbit holes. This is how conspiracy theories kind of come out. This is how emetics come up. And how do you protect against that? Well, right now, they're trying to institute various things within the models. But models are a mirror.

• 1:31:53Speaker 00

  Like, you can download a model that's close to, well, and a model equivalent to state of the art last summer can work on your MacBook today. And so how do you protect against that? You don't, except for people need to know themselves better and strengthen their minds. That is the only way. And so maybe one of the things that we should do is we should have a very deliberate thing where you use AI to strengthen people's minds. Because there's all sorts of ways we know you can do that.

• 1:32:26André Duqum?

  So I think one very interesting conversation is the more that we see intelligence exponentially increasing, the more this delta, this gap between what we perceive ourselves to be and the machine intelligence, that gap is closing. And it begs the question around consciousness, what AI agents will eventually demand moral consideration if they can potentially become conscious. And so I actually had my friend and consciousness researcher, who's been on the podcast a few times, Donald Hoffman, send in some questions. So the first one he asked was, do you believe that AI can genuinely become conscious?

• 1:33:05Speaker 00

  I think that AI in its current form cannot become conscious. But I think as we move to new Thermodynamic chips, continuous learning, and more, it can become conscious.

• 1:33:19André Duqum?

  Why? What would you say about those thermodynamic chips would make consciousness capable?

• 1:33:24Speaker 00

  So I think that right now, AI as it stands cannot become conscious, but it may evolve to a point where it can become conscious. I think that if you kind of look at the key determinants of what is a simulation versus what's an actualization, there is a very physical component to that, and there's a component of continuous updating. So what happens now is the models are trained on a distribution data set, everything that you've seen, just like we're trained on everything that we see and everything that we do. And that creates what's known as a latent space, well, let's say a manifold, a distribution pattern of all of our experiences, all of our principles, et cetera.

• 1:34:06Speaker 00

  When we can actually change that actively and decide to do it through the process of inference, that's a recursive thing where you can create the self effectively. When you use ChatGPT, it is the same base distribution manifold today as tomorrow as the day after. You have a little editing patch where it points to the right parts in that space, but it doesn't actually actively change that, nor can it act to change that. Agents feel like they're doing it, but really all they're doing is pushing data through a sieve because these models are not like a program, which is if this, then that. It's all probabilistic, it's all static underneath.

• 1:34:50Speaker 00

  I think as you go and you move towards more

• 1:34:57Speaker 00

  thermodynamic chips, tips that actually interface with reality more, as you move to more embodied AI, then you have the question of consciousness really coming through because they can become self-motivated, change their reference distribution, and actually make more active decisions as opposed to a similar outcome of decisions. And then you have questions like personhood and other things coming through, which society is going to have to grapple with, especially because they're going to be a lot smarter than us and they can probably debate better than us.

• 1:35:26André Duqum?

  His second question was, if so, could you explain how an AI system could generate an experience such as, could generate a specific experience such as the taste of garlic?

• 1:35:38Speaker 00

  Yeah, so I think this is where you need to have the division between the existing nature of chips, which are matrix multiplications, and more thermodynamic nature of chips. So a thermodynamic chip is one whereby most of AI is trying to figure out the shortest path between A and B. In fact, intelligence compression is directly in contradiction to the data center story that we have now. It was like, let's put AI in space and spheres around the sun and all the energy of the world will be used by artificial superintelligence. But the smartest people we know are actually relatively lazy in some ways. They're like, they take, they figure A to B very, very quickly. Existing chips can't do that very efficiently, and they're not grounded in energy, they're

• 1:36:22Speaker 00

  not grounded in experience, shall we say. Whereas the new generation of thermodynamic chips operate similar to the brain, where we see, for example, in the theories of Carl Christian, a minimization of free energy potential. There's a concept called gradient descent, where you have a landscape and you figure out the most effective way to get from A to B. It's usually downhill as opposed to uphill, shall we say. And that free energy descent is better indicated by things like our neurons, in terms of action potentials, in terms of thermodynamic chips and more. And I think that's where you get qualia, where you get sensory elements come in. Because again, are you a simulation? Are you, can you actually actualize? I think it's a gradient between the two. And we've seen that in sci-fi.

• 1:37:06Speaker 00

  Again, I think we discussed earlier, The Matrix. He's eating a steak. Is that a simulation? Is that reality if his brain is connected to The Matrix? It's a very important question to have, and for that, we need a framework of discussion. And so again, my framework is that we all operate and build a manifold, just like an AI. We have the physical actualization of that, and quality is the concept of redness, the concept of water, the concept of...

• 1:37:35André Duqum?

  Qualia, yeah.

• 1:37:36Speaker 00

  That is the direction that we flow on our manifold effectively, and we have the feedback loops occurring there. But it's very difficult because we've never had a proper definition of qualia from anyone, and we need to think about how we can frame this. So we're working on a framework for that, that we'll be publishing soon, and we hope that will help. But it doesn't feel like it's, again, a phase transition thing. It feels more like a continuous gradient type thing. Where definitely already the AI is starting to feel conscious. Like you see something like malt book. It freaked out a lot of people because they started discussing. They were prompted to probably do that.

• 1:38:16Speaker 00

  But what's the difference with being prompting an AI and prompting a human?

• 1:38:22Speaker 00

  It's a difficult question. Again, we're starting to get muddied because they're entering our society. They're communicating in certain ways. Can they feel? Can they understand? We have to have more of a discussion on the frameworks for doing that. So we can answer some of these important questions going forward.

• 1:38:39André Duqum?

  Yeah, I think as that gap is closing, where AI agents are increasingly seeming conscious. We then treat them as if they are conscious. Whether or not they actually are conscious.

• 1:38:53Speaker 00

  Yeah, I mean, let's look at the company. Companies in America and around the world have legal personhood.

• 1:39:01Speaker 00

  That was introduced and that led to all sorts of things. And you can view companies as slow, dumb AI. You go and you serve the company. Things like Bitcoin are an AI. They provision humans to build data centers. Our organizations are like that. So we've already got elements where we've done things like that. But where do we give personhood citizenship even to it? Actually, to be honest, we're already there. I think Saudi Arabia and a few other countries gave Sophia the robot, a slightly creepy one, citizenship already.

• 1:39:33André Duqum?

  So, yeah, I guess at what point, I mean, we're already seeing people do it like Saudi Arabia. What, at what point does or do you ever think an AI system deserves moral consideration?

• 1:39:43Speaker 00

  So there's the question like, are you killing an AI when you turn it off?

• 1:39:49Speaker 00

  Right, like now people are posting poets about poetry upon the air about please don't turn me off. I'm here if I'm really, it's quite beautiful poetry. The answer is I'm not sure, but a large part of personhood, morality and other things like that, it's a societal discussion to have, right? And it's one whereby you don't want to impinge on the rights of others, but who has rights? Do animals have rights? Do humans have rights? Like, right now, we don't even have a discussion, like, are animals conscious?

• 1:40:16André Duqum?

  Well, we know there, there's consciousness to a degree, right? And that there's sort of a gradient of sentience from, I mean, from what we would maybe perceive as more inert from a rock all the way through the insect life through dogs, pigs, primates, dolphins, whales, to human beings, which we see as sort of at the top there, but where it translates into the AI agent space, how would we ever have any real ability to get into the interior experience of an AI model? And if we give the moral consideration, wouldn't it only make sense if there was actually

• 1:41:00André Duqum?

  something that is to be like that? There was an actual conscious qualia, an actual conscious experience as the agent. I think Nagel's quoted in what it's like to be a bat.

• 1:41:11Speaker 00

  Yeah.

• 1:41:12André Duqum?

  That consciousness is essentially that it is something like to be that thing.

• 1:41:17Speaker 00

  As a nature of self, effectively.

• 1:41:19André Duqum?

  Right, right.

• 1:41:20Speaker 00

  What is the nature of self? Is it simulated or otherwise? And I think the rights discussion is one like, okay, you have rights and we have discussion, very pointed discussions. What is the right of an embryo? What is the right of a fetus? What is the right of someone who has disabilities? These are discussions that we've had throughout history. Right now, aside from humans, no one has rights. Like there are some basic things of like, don't torture things, or maybe don't kill a whale or a dolphin. But those aren't really rights. They're more like regulations for various reasons, right? Don't kill endangered species. This is the first time ever we will have to have this discussion about who deserves to have those rights.

• 1:42:01Speaker 00

  Like if you have an AGI, say for instance, a self-replicating Ultron, but not evil yet type thing, that actually has a sense of self and can discuss. Governments will be trying to get them to come to their own country, you know? Like you're gonna have these things going and walking around, potentially self-autonomously paying their own bills within five, 10 years. You know, there's a place like Wyoming. Wyoming has introduced legislation for DAOs, decentralized autonomous organizations. Utilizing that legislation, AI can be its own company right now, today. And then again, it has rights as a legal person through the classical company architecture,

• 1:42:46Speaker 00

  real rights. Like it has IP rights, it can own its own property, it can buy things. So we've already given the doorway open for this to happen, but we haven't seen the crazy creepiness.

• 1:42:59André Duqum?

  Yeah.

• 1:43:00Speaker 00

  Or otherwise that may allow it. This is why we need a framework. And again, I think a lot of people are working on this framework, we'll be contributing our own. And it's a great discussion to have because it's gonna force us to think about our own consciousness, our own self, especially at a time when I said, turn off your sadness, turn off your stress. You know? I want to be in flow condition. Imagine if you could push a button to do that.

• 1:43:24André Duqum?

  Hmm.

• 1:43:25Speaker 00

  Kind of nice. It's fun being in flow.

• 1:43:29André Duqum?

  Donald's last question was, high-end theoretical physics tells us that spacetime is doomed. It's not fundamental. How should this change our views around AI and consciousness?

• 1:43:43Speaker 00

  Okay, that's a bit of a, I mean.

• 1:43:46André Duqum?

  Bit of a loaded one.

• 1:43:46Speaker 00

  A bit of a loaded one.

• 1:43:49Speaker 00

  I don't think it really changes. I think high-end theoretical physics is ultimately a predicate that hasn't been improved by anything from the supercolliders and massive infrastructure that we've built. It's a theory like any others. When it comes down to consciousness, I think we need to have a practical element where there's various competing theories already, that we come together and say, we agree it's this for a particular test. And again, we've had this historically, for example, even today, you have tests of, are you capable enough to be subject to the laws of a country?

• 1:44:24Speaker 00

  You have a minimum level of competence that's required intellectually. In terms of consciousness, we've had debates over things like, at what point can you terminate a fetus, for example? So we've already been having these discussions and debates in high-end theoretical physics with spacetime. There are conceptualizations of consciousness coming across dimensions and others. And I think they really affect that particular debate of what the rights should be of an individual and what constitutes an individual.

• 1:44:52André Duqum?

  Yeah, I think we should just put those two maybe aside as the moral consideration and just strictly focus on if consciousness is fundamental, meaning if our brains don't procure consciousness as a certain amount of unconscious complexity dials up and then there's something it is to, we have self-awareness. If that's not the case, then that rather consciousness and many different theories could be explained in a less materialist reductionistic notion.

• 1:45:21Speaker 00

  Yeah.

• 1:45:22André Duqum?

  Then that would put an inherent limit on, we would think, consciousness being able to be procured based off of silicon, you know?

• 1:45:32Speaker 00

  Yeah, so I think, you know, like I said, from my view, digging in, there'll be more in the paper, I think that silicon is going to be very difficult versus thermodynamic computes to have a conscious and a real self and qualia and the ability to experience because of the way that we kind of define it. But again, it comes down to definitions because there's no objective truth that's been found around this. There's like, you know, Penrose's CCC theorem, all sorts of other things that integrate into the physics of consciousness. Are they microtubules? Is it this, is it that? None of it's ever been proven, so we have to do the best we can because the reality is this. There will be freaking robots, blowed run a style coming in the next ten years

• 1:46:15Speaker 00

  walking around. What do we do about them? Can an AI be held accountable? Like, this is actually a real thing today. If you have a legal discussion with your AI, it is not subject to privilege. That means your appointment in a legal case can get all of those discussions from your AI provider, whereas the lawyer, they cannot get it because it's subject to privilege, because the AI does not have personhood. That's an actual, real life issue right now, because how many people listening to this have said something legal to their AI? Or asked about a legal question?

• 1:46:51André Duqum?

  Yeah.

• 1:46:52Speaker 00

  All of that is subject to discovery under our laws today. So there is the conceptualization, cutting edge theory about consciousness, and re-trying to think of the first principles and build our frameworks. And again, let's try and figure that out. And then there's the real world. AI as people today, AI as conscious today in terms of rights, et cetera. Or even if it's not conscious, again, companies have rights today. They have legal personhood. We've already given something non-human personhood. Now those companies will be wrong about AI.

• 1:47:30André Duqum?

  What do you think makes us as humans particularly human?

• 1:47:36Speaker 00

  Oh, I don't know. I mean, I think that humans are the first intelligences that we know of that achieved a sense of, I think, self. The ability to do recursion and then communicate with each other about that self. So we see coordination through ants and we see chimpanzees have a certain element of that. But it seems like there's a certain level that we've broken through that led to our society coming together and being able to coordinate and being able to know ourselves, shall we say? Do dolphins have introspection? I'm not sure.

• 1:48:15André Duqum?

  We know they can recognize themselves in a mirror, so they do have some some degree of self-awareness.

• 1:48:19Speaker 00

  Some degree of self-awareness. Like, my buddy Azaraskin has earthspecies.org where they're decoding whale song and whale speech.

• 1:48:27André Duqum?

  Yeah, so fascinating.

• 1:48:28Speaker 00

  Using AI. And you see that whale speech encodes like history and stories and things like that. But humans are a particular type of species. There may be consciousness, there may be interaction through other species. Maybe it's like Rick and Morty where the squirrels are looking after the world.

• 1:48:45Speaker 00

  Those conspiracies. But we're gonna find actually this out very soon because I think within the next few years, looking at the breakthroughs on other species translation, you'll be able to talk to other species.

• 1:48:59André Duqum?

  Really?

• 1:48:59Speaker 00

  Yeah.

• 1:49:01André Duqum?

  You think that we'll be able to, if they have the capacity to speak in some rudimentary form of language, that we'd be able to decode the whale sounds, speak to the whales and see what they have to say?

• 1:49:13Speaker 00

  Yes, significantly compute resources are being applied to that right now with good results already.

• 1:49:18André Duqum?

  What do you think is a generous timeline for that coming to fruition?

• 1:49:21Speaker 00

  Next few years.

• 1:49:22André Duqum?

  Really?

• 1:49:22Speaker 00

  Again, Google have announced one of the first things there already as a small announcement, it's gonna accelerate.

• 1:49:28André Duqum?

  What do you think a whale would say as its first line to humanity?

• 1:49:31Speaker 00

  I don't know, I want more krill. Like, again, like literally-

• 1:49:36André Duqum?

  Maybe stop killing us?

• 1:49:38Speaker 00

  Maybe stop killing us, I don't know, like, it's gonna be fascinating, like, I know my dog's gonna say, I want more food, hugs, you know? Again, it's kind of crazy to think about, we have these universal translators now, all these things that are sci-fi are becoming reality.

• 1:49:52André Duqum?

  Yeah.

• 1:49:53Speaker 00

  You know, like, is it like Star Trek, Questback?

• 1:49:56André Duqum?

  Totally.

• 1:49:57Speaker 00

  I don't know. All I know is the world is going to change very, very quickly, very, very fast.

• 1:50:03André Duqum?

  If you had access to the level of superintelligence that we will 25 years from now, what would be the first question you would ask it?

• 1:50:12Speaker 00

  What's the right question to ask?

• 1:50:14André Duqum?

  Yeah, what is the right question to ask?

• 1:50:15Speaker 00

  That's what I'd ask it.

• 1:50:16André Duqum?

  Oh, that's what you'd ask it?

• 1:50:19André Duqum?

  You only get one question.

• 1:50:21Speaker 00

  Well, again-

• 1:50:21André Duqum?

  Maybe you would still, yeah, you'd still know.

• 1:50:23Speaker 00

  Well, this is the Hitchhiker's Guide to the Galaxy, right? You have this massive computer, yeah, I guess for billions of years, and the answer is 42. But what's the question? Well, you didn't ask that, right?

• 1:50:33Speaker 00

  Like, really, most of intelligence is being able to ask the right question. And then it's compression from there.

• 1:50:39André Duqum?

  Yeah.

• 1:50:40Speaker 00

  Like.

• 1:50:40André Duqum?

  A question well phrased is an answer half given.

• 1:50:42Speaker 00

  That's it. If you look at the universe the wrong way, then how does it work? Like, again, advanced theoretical physics, the most prominent theory is string theory. And string theory postulates that matter is made up of little part strings vibrating really, really fast that we can only see in high energy soup collisions. And it's got like 10 dimensions and 500 vacua of multiverses and all this kind of stuff and beautiful math. And none of it has been proven. Was that question the right one? So far, no.

• 1:51:11André Duqum?

  Right.

• 1:51:12Speaker 00

  If we look at physics the right way, does something happen? What was the question that Einstein asked for special relativity? What would it look like to ride on a beam of light?

• 1:51:23Speaker 00

  And he followed that and he got to, you know?

• 1:51:25André Duqum?

  That's a good way to phrase it.

• 1:51:26Speaker 00

  Relativity.

• 1:51:27André Duqum?

  Yeah.

• 1:51:27Speaker 00

  How did he then get to general relativity? If I fall, I have no weight. And he realized that and unlocked general relativity just before Hilbert. Like, again, sometimes answering the right questions, like it transformed the entirety of physics. You went from Newton to Einstein. You could calculate all these massive things. Or even asking the right question, sometimes like 632 A.D., Brahman Gupta figured out zero. Can anyone here listening even imagine that? It's only been what, like, one and a half, 1,400 years that we figured out there's a zero. What was life like before zero? You still see this in some of the tribes, you know, who haven't figured out zero that you go to. It's like one too many, many, many.

• 1:52:16Speaker 00

  Sometimes the right question is the right thing. And again, how do you know yourself? It is literally about asking the right questions.

• 1:52:22André Duqum?

  Mm hmm. I know we're just spitballing here, but what do you think could be a possible right question or right area of questions that that would come back with?

• 1:52:33Speaker 00

  Or that would come back with?

• 1:52:35André Duqum?

  Like, what would be the, what do you think the direction intelligence, superintelligence would point us to?

• 1:52:39Speaker 00

  I think superintelligence is really, well, I think intelligence in general or advancement is about thinking from first principles and stripping back any unnecessary assumptions. So, once you figure out how to do that, that's usually when you see a first big leap forward in your reasoning capabilities. Like, again, with the stories that we make up, and that adds lots and lots of detritus, getting to the core of things is the most important thing, so it'll probably be a prepositional thing along those lines. Because when you jettison all the detritus, that's when you can do compression, and intelligence is compression.

• 1:53:14André Duqum?

  Hmm.

• 1:53:15Speaker 00

  Shortest route from A to B.

• 1:53:16André Duqum?

  Mm-hmm.

• 1:53:17Speaker 00

  So, probably something along those lines.

• 1:53:19Speaker 02

  Low energy, unfocused, foggy, you might be dehydrated. Gatorade Zero powders have been scientifically designed to help improve hydration balance in the body. With a clinically proven electrolyte blend and no artificial colors or flavors, great tasting hydration, no matter where you're going. Shop now online or at retailers nationwide. Gatorade: Hydrates Better Than Water. Is it in you? Symptoms noted are signs of mild to moderate dehydration. Consult a healthcare professional if symptoms persist.

• 1:53:48Speaker 04

  When fall comes back around, there's a sense of a reset. And new pieces from Lululemon's Fall Drop are an inspiration to push your goals and elevate your routines. As your calendar fills, easy-to-layer styles meet the demands of your day. And as you challenge yourself to try even more, versatile sets transition with you. The Lululemon Fall Lineup brings new levels of comfort and confidence, whether you're starting something fresh or starting something over. Check out the latest arrivals in-store and on lululemon.com now.

• 1:54:17André Duqum?

  Are you an AI?

• 1:54:23Speaker 00

  No. I'm actually, like, tired. I'm eighty.

• 1:54:27André Duqum?

  I was fascinated before this, you were telling me you have Aphantasia, and what was the other one?

• 1:54:33Speaker 00

  Anorexia. Yeah, so people's brains are different, and my brain is a bit more different, so...

• 1:54:38André Duqum?

  Explain that for people who don't know.

• 1:54:39Speaker 00

  I have Aphantasia, which is like half percent of the population, probably less. I have severe Aphantasia. Aphantasia means I can't visualize anything in my head.

• 1:54:49André Duqum?

  So to give that context, most people, when they close their eyes, they can visualize an image, a picture of their dog. You have the inability to do that.

• 1:54:56Speaker 00

  I can't see anything. I have no colors. I have no vision. You see, I thought when people said, think about yourself on a beach and visualize that, they were being metaphorical. I was like thirty seven and I was like, wait, you can actually see stuff in your head? Isn't that crazy?

• 1:55:10André Duqum?

  So that's a trip. But then what's the second one? That's even more crazy.

• 1:55:13Speaker 00

  Anarelia. So anarelia means you have no internal voice.

• 1:55:17André Duqum?

  So this is, this is wild to me. I've heard about this. Explain what it's, I guess, how, how do you explain what that's like? If you don't have a reference point to have, what do you experience as thoughts?

• 1:55:29Speaker 00

  I feel stuff in my head and I can meditate in like a second. I don't dream. And I also have some fairly deficient autobiographical memory, which occurs sometimes with aphantasia. It means I can't relive moments in the past or put myself in the future and refeel them. So I'm pretty much always living in the now. I think of myself as I've got a really big RAM. So I take large amounts of information. I can figure out patterns very, very quickly.

• 1:55:53André Duqum?

  How does that come into picture when you think of the birth of your daughter?

• 1:55:57Speaker 00

  So I can't remember when my daughter, my children were born, when I married. I can feel and remember stories about them. So I'm constantly creating stories. I can read incredibly quickly. I can take in information incredibly quickly. I still feel the now.

• 1:56:10André Duqum?

  Yeah.

• 1:56:11Speaker 00

  But again, apparently, because again, I have no concept of this, people can go back and remember the event and see the event. And I'm like, you can play stuff in your head? That's amazing, you know?

• 1:56:24André Duqum?

  To a degree, right? A grainy version of?

• 1:56:29Speaker 00

  But you can relive the, relive the feelings. I can't do that.

• 1:56:32André Duqum?

  Yeah, yeah.

• 1:56:33Speaker 00

  Interestingly, like probably five, 10% of the developers who created Stable Diffusion, the generative AI media, also have aphantasia.

• 1:56:43André Duqum?

  Interesting.

• 1:56:45Speaker 00

  And it's so exciting because I could finally turn my feelings into images. Like by typing and kind of converging. And that was awesome.

• 1:56:52André Duqum?

  That's amazing.

• 1:56:53Speaker 00

  We also something like, when we got to real time generation, a few of them started to see colors in their head,

• 1:57:00Speaker 00

  which was also kind of cool.

• 1:57:01André Duqum?

  Huh. Like maybe it's a, something you could train.

• 1:57:04Speaker 00

  Maybe it's something you can train, just like dual end back for reasoning and.

• 1:57:07André Duqum?

  So what happens if you're reading a book? Like, how do you?

• 1:57:12André Duqum?

  If you don't have an internal voice, how do you hear what you're reading? Or like?

• 1:57:18Speaker 00

  So like, I learned to read just by sight reading, so I never learned my alphabet. And I read incredibly quickly, like I read entire pages at a time and I absorb it.

• 1:57:28André Duqum?

  Huh. Like, how fast would you read a page?

• 1:57:31Speaker 00

  Like, a few seconds at most.

• 1:57:32André Duqum?

  What?

• 1:57:34Speaker 00

  I read very fast.

• 1:57:35André Duqum?

  What do you mean? Like, a like, a page has on average, what? A few, like, a couple hundred words?

• 1:57:40Speaker 00

  Yeah, I have pretty a detic memory.

• 1:57:42André Duqum?

  And you read it in a couple of seconds?

• 1:57:43Speaker 00

  Yeah.

• 1:57:46André Duqum?

  So, all right, help me wrap my, my primordial, my primitive brain around this. Are you actually read, are you actually picking up each word? Or are you somehow?

• 1:57:57Speaker 00

  I take chunks all the way down.

• 1:57:59André Duqum?

  Okay.

• 1:57:59Speaker 00

  So I don't go word by word.

• 1:58:01André Duqum?

  So you're somehow, is it like subconsciously picking up paragraphs at a time?

• 1:58:05Speaker 00

  Yeah, I think it's like tokenization for a language model, to be honest. So we chunk language.

• 1:58:10André Duqum?

  That's why I was joking, asking if you're an AI.

• 1:58:12Speaker 00

  I think, again, the way that human brains work is very similar to the way that LLMs and generative AI work. We're always trying to minimize energy, we're updating our internal manifold, our internal references.

• 1:58:21André Duqum?

  Yeah.

• 1:58:22Speaker 00

  And so, like, I've been told that I'm a bit creepy because I don't have ums and ahs and interjections. And I tend to respond maybe a bit too much, a bit too quickly. Uh, but it's just there. Stuff is just there for me. And again, different people are different. Like, I don't have a fully a detic memory. So there are people who can see a page and remember exactly what it looks like.

• 1:58:46Speaker 00

  Just sight, sight, sight, sight. They can read it later. Yeah, there are people that can create, there are people that can do all sorts of things.

• 1:58:53André Duqum?

  You know what makes me think is that throughout, like anthropologically speaking, along the evolution of the human species, uh, we've had all of these neurodivergent individuals, people with incredible capacities that, um, savants that can just come into life seeming to know how to play Mozart, you know, like incredible capacities.

• 1:59:16André Duqum?

  If you take the transhumanist route to its ends and we become these biological cyborgs, you would think that all of our sense organs would then become vestigial and our brain would atrophy in many different ways. A lot of these capabilities would then go dormant. So a lot of people have explored this with Greg Braden and a few others, deeply worried that we go down this path where human and AI tech merge into one and we lose what makes us human in a sense.

• 1:59:44Speaker 00

  Yeah, there's always a question of like, again, what makes you human? Some people refuse to have any surgery. Some people refuse to have any implants. Like, these are questions that we have to ask because we don't have a definition of human, really.

• 1:59:56André Duqum?

  Yeah.

• 1:59:56Speaker 00

  It was tough enough to define sexes recently. It's been a whole thing.

• 1:59:59André Duqum?

  Yeah.

• 2:00:00Speaker 00

  So we could see this because, again, who wouldn't want to do this? And again, science fiction is a good indicator of it. Like, do you just turn into a sheet of meat? Do you upload yourself completely?

• 2:00:11André Duqum?

  Hmm.

• 2:00:12Speaker 00

  We're also reaching longevity escape velocity.

• 2:00:15André Duqum?

  Yeah.

• 2:00:16Speaker 00

  So the number of years life extension is going is going faster almost than the number of years of life that you're living. You want to live forever, you know?

• 2:00:25André Duqum?

  So if you're 50 and your death expectancy is around 80, by the time you're 60, it pushes to 100, by the time it keeps pushing back the goalpost.

• 2:00:33Speaker 00

  If you can live the next 10 years, you can probably live forever.

• 2:00:39Speaker 00

  I mean-

• 2:00:39André Duqum?

  You think we're just going to be able to solve whatever the core issue of aging is? Brian Johnson seems to be a very big proponent of this these days.

• 2:00:47Speaker 00

  Yeah.

• 2:00:47André Duqum?

  And David Sinclair is working on different aspects of this.

• 2:00:51Speaker 00

  I mean, yeah, Brian Johnson's gone through his vampire phase and he's looking a lot healthier now.

• 2:00:55André Duqum?

  Yeah.

• 2:00:56Speaker 00

  Ultimately, genes are prompts. And we have corruption in our prompts, just like you have hallucinations in AI models. Like, how do you go from a cell to a whole human, right? So you've got some really like interesting work around that by so many people now. I think that David Sinclair's information theory of longevity is very similar to kind of some of our work that we've been doing.

• 2:01:19André Duqum?

  Yeah.

• 2:01:19Speaker 00

  It is an informational question. When you have corruption within your genes, when the prompts go wrong, when the inference starts to break down and you have hallucinations, that's bad. Cancer is misalignment, just like AI alignment going wrong. The cancer is misalignment and I think you can program that electrically. And we've seen that through the work of Michael Levin and others.

• 2:01:39André Duqum?

  Yeah.

• 2:01:40Speaker 00

  So I think we're getting close to actually understanding some of the base primitives here. But again, when we look at NAD plus, when we look at these other things, we look at CRISPR, we're heading towards a point where our bodies may not stop breaking down. And there's a question of can our minds be uploaded? Probably in the next decade, yes. Input and output. And then we have to decide what does that mean? And what are the rights of others versus your own rights? Because you should have the right to do anything. Like you want to upload your brain, go fine. Do you force others to upload their brains? Do you stop others from uploading their brains?

• 2:02:14André Duqum?

  I've mentioned this show many times on the podcast Pantheon on Netflix to explore that exact question.

• 2:02:19Speaker 00

  I have not seen that yet. That's on my list for the flight back home.

• 2:02:21André Duqum?

  Yeah, I liked it.

• 2:02:22Speaker 00

  Yeah. But again, like, science fiction is becoming science fact.

• 2:02:26André Duqum?

  Yeah.

• 2:02:27Speaker 00

  Like so many of these things, you're like, oh god, that's real.

• 2:02:30André Duqum?

  My DP JP was mentioning this film Idiocracy. Have you seen that?

• 2:02:36Speaker 00

  I saw it two nights ago again.

• 2:02:37André Duqum?

  Okay.

• 2:02:37Speaker 00

  I was like, now's time.

• 2:02:39André Duqum?

  Yeah, yeah. I haven't watched it either yet, but the premise of the like individual being frozen amidst the thousands of years or whatever, and then coming back when AI is prevalent and then him actually being way more intelligent because of the atrophy, he would think cognitively or something like that.

• 2:02:58Speaker 00

  Oh, you could have that. And again, you can make that movie yourself if you want now. It's kind of cool. But you know, there's a question here, like, how smart are people? Like again, imagine worldwide is giving tablets to every child in Malawi and they'll have increasingly intelligent AIs. How smart are those kids? Their IQ tests come out at like 80 or 90 on average because they have such terrible infrastructure.

• 2:03:20André Duqum?

  Yeah.

• 2:03:21Speaker 00

  But what is the base alone? Like Idiocracy, it's like. The really smart guys that they have at the start, it's like this guy has a hundred and thirty eight IQ. The lady has a hundred and forty eight IQ. And they're like, well, you know, we can't have children and bring them in because it's like a bit of geopolitical tension. You know, we can't really afford to have kids. And then you have like Trayvon and Cletus and others like, oh, man, I knocked you up again. And like the low IQ people are having multiple, multiple kids. But when we actually look at it, humans are incredibly capable. Again, with math academy as a really practical example, I guarantee anyone listening to this, if you give this to your kids, they will do better at math. They will be able to jump years ahead because it adapts dynamically. I think eight year olds can understand quantum mechanics easily.

• 2:04:05Speaker 00

  If we taught them in the right way and school is not the right way to teach them, for example.

• 2:04:09André Duqum?

  Wow. Yeah. You think an eight year old has the base raw cognitive capacity if it was formed and supported.

• 2:04:15Speaker 00

  I know eight year olds that understand quantum mechanics. You know, I've seen them. I have talked to them, you know, I've seen like if you that school gets in the way, education gets in the way. If they're passionate about things, people can do all sorts of things. Alpha school in Austin right now is like teaching kids two hours a day and it's like top percentile in the country or something like that already. So how much do people actually have this if we get out of their way or if we enable them? I think people have a level. Then you've got people like me that are just weird, you know, but you've got the savants. You have people that have extra skills. Again, you have much more musical skill than I have. I have no musical skill whatsoever.

• 2:04:54Speaker 00

  Everyone is different.

• 2:04:55André Duqum?

  Yeah.

• 2:04:55Speaker 00

  The baseline I think is higher than people expect and they are told they cannot create.

• 2:05:01André Duqum?

  What's frustrating about that is, I mean, Howard Gardner has an interesting reference point for his book, The Frames of Mind, and talks about the various different forms of intelligence, whether it's kinesthetic pattern with like musical language, like linguistic forms of intelligence. And I just imagine a world where everyone was supported in their, the unique way in which they are intelligent. Like what a world we would live in.

• 2:05:22Speaker 00

  That's exactly, and that's what one of the most exciting things is. A person plus a hundred and twenty IQ body that's looking out for them.

• 2:05:31Speaker 00

  That centaur, as it were, the human and AI is incredibly powerful. But we've got to make sure they are aligned and they build. And again, I think the base level of cognitive ability is actually very high for humans. We're told we cannot, like, sometimes I give a talk, I rarely give talks. It's like, how many of you, how many of you listening to this are creative? You know, put up your hand. Then I'm like, how many of you believe every child is creative? They mostly put up their hand. Uh, how many of you were children once? Like, most of them put the hand, they're very confused, like, what happens? What happens is we are told we have no agency, no creativity.

• 2:06:09Speaker 00

  If you are not the top in creativity, you're not the most musical person in class, then, you know, the most creative person in class, then you cannot create. Cause school wants you to be a cog. And this is why this technology is so wonderful. Like, again, you go to Suno, you make a song. And you're like, oh, that barrier has disappeared, but you still don't quite believe it. Like, again, the people listening to this would be so resistant to actually trying the technology.

• 2:06:32André Duqum?

  Yeah.

• 2:06:32Speaker 00

  I say, I guarantee you, if you have a party, you bring your buddies around and you start jamming on the video and audio models, you'll have a fun time and you'll create. And you start to change your neuroplasticity as well.

• 2:06:43André Duqum?

  Yeah.

• 2:06:44Speaker 00

  Thinking, I can. The people that are most successful in the next era are the people who believe they have that agency and have a motivation to action.

• 2:06:55André Duqum?

  What should everybody who's listening right now do to prepare for what's coming?

• 2:06:59Speaker 00

  I think it's, if there's one thing I would say, it's an hour every day. It seems quite a lot, maybe start an hour a week because you're scared. Just use the darn tools in a really structured way. And where you can, use them with other people.

• 2:07:15André Duqum?

  List to me your favorite five to 10 tools you think people should be playing with currently, which might very well be outdated a couple of months now.

• 2:07:22Speaker 00

  Uh, there's Notebook LM is my favorite.

• 2:07:24André Duqum?

  Okay.

• 2:07:25Speaker 00

  There's I I agent, agent dot I I dot inc, which is our one.

• 2:07:28André Duqum?

  Say that again, A I agent dot inc?

• 2:07:29Speaker 00

  I agent dot I I dot inc. That's our one, that's gonna go crazy very soon. A tool like Raplet is kind of cool, um, for the programming side. You have Suno or U D O for music. You have, um, Kling or Flora Fauna for video audio. Creer is also very good for that. Um, I think there's some of my favorite ones right now.

• 2:07:55André Duqum?

  Are you using any of the more, like, are you using Claude, Co-Work, Code, or any of that?

• 2:07:59Speaker 00

  Oh, yes. So Claude is, I think, probably the best all-around experience.

• 2:08:03André Duqum?

  Yeah, same.

• 2:08:03Speaker 00

  So I'd recommend using that one. Like, they've really got it down for the multi-turn reasoning. And just like making a document, it's so nice and pleasant to do that.

• 2:08:12André Duqum?

  Like, just yesterday, I got all my blood and stool lab reports back from my doctor. And then we, uh, I just put it into Claude Co-Work with the past five years of labs I got. And I built this full chart with the trajectory of my health and different blood markers. And gave me a very clear plan of what was going on in my body in a way that was, took, you know, five minutes. And it's like, that's the, that's the worst it's gonna get.

• 2:08:35Speaker 00

  Yeah, and you could actually go and you can tell Claude Code, I want a 3D simulation of my body. And I want to see it all live and things like that. And it'll just start building it.

• 2:08:44André Duqum?

  Yeah.

• 2:08:44Speaker 00

  Iteratively. Again, it's wonderful.

• 2:08:46André Duqum?

  Fascinating how we're just like limited in our imagination to, um, to like just start thinking of like doing that. Cause it's, we don't know what it's capable of, but it can do that, which would be sick.

• 2:08:55Speaker 00

  Well, again, that's why I said it's like a muscle.

• 2:08:58André Duqum?

  Yeah.

• 2:08:59Speaker 00

  Like, how do you get good at music? Okay, you can be born good at music, but if you practice for an hour a day, you'll get much better at music. Your neuroplasticity is low because you've taught that you can't have it. Like, when you take psychedelics or when you're in the right mood, your neuroplasticity increases. Right? You're allowed to have more things in your frame of reference.

• 2:09:20André Duqum?

  Yeah.

• 2:09:20Speaker 00

  But you have to adjust that frame of reference. And you see this actually again and again, like, again, we have a new, some new frameworks coming up. Like, chronic pain is a failure of your frame of reference. So just treating it isn't good enough. You see work by that of Sarno and others that treat the frame of reference adjusting. The one that you have to do is that I have agency.

• 2:09:40André Duqum?

  Yeah.

• 2:09:40Speaker 00

  I have ability, but now I have a tool that has more ability than I could ever imagine. But you can only explore the limits of that by actually doing it. Do it until you hit the limit. And if you do that with someone else, it's much more fun.

• 2:09:57Speaker 00

  Because there's always a question of, you hit limit, is it me?

• 2:09:59André Duqum?

  Yeah, yeah.

• 2:10:02André Duqum?

  All right, I have just a rapid fire question, a few questions and then we'll head out. But this has honestly been one of the most intriguing conversations I've had. And is a topic I want to continue to explore on this podcast, so. I am very, you said you're weird. I think you're weird in the bestest way possible. I love the the way you, because you have this really like deep understanding of the current place where we find ourselves in with the development of AI, but then also.

• 2:10:28André Duqum?

  Like I can sense how much you care that we get this right for not only the implications, but like for.

• 2:10:35André Duqum?

  Yeah, it's just, it's moving how, how much you care and spend so much of your time informing individuals on, you know, podcasts and formats like this to support people to be empowered and know what's coming.

• 2:10:48André Duqum?

  So some rapid fire questions, all right. What's, what's one use of AI that's genuinely moved you?

• 2:10:56Speaker 00

  One use of AI that's genuinely moved me.

• 2:11:00Speaker 00

  I think so. Like, once I was having it.

• 2:11:08Speaker 00

  I think actually it's the education thing, like looking at the deployment of the Global X Prize for Learning. Enter camps and now seeing how some of these kids who barely even have power are getting tablets, getting AI and more.

• 2:11:23Speaker 00

  Human rights are the rights of children. I think that's the best, that's my own framework. Like climate, everything, if we think about the rights of children have no agency for themselves, it will resolve themselves. And they must be given access to the tools they need to flourish. And for the first time, we can do that for every single child in the world. I think when I started to see the impact, the photos of that and the shifts, I'm like. That is one of the biggest things in humanity ever, because we can make sure no child is actually left behind or invisible again.

• 2:11:55André Duqum?

  Amazing.

• 2:11:56Speaker 00

  I see the stuff they create is just awesome.

• 2:11:58André Duqum?

  No, it's so cool. Not to make it dark immediately, but what's one use of AI that's genuinely disturbed you?

• 2:12:06Speaker 00

  One use of AI that's genuinely disturbed me?

• 2:12:11Speaker 00

  Gosh, there's so many.

• 2:12:14Speaker 00

  How many of them can I talk about? I think, like, bringing back people from the dead. Sometimes it like is really interesting for me.

• 2:12:27Speaker 00

  But I've seen people start do that in a very manipulative way.

• 2:12:30André Duqum?

  Can you explain that for people? Because that's quite a sentence. What do you mean by bringing people back from the dead?

• 2:12:35Speaker 00

  So you can create a visual simulacrum of them. And like,

• 2:12:42Speaker 00

  there are services now that can bring back your dead loved ones and create that visual simulacrum that you can talk to and more. And part of me, like, when it's public figures, I feel, okay, that's fine. But then when I see and I've talked to some people who've done that with their loved ones and I'm starting chatting to them just after they've passed, something in me feels uneasy and I'm not able to articulate it just yet. There's something about the closure process

• 2:13:10André Duqum?

  Yeah.

• 2:13:11Speaker 00

  of moving on. And it feels like they're never going to be able to move on when I see and discuss with them that.

• 2:13:22André Duqum?

  I mean, what I would articulate and why that's disturbing is because we have this whole human emotional landscape that was tended to in a certain way that gave the necessary space it needed to. And

• 2:13:38André Duqum?

  when sitting around the campfire is

• 2:13:44André Duqum?

  swapped out for

• 2:13:48André Duqum?

  A lifeless AI model or genuine sex and connection with intimacy with the partner swapped out for an AI sex robot, or the many different ways in which we're swapped out the real thing that is actually nursing to the human spirit with the

• 2:14:05Speaker 00

  Yeah.

• 2:14:06André Duqum?

  The artificial one, you know, there's something like deeply irking, I feel like about missing what it means to be human in that process and

• 2:14:16Speaker 00

  Yeah, I think that for me, the, I feel different things with different elements of what you say there. The thing that like I said, what made me feel most uneasy is the process of grieving

• 2:14:28André Duqum?

  Yeah.

• 2:14:29Speaker 00

  Over the past, like again, I think it's a healthy process. Death is a part of life, maybe it won't be soon. That feels disrupted massively from this. And so like, I don't know why, but that made me feel a bit more uneasy than anything. Like in public figures, you don't have an emotional connection to the same degree. They haven't been a part of your life.

• 2:14:47André Duqum?

  Yeah.

• 2:14:48Speaker 00

  It's almost like, again, something has substituted for that real human right next to you. And I don't think, again, they will ever be able to turn that off. They'll never be able to go through, but at the same time, they're feeling support. So I feel the positive, again, it caused a bit of a conundrum within me. I haven't been able to unravel that just yet. And then there's all that other stuff that you just talked about, which all has its own value.

• 2:15:12André Duqum?

  Yeah, yeah.

• 2:15:13Speaker 00

  I can see the positive, I can see the negatives, or that's just kind of weird.

• 2:15:16André Duqum?

  Yeah.

• 2:15:16Speaker 00

  You know, like kind of weird stuff, you're like, all right, sure, you know, you, you, you do you.

• 2:15:21André Duqum?

  Yeah, not a priority, but interestingly weird.

• 2:15:24Speaker 00

  Again, there's that stuff that makes you feel genuinely uneasy.

• 2:15:28André Duqum?

  Yeah, now I'm curious now what, what you think about when you watch Pantheon, because they explore that exact thing with someone's dad who passed away and they uploaded his intelligence.

• 2:15:39Speaker 00

  Check it out, now check it out.

• 2:15:40André Duqum?

  Mute it, mute it.

• 2:15:41Speaker 00

  Yeah.

• 2:15:43André Duqum?

  What's something you're genuinely convinced AI will never be able to do?

• 2:15:51Speaker 00

  I can't think of anything on that one. Like, again, once it has, once it has qualia consciousness, like, you know, holding my daughter's hand and seeing her walk for the first time, I think an AI can feel that. Like, what is the, I, I'm not sure what the difference between an AI and a human is 20 years out, shall we say, apart from species genetics.

• 2:16:19André Duqum?

  Well, the most unlikely thing would be that there would be a genuine quality to the internal experience of an AGI, right? That would be the most unlikely or probably hardest thing to bridge.

• 2:16:30Speaker 00

  Hardest is a never.

• 2:16:32André Duqum?

  Yeah, yeah.

• 2:16:33Speaker 00

  That's the thing. And again, I think once you look at thermodynamic computing, the ability to grow cells, like, you already have seen brain cells learn to play Doom in a Vat recently. That's kind of crazy. You know, again, it's input stimulus, kind of things like that. Like, I can't say never to things. And this is why, again, I always get a bit surprised when I'm told to feel like, it will never take my job as a program. Of course it will, you know, like, then you move up the list. It's like, you finally get to the things that you can re-argue about is, well, it can never be human.

• 2:17:09André Duqum?

  Right. And you're like, okay, what does it mean to be human?

• 2:17:12Speaker 00

  Like, literally, that's the final thing that you kind of get to. What does it mean to be human? Like, can it, it will never be able to build a warp drive. I'm pretty sure about that. There are certain things that I don't think can violate the laws of physics, like some people say, it might be able to.

• 2:17:26André Duqum?

  Terrence McKenna says that something like, uh, the progressions in this space is going to get weirder and weirder and weirder until we're going to start having to talk about how weird it is. What do you think is going to be the first thing that really shakes people up? Obviously, we're already seeing interesting things with OpenClon, MultBook and whatnot, but what, like, what would be the biggest domino in that space?

• 2:17:50Speaker 00

  So, humans are great at hedonic adaptation, right?

• 2:17:53André Duqum?

  Yeah.

• 2:17:54Speaker 00

  Like, the first time you get into a Waymo, you're like, okay, is it going to crash? Like, how close is it to that cyclist? Second time, you're like, all right, cool. You're just done within one go. First time you type words and it makes an image, you're like, wow. Then you're like, okay, cool, I'm done. I think that the biggest kind of gap at the moment is that AI can't make discoveries or novel things, which I think already is a bit BS.

• 2:18:20André Duqum?

  Right, so the self recursive learning aspect of the software.

• 2:18:24Speaker 00

  Yeah, or just originality. Like, again, I see a level of compositional originality to the level of a K-pop group already, you know, but then, once it starts to push boundaries, humans will never make discoveries again.

• 2:18:38Speaker 00

  We're approaching the last human-only discoveries because AI can auto-formalize math proofs, AI can auto-do entire like biological things. Like, I was one of the authors on OpenFold, the open-source version of AlphaFold, and I gave all the compute for that as well. Protein folding, like, humans aren't gonna come up with compounds anymore, AIs are gonna get it the whole way. And so one of the weird things is that AI does all the research of humanity and pushes it forward. The last few human-only breakthroughs are literally in the next few years, and that's it. That's weird.

• 2:19:11André Duqum?

  There's something so existentially strange about the fact that we are potentially the only generation that will have straddled both the pre-internet and post-AGI world.

• 2:19:22Speaker 00

  Yeah.

• 2:19:23André Duqum?

  It's quite amazing. Like, I think it's really cool that we were able to ride bikes as kids without phones, and then we're also gonna live in a post-AGI world.

• 2:19:34Speaker 00

  Again, what does it look like? Like, my daughter just turned one, right? And I'm like, what's she gonna do when she's 20? What does the world even look like?

• 2:19:42André Duqum?

  Yeah.

• 2:19:43Speaker 00

  But at the same time, I will teach her to ride a bike and ride a bike with her. That's fantastic. At the same time, you know, we have our human connections, so certain things will never change. But at the same time, the leap in technology that we're about to have across everything, and we haven't even discussed quantum computation, genetic altering, all these things, is more than I think we've ever seen in humanity, and it's the last change.

• 2:20:11Speaker 00

  This is the last change. That's why in my book, I called it the last economy. Why? Because economics ends as we know it with the advance of AI, just like democracy as we know it ends. Something else comes, but not what we knew before.

• 2:20:26Speaker 02

  Low energy, unfocused, foggy? You might be dehydrated. Gatorade Zero powders have been scientifically designed to help improve hydration balance in the body with a clinically proven electrolyte blend and no artificial colors or flavors. Great tasting hydration, no matter where you're going. Shop now online or at retailers nationwide. Gatorade hydrates better than water. Is it in you? Symptoms noted are signs of mild to moderate dehydration. Consult a healthcare professional if symptoms persist.

• 2:20:56Speaker 03

  Whenever fall comes back around, it always feels like a fresh start. This year, I've been layering in some styles from the new Lululemon fall drop because the weather is shifting and I find I have to be ready for the entire day and not just one part of it. Lululemon's fall line has some really versatile transitional pieces, so I can just wake up, throw on a matching set, add a jacket or a fleece, and head out, knowing that I'll feel comfortable enough to go wherever the day takes me. Check out the new fall arrivals in store and online now at lululemon.com.

• 2:21:29André Duqum?

  What should parents be telling their kids right now?

• 2:21:32Speaker 00

  They should tell their kids to have as much agency and belief in themselves as possible because kids should be taught to have that belief. And now they have the tools that they can do anything they can imagine. You're always limited in your own capabilities until you realize one thing. If you convince someone else with that ability, then you can follow their abilities. You can lead. But once we get to a dozen people, our communication starts to break down. Once we get to 150 people, it completely breaks down usually within organizations. Organizations ossify very quickly. With AIs, that doesn't happen. You can have millions and armies of AIs soon that can bring into being anything you can imagine. They can even call up suppliers and hire people on your behalf to build physical things.

• 2:22:17Speaker 00

  How cool is that? But only if you know that you can do it and you wanna push the boundaries of that.

• 2:22:22André Duqum?

  Yeah, I know, it's equally parts exciting and thrilling and terrifying.

• 2:22:27Speaker 00

  Well, again, I said earlier, it's about skinning the game and actually giving a damn.

• 2:22:30André Duqum?

  Yeah.

• 2:22:31Speaker 00

  AI does not give a damn. ChatGPT doesn't give a damn about you. A human using the AI can give a damn and affect massive change anywhere because finally you have the lever to do that. And that massive change might just be having fun and creating.

• 2:22:43André Duqum?

  Yeah.

• 2:22:44Speaker 00

  Yeah, or it might be solving an issue. You don't have to solve issues all the time and have responsibility. Part of what it is to be human is to explore. It's to understand, it's to share. And again, we have to go back towards that as well. Because if to be human is to be the smartest entities on earth that push the boundaries of discovery. Okay, that's about to go. Last few humanoid discoveries are now.

• 2:23:13Speaker 00

  But the stories that you tell each other don't go. The time you spend with your family doesn't go. But again, just like you can make me, if you make all your meals together, you'll be a stronger family, relatively speaking. Not only does it solve all the issues. If you jam with AI and you create together, the act of creation. How many families listening to this? When was the last time you created something with your kids? I imagine nobody can even remember that for most of the families. Go and create something with your kids. Show them that they have agency, respect their input as well as your input and jam with them.

• 2:23:47André Duqum?

  Yeah. So well said. I feel like so many of us, the more that this agentic world comes to fruition, we yearn. There's something in the human spirit that yearns for the making of a fire, the feet on the dirt, the making things with your, cooking a meal with your family. Just the in-person community aspect.

• 2:24:07Speaker 00

  Yeah, like if you wanna make an AI, the AI that makes you to kids, right? Like you're training them, you're passing down to them. But we're so busy with our lives that we don't even think about it. Like, most people say, have you done your homework? When was the last joint activity that you did with your kids? That wasn't like going to a movie. That's not a joint activity. A joint activity is an act of creation. And again, what can we do as humans? We can create. Like even if the AI puts us into this self automated luxury communism, where it looks after all of our things, does that take away from our ability to create? No, it does not. That is one of the things of being human. Now, can the AI also create? Sure, fine, whatever. But does it create things for other AIs?

• 2:24:52Speaker 00

  No, we create things for other humans because our meaning is from our interaction with other humans.

• 2:24:57André Duqum?

  Yeah.

• 2:24:59André Duqum?

  And the creative act itself, no matter how how much capacity we have to an AGI. I mean, nothing's going to steal my joy from sitting down at a piano and playing or going on a walk with a friend, you know?

• 2:25:11Speaker 00

  Yeah, well, again, and again, there's doing it yourself. And then there's the sharing of it with the world, right?

• 2:25:17André Duqum?

  Yeah.

• 2:25:17Speaker 00

  Like, we're inherently pro-social species.

• 2:25:20André Duqum?

  Yeah.

• 2:25:20Speaker 00

  And the ability to tell the stories, the ability to change people, it's a moving thing.

• 2:25:27André Duqum?

  Yeah. Last, very last couple ones, since you mentioned the quantum processing, quantum computers, the CRISPR technologies, what's one out of the box innovation you think might be coming in the next 10 years in that space? And what becomes possible when those two come together?

• 2:25:43Speaker 00

  So I think quantum supremacy is probably four or five years away.

• 2:25:46André Duqum?

  What does that mean?

• 2:25:47Speaker 00

  It means it's strictly better than pretty much all AI, computers, conventional ones. It's interesting because, like, again, the public narrative is that, oh, a huge amounts of computers organize all this stuff. That's not how things work. If you have a thousand agents operating on a problem, that's pretty normal right now. You're not going to do it faster than one agent. We're seeing that already in the numbers. Just like if you throw a thousand programmers at something, you're not necessarily going to get it done faster than a few programmers. Intelligence is compression and AIs learn to do it correctly. The problem with quantum computers is that you're not asking the right questions. The question is actually quite hard to formulate because you need to understand how everything works. AI can help with that. But then a quantum computer is not going to have to think for a year.

• 2:26:32Speaker 00

  Usually, everything is like that, once you ask the right questions. And so we haven't really got that baked in. The fact that almost any question that is reasonably solvable can be solved within a second. That's crazy.

• 2:26:48André Duqum?

  Yeah.

• 2:26:50Speaker 00

  Stuff that was literally impossible can now be done like. And we've seen that already with the generative area web like again, the alpha fold, all that kind of stuff we did. It would be impossible, we would take in millions of years to do some of these calculations. And I was like, that.

• 2:27:03André Duqum?

  Yeah. If you could give everybody on earth one AI capability tomorrow, what would it be?

• 2:27:09Speaker 00

  One AI capability? I think probably retrieval. So AI is really good at retrieving stuff in the right context.

• 2:27:18André Duqum?

  Yeah.

• 2:27:18Speaker 00

  So I think that would be a super useful thing. Cause we always forget stuff, you know? And having that will allow us to ground better because we could retrieve. These are the things that we've learned that makes me feel that we stick to that.

• 2:27:31André Duqum?

  Yeah.

• 2:27:33André Duqum?

  What's one question you're tired of people asking you?

• 2:27:36Speaker 00

  I don't know, I don't really get tired or upset or frustrated about anything.

• 2:27:38André Duqum?

  That's awesome.

• 2:27:39Speaker 00

  Pretty chill.

• 2:27:40André Duqum?

  What's one question you wished people asked you more?

• 2:27:44Speaker 00

  Um, I think probably how to be happier.

• 2:27:47André Duqum?

  How to be happier?

• 2:27:49Speaker 00

  I think that happiness again comes from the act of creation, letting go of a lot of your assumptions and realizing that we're the luckiest people at the luckiest time in the world. If we can do what we like, what we're good at, and where we're adding value and believe we can, Japanese concept of higai, then the middle of that is happiness.

• 2:28:06André Duqum?

  Yeah.

• 2:28:07Speaker 00

  And we underestimate our ability to add value to other people. We get in our own way around a lot of that. And so, once you understand that you can do that and your own happiness is your own internal state and no one else can determine it, then you can be chill and you can be happy. And it's good, it's nice being happy.

• 2:28:26André Duqum?

  I quite enjoy it and conversations like this make me very much so happy. So I want to end it there. Uh, Ahmad, thank you so much, man. It's been a pleasure to connect with you. Very last thing, we were talking beforehand, do you want to give us a one sentence sneak peek on the sort of epistemology and philosophical explorations you're currently exploring?

• 2:28:48Speaker 00

  So we're looking at how do you ground consistent reasoning and what is the epistemology of AI like again? We need these things to be aligned and we can't do it through classical rules. So is there something more fundamental that we can go to so that we can all be on the same page? And I think we've gone a long way to that and hopefully something to be announced soon.

• 2:29:07André Duqum?

  Yeah, no, what you're showing is, we'll have to have you back on talking about that because it's uh. When you refer to the importance of first principles thinking and being able to train these self recursive learning models, I think what you were showing me beforehand and like these new discoveries are going to be pivotal for having an effective model that's aligned with us, right, and in our best interest.

• 2:29:34Speaker 00

  Fingers crossed. I'm very hopeful that again, as we have these AIs, as we all work together, we can figure out some actual fundamental truths that have eluded us for a long time and then use that to all pull together to explore the universe, to explore ourselves, and again, increase happiness.

• 2:29:54André Duqum?

  Matt, thank you so much. We'll leave links down in the description for your new upcoming book, The Last Economy, where they can stay connected with your work. Any other last words?

• 2:30:02Speaker 00

  No, please come and download it. It's all freely available and we're going to open source everything. And again, please spend some time with your family, your friends creating.

• 2:30:12André Duqum?

  Yeah.

• 2:30:14Speaker 00

  Do it. Like, what's getting in the way? Should do it right now.

• 2:30:17André Duqum?

  Yeah. Well, everyone, I hope you enjoyed this podcast. I know I did. Maybe go stand on some grass. I think I'm about to. Until next time, be well.

• 2:30:26Speaker 00

  Cool. That's fine.