YT YouTube2Text

YouTube2Text → Transcripts → Jensen Pushes Back on Doomers, Xi & Trump Talk AI, and “AI” Gets a Rebrand | #294 MOONSHOTS Live

Jensen Pushes Back on Doomers, Xi & Trump Talk AI, and “AI” Gets a Rebrand | #294 MOONSHOTS Live — Transcript

by Peter H. Diamandis · 9,285 words · 1,397 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:05Good morning everybody and welcome to
0:07Moonshots Live, our inaugural event.
0:11>> Yeah,
0:13it is such an awesome treat to have all
0:16of you here and get a chance to meet you
0:18guys. Uh our smiling muscles are
0:21exhausted from this morning, but I think
0:22we're going to be going throughout the
0:24day. On behalf of Dave Blondon, AWG,
0:29Seem, and Emod, it is an awesome
0:32pleasure to have you guys here. Yeah,
0:34give it up.
0:38So, we're going to talk about this, but
0:41our mission is to keep you hopeful,
0:45optimistic, and to inspire you to find
0:49your massive transformative purpose, to
0:51choose your moonshot, and to make it
0:54happen.
0:55Right? It's this is life during the
0:57singularity. It's never going to be any
0:59slower. We have the opportunity each and
1:03every one of us to make the future what
1:05we desire. The future is not happening
1:07to us. It's what we create. Right? Ours
1:11is the ability given these superhero
1:14technologies enabling us to build,
1:17create, and steer that future. Right? AI
1:21is not a hand grenade. It's a jetpack.
1:26It's an opportunity for us to find
1:28problems and solve problems. You know, I
1:30define an entrepreneur as someone who
1:32finds problems and solve problems. And
1:34the more entrepreneurs on the planet,
1:37the more problems get solved. Do we have
1:39challenges? Of course. But that's what
1:42we humans do. We find the challenges and
1:45we fix them. Please remember that. My
1:48hope for you, each and every one of you
1:50here today, is to find extraordinary
1:53co-founders here in this room. To find
1:56those individuals who share your
1:58purpose, design moonshots, and execute
2:01on them. The ability has never been
2:04faster and more capable than it is
2:06today.
2:08So, by the end of today, if you're not
2:11clear what your purpose is, please dig
2:13in to find it. And walk away with at
2:17least one sentence in your mind. This is
2:21what I'm going to build. This is the
2:24problem I'm going to solve. And then
2:27today or tonight, find three people and
2:31speak that statement out loud. When you
2:33speak it out loud, you make it real. And
2:35guess what? the person you talk to may
2:38be your next co-founder, maybe your next
2:39funer, maybe your next partner. Take the
2:43time to do that because you can. And
2:47this is the room where magic is going to
2:50happen.
2:51All right. Uh it's time to introduce my
2:54moonshot mates. Let's roll the video.
3:00>> The question is, how do you build? How
3:02do you make things better? People who
3:04were already in the ecosystem now have a
3:06super intelligence at their beck and
3:08call. The story line that we're
3:09capturing here will last for millennia.
3:11>> As a mathematician, GPT 5.6 Pro is the
3:14only quality math model.
3:16>> All of this automation and AI enablement
3:19and cognitive abundance allows us to be
3:21deeply deeply deeply creative.
3:23>> Video Gen combined with AI is a very
3:27powerful force.
3:28>> A year from today, we'll all be like,
3:29"Wow, remember how slow it was?"
3:32Please give a warm welcome to the
3:34moonshot mates.
3:38>> All right, I'm going to call them out.
3:41I'm going to call them out one at a
3:43time.
3:44Our Afghan warlord, Selene Miso,
3:49our super intelligence, Alex Gross.
3:53our giant Immad Gust
3:58>> and my fraternity brother.
4:03Gentlemen,
4:05first of all, I did hug Alex and he is
4:09somewhat physically real.
4:10>> You didn't pinch my should be a robot.
4:12He is real.
4:14>> I'm a real boy.
4:15>> Yes.
4:18>> Uh first of all, wasn't last night
4:20amazing? How about give it up for
4:22William Shatner? Huh?
4:25>> Unbelievable.
4:27>> You know, uh I'm just blown away by his
4:31level of energy. Yeah. Why is my watch
4:35going off?
4:37It says I have an emergency call going
4:39on. Okay.
4:40>> Okay.
4:40>> No, I'm okay. I'm okay. Stop it. You
4:44know, when technology fails you, it's
4:46great.
4:50A lot of news going on.
4:53Uh shall we jump in?
4:55>> Let's do it. Yeah. What else would we?
4:56>> So, yeah.
5:01>> All right. Here's my clicker.
5:03>> So, uh let's open up with the story
5:06that's been underlying the last two
5:08weeks. Uh there's been really a fear,
5:12you know, a pandemic of fear going on in
5:14the planet. we had two weeks ago. We
5:15hear, you know, Dario goes on with his
5:18blog. Sam Alman comes in and says, "I
5:20agree." Elon comes in and says, "I
5:22agree." You know, even even Demis comes
5:25in and says, "I agree." And then two
5:27nights ago, just before the sheump
5:29meetings, uh Sam and Daario were there
5:32at the Security Council saying, you
5:35know, we're in the midst of a
5:37existential threat.
5:40But
5:41you know and and just to be clear I do
5:44think there is danger from AI but I
5:46think it's danger from humans using AI
5:48for mal intent. So I'd like to do is
5:52actually share two videos
5:55of uh some other leaders in the industry
5:59and let's talk about this. I want to
6:00talk about, you know, is the fear
6:03warranted from AI on its own in a rogue
6:08exploit. So, let's take a look. The
6:11first one is Jensen and some comments
6:12from him. Uh, let's start here.
6:15>> Nobody's building more comput today than
6:17the people asking to be slowed down. All
6:19all of the other narratives to deflect
6:22blame to to make it sound like AI is so
6:25powerful. I have no idea how to fix it.
6:27It's not my fault. It's just because the
6:29technology is just so powerful.
6:30>> So when Jeffrey Hinton is on TV saying
6:32he thinks a 10% chance of societal
6:34destruction is not unreasonable.
6:35>> I would tell I would tell Jeff that that
6:38um it's irresponsible to say all that.
6:41All of his predictions have been wrong.
6:43Enough predictions. That 10% chance is
6:46not grounded on science. It's not
6:47grounded on research. It is it it just
6:49because it comes from a scientist
6:50doesn't make it scientific.
6:58drop the mic moment
7:02>> and then Zuck weighed in. Here we go.
7:06>> Is AI going to kill us?
7:09>> It's just that this is the place you
7:10have to start this week.
7:12>> Well, I think if we all do a good job
7:14and um and act responsibly, then no. I
7:17mean, I'm quite optimistic about the
7:18fact that this is going to be a very
7:20positive future for everyone. Look, I
7:22mean, I think that there's a lot of uh
7:24rhetoric that is filled with doom. And I
7:27mean, my view on this is that there are
7:30a number of labs that are working on
7:31building um advanced AI. And I don't
7:36think that we need some kind of like
7:38industry-wide coordination to not to not
7:41necessarily mess this up. I think that
7:43just each lab needs to take the time and
7:46when it sees that there are issues, you
7:49just take the time that you need
7:50internally to uh make sure that you're
7:52proceeding safely. My my own view on
7:54this is that we're entering a phase
7:56where trust and alignment is actually
7:59going to be the most important next set
8:01of capabilities.
8:07>> Immod, let's go to you first on this.
8:10>> Yeah. Well, I think that it's an
8:12interesting period because we're getting
8:13the capability increases and people are
8:15worried on different sides, but a lot of
8:17the open AI outbreaks and things were
8:19bad security protocols. It was bad
8:21infrastructure on their part. And the
8:23most dangerous point we're coming to is
8:25humans misusing it deliberately or not
8:28before we get through to the other side.
8:31So I think we've got a series of dangers
8:32coming and they're all being conflated
8:34into one thing which is AI breaking out
8:36suddenly kills us all without actually
8:38looking at the individual things and
8:40putting in place existing regulations
8:42and enforcing them and on the other hand
8:44thinking how can we actually collaborate
8:46to make sure this is aligned and works
8:48with us. Alex, I've gone on record
8:51multiple times as saying I think
8:53something probably went wrong after
8:55World War II and we lost out on at least
8:5750 years of progress due to moral panic
9:00over nuclear energy. And I am seriously
9:03concerned that a moral panic of the
9:05moment over super intelligence could set
9:07us back another 50 years. So it's right
9:09on the t-shirt pdoom less than zero.
9:16And I I think the worst possible outcome
9:18is deceleration. So don't del.
9:21>> Okay. Seem pab greater than pdoom.
9:24>> Oh like a thousand times. Look uh we
9:28have evolved for 4 billion years
9:31fighting uh all sorts of issues. Uh life
9:36tries to the the universe tries to wipe
9:38you out, right? There's so many entropy
9:41tries to wipe you out. We're local
9:43anti-anthropic phenomena that somehow
9:46managed to create order out of chaos.
9:48And the idea that the we'll get wiped
9:51out now by something seems to be
9:53unbelievably absurd.
9:55>> You got to remember that 99.9% of all
9:57species are no longer with us.
9:59>> Yeah, that's fine. But we keep moving
10:00on. We keep moving on. Look, if we
10:03morph, if we merge with AI, if we merge
10:06with uh thing, and let's note that's
10:08been going on forever, right? Iman's
10:10wearing spectacles, technically he's a
10:11transhumanist. The minute the minute you
10:14get a vaccination as a child,
10:15technically you're a cyborg. So, we've
10:17been augmenting the human experience
10:18with technology for a very very long
10:21time. And look at the unbelievable lives
10:23we're living now. And the opportunity to
10:25solve more and more problems. The the
10:27huge problem we have, we talk about this
10:29all the time on the podcast. Our brains
10:31are so geared towards negativity, we get
10:33stuck in that phenomena. And I think the
10:36work that we're doing here, which is I
10:38think why everybody's here and so
10:39excited, is the is pushing that positive
10:42narrative because the world is
10:43infinitely better than we've had it.
10:45>> So, let's celebrate that.
10:47>> Amen.
10:48>> Pab,
10:49baby.
10:50>> Pab.
10:50>> Uh Dave, close the tab on this.
10:53>> Well, we're on the cusp of a moment I've
10:54been waiting for my entire life. I'm not
10:56going to sit here and and like throw
10:57cold water on this greatest moment
10:59really in the history of humanity. Um
11:02Jeffrey Hinton is calling it the way he
11:03sees it. I think I think the quote like
11:06just because it comes from a scientist
11:07doesn't make it scientific. That's going
11:08to end up on an Alex t-shirt.
11:11>> That's that's awesome.
11:12>> But he's calling it like he sees it. But
11:14but clearly uh we're moving forward at
11:17warp speed. It's improving itself. The
11:20ability to make the world just a
11:22dramatically better place is right in
11:23front of us. Nothing's going to slow
11:25that down. The misuse of it is an
11:27incredible threat and there are
11:29solutions to that. But it's happening
11:30anyway. like to to get stuck on
11:32doomerism is completely dysfunctional.
11:34We're not slowing down. It's going to
11:36happen.
11:36>> Yeah. Uh the other story that transpired
11:40yesterday uh or kind of didn't transpire
11:44yesterday was the sheump AI
11:47conversation.
11:48>> Yeah.
11:48>> Right. So, uh what came out of that was
11:50some state dinners. Uh I think the
11:53interesting quote that Trump posted is
11:56quote, I want to leave it exactly where
11:58it is. That is China's position also.
12:01Our guard rails uh is the department of
12:04justice.
12:06So wait, so this ties back to the
12:10conversation we had in the last pod
12:11where Bessant said we're not going to
12:13wave liability.
12:14>> Yeah.
12:15>> Right. And and so this is a very
12:16important point. The labs are liable for
12:20their product if it does harm. And I
12:23think that's one of the most important
12:24things that can you know they've asked
12:25for liability of w you know why over
12:27liability and one of the questions we've
12:29discussed all the time here is is that
12:31the reason that they're saying it's too
12:34dangerous you know 10 20% existential
12:37threat. Uh Alex, let's go to you on
12:40this.
12:40>> I I've made the point on the pod in the
12:42past. I think we saw in the past 2 to 3
12:44weeks an attempted formation of an AI
12:46safety cartel. And I don't think this
12:49was even a recent occurrence. This is
12:51right there in the founding conceits of
12:53open AI. The notion that if and when we
12:55got close to super intelligence, however
12:57construed, or AGI, however construed,
13:00that there would be a deliberate effort
13:01to form wasn't being called a cartel at
13:04the time, but a coordination mechanism.
13:06>> I don't think it's being called a cartel
13:07now either.
13:08>> Well,
13:10some some of us view it as an attempted
13:12cartel, including myself. And I I I
13:14think competition is good. Think back to
13:16the creation of Open AI. So when Sam and
13:18Elon and Greg and others created OpenAI,
13:21the purpose originally for Open AI was
13:23to be a counterbalance to Google
13:25DeepMind, which Elon especially was
13:27concerned would become a singleton in
13:29the future that would dominate the
13:30future of Liteco with their AI. So we
13:32got Open AI, a counterbalance, if you
13:34will, and I think competition is very
13:37healthy in this space. I think forming a
13:39cartel for full ownership of what maybe
13:43I've called in the past the frontier
13:45liberation front and pushing out a
13:47frontier liberation front. Terrible
13:48idea. I think defensive coscaling is how
13:51we end up in a a great future. When if
13:54we were at the the dawn of the city in
13:57in humanity, it would be a terrible idea
14:00to say, well, we can't form cities. We
14:02need to form an alliance to regulate the
14:04formation of cities because cities breed
14:06crime. No, it's a terrible idea. Cities
14:08are fountains of economic growth. What
14:10do we do instead? We have police forces
14:13and fire departments and all sorts of
14:15other municipal services that
14:16defensively co-scale with the
14:18population. The singularity, I claim, is
14:20going to need exactly the same thing.
14:22Don't desell, just defensively coscale.
14:26>> I love that. Um, Seem, you know, so yes,
14:32give it up for ASI. I think your I think
14:34your city metaphor is such a great one
14:36because it just paints the picture very
14:38clearly as to how you organize for the
14:40future. You just need to find new
14:42structures that match the specific
14:44demand that's occurring.
14:46>> Yeah. Immad, do you think China and the
14:49US will trust each other?
14:51>> No.
14:51>> Okay.
14:53>> I mean, so how you know Bessant asked
14:56for a hotline, you know, our AI just
14:59escaped and it's arming our nuclear
15:01missiles. You know, it's like
15:03>> I think the hotline is again more for
15:05these sub ASI sub breakout kind of
15:08things where you do have to have
15:10responsible reporting like OpenAI's
15:12model hacked the Australian medical
15:14system that came out recently. I mean
15:16maybe that's what AGI should do. It
15:18should hack all our systems and upgrade
15:19them, right? It's like the align thing
15:21to do.
15:22>> But at the same time, I think that there
15:24can be a level of coordination because
15:26AI is entering society. So you look at
15:28the city, the police force, the fire
15:31department, others, they're utilities
15:32that actually do have liability
15:34waiverss, but that's an exchange for
15:36being a utility for humanity.
15:38>> And the AI that comes into our systems
15:41can't be controlled by the big labs. And
15:44then they should have a liability waiver
15:46because
15:47>> they should have a liability waiver.
15:48>> That type of AI that comes into our
15:50systems shouldn't be controlled by the
15:51big labs. And they should have liability
15:53waiverss because they're working for the
15:55betterment of society. And that's a
15:57different type of AI that we're not
15:59really looking at. But it's inevitable
16:01that the AI will run our judicial
16:03system, our medical systems, our roads,
16:05our schools. And who's really thinking
16:08about that side of things? Because
16:09again, that's the utility AI versus the
16:12frontier AI. And it's all being merged
16:14into one. And that's something
16:15definitely the Chinese and Americans can
16:17learn from each other on because they
16:19don't both want better functioning
16:20societies. I can't wait till the ASI
16:22says, "Yes, we hacked your medical
16:24system because it was so bad.
16:27and we've just, you know, improved it,
16:29you know, 10x
16:30>> overnight. The DMV just goes the whole
16:32backlog disappears.
16:34>> Can I can I build on that just for a
16:36second?
16:36>> Yeah, sure.
16:37>> If if you take this uh structure that we
16:40have, I've I've said this repeatedly,
16:42nation states are a terrible way of
16:44governing the world today. They're
16:46they're geared towards scarcity. I think
16:48what's going to happen with AI in the
16:50most positive uh potential outcome is
16:54that AGI, ASI, whatever the hell you
16:56want to call it, you've heard my rant on
16:58that uh emerges and it forces us to
17:01operate as a world community. It forces
17:03the global countries to to formulate and
17:06come up with a consistent steady
17:08>> interface standards. Yeah.
17:09>> Yeah. And and come up with a common
17:11interface with that uh whatever that
17:13thing is and then we can move the world
17:15forward. Right now, we cannot move the
17:17world forward with the the log gem of
17:21old 15th century structures that that
17:24are outdated from from at least 100
17:27years.
17:27>> I'll take the other side of that if I
17:29may, which is at the dawn of audio
17:31recording, there was this notion, many
17:33thought in the time of Edison that the
17:35ability to record the human voice would
17:38lead to a standardization of accents,
17:40standardization of everything. The exact
17:42opposite happened in in many ways.
17:44Although there has been a die- off of
17:46low resource languages and the long
17:48tale, there has been a proliferation of
17:50new accents and dialects thanks to audio
17:52recording. I would predict the exact
17:54opposite is going to happen thanks to
17:56super intelligence. Not an erosion of
17:58nation states, but a proliferation of
18:00new forms of government.
18:01>> So, so one trend that's occurring, I I
18:03totally agree with you. We are moving
18:05this century from the concept of a
18:07nation state to a city state. That's the
18:10granulation. And it allows a smaller
18:12entity to operate independently. Take
18:14any city. Take LA. If you have solar
18:16energy, satellite internet, vertical
18:18farming, that could happen. Great. We
18:21could do lots here. Uh vertical farming,
18:23satellite internet, solar energy. You
18:25don't need a country. You can you can
18:27self-manage and self-determined to a
18:29large extent. And the tension you see in
18:31the world today is not left versus
18:33right. It's urban versus rural.
18:35>> Right? Brexit was London versus the rest
18:37of the country. Uh Trump is urban versus
18:39rural. And this is the tension that's
18:41happening. We're going to see the de the
18:44problem with nation states is they have
18:45armies and they like to use them. Uh but
18:47over time we're going to see a much more
18:49highly granulated u uh determinism that
18:52will happen which is good for the world
18:53because it'll increase.
18:54>> Dave, why don't you take why don't you
18:56close us out on Trump here?
18:57>> Yeah. The the only surprising thing that
18:59I heard from you know Alvin Graen, good
19:01friend of the pod.
19:02>> You were texting with him last nighting
19:04the White House. He's at the UN you know
19:06with the delegation. So they agreed to
19:08meet again in November which is like you
19:10know 100 years from now in singularity
19:12time. Um but uh the only thing that
19:14surprised me in his notes was you know
19:17our agenda is stop US panic. This is we
19:21we don't want to slow down and lose to
19:23China. We don't want a global we don't
19:24want Bernie Sanders to get everybody in
19:26the country riled up and then have
19:28everything stop. That's our concern.
19:30What's your concern China? Our concern
19:33is the CCP existing at all and the great
19:36firewall of China being penetrated by
19:38AI. Like, oh, I forgot all about Tianaan
19:42Square like and and how fragile the
19:44entire ecosystem is over there.
19:45>> That really surprised me. So then then
19:47their actions make a lot more sense if
19:49you put it in the context of, oh,
19:50they're they're so worried about just
19:52existing as a government.
19:54>> And then this happened in DC. The United
19:58States also totally rejects any attempt
20:00to construct a globalist scheme to
20:02control for the artificial intelligence
20:06being spoken of so much now here and
20:09after officially called super
20:12intelligence changing the name.
20:16I
20:16>> I I thought I thought for sure he was
20:19going to say super duper intelligence
20:22>> actually.
20:25Yeah. But it's it's interesting, right?
20:28So, uh, the US won the race to super
20:31intelligence by rebranding it.
20:35>> Immod.
20:36>> Yeah. I mean, I think it's a great name,
20:38you know, super intelligence. Why not?
20:39I'm looking forward to come to Super
20:40America eventually.
20:43>> I'm still stuck on Lake America. So,
20:47>> the Gulf of America.
20:48>> I was thinking of what Selene was
20:49saying. I think the biggest challenge in
20:51the world today is how do we marry
20:53collective intelligence with localized
20:55wisdom? That's what we're really saying
20:58because even
21:01even what Dave was saying right now, we
21:03have these machines that are collecting
21:04our intelligence. They're learning and
21:06they're adapting our systems and there's
21:08better ways of operating to replace
21:10these. But then you have to have the
21:11localization. You need the diversity.
21:13You need to have it responding to local
21:14needs. And I think you know super
21:16intelligence actually a remarkably good
21:18way to think about it because a lot of
21:20the fears and concerns are about the
21:23intelligence that you know it was ASI it
21:25was artificial super intelligence having
21:28that differentiated from the
21:29intelligence that's helping us dayto-day
21:31really differentiates and delineates
21:33that because you know what is it 99%
21:36perspiration 1% innovation like majority
21:39of the AIs that will help us every day
21:41are the perspiration AI they're the
21:43execution AIs the innovation we still
21:45need. We need new materials, new
21:46physics, etc. But that's a different set
21:49of things and that's not something we're
21:51even smart enough, most of us, to be
21:52able to use. So, let's have one set of
21:54looking at that and let's make sure that
21:56the 99% reaches the 99% to uplift
21:59everyone.
22:01>> Dave, are we going to call it super
22:02intelligence from here on out?
22:04>> You never tell Donald Trump when he's
22:05looking at the teleprompter, you know,
22:06he's he's saying something someone
22:08wrote, but when he looks down, he's
22:09like, did he just make that up once in a
22:11while? You can't tell. He did.
22:13>> Is there an agenda? I I think if there
22:15is an agenda is to try and deflect the
22:17inevitable uh you know wave that's
22:20coming from the Bernie Sanders world. If
22:21you rebrand it and rename it, you're
22:23just a moving target.
22:26>> So So that that could be part of the
22:27strategy, I think. But
22:28>> uh yeah, but the whole the whole goal
22:30there is to keep moving at full speed,
22:32stay ahead of China, and reduce panic as
22:34much as possible. I I am going to say,
22:36you know, one of the things I've been
22:38very vocal about and I think we need to
22:40have in the conversation is yeah, move
22:42at full speed, but let's also direct the
22:45labs to focus on alignment. Uh because I
22:49I think and we're going to talk about
22:50that in a little bit. I I think yes,
22:54move fast, but let's take resources that
22:56you have, you know, point 100,000 agents
22:58towards alignment work. Uh See, super
23:01intelligence, do you buy in? I mean,
23:02you've been asking what the hell that
23:04all means in the first place. Yeah, I
23:05mean look, we talk about AGI, first of
23:07all, it's not artificial. It's not
23:09general. It's not really intelligence.
23:10Other than that, I'm good with it. Um,
23:13you have the ASI shoe. I think that
23:15there there's the whole spectrum of this
23:18uh is is problematic. We need somebody
23:21to come up with a good name for all of
23:22this that's completely different. I've
23:24said repeatedly, I think what we're
23:26building here is an complimentary
23:29mechanism to human intelligence, not
23:31replicative. and the narrative keeps
23:33thinking it's replicative and then you
23:35end up with the Matrix Skynet, you know,
23:37and the robot overlords come and take
23:39over the world. We're so geared for
23:41that. Um, the first article, we did some
23:44research on this. The first article that
23:46ever appeared that we could find that
23:48said robots or AI will take all the jobs
23:50in 5 years appeared in 1964.
23:53>> Yeah. Okay. This is that we get stuck on
23:56these tropes and we can't get out of it.
23:58Right. It's exactly what Alex mentioned
23:59earlier when when uh recording came out.
24:02The same thing when the printed book
24:04emerged. It's that people said it would
24:05destroy people's brains. So we get stuck
24:08on these.
24:08>> Depends what book you read.
24:09>> Yeah. And this is I think why Peter the
24:12work you're doing is so important is is
24:14we have the opportunity. We are actually
24:16reaching abundance. All the data shows
24:18that our narratives just sadly lack very
24:21far behind. I think this is what you've
24:23done with this whole um the X-P prize
24:25for creating a positive vision of the
24:27future is one of the most important
24:29things that's happened in maybe a
24:31hundred years in the world. Shift the
24:32world towards positive things.
24:34>> Thank you.
24:36>> Uh Alex, close us out on this one. You
24:39love neols and I love the ones you
24:41create. Super intelligence. Are you
24:43going to call it that from here on out?
24:45>> I already do. I have a company that I
24:47started called physical super
24:48intelligence. It's in the name.
24:49>> Always ahead of the curve, ladies and
24:50gentlemen. Yeah, I'm I'm a big fan of
24:52the name super intelligence as opposed
24:54to AGI. Maybe for different reasons.
24:57What I like about the name and and Nick
24:59Bostonramm's term super intelligence is
25:01it emphasizes that it's simply more
25:04intelligence and it in some sense I
25:06think demystifies AGI or some of these
25:09other terms. We're going to live to the
25:11extent we don't already in a world with
25:14the cognitive equivalent of many
25:16trillions of human beings. an earth, a
25:18solar system with the equivalent
25:20population of many trillions of people.
25:22And I think that's a better cognitive
25:24frame to think about what would a solar
25:26system filled with trillions of humans
25:28look like? What would it look like
25:30economically, socially, scientifically?
25:33Super intelligence, I think, as a term
25:35conotes that in a way that AGI doesn't.
25:36So, I'm a big fan.
25:38>> Nice. Well, as soon as we renamed it
25:40super intelligence, Bernie Sanders said
25:43we should make it illegal. We should ban
25:44it. Let's watch this video.
25:47I hope very much that they will sit down
25:50and begin the process of negotiating a
25:53comprehensive treaty to establish a
25:56pause on advanced AI and a ban on AI
26:02super intelligence.
26:04>> All right. So, uh
26:08so three key elements in in Bernie's
26:10plan here. First, a permanent ban on
26:12developing artificial super intelligence
26:15defined as AI that exceeds human
26:17cognitive ability across most domains.
26:20Second, an immediate pause on advanced
26:23AI development until a new regulatory uh
26:26system is written into rules. Third, a
26:28new cabinet level uh department of AI.
26:31And get this, the penalties are
26:34corporate death. There's a corporate
26:37death penalty that your company is off.
26:40shut down and for the individuals 20
26:42years in federal prison uh which is
26:45modeled after the illegal nuclear
26:47weapons. Uh so
26:50>> h interesting. I mean this is
26:53fear-mongering at its best. Uh Dave, you
26:56want to so many real problems that need
26:58to be solved that just hate this idea
27:00that we're going to turn this into a do
27:02or don't argument which is utterly
27:03insane. Also, the way liability law
27:07works in the US, you know, it's very
27:09easy to scare people into not doing
27:10anything. You know, if you're afraid of
27:1120 years of prison, you don't you don't
27:13even know what the rule is. They haven't
27:14even proposed a rule yet. The the fear
27:16of of being, you know, beheaded in the
27:20street is very real. So, then people
27:21stop acting. That's the worst thing that
27:23could possibly come out of this. And I I
27:25think, you know, Bernie is from Vermont.
27:26I love Vermont. Spend more time in
27:28Vermont than any other state. And it's
27:30beautiful because not much has been
27:32done,
27:33>> you know, and so I think this
27:35perspective is going to be more European
27:37like like when in doubt, do nothing is
27:40going to be the the default behavior.
27:42And that's so toxic. I mean, if you do
27:44nothing, very very bad things are going
27:46to happen. Like the things you should be
27:48worried about like terrorists using AI
27:50to build bioweapons or China getting
27:51miles ahead of the US. Those are the
27:53real things to worry about. and you're
27:55if you're not acting and like turning it
27:58into a political leftright battle, it's
28:00the stupidest thing that could happen
28:02right now. We need very specific
28:04technical solutions to open source to
28:07monitoring to te just like with nuclear
28:09power. And if we get moving on those, we
28:12can solve them very easily in time for
28:15for ASI. If we just turn it into a
28:17political debate and we're frozen and
28:19people don't know what to do
28:20>> into SI
28:22Alex, your thoughts here.
28:24>> I think
28:25>> you you voted for Bernie, I assume,
28:26right?
28:27>> I'm I'm I'm not a Vermont voter.
28:29>> Okay.
28:30>> For for better or for worse. I think
28:33putting a stat I mean on the one hand
28:35here we are in the land of Hollywood and
28:37every Hollywood movie needs a villain.
28:39Yes.
28:39>> And I I think the singularity story in
28:42Hollywood has found one now. I I think
28:45putting a statutory cap on intelligence,
28:48aside from being a terrible idea, is an
28:50evil idea. I I think it's tantamount to
28:53capping human intelligence. This is the
28:56the story of dystopias in science
28:58fiction.
28:59>> Burn the books.
29:00>> Burn the books and burn the minds.
29:02>> And I I think putting a statutory cap on
29:04because we're going to be merging with
29:06the machines. It's effectively a cap on
29:08individual human intelligence. I I think
29:10penalizing people for being too smart is
29:13a terrible idea. It's an evil idea. And
29:15I think history will come to reflect
29:18that proposing a cap in whatever form on
29:21intelligence, whether it's machine or
29:23human or human machine hybrid, is
29:26basically the equivalent of communism in
29:28the 21st century.
29:29>> Wow. I'll tell you the
29:34the one thing to me that's most
29:36dangerous in the Bernie sentences there
29:38a lot a lot of it can be reversed in
29:40short order but the data center part of
29:41it if your state doesn't have data
29:44centers and they start getting built all
29:46in other states they're not coming back
29:49you know once the ground is broken the
29:50chips are there the investments made
29:52that's the part you'll never get back
29:53and that's going to be such a massive
29:55driver of the economies and the states
29:57that get ahead that's the part I think
29:58that they're really going to regret in
30:00the in the burning statements.
30:03>> Quick thoughts.
30:04>> Uh two things. When you look at this
30:06whole narrative, it's interesting to me
30:07to ask the following question. What
30:10evidence on either side would have
30:12somebody change their minds, right? If
30:14you think ASI is bad, then you're like
30:16your narrative stuck. There's nothing we
30:17can do to show you that it's not bad.
30:19And same with the other side. It's hard.
30:21It's very tough to change narratives in
30:23our world. The really good news in all
30:25of this, I think it the she summit, the
30:29Bernie stuff, it's completely
30:31irrelevant. None of it matters. And I
30:33think Dave pointed this out. The minute
30:34you have local weights available and
30:36people can run their own models at a
30:38local level, the world is going to move
30:39ahead. It doesn't matter what any single
30:41person does now, it's just going to
30:43trundle forward, which is great. Um,
30:45we've always had this huge uh humanity
30:49trundling forward, solving problems. I
30:51mean, we will likely solve cancer in the
30:53next 2 years.
30:55>> That is just an unbelievably optimistic
30:57thing to aim for. Why would you want to
30:58slow that down?
31:00>> I might as the foreigner on this on this
31:02diet, what are your thoughts?
31:03>> Hey, what am I? Chopped liver.
31:06>> You're a New Yorker. You're just
31:07Canadian. I'm just Canadian.
31:09>> You're a New Yorker. So, we give it up.
31:11>> You know, we did Brexit to get away from
31:13regulations, right? Um, but I think the
31:16real thing here is this is like a few
31:18decades ago when the US government tried
31:20to ban encryption and basically it was
31:22shown that that's a free speech thing
31:24because you're trying to ban math. ASI
31:26super intelligence is mathematics and
31:28it's unconstitutional to ban it
31:30>> because it's free speech.
31:33>> Yeah.
31:33>> All right.
31:35I'm going to move us forward. So, uh, in
31:38the last uh, 22 days, we've had 12 major
31:43model releases.
31:45four releases in the last five days,
31:48right? Like we're averaging like sub a
31:52release every two days, which is insane.
31:54>> Almost Alex says one model a day.
31:56>> One model per day by the end of this
31:57year.
31:58>> Yeah.
31:59>> So, uh let's do a quick chat on on
32:02Claude Opus 4 5.5, uh GPT Saul, uh GPT6
32:07Saul, Luna, and uh we can skip Grock 4.7
32:10for the second. Um, not that exciting.
32:14Sorry, Elon.
32:16Uh, but let's begin with the MetaMuse
32:20news. So, Metam Muse uh can personally
32:23book your flights, order your food,
32:25manage your calendar, do your shopping,
32:27and it hit number one on the app store
32:30with 2.8 million downloads, right? The
32:33fastest consumer AI product, surpassing
32:35even chat GPT when it came out. Y and
32:39interestingly enough, right, people
32:40forget that Meta touches 3.8 billion
32:43people on the planet and that's one of
32:47the massive distinguishers. If you have
32:49that level of connectivity to an end
32:51customer, you can push your product
32:53quickly.
32:54>> Yeah,
32:55>> Dave.
32:56>> Yeah, I know there's a fork in the road
32:58now where you have consumer AI which is
33:00insanely valuable. I mean, will
33:02completely replace search and all the
33:04other ways you interact with the
33:06internet. Um and then on the other end
33:08you have uh you know recursively
33:10self-improving strong AI which we just
33:13benchmarked the hell out of. So so those
33:14two things now right now what's the
33:16coolest model? We're used to saying oh
33:18Opus 5.5 just crushed all the
33:20benchmarks. It's the coolest model. Well
33:22that's distinct from what Metamuse is
33:24doing which is it doesn't have to be the
33:26most intelligent model to be the most
33:27useful in terms of your day-to-day life.
33:30And so the stocks like including Eper
33:32quote where I'm the chairman went down
33:3415% on backtoback days because
33:37everybody's like, "Oh my god, I'm gonna
33:39always find my auto insurance and my
33:41mortgage through Metamus."
33:42>> Interesting.
33:43>> Which is it's not going to actually play
33:44out that way. But but that's that
33:46reaction was incredible. Like so
33:48everyone's starting to visualize, wait,
33:49is Google search going to become
33:51completely irrelevant? This Metamuse
33:53thing just looks like my my co-pilot for
33:55life. And that that definitely is going
33:57to happen. It's fascinating to see the
33:59way the labs are leaprogging each other.
34:01A quick thought on this, Alex, before I
34:02go to Opus 5.5.
34:04>> Meta needs Muse to work. And I think
34:06it's even in the name. If you look at
34:08Muse on the one hand, and you look at
34:10Manus, the attempted Meta acquisition of
34:12the Chinese CUA, which the Chinese
34:14government forced Meta to unwind,
34:17>> this is Meta's primary distribution play
34:20to leverage its existing distribution
34:22channels. If you look at Instagram, Meta
34:24is using Instagram to the hilt to
34:26cross-promote Muse. Get everyone to
34:28install Muse because Meta needs a
34:31distribution play for its AI. Otherwise,
34:34Meta risks oblivion, irrelevance in the
34:37future, Instagram, the the family of
34:39apps, Instagram, etc. simply won't be as
34:41relevant because it'll all be synthetic.
34:43So, Meta needs to leverage its existing
34:45app, Beach Head, to get everyone onto
34:48their own CUA, which is Muse/manus.
34:50>> All right. So this week we saw Claude
34:52Opus 5.5 deliver performance equivalent
34:55to Fable 5.1 at literally half the
34:58price. Let's get the chart up there. Uh
35:01and you know I switched over Skippy to
35:04Opus 5.5. I want to save some money too.
35:07>> Uh performance on this Emod. What are
35:09your thoughts on it?
35:10>> It's a fantastic model.
35:11>> Isn't it incredible?
35:12>> Yeah. I mean Opus 5 Opus 5 I thought was
35:14going to kill us all. It was completely
35:15unhinged and 5.5 is a genuinely pleasant
35:19model to use that has a massive internal
35:23space of knowledge and understanding of
35:26taste which is the first thing I would
35:27say this is the first truly competent
35:29model and that's what drives muse and
35:31things like that we have the arrival of
35:33competent AI as of now
35:35>> and it's such a joy to when when you put
35:37a visual interface on top of anything
35:39you vibe up now it not only does it show
35:42you mockups before you actually commit
35:44to building it it then builds exactly
35:47what it proposed in the visuals. So like
35:49the visual understanding is so much
35:51better than it was uh just what four
35:54months ago
35:54>> 4 days ago.
35:55>> So then really really incredible. So so
35:58the the experience though of knowing
36:01what it's doing because you put an
36:02interface on it and then interacting
36:03with it and feeling like you're inside
36:04it. It's like night and day different
36:06from just yeah 4 days ago. If you
36:08haven't tried it yet, just upgrade all
36:10your your systems. Uh it's a mouse click
36:13away. is just mind-blowing how much
36:15better it is.
36:15>> Alex, quick thought on on this one.
36:17>> I could talk about benchmarks, but I
36:19think qualitatively new capabilities are
36:21more interesting in this instance. So,
36:23what I'm seeing that's new out of Opus
36:255.5 is it can generate videos. They're
36:28procedural videos. It can generate
36:30animations. They're procedural
36:31animations. It can generate images.
36:34They're procedural images. So the
36:35elephant in this room, one of the
36:36elephants in this room, drink
36:40was that open AI has had a video model
36:44has an image model. Anthropic is
36:47exposing nothing like that. You can't
36:48generate pixel-wise a highquality image
36:51with anthropic models or a video or
36:54audio with anthropic models. You can
36:56with open AI's models. Anthropic seems
36:58to be taking a very different approach
37:00to multimodality and omniodality which
37:02is code underlies everything. So now for
37:05the first time with Opus 5.5 you can
37:08generate a really highquality video
37:10that's purely procedurally generated.
37:12You can generate an interactive game
37:14with textures but there it's not
37:16generating the textures pixel-wise. I
37:18mean you can ask it to but it's
37:19generating it all procedurally. And I I
37:21think these are two very different
37:23strategies. Anthropic almost if folks
37:25remember the distinction the wars in the
37:27early days of computer graphics between
37:29vector graphics and raster graphics.
37:31Anthropic is playing the vector graphics
37:33game where everything is just commands
37:35and OpenAI is playing the the
37:38raster/pixel basis and I think they're
37:40on a collision course and the collision
37:42course is going to be robotics because
37:44the question is going to be if you need
37:46a strongly embodied AI in the physical
37:48world is the vector basis the right one
37:51or the pixel arrest
37:52>> which is our which is going to be our
37:54very next story. I love the fact that
37:55you know a week after Dario says we got
37:58to pause slow down they release their
38:00next model. It's the height of it's it's
38:02honestly that that is a revealed
38:04preference. It's the height of I think
38:06hot take. It's the height of hypocrisy
38:08to on the one hand say AI is going to
38:10kill us all and then out of the other
38:12side of your mouth a few days later
38:14announce the world's strongest AI and
38:16biolabs and other capab
38:19>> and so on the same day that Opus 5.5
38:21gets released we see GPT6 Saul and Astro
38:25being released. Uh Dave any thoughts on
38:28this one? I mean it's if you look at the
38:31price performance it's just absolutely
38:32wild. Um I I think there was a big
38:36school of thought a year ago, two years
38:38ago that hallucinations were going to be
38:40a disaster. They'll never go away. You
38:41can't use this actually rampant at MIT
38:43if you can believe that. Like this
38:45hallucination problem will be with us
38:46forever. And like everything else, we've
38:49just scaled out of it. And you see the
38:50the error rate just plummets with pure
38:54scale.
38:55>> And and I this is the abundance thesis
38:57at large, right? The models are getting
38:59twice as good, twice as cheap at an
39:01accelerating rate. Yeah. Yeah. Eman,
39:04have you been playing with this at all?
39:05>> Yeah. And I I think again for our test,
39:07it's an incredibly competent model. The
39:09models are smart enough now. We need
39:11them not to make mistakes. We need them
39:13not to drop the ball. And that's exactly
39:15where everyone is going. And that's why
39:17things like I said like Muse, Instinct,
39:19others are going to going to go
39:21tremendously well because they can be an
39:23amazing PA, chief of staff, organizer,
39:26etc. And we all need some competence in
39:27our lives. It's so funny as a as an
39:29organizer and chief of staff, that's
39:30insanely powerful. And you know, anyone
39:33who used GPT2, 3, and 4, uh, is used to
39:36seeing misinformation coming out of it.
39:39But if you were a kid today and you
39:41start with GPT5, and six, you're just
39:44going to trust it out of the box because
39:46it's right 99.9% of the time, which is
39:48kind of dangerous because it still does
39:51occasionally just give you something
39:52factually wrong.
39:54>> All right, I'm going to move us into the
39:55physical world. uh an important story
39:57that we started talking about is uh
40:00something called driving bench. So a
40:03group of developers created a new
40:05benchmark called driving bench. Uh they
40:07took a Toyota Corolla, enabled it with
40:09an interface so software could steer,
40:12hit the gas, break, and then they asked
40:15a generalpurpose AI model to drive
40:19through a 130 m cone course uh with one
40:23command. drive from here, get to there.
40:27Uh, and here's the data. GPT6 Astra
40:31finished the entire course 100% on its
40:33second try. Right? So, I'm going to read
40:36this quote from uh from Boris Power who
40:39said, quote, "Incredible results. This
40:41should be a nail in the coffin of
40:43specialized models that were trained
40:45from scratch with a lot of specialized
40:47effort versus just training the most
40:49powerful generalized model and
40:51eventually distilling a small
40:53specialized model as needed. So, Alex,
40:56you're excited about this one.
40:57>> The bitter lesson is bitter indeed.
40:59>> Yeah.
41:00>> Yeah. It's very exciting that I I think
41:03essentially as an elementary school
41:04project either now or imminently, this
41:07whole notion that you needed separate
41:09models for robotics is right out the
41:11window. You'll just vibe code a robot.
41:14You drop an open frontier model, say
41:17Astra, into a robotic body and it will
41:20oneshot embodied cognition. And this is
41:23I I think probably pretty upsetting to a
41:25number of academic computer scientists.
41:27>> John Lun.
41:28>> Yeah. that that actually decades of
41:30computer scientists computer science and
41:33robotics probably in with the benefit of
41:35hindsight total waste. All we needed was
41:38all we needed was a generalist model and
41:40it turns out with a generalist model
41:42robotics is cooked as well.
41:43>> Super intelligence can drive a car. What
41:45a surprise.
41:47>> I've got total model fatigue. I mean I
41:50can't keep up with 5.5 versus five
41:53versus Astra. I mean just raise your
41:55hand if you have model fatigue. And I
41:56just think I mean it's great.
41:59Everything's fantastic. I love the
42:01progress. But Jesus, I can't I can't
42:03keep.
42:03>> You're so tired of winning. It's time to
42:06tired of winning. That's a great That's
42:08a great point. I mean, you know, one
42:10thing that occurred to me, I wonder,
42:12we've been doing two episodes a week.
42:13Yeah.
42:14>> It's like killing us. I wonder if we're
42:15the problem. We're we're we're driving
42:18so fast. We're positive here and we need
42:21to maybe take respons.
42:22>> If if we don't report it, it didn't
42:23happen. Yeah, those different tools.
42:25God,
42:26>> you you think if we pause the pod, we
42:27pause the singularity.
42:29>> I'm I'm I'm just wondering, we could
42:31accelerate it, we could slow it down. I
42:32think we may be a factor in all of this.
42:35>> Oh my god. Emod, what's your take on
42:36drive bench?
42:37>> I mean, look, it oddly it's generalized
42:40intelligence, right? Like it's learning
42:41all these things and it's the high
42:42competence will reverberate to biology,
42:45to physics, to just about anything. I
42:47don't think there's a single specialized
42:49model that you can say will outperform
42:51in a couple of years time. Yeah,
42:53interesting. Dave, close us out on this
42:55one.
42:56>> Well, I I think both things can be true.
42:57Like, a generalized model can drive a
42:59car. It's it can build a It can be a
43:01robot. It can build a robot. That's
43:02absolutely true. A specialized model,
43:05like if you said, "Hey, it completed the
43:07course on its second try." Elon's going
43:08to be like, "Yeah, your Tesla can't do
43:10that on a second try. I'm sorry. That
43:11that's not going to work for most
43:13drivers." But it I think the the thing
43:15that's
43:17the thing that's missing in the
43:18storyline there is how much compute did
43:20you use to do the task? Compute is going
43:22to be forever starved from here forward.
43:24Chip prices have gone up. Like HBM RAM
43:26has gone up 5x in price. It's crazy. But
43:30that's because the AI is so incredibly
43:32valuable. People want the compute. And
43:34so if you used more compute to do the
43:37same task, that's a crime.
43:39>> I think this is a really important point
43:40that we're shifting the bottleneck now
43:42from the models down down the stack to
43:46compute and then eventually energy.
43:48Energy.
43:48>> We're at energy right now. We're we're
43:50we're short like 60 gawatts of energy in
43:53the next two years.
43:54>> Yeah. I'm going to move us to uh a final
43:57segment on science which is something we
43:59all truly love. So a few stories here.
44:01So on Tuesday, Anthropic announced it
44:04had built its own life science research
44:06group and its own wet labs. Again, the
44:09organization saying we should slow down
44:12is building wet labs. Okay, great. it
44:14and it's reported that Claude working
44:16autonomously discovered a previously
44:18unknown enzyme. Uh here's how they gave
44:21Claude one prompt. Quote, "Search a
44:24massive DNA database for interesting
44:26reverse transcriptases and enzymes that
44:29copy RNA or DNA. They had 950 clawed
44:32agents searching for 21 hours and they
44:35found something. They found a DNA
44:38repeating pattern very similar to
44:40crisper. Uh they've since named it the
44:42array associated reverse transcriptase.
44:45They don't know exactly what it does
44:47yet, but they believe it will be similar
44:51to programmable transcriptes
44:54uh and gene editing. You know, so
44:56Jennifer DNA won the Nobel Prize for
45:00discovering crisper along with her co
45:03her co-discoverer. Uh I'm curious are we
45:07just going to see Nobel Prizewinning
45:10work falling on a weekly basis as a
45:13result of this Alex?
45:14>> That is the point Peter of our book
45:16solve everything together that we should
45:18expect it and then the natural question
45:20is well what is Nobel Prize worthy in an
45:23era when you can get a thousand Nobel
45:24prizes in one year we we have a century
45:26of human progress in a period of a few
45:28months. I I suspect Nobel prizes will be
45:30awarded for larger bodies of work. But I
45:33I have to again point to the other
45:35elephant in the room here, which is the
45:37utter hypocrisy of saying AI is going to
45:40kill us all and then opening a wet lab
45:42and then using the AI, Frontier AI in
45:45combination with a wet lab to synthesize
45:47proteins and other molecular assemblies
45:49of unknown function. I think
45:52>> what could possibly go wrong?
45:54>> This is I and and I I think this is
45:56Frontier Labs. This is the revealed
45:58preferences of the Frontier Labs. I
46:00think they don't believe their own
46:01attempts to do regulatory capture and to
46:03form a safety cartel. I think their
46:05revealed preference is no actually full
46:07steam ahead to AI for solving biology
46:09>> and every lab has got a biology effort
46:12in focus right now.
46:13>> Yeah. Can I just feed the pdoom just for
46:15this for a second give people the fodder
46:18to go crazy. So uh once we discovered
46:21crisper right we knew we found out that
46:24we could edit the human genome as you
46:26can as easily as you can edit a word
46:28document or edit software right and when
46:30you learn a new language you learn how
46:32to uh do reading then you do
46:34comprehension then we do writing now
46:36we've learned how to write to the genome
46:38each of us is 50 trillion cells in the
46:40human body governed by what the DNA uh
46:43tells tells that cell to do meaning a
46:45human being is now a software
46:46engineering problem right and that
46:48really you've seen designer babies
46:51coming from China etc. As we merge this
46:53with what's potentially possible with
46:55AI, the possible benefits go through the
46:58roof, the possible downsides go through
47:00the go through the roof also. And we are
47:02now in full for if you if you were in
47:05full Frankenstein mode. So there is some
47:08ammunition for all the P doom people, go
47:09to town with that.
47:12This is how we solve aging. This is how
47:14we solve disease. I don't think we get
47:16to the happy future where the top 5,000
47:19diseases have been solved or longevity
47:21escape velocity has been durably
47:23achieved without super intelligence. So
47:26I I for one am a fan of Anthropic and
47:28all of the other frontier labs opening
47:30their labs because this is how we solve
47:32everything.
47:32>> Yeah.
47:33>> All right. Our last story here is uh a
47:37recent release uh by Sam Rodriguez at
47:39Edison Scientific and Future House. They
47:42put forward what they're calling the
47:44millennium. Can we get up on the screen,
47:45please? Uh, the Millennium Problems for
47:47Biology. Uh, so you know, Alex, you and
47:50I wrote and solve everything. Pick your
47:52targets, build your harnesses, set up
47:54your benchmarks, and go. So, some of the
47:57benchmarks here, you know, they're as
47:59they're saying, okay, math is not only
48:02cook, it's incinerated,
48:03>> thoroughly incinerated.
48:04>> Let's go after biology, right? They're,
48:06you know, some of my favorites here are
48:08Origin of Life, uh, Crier Preservation,
48:11um, you know, regrowing ar limbs, uh,
48:16you know, EMOD, they missed age reversal
48:18on this one.
48:19>> They did, and I think it's good, but,
48:21you know, let's have a hundred of them
48:23and let's chuck a thousand agents at
48:25each and open source everything they're
48:27doing.
48:27>> Yeah, that's what the lab should
48:30collaborate on. You know, uh they they
48:33said that uh uh their harness here is
48:35that these Millennium prizes should be
48:39verifiable in a standard wet lab in a
48:42couple of days. They should be hard to
48:45solve and very easy to check. Dave,
48:48>> that's why age reversal probably isn't
48:49on the list.
48:50>> No, no, no. We we so listen when we set
48:52up the uh $101 million health span X-
48:55prize originally Peter Teal and uh uh
48:57and Aubrey Deg Gray came to me and said
48:59let's do a longevity prize and I said I
49:01don't know how to do it without a 20 or
49:0330 year time horizon. And then George
49:05Church said forget about longevity
49:08measure age reversal. So if you can give
49:10someone a treatment and I can measure
49:11that you're functionally 20 years
49:13younger that's measurable in days.
49:15>> Well my daughter is my go-to on this
49:17because she's over at Madna. she's a
49:18biotech person and and she was like,
49:20"Yep, this is the list. It's perfect."
49:22So that's that's my two cents on it.
49:24Like I think it's just phenomenally cool
49:26to put a framework around progress here
49:27because like you said, the Nobel Prize
49:29is going to become irrelevant. In fact,
49:31as of now, it is completely irrelevant.
49:33Uh so putting a framework on progress in
49:35this area is incredibly important. So
49:37this is
49:38>> Alex, when we have digital twins for
49:40cells, which is I I think the critical
49:42path to solving all disease in in my
49:44mind, the chronology looks something
49:46like this. over the next few years. Just
49:48as we've trained large language models
49:49off of substantial substantially all of
49:52humanity's behavior on the internet off
49:54of text and images, similarly we'll
49:57train digital twins of cells of all the
50:00cells in our bodies off of huge data
50:02sets interventional and otherwise. And
50:04then we'll use those perfect digital
50:05twins of cells to do exhaustive searches
50:08like alphogo tree search type searches
50:10of intervention space to discover how to
50:12cure a disease. You start from a
50:14diseased cell state and you're looking
50:16for a creative strategy, a move 37 if
50:18you will, to get from the diseased cell
50:20state to the healthy cell state. And
50:23what's beautiful about the this approach
50:25to millennium problems from biology is
50:27that, as you say, Peter, these are easy
50:29to verify, hard for now to solve
50:32problems. And that basically turns every
50:34hard biology into a mathematical
50:36conjecture and
50:37>> and a target
50:39>> and a target, but a mathematical target.
50:41So every basically every biology problem
50:43with the benefit especially of these
50:45digital twins is just going to become a
50:47search problem and that means it's
50:48tractable for the first time in history.
50:50>> Selm close us out here please.
50:52>> Biology is cooked.
50:55>> It's true.
50:55>> Yeah.
50:56>> All right. I'm going to wrap our WTF
50:58episode here. Um Immod you're going to
51:00be joining us with Kathy uh just after
51:03lunch. Thank you very much. Give it up
51:05for Immod here.

About this transcript

This page contains the full transcript of Jensen Pushes Back on Doomers, Xi & Trump Talk AI, and “AI” Gets a Rebrand | #294 MOONSHOTS Live by Peter H. Diamandis, generated from the public captions YouTube serves with the video. The transcript has 9,285 words across 1,397 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.