YT YouTube2Text

YouTube2Text → Transcripts → Jensen Huang: The Doomer Hoax, Superintelligence is Here, and The Future of AI (ft. President Trump)

Jensen Huang: The Doomer Hoax, Superintelligence is Here, and The Future of AI (ft. President Trump) — Transcript

by All-In Podcast · 9,018 words · 1,363 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Some people call it vision. Vision is an
0:03awfully big word to me because I I
0:04believe first of all vision matters.
0:07[music]
0:09>> We preempted the weekly show. And
0:12there's only three people we preempt the
0:14show for. President Trump, Jesus, and
0:17Jensen. [laughter]
0:18>> The number one [music] podcast in the
0:20world.
0:20>> That's Jensen Wong.
0:21>> He's the founder, president, CEO of
0:22Nvidia.
0:23>> Whether you know it or not, his
0:24decisions are shaping your future.
0:26>> Nvidia is the most important stock in
0:28this market. Jensen is arguably the best
0:30executive in history.
0:31>> Revenue [music]
0:32exploded 97% year-over-year.
0:34>> Not only is demand already strong, is
0:36actually accelerating. Nvidia is the
0:38only computing platform that is a full
0:41stack AI [music] factory. A GPU is like
0:43a time machine because it lets you see
0:45the future sooner. And if we could see
0:47the future and we can predict the
0:49future, then we have [music] a better
0:50chance of making that future the best
0:53version of it.
0:55>> Please welcome Jensen Hang.
0:58>> [music]
1:01>> Oh, we got a standing O on the way in.
1:04>> Oh, come on.
1:05>> Standing O.
1:07>> Standing O on the way in.
1:10>> There's our guy. [laughter]
1:12>> Ladies and gentlemen, GPU Jesus.
1:19>> They love you.
1:20>> They love you.
1:21>> Thank you. I love you back. Number one
1:23podcast in the world.
1:25>> In the world.
1:25>> Absolutely.
1:26>> Wow. We like the new jacket.
1:28>> Well, you know, you auctioned the open.
1:30>> I just I felt you guys needed some
1:32energy.
1:33>> Yes. This is the
1:34>> I know we're talking about serious stuff
1:36here, but we need to talk about it with
1:38energy. Yes.
1:39>> Let's uh let's start with this uh essay
1:42from this weekend.
1:43>> Which one? [laughter]
1:46>> Let's start with Daario's essay because
1:47>> was Hemingway involved? [laughter]
1:50>> Actually, did anybody run it through
1:51Pangram? I don't even know how much of
1:53it was AI helped, but that was a pretty
1:57incredible thing. And then I think what
1:59a lot of people were surprised by was
2:01the coalesing of the frontier labs
2:03around the essay itself. Just Jensen
2:05unpack what happened, how you read it,
2:08how you interpreted it, and then we'll
2:10get into some details that were inside
2:11of it. But maybe just the highle
2:13thoughts to kick it off.
2:14>> Well, first of all, there were a lot of
2:15stuff in there.
2:16>> Yeah. and and uh uh first there is a
2:20there's a part about about safety which
2:22we have to take very seriously. Safety
2:24is paramount. Obviously um uh safety and
2:28leadership are not false. They're false
2:30choices. You're you're able to innovate
2:33quickly. You're able to execute quickly
2:35and and uh America's able to lead and to
2:39do it safely. I think those are those
2:40are false choices but safety is
2:42obviously important. Uh there's a matter
2:44of of internal control that I think he
2:48was speaking to. Uh obviously the the
2:50coxin uh whistleblower is very serious
2:53matter. When whenever you have a
2:55whistleblower, you got to take it very
2:56seriously. I thought Coxin
2:58[clears throat] had great courage uh to
3:00uh put out put out uh uh what his
3:03concerns were. Um and even then there
3:05were some issues that were kind of
3:06conflated within that. Uh I think the
3:08the the whistleblowing is is fine. I
3:11think the the uh scientific prediction
3:15uh about the future uh is less aligned
3:18because it's not grounded on science
3:20obviously and um uh it it was expressed
3:23by a scientist but it was obviously not
3:24grounded on science and and so I take I
3:26take issue with that but obviously the
3:28whistleblower part of it uh [snorts] you
3:30know I I think there's just a whole
3:31bunch of stuff uh pausing uh uh pacing
3:38those are all the voluntary things that
3:40they could do if they feel that their
3:41company is out of control. Uh if Coxin
3:44saw something, you know, obviously we
3:46didn't we don't know what Coxin saw,
3:48>> but he if he saw that the company was
3:51out of control
3:52>> and maybe it's a transition from uh uh
3:56research to engineering. As you know,
3:58these labs are research are
4:00transitioning from research to
4:01engineering. Extraordinary talent,
4:03extraordinary engineering. Um but
4:05obviously engineering is different than
4:07research. Maybe that transition is
4:09clumsy. you know, I don't we don't know
4:10what what he saw and ultimately only he
4:12knows. Um, but if there was a a matter
4:15of lack of control, that's a different
4:16topic. Um, how should the government
4:18deal with it? Now all of a sudden, uh,
4:20regulation and reg I mean it just covers
4:23everything in one blog.
4:24>> Can you just help us sort of unpack? We
4:25we play we tried to play this game
4:27actually this week on the pod and it was
4:28difficult which is how do you describe
4:31like you know my mom calls me and she's
4:33like Jimoth what is this whole
4:35civilizational death thing? I don't know
4:37how to explain it to her. So when you
4:39have very smart people like that
4:41quantize it and quantify it, I think
4:43that's probably what's perturbing to
4:45some people. They're like, "What does
4:46that mean, 10% of extinction?" Nobody
4:49knows how to explain that to the average
4:51person how that's even possible. Well,
4:53first of all, we shouldn't
4:56uh because it's made up. Uh first of all
4:58I think that [laughter]
5:01[applause]
5:03we shouldn't because we it's made up and
5:05these are these are well educated uh
5:08they're called researchers um obviously
5:12they're working in a lab and so the
5:14confluence of these words and then and
5:17then and then the prediction is alarming
5:19and troubling and it shouldn't be done.
5:21It's it's it's irresponsible. Now the
5:23fact of the matter is let's go back and
5:24look at the real facts. The facts are uh
5:27there was a prediction that in 5 years
5:29time radiology will be completely taken
5:32over by artificial intelligence and
5:33there'll be no radiologists in the
5:34world. That has proven to be exactly the
5:37opposite. We need more radiologists than
5:39ever in the world. However, AI has taken
5:41over radiology completely which is great
5:44is automated scan reading which is
5:45great. Um uh there was a prediction that
5:48within 6 to 12 months, wasn't it just
5:50last year? Within 6 to 12 months, uh 90%
5:54of code would already be generated uh by
5:57AI. That has turned out to be wrong. Uh
6:00within 6 to9 months, that was predicted
6:02last year, 50% of entry jobs will be
6:04wiped out. That has proven to be wrong.
6:07Uh let's see what else. What else has
6:08proven to be wrong? I mean, all of these
6:10predictions have been wrong,
6:12>> right? Well, that GPT2 would be too
6:16unsafe to release. That llama 3 would be
6:18too unsafe to release.
6:19>> Oh, one.
6:20>> Yeah, we've heard the
6:21>> half of white collar jobs would be gone
6:23next year.
6:23>> The jobs apocalypse. Yeah.
6:25>> We have to take accountability. We have
6:28to take account for all of the stupid
6:30predictions that were made,
6:32>> right?
6:33>> Somebody has somebody has to take Yeah.
6:36[applause]
6:37>> And and so we ought to just keep track
6:39of all that. And of course people do and
6:42remind us that those those predictions
6:45are inconsistent
6:47with ultimately America winning the AI
6:50race.
6:51>> The short form for that is some people
6:52are saying you know they say trust the
6:54[clears throat] experts and they used
6:56the analog of co which again started
6:58with people that were researchers
7:00educated people that had an asymmetric
7:02awareness of the thing that the rest of
7:03us did not saying things that ultimately
7:06turned out we find out in facts uh not
7:09to be true. Um, and so there's this war
7:12that's happening right now between the
7:14trust the experts movement and the, you
7:16know, well, let's just look at the
7:17actual history of these predictions and
7:19let's just think more methodically.
7:22Where is this coming from? Because it's
7:25coming from inside the places that's
7:26actually making it. Like what do you
7:28think is the psychological makeup or
7:29what is the real incentive? Maybe it's a
7:31business incentive, maybe it's a
7:32political incentive. Can you just maybe
7:35guess or how do you how do you think
7:36about what's why they're doing this?
7:38Well, first of all, I got to tell you
7:39these [clears throat] are some of the
7:41most consequential companies in history.
7:43Uh, uh, extraordinary engineers,
7:45extraordinary researchers, uh, really
7:47fantastic work. Um, [clears throat] uh,
7:50on the one hand, uh, I work very closely
7:52with them as companies to companies. Uh,
7:55on the other hand, uh, we have to have
7:58conversations like this in public. And
8:00it's really unfortunate. And I I think
8:02that that these these companies um
8:05really ought to be built the way that we
8:07used to build companies, which is in
8:09silence,
8:10>> right? You know, and
8:13so wait, wait, Jensen, you don't allow
8:15anybody in your organization to speak
8:17for the entire organization, especially
8:19when they're having like a bad weekend
8:21or they rage quit. They're they're not
8:22allowed to tweet on your behalf and the
8:24organization's behalf. No, because well
8:27that's that's what they decided when
8:29they came to work for us and we told
8:31them uh these are this is the way you
8:33behave when you work in our company and
8:35and uh if you would like if you like the
8:37culture of our company um uh which as
8:40you know the NVIDIA culture and the
8:41NVIDIA employee base uh incredibly
8:44happy. Yeah. uh they like the fact that
8:46the company is consistent, that we're
8:48stable, that our core values are
8:50consistent with taking care of the
8:52families and creating the conditions by
8:53which they can do their life's work. Uh
8:56that we do meaningful work, we do it we
8:58do it as quietly as we can and uh we
9:01contribute to everybody else's success,
9:02which we're very proud of. And so those
9:05kind of core values people are attracted
9:06to. Um but when you come and work in our
9:09company, there are also some things that
9:11we don't appreciate that you do. Like
9:13for example, we don't welcome uh
9:15political discourse in our inside our
9:17company. Take it home. You guys talk
9:19about politics outside the company. Um
9:22we
9:23>> Yeah. [applause]
9:27>> Uh we we are um the company is an
9:30a-olitical company. You know, we're
9:32bipartisan. We want America to succeed
9:35and and um uh we want we want uh
9:38whatever uh government is in place uh
9:41we'll do everything in our power to help
9:42America succeed. And so so the the
9:45discourse about about uh about race and
9:50religion and politics and all of that
9:53stuff we tell people do it outside the
9:56company. It's not not for us.
9:58>> In terms of u maybe AI regulation then
10:00more narrowly. Um Satya was here this
10:02morning and what he said is you know
10:04before we talk about regulation that
10:06could really styy things why don't we
10:07just get some basics right? Why don't we
10:09get measurement right? Why don't we get
10:10standardization right? Right. Um where
10:13do you land on
10:14>> get engineering right?
10:15>> Get the engineering right. Right.
10:16Translate the research in a more
10:17predictable way so that we're not
10:18fear-mongering. Keep it inside until
10:20we're ready to expose it. Um what do you
10:22think the right response is? You know
10:24Demis had a proposal which was sort of
10:25this more FINRA like organization. It's
10:28not clear what Daria wants. This
10:30transnational mutated thing that has
10:33some sort of control. Where do you land
10:35on this? The sort of perspective of what
10:37what do we need right now?
10:38>> You know, regulation should solve actual
10:41problems.
10:42And so the question is what actual
10:44problems have we enjoyed,
10:45>> right? And and um if you look at look at
10:49the actual problems um all of the actual
10:52problems so far have come from the labs.
10:55And the reason for that, the reason for
10:57that and and just in their defense, the
10:59reason for that is because they have the
11:00most compute,
11:01>> right?
11:02>> And the reason for that is because
11:03they're trying to solve uh the frontier
11:06problems. And so in their defense and so
11:09it's sensible that that um the labs, the
11:13frontier labs will be where the most
11:16danger come from. It is unlikely that a
11:19high school student uh did something
11:21because they just simply won't have
11:23enough compute,
11:24>> right? And so uh it's unlikely that a
11:26startup will be the reason because they
11:28won't have enough compute. It's it's uh
11:30they you know in fact you could look
11:32across the planet and everybody won't
11:33have enough compute with the exception
11:35of the frontier labs. And so so now the
11:38question is if you look at what actually
11:39happened um and they're doing pioneering
11:42work. It's really very hard. Um they're
11:44transitioning from research to
11:46engineering. Um, I could imagine and
11:48they're they're they're they're
11:49obviously building some of the most
11:51consequential technology and companies
11:53in the world. Uh, they're building their
11:55company, they're building their culture,
11:57they're building the technology, they're
11:58building engineering, they're building
11:59products all at the same time. And so I
12:01I can understand it's a little bit hair
12:03on fire. Um, uh, but nonetheless,
12:07the four incidents from one lab, the one
12:09giant incident from the other lab, um,
12:12the first thing that you have to do is
12:14just root cause the problem from an
12:15engineering perspective. what happened,
12:18>> what could we have done differently and
12:21what are we going to in to implement and
12:23institutionalize whether it's technology
12:25or methods or processes and make sure
12:27that we don't let it happen again. Now,
12:30I would bet you money that in every
12:32single one of those cases is within
12:34their control
12:36in the future to prevent it
12:39because the alternative if it's not in
12:42their control and I'm sure that they are
12:44I'm sure I'm I'm sure that I'm sure
12:45those four four incidents won't happen
12:47again. Um they I'm sure they root caused
12:49it and fixed it. I'm sure uh they have
12:52now technology
12:54for you know sandboxes and run times and
12:57monitors and continuous in continuous
13:00monitors and you know and so I'm I'm
13:02certain they have much much better
13:03technology now the alternative is also
13:07unlikely which is for them to say look
13:10we had these incidents after we're done
13:14analyzing it we came to the conclusion
13:16we don't know anything that happened and
13:18we have no idea how to control it and
13:21we're asking society for help.
13:23>> Yeah.
13:23>> Now, if that's the case, then we ought
13:25to, you know, a bunch of bunch of
13:27companies with engineers ought to send
13:29engineers in. I mean, and we should
13:31advise them if we can, but I doubt it. I
13:33think they they have extraordinary
13:34people. They got this handled.
13:35>> But we we're not operating in a vacuum.
13:37[clears throat] David, last night you
13:38informed me that there is a Chinese lab,
13:40the makers of GLM, who are going to put
13:42three billion towards a recursive
13:44self-improvement run. So, maybe you
13:46could tee that up for J.
13:48>> Well, that's what was announced. Yeah.
13:49zpoo.com the founder just raised 5
13:51billion and said that one of their
13:52priorities is going to be trying to get
13:54to recurs you know uh AI that trains the
13:59next AI and to try and automate as much
14:01of that as possible. Um yeah I think
14:04that I mean
14:04>> well this is the new sexy phrase but as
14:06[snorts] you guys know RSI is a
14:09combination of a system of ideas.
14:13It's um it starts everything with in
14:15context stuff. It starts with skills. It
14:18starts with reflection. It starts with,
14:20you know, reinforcement learning and
14:21synthetic data generation. And these are
14:23all very sensible ideas that causes AI
14:27to get better at solving a problem, you
14:30know, over time. And you could also have
14:33uh low rank, you know, all of that stuff
14:36doesn't include the weights. Uh you
14:38could actually improve the weights and
14:39it's called Laura. uh Laura could be
14:42could be improved in synthet synthetic
14:44data generation reinforcement learning
14:45enhance it without training the the base
14:48model itself and then over time uh you
14:50could train the base model again with
14:52all of that experience and and so I I
14:54think I think it's a sensible thing that
14:57that you're going to use the technology
14:59uh to enhance productivity of all kinds
15:02of tasks including building AI. I think
15:05that's a very logical idea and and I'm
15:07I'm certain that everybody is using it
15:09in some degree. It's just this phrase is
15:12now being used um to weaponize the
15:16technology in some way and maybe to turn
15:18the
15:18>> as if it's going to spiral out of
15:20control is the impression they're trying
15:23to give. But you don't believe that's
15:24real?
15:25>> No. No, of course not. And the reason
15:26for that is because you could RSI all
15:28day long inside your company, but when
15:31you release a product, you've got to
15:33evaluate it, don't you? You have to test
15:34it again, don't you? You have to make
15:36sure that there's no regression, right?
15:38And so the basic process of control.
15:43These labs are going to as they move
15:44from labs to engineering, they will have
15:46much much better control,
15:48>> right? And when they have much better
15:50control that and control comes from
15:52methods and knowledge and practice and
15:54tools and technology all of those things
15:56that leads to better control
15:58verification and evals
16:00>> it's going to enable RSI to be done
16:03inside the company and for good products
16:05to be released outside.
16:05>> Let's talk about uh open source for a
16:07second. I mean this hugging face we we
16:09were communicating about this and I said
16:11it's going to be one of the most
16:12consequential
16:13um
16:15acquisitions. I don't even want to call
16:16it a transaction because I think it's
16:18more important than that. Um, give us
16:21your first principles explanation of
16:23open source versus closed source versus
16:25open weights and how the ecosystem
16:26should fit together over time.
16:28>> The world needs both closed models and
16:31open models. Um, you want you want to
16:33use I use as much closed models as I
16:35can. This weekend I I used four of them
16:37and and [clears throat] uh they work
16:39terrifically. They're frontier. They're
16:40great experience. They right they work
16:42incredibly well. They're getting better
16:43all the time. Uh, and and the way I
16:46think about closed closed closed models
16:48is kind of like bottled water. You know,
16:52water is free, you guys. I don't know if
16:54I've told you guys, but water is free. I
16:56I don't want to, you know, burst
16:57everybody's bubble, but water's free.
16:59And this morning, I used a lot of free
17:00water taking a shower. And so, you use
17:04the right water in the right places. And
17:06this is no different than electricity.
17:08This is, you know, this is no different
17:09than all kinds of commodities that we
17:11use in the world. You need both. Now in
17:13the case of open the reason why that you
17:16need it is because it could be for
17:18sovereignty reasons, privacy reasons, um
17:20proprietary technology reasons. Look at
17:23the facts. The facts are in the last 6
17:26months
17:27$400 billion of venture funding went
17:30into AI native companies.
17:32>> 80% of them use open models. If not for
17:35open models, how could they build their
17:38dream,
17:39>> right?
17:39>> Because their dream could be different.
17:41Obviously, it'll be different than the
17:43labs, the frontier labs dreams. And
17:45there's America has so many different
17:47ways to innovate. That's one of our core
17:49strengths. Great ideas just coming out
17:51of the fountain. And and so open models
17:54enables that. Open models enables every
17:57single if we want to win the AI race.
17:59It's not about a few technology
18:01companies winning the AI race. It's
18:03about every company in America. Every
18:06comp, every company, every industry,
18:08every researcher, every teacher, every
18:12student, every startup, everybody wins.
18:16Some of them will use closed models. A
18:18lot of them will use open models.
18:20There's 10 million
18:22>> Does it matter?
18:23>> Well, let me just ask, does it matter if
18:24the model the open models come from
18:26China or the US?
18:28>> Well, we're doing everything we can um
18:31to make a contribution in open models.
18:34However, the moment you download, like
18:37for example,
18:39probably the vast majority of the
18:41world's contribution to open source
18:43today is coming from China. They just
18:45have a lot more engineers. They produce
18:47everything in large scale because it's a
18:49larger country. And so they produce
18:51science and math students in volume,
18:53>> right?
18:54>> That's one of our disadvantages, right?
18:55>> They're manufacturing them through
18:57amazing universities like Chinua
18:59University in high volume. Well, they
19:01contribute to open source today. We
19:03download Linux. We download Kubernetes.
19:06We download all the software. A lot of
19:08it has been touched by Chinese. And once
19:11you download it, it's yours. We fork it.
19:13We improve it. We make it ours. And so
19:16we when you download one of these
19:18Chinese models, it just happens to be
19:20made by some really great researchers in
19:22China, but it's now yours. Whatever you
19:26want to do with it.
19:26>> So what exactly is the race?
19:29the race.
19:30>> Yeah,
19:31>> I think that's that's a really good
19:33point. My point is the race is really
19:36about who exploits the technology best.
19:40You know, the last industrial
19:41revolution, all of the inventors were
19:44Maxwell, Volulta, Ampier. None of them
19:47were American.
19:48They were right. The last industrial
19:50revolution came from Europe. But we
19:53exploited it. We took advantage of it
19:55socially better than anybody else in the
19:58world. Look how it turned out for us. I
20:00want to make sure that this next
20:01generation happens just like this.
20:03>> Yeah. Yeah.
20:04>> So why why are the communists getting
20:06their message out so successfully here
20:08right now?
20:11[laughter]
20:14[clears throat]
20:14>> You know I I think first of all the
20:16narrative is much more practical.
20:20The narrative is much more practical.
20:22Nobody's in China is saying that there's
20:24end of this and end of that and you know
20:28cataclysmic this and you know
20:30>> doom or that
20:31>> doom or that.
20:32>> They're much more pragmatic about it.
20:33They see AI as a technology that's going
20:35to advance their economy, advance their
20:37society and they don't have these these
20:40groups who are basically saying it's
20:42going to end civilization
20:43>> and we're making it up. The part that is
20:45frustrating is if it was true if it was
20:47true then we ought to talk about it and
20:49go do something about it, right? Even
20:51even if it's true, we ought to spend
20:54more time doing something about it than
20:56worrying a bunch of people who can't do
20:58anything about it. It's our job to build
21:00it, right?
21:01>> Has there ever been a point in history
21:03where so many people have so vehemently
21:06said something that is so untrue?
21:08>> And they're measurably they're they're
21:10actually demonstrably untrue and it
21:12actually makes sense as untrue. It's not
21:15based on science. It's not based on
21:17research. Everything that's based on
21:18science and research proves otherwise.
21:20Is it a fear of the frontier? Humans
21:21have never been there. We've never seen
21:23it. Therefore, we're scared of it and
21:24therefore it's easy to tell everyone to
21:26be scared of it.
21:27>> It could be life experience as well,
21:28David. Um, so let me give you an
21:30example. When I first graduated from
21:32school, I was an engineer and I didn't
21:35do that much typing. And the reason for
21:37for that is because I was the first
21:39generation before software software
21:40became popular. We had to go build the
21:42computers to make software pop possible.
21:45Could you imagine
21:47in this generation every single engineer
21:49who came into the world of engineering
21:51you spend all your time typing
21:54literally that's what you do when you
21:56get a job they give you a laptop they
21:58give you a chair and you start typing
22:00you you type all day long you type from
22:02the moment you wake up to the m well
22:04there was engineering before typing
22:06>> right
22:07>> and so so can you imagine that the world
22:10has a mountain of engineering work to do
22:12where most of it is not typing anymore
22:14Sure,
22:16we were we had busy engineers before
22:18typing. I think we're going to do a lot
22:20of great engineering after typing.
22:22>> Yeah.
22:22>> When I say typing, I mean coding. I
22:24mean, and so even at NVIDIA when
22:27software engineers talk to me, I I tell
22:29them, you're just typing. I've been
22:32saying that forever, but obviously for
22:34fun. And I I tell them, my favorite key
22:38is backspace.
22:40And and the reason for that is because
22:42the best software is the smallest
22:44software. Yeah. So I I want you to use
22:46backspace software.
22:47>> Let's actually talk about Nvidia. I
22:49let's let's do a little tear down of
22:51Nvidia. So uh tear down meaning just
22:54explain the pieces because there's a lot
22:55of strategy at play. Let's start at the
22:57absolute bottom. So
22:59>> Oh no,
23:03>> this is not planned, but we know who it
23:04is.
23:05>> Oh no.
23:07No.
23:09Mr. President.
23:12Oh, yes, sir. Um, I gotta tell you
23:15something. It I I uh if it wasn't
23:18because of you calling, I would I'm on
23:21stage with the besties. I'm on stage
23:23with the besties. [laughter]
23:25I'm on stage with the besties. I'm on
23:27stage with Sachs. And
23:29>> yeah,
23:30>> you know, the whole group.
23:33Yeah. Jason's here. Chamat's here. David
23:36and David is here. Yeah. I'm sitting in
23:38front of a few thousand people
23:41and we're talking as it turned out we
23:43were we were talking about you.
23:45[laughter]
23:48Good job, sir. Good job. The fact that
23:50you saw through all of that, I mean,
23:52there's a lot of complexity and the fact
23:54of the matter is you saw through all of
23:55that and and I you know, we're all just
23:57really grateful.
24:00>> Tell them I said hi.
24:02[laughter]
24:04>> Do you want to say hi to the crowd?
24:06Jason would like Jason would like to put
24:08you on the
24:08>> even Jason
24:10speaker mode.
24:12>> How do we put How do we put on pus
24:15>> on? Put him on speaker.
24:16>> Speaker. Yeah.
24:17>> Right into the microphone.
24:18>> Here we're going to get a mic.
24:19>> Hang on a second.
24:20>> Hold on, sir. We're getting a
24:20microphone.
24:21>> Mr. President,
24:23>> you're you're now talking to the planet.
24:25>> You see, the great thing about life is
24:28that Jensen can develop the most complex
24:30computer chip in the world that nobody
24:32can copy for 10 years. But he can't
24:34figure out how to put me on SPEAKER
24:36[laughter]
24:39THING. We have to remember this one. So
24:42interesting the AI. It's almost as
24:45conspiracy
24:46and the happiest group is China and
24:49China is very happy. And I could even
24:51say in the country a lot of states are
24:54happy that weren't going to get anything
24:56because they're being uh inundated by
24:58people that want to be there. But now
25:00all of a sudden you see they're building
25:01in Finland. They want to build one.
25:03Google wants to build a big one in
25:05Finland, which I'm not happy about
25:07because they were unable to get
25:08permitting. And I'm telling you, it's
25:10all a hoax. The data centers are great
25:13and they make people wealthy and they
25:14make states wealthy and it's the oil of
25:17the next 20 25 years. It's bigger than
25:19the internet and the AI, you know, much
25:22more so. And uh they're just playing
25:25right into the hands of a lot of people
25:27that don't want to see it happen. And
25:30that could be political people. It could
25:31also be China. And we're not going to
25:34let that happen. It's a It's a hoax. And
25:37>> you're right. We're not going to let
25:38that happen, sir.
25:39>> No, we're not going to let it happen.
25:41The uh the robots are not going to be
25:44taking over the world. And that's not
25:46going to happen. You know, my uncle was
25:48a the top probably maybe the best of all
25:51time, frankly.
25:52professors at MIT for 41 42 years and
25:57can known as being one of the most
26:00brilliant men and he was he was there
26:02for 41 years as the top he was like at
26:06the top top of the ladder top of did
26:08many things Jensen knows all about it
26:10but did many things so I have a little
26:12genetic uh a little genetic strength if
26:15you believe in the resource [laughter]
26:16theory but I do I have genetic
26:18>> that explains why you know so much about
26:21AI I Yeah.
26:22>> Well, I know about AI. I know I also
26:24have common sense about AI. Uh the
26:26robots will not be taking over. Uh the
26:29AI will not be taking over the rest of
26:31the world. The whole thing is a hoax.
26:33Now, with that, we have to be a little
26:35bit careful. We have to very be, you
26:38know, we have to do things and we have
26:40to do them prudently. But that doesn't
26:42mean we're going to stop industry
26:44because, you know, as we work on the
26:45next 10 years about how to destroy it.
26:48So, I'm with you all the way. I didn't
26:49even know how you felt about it. And I
26:50assumed you felt the same way as me.
26:52>> Yes, sir.
26:53>> And we if we're going to lead and I have
26:55an expression, it's whoever wins AI
26:57wins. That's how big it is. It's bigger
26:59than the internet. And whoever wins AI
27:02wins. And we can't let this kind of
27:03stuff happen. And that includes very
27:06much includes data centers. There are
27:08communities that were dying that have
27:10data centers right now. And now they're
27:11wealthy communities. Really wealthy
27:14communities. We're We're going to make
27:15sure that We're going to make sure that
27:17everybody We're going to make sure that
27:18everybody wins in the AI race in
27:21America. Every industry, every company,
27:23every state, every people.
27:25>> Good. Well, I feel strongly about it and
27:27I have the position that can do
27:28something about it. We're not going to
27:29let that stuff happen. So, I have no
27:32idea who's at the meeting. I have no
27:33idea who the hell I'm talking to, but
27:35I'll see. [laughter]
27:38>> Did you Did you hear that? Did you hear
27:40that? Thousands of people are clapping
27:43for you, sir. All I know if you're there
27:46[applause]
27:48to listen to Jensen, but uh he's done an
27:50amazing job and David has done an
27:52amazing job and good luck to everybody
27:54and uh we're going to stay with the
27:57future. The country has never done
27:58better. We have 20 trillion dollars of
28:00investment coming into the country and
28:02that's as opposed to much less than 1
28:06trillion under sleepy Joe Biden and that
28:08was [laughter] for four years. This is
28:10in one year. So, you know, it's it's
28:12really the country is there's ne the
28:14country has never seen anything like it
28:15and we're going to keep it going. And
28:17so, thank you all very much.
28:19>> Thank you, Mr. President.
28:20>> Mr. President,
28:21>> thank you.
28:22>> I'll call you back later. Thank you, Mr.
28:24President. Thank you.
28:26>> Um I was unique.
28:28>> I thought it was a bit. Did you
28:30[laughter] know that was happening?
28:31>> I thought it was a bit. Yeah, that was
28:33>> it was No, it was real. I thought it was
28:35a bit at first when I was like, put him
28:37on speakerphone. [laughter]
28:39Wow.
28:40and he [clears throat] calls you. How do
28:41you how do you think he calls you any
28:43hour of the night, right?
28:44>> Well, we we were we were in the uh we
28:45were in the oval that time when he
28:47called you
28:49sleeping.
28:49>> You were asleep and he like said, "Wake
28:51him up."
28:52>> I felt I felt so bad because he's like,
28:53"Who's coming to this dinner?" And we go
28:55through the list. He's like, "Well, what
28:56about Jensen?" I said, "No, sir. We I
28:58He's on vacation." Cuz he he had to
28:59postpone this vacation for 5 years.
29:02>> And he's like, "Get him on the phone."
29:03[laughter]
29:04>> What's vacation?
29:05>> But what why do you think he sees
29:07through the hoax? It's it's this is the
29:08thing quite an extraordinary thing.
29:10>> It was it's polling minus 80.
29:13>> So for anyone else that's sitting in the
29:15Oval Office. You're going to do what's
29:17popular. You're representing the people.
29:19This is what everyone wants. They want
29:20to shut down the data centers and AI. It
29:22seems to be the popular thing in the
29:24moment. But he says it's a hoax and he
29:27calls it. How does he do that?
29:29>> I got to tell you, I'm not sure. And the
29:30reason for that is because a lot of
29:31people are falling for it. And so the
29:33fact of the matter is it's complicated.
29:35You know, at first, I mean, if you look
29:37at the story, if you look at the
29:38stories, it's all anchored on two
29:40things. The first thing that it was
29:41anchored on was national security. And
29:44recently, that was all blown blown to
29:45bits, right?
29:46>> And so, no, that story is no longer
29:48anchored on national security. Now, it's
29:50anchored on safety. Now, if you want AI
29:53to be safe, um the first thing is we
29:55need to make sure that the the labs that
29:57are building it are in control, that
30:00they're they're good tests for them. uh
30:02if we would like to have third parties
30:04uh uh to to um uh make sure that a third
30:08party evaluator third party evaluators
30:10are available that's no different than
30:12financial control. You guys know we have
30:13auditors
30:14>> and the auditors are quite quite um they
30:17don't have to be as expert as we are in
30:19our business but they just have to ask
30:21the right questions and um I I think I
30:23heard somebody say that it's good to
30:25have uh independent auditors or
30:27evaluators but they just have to have
30:29multiple. I agree with that too. Just as
30:31there's multiple evaluated and auditors,
30:34it makes sure that one company doesn't
30:36become, you know, pilled or somehow
30:38influenced um for for whatever reason.
30:40And so you, you know, there's a lot of
30:42different ways that you could solve
30:43this. Um and so I think the number one
30:45thing is let's build the technology
30:47safely. Let's make sure that the testing
30:50of it is safe. And I I recognize
30:53completely that that what what is being
30:55built is extraordinary. Um but these are
30:58extraordinary companies and and we ought
30:59to hold them to to extraordinary
31:01standards. Um and they want to be and
31:03they want to be
31:04>> I wanted to go back to open source for a
31:06second. [clears throat]
31:07Um a year ago we weren't taking it very
31:11seriously. It was two years 18 months
31:13behind.
31:13>> The one thing that you know one of the
31:15as you guys know one of the challenges
31:17when you're on the call with President
31:19Trump is hard to say something. Um
31:23[laughter]
31:24I'm going to get in trouble for that.
31:25I'm sure he's going to call me up up on
31:26that. But anyhow, uh what I was going to
31:29tell him and and and all of you is that
31:32AI is creating an enormous number of
31:33jobs. The the the thing that he wanted
31:36more than anything at the beginning of
31:37the the administration and that my first
31:39phone call with him, my first time I met
31:41him is that he wants to create jobs in
31:44America. He wants to re-industrialize
31:46the United States. He wants to make sure
31:48that United States has the energy to
31:50support the next industrial revolution.
31:52Without energy, there's no industrial
31:54growth. And so he wants to make sure
31:56that there's energy growth, that there's
31:58job growth, that they're
31:59re-industrializing
32:01the supply chain. Look at everything
32:02that we're doing right now. All of it is
32:04happening right now as we speak. We're
32:06creating more jobs than ever. We're
32:08creating software jobs. We were just
32:10talking about earlier. $400 billion
32:12dollar of venture financing went into
32:15the AI industry just recently. Yeah. 6
32:17months. Well, that's created a ton of
32:20jobs. That's created a ton of jobs. Um
32:22it's created you know obviously enormous
32:25amount of demand for compute which we're
32:26I'm happy about. Um which is also which
32:29is also creating a lot of demand for
32:30data centers and we ought to talk about
32:32that. I think I was just I was talking
32:34to um uh Governor Abbott uh uh of uh
32:38Texas and he was he was uh he wants to
32:40appeal to the industry to make sure that
32:42we are we are empathetic to the small
32:44communities as we're building data
32:46centers all of all across America just
32:49to be better listeners. Let's actually
32:51talk about that for a second.
32:52[clears throat] That's
32:53>> what's incredible about Nvidia if you if
32:55you break down the component parts is
32:57you've effectively had to become the
33:00bank of AI to get the ecosystem going
33:04and you've had to do it at all the
33:05levels. You know, you just did this
33:06thing with Cloverleaf where you're doing
33:07land powers shell. You did this great
33:09thing with Black Rockck and Goldman and
33:12all these folks to to essentially create
33:14the financing capability. walk us
33:16through your capital allocation strategy
33:18like what has to happen to get a broader
33:22ecosystem folks to be able to come in
33:24and underwrite this next phase.
33:26>> Well, we're we're creating as you guys
33:27know this is a new industrial revolution
33:29and and um every aspect of it is true.
33:32Um this new industry
33:34requires manufacturing just as just as
33:37uh the the the um electricity, internet
33:40and now AI. We power anything, we can
33:45find anything. Now with AI, we can ask
33:49and know anything. Isn't that right? And
33:51so that's our future. We tap into the
33:52ether and we can ask it of anything we
33:54want and it could explain it to us. Now,
33:57in order for that to happen, it's got to
33:58produce the intelligence. And so that's
34:00a production process which is the reason
34:02why this infrastructure has to get
34:03built. But once you get the
34:05infrastructure built, the question is um
34:07what about all of the other layers
34:09across the United States? Uh this
34:11industry isn't just about the model.
34:14It's not just about the chips. It's
34:16mostly about the applications on top.
34:19It's mostly about the infrastructure
34:21layer, the data centers and all the
34:23infrastructure, the the the the
34:25construction, the electricity, the power
34:27generation that all of that is involved.
34:30And so I look across the entire
34:32ecosystem and look for bottlenecks and
34:34if there are places where extraordinary
34:36companies are being built
34:37>> constraints
34:38>> constraints extraordinary companies
34:39being built uh maybe it's uh uh uh
34:43supply chain that has to uh get scaled
34:46up so that when we're ready to deploy
34:49compute that they'll be ready for us
34:51land power shell and so this is no
34:53different than looking at the supply
34:54chain upstream. You know, I I probably
34:57uh think about the long-term supply
34:59chain more than most because our
35:01company's really large and and um in
35:03order for us to succeed, a whole bunch
35:05of companies has to support me. You
35:07know, it's got to uh Corning has to, you
35:09know, Wendle at at Corning has to
35:12support me, Lumenum, and you know, TSMC
35:15of course and memory companies and and
35:17so we started working with all of these
35:18companies long before the revolution
35:21that the the growth came so that the
35:23growth could happen. Now I'm got now I'm
35:26doing a downstream.
35:27>> The compet cycle tends to be though that
35:29the earnings over time over long
35:30stretches of time tends to move up the
35:32stack right towards the application
35:34layer where you can over earn for larger
35:36periods of time. Um I mean you bought
35:39hugging face now you're sort of in the
35:41actively in the serving business. I mean
35:43it seems pretty natural that products
35:47like open router make a lot of sense. It
35:49seems pretty obvious that you know there
35:51are better versions of ways to build
35:52things like bedrock. I'm sure you think
35:54about it. What's the natural conclusion?
35:57Because it seems like the folks up here
35:59have no issue trying to move down.
36:01>> Mhm.
36:02>> And you have the best balance sheet,
36:03these incredible engineers, and you have
36:05the proven experience to make it right
36:08and engineer the product and get it out.
36:10So, how do you think about looking up
36:12and saying, "I could probably do that."
36:14>> The the reason why Nvidia runs every
36:16single model in the world, we were the
36:19only It's incredible. Last year about a
36:21year and a half ago the only thing we
36:23ran was open AI.
36:24>> Yeah.
36:25>> And now look at amazing models are
36:27available. The Metamuse is available.
36:29You got gro is available. Grockbots's
36:31incredible. Um we now run Gemini. Uh and
36:35anthropic is is uh scaling up on our
36:37platform as well. Uh since a year and a
36:39half ago, you got all these frontier AI
36:41models that are now open that are
36:43available. So the number of models that
36:45are are are growing. Um there's a whole
36:47bunch of companies that I won't mention
36:49that are building uh frontier models as
36:51well. And the the number of AI labs are
36:54growing. Yeah. The the ineffables, the
36:57uh the reflections, the right the list
37:00goes on. The physical intelligence, the
37:02list goes on. Okay. And so all of these
37:04labs are building on NVIDIA. And the
37:06reason for that is because as a company,
37:09I rather for us to help everybody
37:12succeed instead of taking a slice out.
37:16And so we would go up as far as we need
37:19to but as low as possible.
37:22>> Our strategy is go up as far as we need
37:25to and as low as possible. And the
37:26reason for that is because if I do that,
37:29if I solved the if if not for Nvidia
37:32creating QDNN, all of the frameworks
37:34wouldn't exist. If not for us creating
37:36Megatron uh megatron core, uh then all
37:40of the large scale training wouldn't
37:41have happened.
37:42>> Wouldn't exist. Um, so we we go and we
37:44invent all the technology necessary as
37:46far as we need to and then we let a
37:48thousand flowers bloom.
37:50>> And so that posture allows us to be
37:52quite frankly the only
37:54>> Well, look, let's be honest that I I
37:55agree with you. The push back would be
37:58that it really would be great to have
38:00more competition at the hyperscare
38:02layer. And I think you've done a great
38:04job supporting the NeoClouds. There are
38:06some. And by the way, I think you
38:07introduced me to NBS. Superb, great,
38:09everything. They're amazing. But we need
38:11like 50 of these guys. We need a hundred
38:13of them. We need a thousand of them. And
38:15it just may take some
38:17>> Yeah. You know, it's just I'm
38:20surprisingly uncompetitive
38:23>> really. Yeah. [laughter] That's not my
38:26thing. You know, my thing is kind of
38:28like for example, I'd be more than happy
38:30with five hyperscalers. However, um the
38:34reason I noticed the early customers of
38:36all the Neoclouds, all the what we call
38:38NCPs, all the early customers were the
38:41hyperscalers.
38:42>> Exactly.
38:42>> And the reason for that is because the
38:44hyperscalers plan once a year, but the
38:47market dynamics is so volatile right now
38:50>> that they're always almost wrong. And so
38:53with all these regional clouds who are
38:56agile and they can move fast, um they
38:59know their state or they know their
39:00country, they know their region, they're
39:03securing land, power and shell in a way
39:05that's hard for somebody who sits in
39:07Seattle or sits in Palo Alto to be able
39:09to see the planet.
39:10>> And so we now have basically a largecale
39:13distributed network of companies that
39:15are building securing land power shell
39:17for us. and [snorts] um uh and and now
39:20countries realize it's strategic.
39:22>> Yeah.
39:22>> So many countries are saying I'm going
39:24to take my power and only give it to my
39:26own companies,
39:27>> right?
39:28>> Well, Nvidia is in that country as well
39:30and we could help the Neoclouds in that
39:32country grow and and so whether it's
39:34whether it's Fermas and Australia, we
39:36just did a whole bunch of stuff in
39:37Australia. Um brought on two more
39:39gigabytes. Uh Southeast Asia of course
39:43IOH and others bring on a few gigabytes.
39:45And so we're building gigawatts. So,
39:48we're building, you know, we're we're
39:49scaling up. You know, it's
39:51>> pretty clear, though. I just want to get
39:52this one thing in. It's pretty clear
39:54that you're going pretty high up and
39:56getting very focused on open-source.
39:58Obviously, you have your Neotrons doing
40:01exceptionally well. I use them often.
40:02Hugging face poolside and Laguna uh very
40:06very solid product that you're now uh
40:08aqua hiring, hiring, whatever it is. Um
40:11and then you have your open source stack
40:12for self-driving also uh very
40:15disruptive. So
40:16>> we are the frontier model in five
40:18domains.
40:19>> Yeah.
40:19>> Yeah. And so
40:21>> you don't seem to build products to get
40:23the silver medal. You seem to go for the
40:26gold. So are you going for the gold? And
40:28will you have the best hands-down
40:30open-source model? And then
40:34part B to that is can open source catch
40:36up to frontier models and are you the
40:38person to do it? So the logic the logic
40:40Jason is that that um we will build it
40:44because one uh we can we have the skills
40:47to do it and because our customers need
40:50us to do it
40:51>> right. So, for example, Alpamo is the
40:54world's first thinking self-driving car.
40:57And by by thinking, by reasoning, you
40:59don't need as much data as, you know,
41:01you don't have to train on a few billion
41:04hours of road data because you could
41:06reason about it. Break down the problem
41:08into I've seen this before. It's not
41:10exactly the same, but it's largely the
41:11same as that. Okay? And so, so Alpamo,
41:14why is it necessary? Well, there's a
41:16whole bunch of car companies. Every car
41:17in the world is going to be autonomous,
41:19but beyond that, every ag tech, every
41:23truck, every van, and most of them
41:26aren't big enough in scale to be able to
41:28build that whole stack. So, I'll build
41:30an extraordinary stack for them. They do
41:32last mile adapting for their
41:34application. Now, everything that moves
41:36in the future could be autonomous. If
41:38not for us building uh some of the some
41:41of the biology models, the world
41:42wouldn't have it. uh the the ESM2 uh
41:45protein found language model we created
41:47that ESM fold open fold alpha fold 2 um
41:51all the stuff with with coup equavariant
41:53um all of that stuff technology wouldn't
41:55have existed if we didn't build it uh uh
41:58one of my favorites uh proteina complexa
42:01uh is you know synthesizing next
42:03generation proteins and it's binding
42:05it's groundbreaking stuff we built that
42:07and so we'll build that because Lily
42:08needs it and and uh you know Merc needs
42:11it and others need it and they don't
42:12have the capab ability to do it or they
42:14they're not yet there and so we can make
42:15a real contribution. So I do everything
42:18out of need. I'm not trying to disrupt I
42:21mean we don't wake up in the morning try
42:22to disrupt anybody.
42:23>> We just wake up in the morning try to
42:25help everybody.
42:25>> Jensen, what about what about
42:27competitive threats that might be
42:28emerging to your core business? Can you
42:30just comment?
42:31>> Just so nice.
42:32>> Yes. Well, I know this is Well, I I
42:34actually want I want to just get your
42:36let's just call it a take. What's your
42:38take on Terraab 100 million square foot
42:40facility Elon's announced and um
42:43>> if anybody could do it he can and the
42:45two of us were on a flight together to a
42:48country and um
42:51>> with a person who sometimes calls you on
42:55the phone. It was it was a nice plane
42:57and and and we had like you know and you
43:02know Elon likes to talk about these
43:04things and and so uh we spent a lot of
43:06time talking about it.
43:08>> I I is that anybody could do it because
43:10I mean you could you design chips you
43:12don't fab them. Could your chips be fab
43:14there or is it
43:15>> Well, we know we we know a lot about
43:17process technology because we're pushing
43:19the limits of everything,
43:20>> right? And you [clears throat] know
43:21because we scale at such large scale uh
43:24we have incredible memory technology
43:26inside the company. We're the world's
43:27best sis company. You know we got lots
43:29of amazing.
43:30>> So your take is you've talked a lot
43:31about it.
43:32>> So we could just yeah we could talk
43:33about it and and um you can't discourage
43:36Elon from doing it which is one of his
43:38incred that's his superpower and once he
43:40decides to go do something it's hard to
43:42stop him. And so I
43:43>> And can can you give us your take on
43:44where China is with advanced lithography
43:46systems? Um native grown.
43:48>> They're going to get there by 2030.
43:50>> By 2030. [clears throat]
43:51>> Yeah. And 2030 is just around the
43:53corner.
43:53>> Yeah.
43:54>> Also, that's how long will all be dead
43:56at that time. So,
43:58>> and does and for China, does that mean
43:59the switch is flipped and then that's
44:01all going to go into um mainland fabs
44:05>> almost immediately?
44:06>> You know, the the way to think about
44:08China is really good at high volume
44:10production.
44:12And this is just matter of time.
44:15>> Yeah.
44:16>> And I you know I I think in I I think in
44:20decades as well you know I've been
44:22around a long time and you know for
44:24Nvidia I've got to think about what
44:25happens next decade and decade after
44:27that. So two or three years is it's just
44:29a click. It's nothing.
44:31>> And so as far as they're concerned
44:32they're already there.
44:33>> They're already there.
44:34>> Yeah.
44:34>> Jensen Elon uh and Gwen
44:37>> we've got to run America. We got to run.
44:40>> Yeah. speedun.
44:41>> So, we got
44:42>> speedun.
44:43>> Slowing down is definitely the wrong
44:45strategy.
44:46>> Well, I mean it [clears throat] it feels
44:48apparent, I think, to most of us in the
44:49industry that we're kind of in the AGI
44:51moment. And it's a definition obviously
44:54just as smart as any other human.
44:55>> I think we're already there.
44:56>> We're there, right? And so then super
44:58intelligence is the next way point based
45:01on what you see, based on your customer
45:02base, based on your history here.
45:04>> But Jason, I think we're there, too.
45:05>> You think we're at super intelligence?
45:06>> Yeah. Yeah. When you when you when you
45:09take a narrow segment
45:11a narrow segment I mean my my
45:13self-driving car I don't want you to
45:14make me an omelette I just want you to
45:16drive the car
45:17>> right
45:18>> that is super intelligent
45:18>> super it's better it's better than a
45:20human
45:20>> yeah yeah
45:22>> onetenth the the accident rate
45:24>> exactly
45:24>> uh synthesizing proteins you know uh
45:28doing virtual screening of proteins
45:29we're already there
45:31>> are you having fun being on the frontier
45:33of humanity
45:36>> I like Yeah, [laughter]
45:40>> ladies and gentlemen.
45:41>> Ladies and gentlemen,
45:42>> I like it. I like it. And guys, guys,
45:44it's it's it's great there. The future
45:47is great and we want to get there.
45:50Listen, ride the bike. A lot of us don't
45:53have to work. But I got to tell you,
45:55it's too good not to be.
45:56>> So fun,
45:57>> right? And so, so I want every we I want
46:00to be there. I want all of you guys
46:01there with me. We're all going to be
46:03there. we're going to be enormously
46:05successful together as a humanity. And
46:08um and in the meantime, uh we got to
46:10encourage them, urge them on. They're
46:12doing really, really important work as
46:14you guys know. And I want them to
46:16succeed. Um I also would love for us to
46:19tone down the the the the drama and most
46:23importantly, we need all of America to
46:25come with us. That's how we make it.
46:28>> Ladies and gentlemen, Jensen Long.
46:30[applause]
46:31[music]
46:31>> Thanks, man. Appreciate you. Thank you.
46:35>> Thank you.
46:36>> That was awesome. [music]
46:38Only your part.
46:39>> That was awesome. That was great.
46:41>> Thanks, guys.
46:43That was great, huh? Great time.

About this transcript

This page contains the full transcript of Jensen Huang: The Doomer Hoax, Superintelligence is Here, and The Future of AI (ft. President Trump) by All-In Podcast, generated from the public captions YouTube serves with the video. The transcript has 9,018 words across 1,363 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.