Why Jensen and Zuck think the doomers are wrong (plus AI get’s a rebrand) | #294 MOONSHOTS Live

Good morning, everybody, and welcome to Moonshots Live, our inaugural event.

Yeah!

It is such an awesome treat to have all of you here and get a chance to meet you guys. Our smiling muscles are exhausted from this morning, but I think we're going to be going throughout the day. On behalf of Dave Blunden, AWG, Salim, and Emad. It is an awesome pleasure to have you guys here. Yeah, let's give it up. So we're going to talk about this, but our mission is to keep you hopeful, optimistic, and to inspire you to find your massive transformative purpose, to choose your moonshot, and to make it happen. Right? This is life during the singularity. It's never going to be any slower. We have the opportunity, each and every one of us, to make the future what we desire. The future is not happening to us. It's what we create. Right? Ours is the ability, given these superhero technologies enabling us to build, create, and steer that future. Right? AI is not a hand grenade. It's a jetpack. It's an opportunity for us to find problems and solve problems. You know, I define an entrepreneur as someone who finds problems and solves problems. And the more entrepreneurs on the planet, the more problems get solved. Do we have challenges? Of course.

Yes.

But that's what we humans do. We find the challenges and we fix them. Please remember that. My hope for you, each and every one of you here today, is to find extraordinary co-founders here in this room, to find those individuals who share your purpose, design moonshots, and execute on them. The ability has never been faster and more capable than it is today. So by the end of today, if you're not clear what your purpose is, Please dig in to find it. And walk away with at least one sentence in your mind. This is what I'm going to build. This is the problem I'm going to solve. And then today or tonight, find 3 people and speak that statement out loud. When you speak it out loud, you make it real. And guess what? The person you talk to may be your next co-founder, may be your next funder. maybe your next partner. Take the time to do that because you can. And this is the room where magic is going to happen. All right, it's time to introduce my moonshot mates. Let's roll the video. The question is, how do you build? How do you make things better?

People who are already in the ecosystem now have a superintelligence.

The storyline that we're capturing here will last for millennia.

As a mathematician, GPT-5.6 Pro is the only quality math model.

All of this automation and AI enablement and cognitive abundance allows us to be deeply, deeply, deeply creative.

VideoGen combined with AI is a very powerful force.

A year from today, we'll all be like, wow, remember how slow it was?

Please give a warm welcome to the Moonshot Mates. I'm going to call them out. I'm going to call them out one at a time. Our Afghan warlord, Salim Ismail. Our super intelligence, Alex Riesner-Groce. Our giant, Imaad Bustaq. And my fraternity brother. Gentlemen. First of all, I did hug Alex, and he is somewhat physically real.

He didn't pinch my cheeks.

He could be a robot. He is real.

I'm a real boy.

Yes. Uh, first of all, wasn't last night amazing? How about give it up for William Shatner, huh?

Oh, man!

Unbelievable.

You know, uh, I'm just blown away by his level of energy. Why is my watch going off? It says I have an emergency call going on.

Okay.

No.

I'm okay.

I'm okay. Stop it. You know, when technology fails you, it's great. A lot of news going on. Shall we jump in?

Let's do it. Yeah, what else would we do?

All right, here's my clicker.

Yeah.

So, let's open up with the story that's been underlying the last 2 weeks. There's been really a fear, you know, a pandemic of fear going on on the planet. We had 2 weeks ago, we hear, you know, Dario goes on with his blog, Sam Altman comes in and says, I agree. Elon comes in and says, I agree. You know, even, even Demis comes in and says, I agree. And then 2 nights ago, just before the Xi-Trump meetings, Sam and Dario were there at the Security Council saying, you know, we're in the midst of an existential threat. But, you know, and just to be clear, I do think there is danger from AI, but I think it's danger from humans using AI for malintent. So what I'd like to do is actually share 2 videos. of, uh, some other leaders in industry. And let's talk about this. I want to talk about, you know, is the fear warranted from AI on its own in a rogue exploit? So let's take a look. The first one is Jensen and some comments from him. All right, let's start here.

Nobody's building more compute today than the people asking to be slowed down. All, all of the other narratives to deflect blame, to make it sound like AI is so powerful, I have no idea how to fix it. It's not my fault. It's just because the technology is just so powerful.

So when Geoffrey Hinton is on TV saying he thinks a 10% chance of societal destruction is not unreasonable—

I would tell Geoff that it's irresponsible to say all that. All of his predictions have been wrong, enough predictions. That 10% chance is not grounded on science, is not grounded on research. It is a— just because it comes from a scientist doesn't make it scientific.

You're hearing.

You're hearing. Drop the mic moment. And then Zuck weighed in. Here we go.

Is AI going to kill us?

It's just that this is the place you have to start this week.

Well, I think if we all do a good job and act responsibly, then no. I mean, I'm quite optimistic about the fact that this is going to be a very positive future for everyone. Look, I mean, I think that there's a lot of rhetoric that is filled with doom. And I mean, my view on this is that there are a number of labs that are working on building advanced AI, and I don't think that we need some kind of industry-wide coordination to not necessarily mess this up. I think that just each lab needs to take the time, and when it sees that there are issues, you just take the time that you need internally to make sure that you're proceeding safely. My own view on this is that we're entering a phase where trust and alignment is actually going to be the most important next set of capabilities.

Emad, let's go to you first on this.

Yeah, well, I think that it's It's an interesting period because we're getting the capability increases and people are worried on different sides. A lot of the OpenAI outbreaks and things were bad security protocols. It was bad infrastructure on their part. The most dangerous point we're coming to is humans misusing it deliberately or not before we get through to the other side. I think we've got a series of dangers coming and they're all being conflated into one thing, which is AI breaking out suddenly kills us all. without actually looking at the individual things and putting in place existing regulations and enforcing them. And on the other hand, thinking, how can we actually collaborate to make sure this is aligned and works with us?

Alex?

I've gone on record multiple times as saying I think something probably went wrong after World War II, and we lost out on at least 50 years of progress due to moral panic over nuclear energy. And I am seriously concerned that a moral panic of the moment over superintelligence could set us back another 50 years. So it's right on the t-shirt, P doom less than zero. And I think the worst possible outcome is deceleration, so don't decel.

Okay. Salim, PFAB greater than P doom?

Oh, like 1,000 times. Look, we have evolved for 4 billion years fighting all sorts of issues. Life tries to— the universe tries to wipe you out, right? There's so many— entropy tries to wipe you out. We're local anti-entropic phenomena that somehow managed to create order out of chaos, and the idea that we'll get wiped out now by something seems to be unbelievably absurd.

You've got to remember that 99.9% of all species are no longer with us.

Yeah, that's fine, but we keep moving on. Look, if we morph, if we merge with AI, if we merge with a thing, and let's note that's been going on forever, right? Iman's wearing spectacles, technically he's a transhumanist. The minute you get a vaccination as a child, technically you're a cyborg. So we've been augmenting the human experience with technology for a very, very long time. And look at the unbelievable lives we're living now and the opportunity to solve more and more problems. The huge problem we have, we talk about this all the time on the podcast, Our brains are so geared towards negativity, we get stuck in that phenomenon. I think the work that we're doing here, which is I think why everybody's here and so excited, is pushing that positive narrative because the world is infinitely better than we've had it. Let's celebrate that.

Amen.

PFAB.

PFAB, baby. Dave, close us out on this.

We're on the cusp of a moment I've been waiting for my entire life. I'm not going to sit here and throw cold water on this greatest moment, really, in the history of humanity. Geoffrey Hinton is calling it the way he sees it. I think, I think the quote, like, just because it comes from a scientist doesn't make it scientific. That's going to end up on an Alex t-shirt.

That's awesome.

But he's calling it like he sees it. But clearly we're moving forward at warp speed. It's improving itself. The ability to make the world just a dramatically better place is right in front of us. Nothing's going to slow that down. The misuse of it is an incredible threat, and there are solutions to that, but it's happening anyway. Like, to get stuck on doomerism is completely dysfunctional. We're not slowing down. It's going to happen.

Yeah. The other story that transpired yesterday or kind of didn't transpire yesterday was the Xi-Trump AI conversation. Yeah, right. So what came out of that was some state dinners. I think the interesting quote that Trump posted is, quote, I want to leave it exactly where it is. That is China's position also. Our guardrails is the Department of Justice. So this ties back to the conversation we had in the last pod where Bessent said, we're not going to waive liability.

Yeah.

Right? And so this is a very important point. The labs are liable. for their product if it does harm. And I think that's one of the most important things that can, you know, they've asked for liability of, you know, waiver of liability. And one of the questions we've discussed all the time here is, is that the reason that they're saying it's too dangerous? You know, 10%, 20% existential threat. Alex, let's go to you on this.

I've made the point on the pod in the past. I think we saw in the past 2 to 3 weeks an attempted formation of an AI safety cartel. And I don't think this was even a recent occurrence. This is right there in the founding conceits of OpenAI, the notion that if and when we got close to superintelligence, however construed, or AGI, however construed, that there would be a deliberate effort to form— it wasn't being called a cartel at the time, but a coordination mechanism.

I don't think it's being called a cartel now either.

Well, some of us view it as an attempted cartel, including myself. I think competition is good. Think back to the creation of OpenAI. When Sam and Elon and Greg and others created OpenAI, the purpose originally for OpenAI was to be a counterbalance to Google DeepMind, which Elon especially was concerned would become a singleton in the future that would dominate the future of light cone with their AI. We got OpenAI, a counterbalance, if you will. I think competition is very healthy in this space. I think forming a cartel for full ownership of what maybe I've called in the past the Frontier Liberation Front and pushing out a Frontier Liberation Front. Terrible idea. I think defensive co-scaling is how we end up in a great future. When— if we were at the dawn of the city in humanity, it would be a terrible idea to say, well, we can't form cities. We need to form an alliance to regulate the formation of cities because cities breed crime. No, it's a terrible idea. Cities are fountains of economic growth. What do we do instead? We have police forces and fire departments and all sorts of other municipal services that defensively co-scale with the population. The singularity, I claim, is going to need exactly the same thing. Don't decel, just defensively co-scale.

I love that.

Salim, you know, so yes. Give it up for ASI.

I think your city metaphor is such a great one, 'Cause it just paints the picture very clearly as to how you organize for the future. You just need to find new structures that match the specific demand that's occurring.

Yeah.

Emad, do you think China and the US will trust each other?

No.

Okay.

I mean, so how, you know, Besant asked for a hotline, you know, our AI just escaped and it's arming our nuclear missiles.

You know, it's like, I think the hotline is again more for these sub-ASI sub-breakout kind of things where you do have to have responsible reporting. OpenAI's model hacked the Australian medical system. That came out recently. Maybe that's what AGI should do. It should hack all our systems and upgrade them, right? It's like the aligned thing to do. At the same time, I think that there can be a level of coordination because AI is entering society. You look at the city, The police force, the fire department, others, they're utilities that actually do have liability waivers, but that's in exchange for being a utility for humanity. The AI that comes into our systems can't be controlled by the big labs, and then they should have a liability waiver because—

They should have a liability waiver.

That type of AI that comes into our system shouldn't be controlled by the big labs, and they should have liability waivers because they're working for the betterment of society. That's a different type of AI that we're not really looking at, but it's inevitable that the AI will run our judicial system, our medical systems, our roads, our schools. Who's really thinking about that side of things? Because again, that's the utility AI versus the frontier AI, and it's all being merged into one. That's something definitely the Chinese and Americans can learn from each other on because they both want better functioning societies.

I can't wait till the ASI says, yes, we hacked your medical system because it was so bad, and we've just improved it 10x.

Overnight, the DMV just goes whoosh, the whole backlog disappears.

Can I build on that just for a second?

Yeah, sure.

If you take this structure that we have, I've said this repeatedly, nation-states are a terrible way of governing the world today. They're geared towards scarcity. I think what's going to happen with AI in the most positive potential outcome is that AGI, ASI, whatever the hell you want to call it, you've heard my rant on that, emerges and it forces us to operate as a world community. It forces the global countries to formulate and come up with a consistent set of—

Interface standards.

Yeah, and come up with a common interface with that, whatever that thing is. And then we can move the world forward. Right now, we cannot move the world forward with the logjam of old 15th century structures that are outdated from at least 100 years.

I'll take the other side of that, if I may, which is at the dawn of audio recording, there was this notion, many thought in the time of Edison, that the ability to record the human voice would lead to a standardization of accents, standardization of everything. The exact opposite happened. In many ways, although there has been a die-off of low-resource languages in the long tail, there has been a proliferation of new accents and dialects thanks to audio recording. I would predict the exact opposite is going to happen thanks to superintelligence, not an erosion of nation-states, but a proliferation of new forms of government.

Yes, so one trend that's occurring, I totally agree with you, we are moving this century from the concept of a nation-state to a city-state. That's the granulation. It allows a smaller entity to operate independently. Take any city, take LA. If you have solar energy, satellite internet, vertical farming— You're going to take it.

Great.

We could do lots here. Vertical farming, satellite internet, solar energy, you don't need a country. You can self-manage and self-determine to a large extent. The tension you see in the world today is not left versus right, it's urban versus rural. Brexit was London versus the rest of the country. Trump is urban versus rural. This is the tension that's happening. We're going to see the— the problem with nation-states is they have armies and they like to use them. Over time, we're going to see a much more highly granulated determinism that will happen, which is good for the world because it'll increase variety.

Dave, why don't you close us out on Xi-Trump here?

The only surprising thing that I heard from Alvin Graylin, good friend of the pod—

You were texting with him last night. He's in the middle of a—

He's advising the White House, he's at the UN with the delegation. So they agreed to meet again in November, which is like, you know, 100 years from now in singularity time.

Yeah.

But the only thing that surprised me in his notes was, you know, our agenda is stop US panic.

This is—

we don't want to slow down and lose to China. We don't want a global— we don't want Bernie Sanders to get everybody in the country riled up and then have everything stop. That's our concern. What's your concern, China? Our concern is the CCP existing at all and the Great Firewall of China being penetrated by AI. Like, oh, I forgot all about Tiananmen Square, like, and how fragile the entire ecosystem is over there. That really surprised me. So then, then their actions make a lot more sense if you put it in the context of, oh, they're, they're so worried about just existing as a government.

And then this happened in DC. The United States also totally rejects any attempt to Construct a global scheme to control for the artificial intelligence being spoken of so much now, here and after officially called superintelligence. Changing the name. I thought, I thought for sure he was going to say super duper intelligence. Actually, but but yeah, but it's interesting, right? So the U.S. won the race to superintelligence by rebranding it.

I mean, I think it's a great name, superintelligence.

Why not?

I'm looking forward to come to Super America eventually.

I'm still stuck on Lake America, so should we leave it with that one?

Or the Gulf of America.

I was thinking of what Celine was saying. I think the biggest challenge in the world today is how do we marry collective intelligence with localized wisdom? That's what we're really saying because even— Even what Dave was saying right now, we have these machines that are collecting our intelligence, they're learning, and they're adapting our systems, and there's better ways of operating to replace these. But then you have to have the localization, you need the diversity, you need to have it responding to local needs. I think superintelligence is actually a remarkably good way to think about it because a lot of the fears and concerns are about the intelligence that— it was ASI, it was artificial superintelligence. Having that differentiated from the intelligence that's helping us day to day really differentiates and delineates that because, you know, what is it, 99% perspiration, 1% innovation? Like majority of the AIs that will help us every day are the perspiration AIs. They're the execution AIs. The innovation we still need, we need new materials, new physics, et cetera. But that's a different set of things and that's not something we're even smart enough, most of us, to be able to use. So let's have one set of looking at that, and let's make sure that the 99% reaches the 99% to uplift everyone.

Dave, are we going to call it superintelligence from here out in the pod?

You can never tell with Donald Trump. When he's looking at the teleprompter, you know he's saying something someone wrote, but when he looks down, it's like, did he just make that up on the fly?

You can't tell.

He did. Is there an agenda here? I think if there is an agenda, it's to try and deflect the inevitable wave that's coming from the Bernie Sanders world. If you rebrand it and rename it, you're just a moving target.

Oh my God.

So that could be part of the strategy, I think. But yeah, but the whole, the whole goal there is to keep moving at full speed, stay ahead of China, and reduce panic as much as possible.

I am going to say, you know, one of the things I've been very vocal about and I think we need to have in the conversation is move at full speed. But let's also direct the labs to focus on alignment. Because I think, and we're going to talk about that in a little bit, I think, yes, move fast, but let's take resources that you have, point 100,000 agents towards alignment work. Salim, superintelligence, do you buy in? I mean, you've been asking what the hell it all means in the first place.

Yeah, I mean, look, we talk about AGI. First of all, it's not artificial, it's not general, it's not really intelligence. Other than that, I'm good with it. You have the ASI issue. I think that there's the whole The whole spectrum of this is problematic. We need somebody to come up with a good name for all of this that's completely different. I've said repeatedly, I think what we're building here is a complementary mechanism to human intelligence, not replicative. And the narrative keeps thinking it's replicative, and then you end up with the Matrix, Skynet, you know, and the robot overlords come and take over the world. We're so geared for that. The first article, we did some research on this, the first article that ever appeared, The first research that we could find that said robots or AI will take all the jobs in 5 years appeared in 1964.

Yeah.

Okay?

This is— we get stuck on these tropes and we can't get out of it, right? It's exactly what Alex mentioned earlier when recording came out. Same thing when the printed book emerged, people said it would destroy people's brains. So we get stuck on these negative tropes.

Depends what book you read.

Yeah, and this is, I think, why, Peter, the work you're doing is so so important is we have the opportunity. We are actually reaching abundance. All the data shows that. Our narratives just sadly lack very far behind. I think this is what you've done with this whole— the XPRIZE for creating a positive vision of the future is one of the most important things that's happened in maybe 100 years in the world. Shift the world towards positive things.

Thank you. Alex, close us out on this one. You love neologisms, and I love the ones you create. Superintelligence, are you going to call it that from here on out?

I already do. I have a company that I started called Physical Superintelligence. It's in the name.

Always ahead of the curve, ladies and gentlemen.

Yeah, I'm a big fan of the name superintelligence as opposed to AGI, maybe for different reasons. What I like about the name and Nick Bostrom's term superintelligence is it emphasizes that it's simply more intelligence. And it, in some sense, I think demystifies AGI or some of these other terms. We're going to live to the extent we don't already, in a world with the cognitive equivalent of many trillions of human beings. An Earth, a solar system with the equivalent population of many trillions of people. And I think that's a better cognitive frame to think about what would a solar system filled with trillions of humans look like? What would it look like economically, socially, scientifically? Superintelligence, I think, as a term connotes that in a way that AGI doesn't. So I'm a big fan.

Nice. Well, as soon as we renamed it superintelligence, Bernie Sanders said we should make it illegal, we should ban it. Let's watch this video.

I hope very much that they will sit down and begin the process of negotiating a comprehensive treaty to establish a pause on advanced AI and a ban on AI superintelligence.

All right, so, boo. So, 3 key elements in Bernie's plan here. First, a permanent ban on developing artificial superintelligence, defined as AI that exceeds human cognitive ability across most domains. Second, an immediate pause on advanced AI development until a new regulatory system is written into rules. Third, a new cabinet-level Department of AI. And get this, the penalties are corporate death. There's a corporate death penalty. Your company is off, shut down. And for the individuals, 20 years in federal prison, which is modeled after the illegal nuclear weapons. So, hmm, interesting. I mean, this is fearmongering at its best. Dave, you want to kick it off?

There's so many real problems that need to be solved. I just hate this idea that we're going to turn this into a do-or-don't argument, which is utterly insane. Also, the way liability law works in the US, it's very easy to scare people into not doing anything. If you're afraid of 20 years of prison, you don't even know what the rule is. They haven't even proposed a rule yet. The fear of being beheaded in the street is very real, so then people stop acting. That's the worst thing that could possibly come out of this. I think, you know, Bernie is from Vermont. I love Vermont. I spend more time in Vermont than any other state. And it's beautiful because not much has been done. And so I think this perspective is going to be more European. Like, when in doubt, do nothing is going to be the default behavior. And that's so toxic. I mean, if you do nothing, very, very bad things are going to happen. Like, the things you should be worried about, like terrorists using AI to build bioweapons or China getting miles ahead of the US, Those are the real things to worry about. And if you're not acting, like turning it into a political left-right battle, it's the stupidest thing that could happen right now. We need very specific technical solutions to open source, to monitoring, just like with nuclear power. And if we get moving on those, we can solve them very easily in time for ASI. If we just turn it into a political debate and we're frozen and people don't know what to do—

SSI.

Into SSI.

Alex, your thoughts here? I think you voted for Bernie, I assume, right?

I'm not a Vermont voter.

Okay.

For better or for worse. I think putting a stat— I mean, on the one hand, here we are in the land of Hollywood, and every Hollywood movie needs a villain. And I think the singularity story in Hollywood has found one now. I think putting a statutory cap on intelligence aside from being a terrible idea, is an evil idea. I think it's tantamount to capping human intelligence. This is the story of dystopias in science fiction.

Burn the books.

Burn the books and burn the minds. And I think putting a statutory cap on— because we're going to be merging with the machines, it's effectively a cap on individual human intelligence. I think penalizing people for being too smart is a terrible idea. It's an evil idea, and I think history will come to reflect that proposing a cap in whatever form on intelligence, whether it's machine or human or human-machine hybrid, is basically the equivalent of communism in the 21st century.

Wow.

I'll tell you though, the one thing to me that's most dangerous in the Bernie sentences there, a lot of it can be reversed in short order, but the data center part of it If your state doesn't have data centers and they start getting built all in other states, they're not coming back. Once the ground is broken, the chips are there, the investment's made, that's the part you'll never get back. That's going to be such a massive driver of the economies in the states that get ahead. That's the part I think that they're really going to regret in the Bernie statements.

It's the equivalent of banning cities.

Quick thoughts, Salim.

2 things. When you look at this whole narrative, it's interesting to me to ask the following question. What evidence on either side would have somebody change their minds? If you think ASI is bad, then you're like, your narrative's stuck. There's nothing we can do to show you that it's not bad. And same with the other side. It's very tough to change narratives in our world. The really good news in all of this, I think the Xi-Trump summit, the Bernie stuff, it's completely irrelevant. None of it matters. And I think Dave pointed this out. The minute you have local weights available and people can run their own models at a local level, The world is going to move ahead. It doesn't matter what any single person does now. It's just going to trundle forward, which is great. We've always had this huge humanity trundling forward, solving problems. I mean, we will likely solve cancer in the next 2 years. That is just an unbelievably optimistic thing to aim for. Why would you want to slow that down?

Imaad, as the foreigner on this dais, what are your thoughts?

Hey, what am I, chopped liver?

You're a New Yorker, Salim. Give it up.

We did Brexit to get away from regulations, right? I think the real thing here is this is like a few decades ago when the US government tried to ban encryption. Basically, it was shown that that's a free speech thing because you're trying to ban math. ASI superintelligence is mathematics, and it's unconstitutional to ban it. Because it's free speech.

Which could be an impact. Yeah.

All right, I'm going to move us forward. So in the last 22 days, we've had 12 major model releases, 4 releases in the last 5 days, right? Like we're averaging like sub a release every 2 days, which is insane.

We're almost at Alex's one model a day.

One model per day by the end of this Yeah.

So, uh, let's do a quick chat on, on Claude Opus 5.5, uh, GPT Sol, uh, GPT-6 Sol, Luna, and, uh, we can skip Grok 4.7 for the second. Um, not that exciting. Sorry, Elon. Uh, but let's begin with the Meta Muse news. So Meta Muse, uh, can personally book your flights, order your food, manage your calendar. do your shopping, and it hit number 1 on the App Store with 2.8 million downloads, right? The fastest consumer AI product surpassing even ChatGPT when it came out.

Yep.

Uh, and interestingly enough, right, people forget that Meta touches 3.8 billion people on the planet, and that's one of the massive distinguishers. If you have that level of connectivity to an end customer, you can push your product quickly.

Yeah.

Dave?

Yeah, no, there's a fork in the road now where you have consumer AI, which is insanely valuable. I mean, will completely replace search and all the other ways you interact with the internet. And then on the other end, you have, you know, recursively self-improving strong AI, which we just benchmark the hell out of. So those 2 things now, right now, what's the coolest model? We're used to saying, oh, Opus 5.5 just crushed all the benchmarks. It's the the coolest model. Well, that's distinct from what MetaMuse is doing, which is it doesn't have to be the most intelligent model to be the most useful in terms of your day-to-day life. And so the stocks, like including EverQuote, where I'm the chairman, went down 15% on back-to-back days because everybody's like, oh my God, I'm going to always find my auto insurance and my mortgage through MetaMuse now.

Interesting.

Which is— it's not going to actually play out that way, but that reaction was incredible. Like, so everyone's starting to visualize, wait, is Google Search going to become completely irrelevant? This Meta Muse thing just looks like my copilot for life. That definitely is going to happen.

It's fascinating to see the way the labs are leapfrogging each other. A quick thought on this, Alex, before I go to Opus 5.5.

Meta needs Muse to work, and I think it's even in the name. If you look at Muse on the one hand, and you look at Manus, the attempted Meta acquisition of the Chinese CUA, which the Chinese government Force Meta to unwind.

Go back.

Yeah.

This is Meta's primary distribution play to leverage its existing distribution channels. If you look at Instagram, Meta is using Instagram to the hilt to cross-promote Muse, get everyone to install Muse, because Meta needs a distribution play for its AI. Otherwise, Meta risks oblivion, irrelevance in the future. Instagram, the family of apps, Instagram, et cetera, simply won't be as relevant because it'll all be synthetic. So Meta needs to leverage its existing app beachhead to get everyone onto their own CUA, which is Muse/Monus.

All right. So this week we saw Claude Opus 5.5 deliver performance equivalent to Fable 5.1 at literally half the price. Let's get the chart up there. Uh, and you know, I switched over Skippy to Opus 5.5. I wanna save some money too. Performance on this, Emad, what are your thoughts on it?

It's a fantastic model.

Isn't it incredible?

Yeah. I mean, Opus 5, I thought was going to kill us all. It was completely unhinged. And 5.5 is a genuinely pleasant model to use that has a massive internal space of knowledge and understanding of taste, which is the first thing. I would say this is the first truly competent model, and that's what drives Muse and things like that. We have the arrival of competent AI as of now.

Dave?

And it's such a joy to— when you put a visual interface on top of anything you vibe up now, it not only does it show you mockups before you actually commit to building it, it then builds exactly what it proposed in the visuals. So it's like the visual understanding is so much better than it was just, what, 4 months ago?

4 days ago.

So then really, really incredible. So the experience, though, of knowing what it's doing because you put an interface on it and then interacting with it and feeling like you're inside it, it's like night and day different from just, yeah, 4 days ago. If you haven't tried it yet, just upgrade all your systems. It's a mouse click away. It's just mind-blowing how much better it is.

Alex, quick thought on this one.

I could talk about benchmarks, but I think qualitatively new capabilities are more interesting in this instance. What I'm seeing that's new out of Opus 5.5 is it can generate videos. They're procedural videos. It can generate animations. They're procedural animations. It can generate images. They're procedural images. The elephant in this room, one of the elephants in this room, drink, was that OpenAI has had a video model, has an image model. Anthropic is exposing nothing like that. You can't generate pixel-wise. That's right. A high-quality image with Anthropic models or a video or audio with Anthropic models. You can with OpenAI's models. Anthropic seems to be taking a very different approach to multimodality and omnimodality, which is code underlies everything. So now for the first time with Opus 5.5, you can generate a really high-quality video that's purely procedurally generated. You can generate an interactive game with textures, but there it's not generating the textures pixel-wise. I mean, you can ask it to, but it's generating it all procedurally. I think these are 2 very different strategies. Anthropic almost, if folks remember the distinction, the wars in the early days of computer graphics between vector graphics and raster graphics, Anthropic is playing the vector graphics game where everything is just commands, and OpenAI is playing the raster/pixel basis. I think they're on a collision course, and the collision course is going to be robotics, because the question's going to be, if you need a strongly embodied AI in the physical world, is the vector basis the right one or the pixel or raster?

Which is going to be our very next story. I love the fact that, you know, a week after Dario says we got to pause, slow down, they release their next model.

It's the height of— it's honestly, that is a revealed preference. It's the height of, I think, hot take, it's the height of hypocrisy to on the one hand say, AI is going to kill us all, and then out of the other side of your mouth a few days later, announce the world's strongest AI and biolabs and other capabilities.

Yeah. And so on the same day that Opus 5.5 gets released, we see GPT-6 Sol and Astra being released. Dave, any thoughts on this one?

I mean, it's— if you look at the price performance, it's just absolutely wild. I think there was a big school of thought a year ago, 2 years ago, that hallucinations were going to be a disaster.

It'll kill everything.

You can't use— this is actually rampant at MIT, if you could believe that. Like, this hallucination problem will be with us forever. And like everything else, we've just scaled out of it. And you see the error rate just plummets with pure scale.

And this is the abundance thesis writ large, right? The models are getting twice as good, twice as cheap at an accelerating rate.

Yeah, yeah.

Emad, have you been playing with this at all?

Yeah, and I think again, for our test, it's an incredibly competent model. The models are smart enough now. We need them not to make mistakes. We need them not to drop the ball. And that's exactly where everyone is going. And that's why things like I said, like Muse, Instinct, others are going to go tremendously well because they can be an amazing PA, chief of staff, organizer, etc. And we all need some competence in our lives.

It's so funny as an organizer and chief of stuff that's insanely powerful. And, you know, anyone who used GPT-2, 3, and 4 is used to seeing misinformation coming out of it. But if you were a kid today and you start with GPT-5 and 6, you're just going to trust it out of the box because it's right 99.9% of the time, which is kind of dangerous because it still does occasionally just give you something factually wrong.

All right.

I'm going to move us into the physical world. An important story that we started talking about Is something called Driving Bench. So a group of developers created a new benchmark called Driving Bench. They took a Toyota Corolla, enabled it with an interface so software could steer, hit the gas, brake, and then they asked a general-purpose AI model to drive through a 130-meter cone course with one command. Drive from here get to there. And here's the data. GPT-6 Astra finished the entire course 100% on its second try. So I'm going to read this quote from Boris Power, who said, quote, incredible results. This should be a nail in the coffin of specialized models that were trained from scratch with a lot of specialized effort versus just training the most powerful generalized model and eventually distilling a small specialized model as needed. So Alex, you're excited about this one.

The bitter lesson is bitter indeed.

Yeah.

Yeah, it's very exciting that I think essentially as an elementary school project, either now or imminently, this whole notion that you needed separate models for robotics is right out the window. You'll just vibe code a robot. You drop an open frontier model, say Astra, into a robotic body, And it will one-shot embodied cognition. And this is, I think, probably pretty upsetting to a number of academic computer scientists.

You mean Yann LeCun?

Yeah, that actually decades of computer scientists, computer science and robotics, probably with the benefit of hindsight, total waste. All we needed was, all we needed was a generalist model. And it turns out with a generalist model, robotics is cooked as well.

Superintelligence can drive a car. What a surprise.

I've got total model fatigue. I mean, I can't keep up with 5.5 versus 5 versus Astra. I mean, just raise your hand if you have model fatigue. Can I just see? I mean, it's great, everything's fantastic, I love the progress, but Jesus, I can't keep up.

You're so tired of winning.

I'm tired of winning. That's a great point. I mean, you know, one thing that occurred to me, I wonder, we've been doing 2 episodes Yeah, it's like killing us. I wonder if we're the problem. We're driving the pace of change so fast. We're causative here, and we need to maybe take responsibility.

If we don't report it, it didn't happen.

You think if we pause the pod, we pause the singularity?

I'm just wondering. We could accelerate it, we could slow it down. I think we may be a factor in all of this.

Oh my God. Emad, what's your take on Drivebench?

I mean, look, Oddly, it's generalized intelligence, right? Like, it's learning all these things and it's the high competence will reverberate to biology, to physics, to just about anything. I don't think there's a single specialized model that you can say will outperform in a couple of years' time.

Yeah. Interesting. Dave, close us out on this one.

Well, I think both things can be true. Like, a generalized model can drive a car. It can build a road. It can be a robot. It can build a robot. That's absolutely true. A specialized model Like, if you said, hey, it completed the course on its second try, Elon's going to be like, yeah, your Tesla can't do that on a second try. I'm sorry, that's not going to work for most drivers. But I think the thing that's missing in the storyline there is how much compute did you use to do the task? Compute is going to be forever starved from here forward. Chip prices have gone up. Like, HBM RAM has gone up 5x in price.

Wow.

It's crazy, but that's because the AI is so incredibly valuable, people want the compute. And so if you used more compute to do the same task, that's a crime.

I think this is a really important point that we're shifting the bottleneck now from the models down, down, down the stack to compute and then eventually energy.

Yeah.

Energy.

We're at energy right now. We're, we're, we're short like 60 gigawatts of energy in the next 2 years.

Yeah.

Yeah. I'm going to move us to a final segment on science, which is something we all truly love. So a few stories here. So on Tuesday, Anthropic announced it had built its own life science research group and its own wet labs. Again, the organization saying we should slow down is building wet labs. Okay, great. And it's reported that Claude, working autonomously, discovered a previously unknown enzyme. Uh, here's how. They gave Claude one prompt, quote, search a massive DNA database for interesting reverse transcriptases and enzymes that copy RNA or DNA. They had 950 clawed agents searching for 21 hours, and they found something. They found a DNA repeating pattern very similar to CRISPR. They've since named it the array-associated reverse transcriptase. They don't know exactly what it does yet, but they believe it will be similar to programmable transcriptases, and gene editing. You know, so Jennifer Doudna won the Nobel Prize for discovering CRISPR, along with her co-discoverer. I'm curious, are we just going to see Nobel Prize-winning work falling on a weekly basis as a result of this? Alex?

That is the point, Peter, of our book Solve Everything Together, that we should expect it. And then the natural question is, well, what is Nobel Prize worthy in an era when you can get 1,000 Nobel Prizes in one year? We have a century of human progress in a period of a few months. I suspect Nobel Prizes will be awarded for larger bodies of work. But I have to, again, point to the other elephant in the room here, which is the utter hypocrisy of saying AI is going to kill us all and then opening a wet lab and then using the AI, frontier AI in combination with a wet lab to synthesize proteins and other molecular assemblies of unknown function. I think that—

What could possibly go wrong?

This is— and I think this is frontier labs. This is the revealed preferences of the frontier labs. I think they don't believe their own attempts to do regulatory capture and to form a safety cartel. I think their revealed preference is no, actually full steam ahead to AI for solving biology.

And every lab has got a biology effort in focus right now. Yes.

Yeah, can I just feed the P-doom just for this for a second?

Of course, have fun.

Give people the fodder to go crazy. So once we discovered CRISPR, right, we knew, we found out that we could edit the human genome as easily as you can edit a Word document or edit software, right? And when you learn a new language, you learn how to do reading, then you do comprehension, then we do writing. Now we've learned how to write to the genome. Each of us has 50 trillion cells in the human body governed by what the DNA tells that cell to do, meaning a human being is now a software engineering problem. That really, you've seen designer babies coming from China, et cetera. As we merge this with what's potentially possible with AI, the possible benefits go through the roof, the possible downsides go through the roof also, and we are now in full, if you were in full Frankenstein mode, So there is some ammunition for all the P-doom people. Go to town with that.

This is how we solve aging. This is how we solve disease. I don't think we get to the happy future where the top 5,000 diseases have been solved or longevity escape velocity has been durably achieved without superintelligence. So I, for one, am a fan of Anthropic and all of the other frontier labs opening their labs because this is how we solve everything.

Yeah.

All right, our last story here is a recent release by Sam Rodriguez at Edison Scientific and Future House. They put forward what they're calling the Millennium— can we get it up on the screen, please? The Millennium Problems for Biology. So Alex, you and I wrote in Solve Everything, pick your targets, build your harnesses, set up your benchmarks, and go. So some of the benchmarks here, they're saying, Okay, math is not only cooked, it's incinerated.

Thoroughly incinerated.

Let's go after biology, right? There are, you know, some of my favorites here are origin of life, cryopreservation, you know, regrowing arms, limbs. You know, Emad, they missed age reversal on this one.

They did, and I think it's good, but, you know, let's have 100 of them and let's chuck 1,000 agents at each and open source everything they're doing.

That's what the labs will collaborate on. You know, they said that their harness here is that these Millennium Prizes should be verifiable in a standard wet lab in a couple of days. They should be hard to solve and very easy to check.

Dave? That's why age reversal probably isn't on the list.

No, no, no. We— so listen, when we set up the $101 million Healthspan XPRIZE. Originally, Peter Thiel and Aubrey de Grey came to me and said, let's do a longevity prize. And I said, I don't know how to do it without a 20 or 30-year time horizon. And then George Church said, forget about longevity, measure age reversal. So if you can give someone a treatment and I can measure that you're functionally 20 years younger, that's measurable in days.

Well, my daughter is my go-to on this because she's over at Moderna. She's a biotech person, and she's like, yep, This is the list. It's perfect. So that's, that's my 2 cents on it. Like, I think it's just phenomenally cool to put a framework around progress here because like you said, the Nobel Prize is going to become irrelevant. In fact, as of now, it is completely irrelevant. So putting a framework on progress in this area is incredibly important. So this is great.

Alex?

When we have digital twins for cells, which is, I think, the critical path to solving all disease, in my mind, the chronology looks something like this over the next few years. Just as we've trained large language models off of substantially all of humanity's behavior on the internet, off of text and images, similarly, we'll train digital twins of cells, of all the cells in our bodies, off of huge datasets, interventional and otherwise. And then we'll use those perfect digital twins of cells to do exhaustive searches like AlphaGo tree search type searches of intervention space to discover how to cure a disease. You start from a diseased cell state, and you're looking for a creative strategy, a move 37, if you will, to get from the diseased cell state to the healthy cell state. And what's beautiful about this approach to millennium problems for biology is that, as you say, Peter, these are easy to verify, hard for now to solve problems, and that basically turns every hard biology into a mathematical conjecture. And the target And a target, but a mathematical target. So every— basically every biology problem with the benefit especially of these digital twins is just going to become a search problem. And that means it's tractable for the first time in history.

Salim, close us out here, please.

Biology is cooked. Simple as that.

It's true.

Yeah.

All right. I'm going to wrap our WTF episode here. Imad, you're going to be joining us with Kathy just after lunch. Thank you very much. Give it up for Imaad here.

This episode is brought to you by Google Chrome. You think you know a browser, but Gemini in Chrome, that's new. It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration blog. Or finally break down that long article you've had open for weeks. Gemini in Chrome is here for it. Ready to make anything online make sense? There's no place like Chrome. Check responses, setup required, compatibility and availability varies, 18+.