YT YouTube2Text

YouTube2Text → Transcripts → Extended interview: Nvidia CEO explains why he disagrees with AI warnings

Extended interview: Nvidia CEO explains why he disagrees with AI warnings — Transcript

by CBS Sunday Morning · 7,570 words · 1,117 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Jensen, thank you for doing this today.
0:02>> I'm happy to do it. Nice to see you.
0:04>> It's nice to see you, too. Thank you for
0:05your time. My first question, we've
0:08heard from some of the top CEOs, Dario
0:10Amadeday, Sam Alman, Elon Musk saying
0:13that we need to pace the development of
0:15AI. You've got whistleblowers saying
0:17that the end of humanity could be within
0:2110 years. Do you agree with this?
0:24Well, first of all,
0:26[clears throat and cough]
0:26uh AI technology is is uh uh incredibly
0:30powerful. It could of course do
0:32incredible good and it could be applied
0:34to uh uh terrible applications as well.
0:38And so so we we have to make sure that
0:40when we're developing this technology,
0:42we develop it safely. And the industry
0:45is transitioning from uh developing an
0:49AI technology that was that was that is
0:52becoming capable meaning that uh this
0:55this first part of this year we went
0:56through a transition where AI became
0:58useful and it became useful uh the
1:02demand for it is incredible. Uh token
1:05generation of these AI products are now
1:07profitable and actually incredibly
1:09profitable. So many companies are
1:11adopting it. Uh, OpenAI's latest Astra
1:13is incredible. We're using it all over
1:15the place. We use cloud code. We use
1:17cursor. Uh, SpaceX SpaceX AI's product.
1:22Uh, we use um, uh, cognition. And so the
1:25capability of the technology has now
1:27reached extremely useful state. And so
1:31these companies are transitioning from
1:34being a lab trying to get the technology
1:36to work to now a products company. And
1:40so in fact you know even calling it a
1:42lab kind of tells you something. They
1:43were in research for the first 1015
1:46years and now all of a sudden in the
1:48last six months practically the industry
1:51has shifted to high production ramp
1:54mode. And so so I think that all of the
1:57all of the um the reactions um that are
2:01happening are part of that transition at
2:04its core.
2:05>> Are they correct assessments of what
2:08could happen? Are we all going to be, in
2:10the words of some, gone in a decade?
2:14>> No, that's that's completely false. And
2:16that's that kind of narrative um uh is
2:19dramatic. It gets a lot of attention. Of
2:22course, obviously, it's gotten the
2:24attention of of the press. And um I I
2:28think at its core, they care deeply that
2:30the products are built safely.
2:32>> They're transitioning from a lab to
2:34engineering.
2:35um they have to dedicate a lot more of
2:38their research researchers, a lot more
2:42of their compute to safety. That's
2:46that's um in in the case of developing
2:48products, there's the engine, there's
2:49the design part of it, the capabilities
2:51part of it, and there's the verification
2:53part of it. In the beginning of any
2:55industry, most of the engineering goes
2:57towards making the product work. And
3:00then as the product becomes mature and
3:02it rolls out into society, most of the
3:05engineering is actually dedicated to
3:06verification, testing, evaluation,
3:10verification, benchmarking,
3:12um uh attesting to standards, uh making
3:15sure that that the product that they
3:17ship uh is reliable, um it's functioning
3:21as expected, uh delivers on the promise
3:24of the product. And so it's very
3:26sensible that right now during this
3:29transition they have to move a lot of
3:31their capabilities to AI safety. I think
3:34that that's really where the the crux of
3:36it. Um
3:39any anything all the messaging around
3:41the drama of it is unnecessary.
3:44It's is quite frankly um uh too much
3:48drama irresponsible. It is now grounded
3:51on facts. It's not grounded on science.
3:54>> So it's wrong. Well, it's f it's wrong
3:56because it's not grounded on science.
3:58Their their concern is not wrong. I just
4:00want to make sure that at the core of
4:02it, the concern is not wrong. This is no
4:04different than I just landed from a
4:06plane. I'm really glad that the the the
4:08the entire ecosystem and industry of air
4:12travel takes safety seriously and all
4:15the people who are working around the
4:16clock, they're more people working on uh
4:19airline safety and aircraft safety and
4:22air travel safety than they are working
4:24on making uh new types of aircrafts. And
4:27so it makes sense and and we're we're
4:29just going through the same thing.
4:31>> You've said we don't need any new laws.
4:34We don't need any new regulations.
4:38But you just said here that you support
4:42safeguards in major industries. Why not
4:45support the equivalent of an FDA or of
4:47FAA for AI?
4:49>> Well, before we come up with new new
4:51laws and new regulations, let's apply
4:54the current laws and the current
4:55regulations. There are all kinds of laws
4:58and regulations. You you know we live in
5:00you know the land of laws and is the
5:03reason why we feel safe and secure and
5:05governed and and we have all kinds of
5:08laws and regulations applied to each and
5:11every one of those applications whether
5:13it's food or um air travel or any
5:16>> specifically for AI.
5:18>> Yeah. So what's what's the what's the
5:19missing part? Tell me the missing part.
5:22I'm curious to know your opinion why you
5:24think there are no missing parts when
5:26someone like a Dario Amade or a Sam
5:28Alman is saying we'd like to see
5:31regulation.
5:33>> I think I think the what I was
5:35responding to is in the case of both of
5:37the labs um one lab had four incidents
5:42cyber security incidents uh triggered by
5:45them. Another one uh just had a couple
5:48more. And so so uh before both in both
5:52of those incidents has plenty of laws to
5:54deal with. You have cyber security
5:56unauthorized entry uh uh uh laws. You
6:00have um you have damage uh damage
6:03liabilities. You have you know all kinds
6:06of liabilities associated with cyber
6:08security. Uh you have all kinds of
6:09liabilities associated with products.
6:12And so if your product does harm, if
6:14your product doesn't perform as as you
6:15promised, and if you entered into
6:17contracts with companies for service
6:19level agreements, there are plenty of
6:21laws associated with that. My point is
6:23apply those first and don't let don't
6:26let this this doomsday narrative cause
6:31somebody to relieve them of the laws
6:34that currently exist. Now, go and go and
6:38read between the lines. There's a lot of
6:40there's they're actually not asking for
6:43more laws. They're asking to be relieved
6:46of the laws we do have. And I I think
6:48that that's a problem. We have plenty of
6:50laws. So, let's apply those first.
6:52>> So, do you think then is are these
6:53companies asking to be regulated in a
6:56sense because they don't want to have to
7:00handle this themselves and take their
7:01own responsibility for what they're
7:04building. Well, they're moving from a
7:06lab to a products company,
7:08from a from a uh from a a non uh
7:14customer revenue producing to a customer
7:18revenue producing business. And and that
7:21transition must be extraordinary because
7:24they're doing it so fast. And these are
7:27two of the most consequential technology
7:29companies in history. Uh they're
7:31incredible, incredibly talented
7:33researchers there. They're doing
7:35excellent work, but it doesn't change
7:37the fact. The fact is they're
7:39transitioning from a laboratory into a
7:42product service company generating tens
7:47of billions of dollars of revenues,
7:49expecting it to be hundreds of billions
7:51of dollars of revenues. both of them
7:54increasing their compute capacity at at
7:56a rate no companies have ever seen in
8:00human history. And uh and and and
8:03deservedly so because their products are
8:06good and they're and they're useful and
8:07people want more of it and they like to
8:10generate more revenues and their
8:11investors want to see their growth and
8:14and all of that is good. But there are
8:18laws. There are product liability laws
8:20and those product liability laws ought
8:22to be tested against the current
8:25incidents and um and the company well
8:29then we'll see what happens next. Then
8:30we'll see.
8:31>> Yeah. Yeah. Yeah. Yeah. Do the do the
8:33obvious first. Do the practical first
8:36and then we can talk about the
8:38theoretical later. But don't talk about
8:39the theoretical without doing the
8:41practical.
8:43>> In making the argument for no additional
8:45government regulation. I'm curious to
8:47know.
8:47>> So I'm I'm arguing for exercising
8:50>> the current. Right. So but you are
8:52saying you're not you don't advocate at
8:54this moment for additional government
8:56regulation
8:57>> to do what?
8:58>> Right. So if that's the premise
9:00>> until until we know what's missing.
9:02Until we know what's missing, why add
9:05more laws? So on that note,
9:08>> curious, do you have more faith in the
9:12four or five CEOs who control this
9:15technology than in additional
9:18regulation? Do you trust them to do the
9:21right thing and to
9:25allay some of the fears here?
9:28>> They're surely uh capable of doing the
9:30right things. what they're going what
9:32they're transitioning from and to is an
9:36understandable thing. First of all,
9:38they're all young CEOs, first- time
9:40CEOs. I've been doing this for 33 years,
9:42and I get the benefit of, you know,
9:44hindsight and and to to sit here in the
9:46back seat and pontificate about about
9:48their business. Um, but the fact of the
9:50matter is they're going through this
9:52transition. It's not easy building a
9:53company, building organizations,
9:56building products, building industries,
9:59building a business, going public. All
10:01of that is new and it's all happening at
10:03the same time.
10:04>> And so I'm empathetic to the enormous
10:07the enormity of the challenges that
10:09they're dealing with. And I'm also I
10:13also have a lot of confidence in in um
10:16uh the capabilities of the teams that
10:17they have. These are extraordinary
10:19people. So you trust you trust
10:22>> uh I believe in them. I believe in the
10:24people. I also am empathetic of the
10:26circumstances.
10:28I don't approve of uh deflections and
10:32distractions about the end of the world.
10:36Um and and I rather them focus on the
10:39things that I know they're focused on.
10:41Their engineers are absolutely uh uh
10:44able to understand that that that the
10:48incidents they had requires them to
10:51create more secure sandboxes and
10:55isolation and containment of their
10:57experiments. Um I am certain that they
11:00have learned that they need to monitor
11:02much more closely all of their software
11:05as it's operating during testing and
11:07eval. I am certain that if they were uh
11:11uh if their if their uh software were to
11:14interact with external from their cloud
11:18interact with external clouds uh and
11:21other services that they should let
11:23people know and if incidents were to
11:26happen uh to make sure that they share
11:28it publicly so that the whole industry
11:31can learn from it and we can all improve
11:33ourselves. This is how the cyber
11:35security industry works. You know I just
11:38so you know right now all over the world
11:40there's cyber attacks right
11:42>> and there's all over the world engineers
11:45are working feverishly to patch their
11:47software improve their software improve
11:49vulnerability. These are incredibly
11:51complex systems. The fact that we have
11:53open software and we have uh open-source
11:56software so that all the companies could
11:58defend themselves and upgrade their
11:59software is one of the reasons why the
12:02industry works and it's working right
12:03now as we speak and we watch each each
12:05other's back. there are hundreds of
12:07companies that work together as a
12:09community and um and you know we have
12:12the as asymmetric benefit of having a
12:16lot of defenders against attackers and
12:19and so there's there's a lot of stuff to
12:21to uh to unpack about the incidents they
12:25had but it doesn't change the fact that
12:27they're currently doing they're back to
12:30grounding themselves in engineering.
12:32They're going to improve their
12:34engineering. They're going to invest a
12:37lot more in evaluations and benchmarking
12:40and testing and that's just a good
12:43transition for them.
12:46>> I'm curious, you know, as Americans look
12:47at you, you're the CEO and founder of
12:49Nvidia, this incredibly the most
12:52valuable company in the world. Why
12:55should on this specific issue of safety,
12:58>> why should Americans trust your
13:00perspective when your company's success,
13:03your net worth is directly tied to
13:06selling chips?
13:09>> Our company's success is directly
13:12connected to the safe deployment of
13:16products and services. If the computer
13:19industry doesn't provide value to the to
13:22the world safely and productively um
13:26provide enormously positive impact as we
13:28have so far if we don't continue to keep
13:31do if we don't continue to do that our
13:33value would would be diminished and so
13:37it is in our very personal and
13:40industrial benefit that this new class
13:44of applications called AI is able to be
13:46produced and deliver to to the world
13:49safely. We have to do this, you know, we
13:52have to we have to balance this
13:54responsible optimism. The reason why we
13:57do what we do for a living is because we
13:59believe it's great for the world. We
14:02believe that we want our children to be
14:04part of it. My children are part of it.
14:06Um we want all of our families to be
14:08part of it. We want this future where we
14:11have less human suffering, where
14:12companies could be more productive,
14:14where we could discover knowledge more
14:16quickly. We want all of this in the
14:18future. We We want to deal with aging
14:20populations. We want to have sustainable
14:22energy and use AI market forces to drive
14:26the world into sustainable energy. We we
14:29want all of these things, but we have to
14:32do it responsibly and we have to do it
14:34safely. And we have to do this all with
14:38maturity so that we don't scare other
14:40people out of scare scare people uh with
14:43dramatic and historical um prediction of
14:47the future. That is just not true.
14:50>> The elephant in the room here is the
14:52race with China. H
14:55>> the argument has been we can't slow down
14:58AI progress lest we lose to China.
15:03>> Do you think that's a valid argument?
15:08>> Um
15:11as an advice to every single CEO in
15:14every industry, every student, every
15:15professor, every member of the American
15:18society, we want America to be as
15:21prosperous as we can.
15:23The more prosperous we are, the richer
15:26we are as an economy, the more we could
15:29provide for social security, the more we
15:32could provide for national military
15:35security, the the better we could be a
15:38citizen of the world and support other
15:41people's challenges and trials and
15:44tribulation and pain and suffering, um
15:46the more philanthropic we can be as a
15:48nation. And so I would advise that every
15:51single American citizen, everybody in
15:54America pulls in the same direction as
15:57hard as we can.
15:58>> So should we slow down and would that
16:01affect the race with China? The American
16:04standing in
16:05>> we should go as fast as we can
16:07irrespective of anybody else. It's not
16:10um if China were to discover a cure for
16:13cancer, it's great for us. If China were
16:16to uh invent a new technology for
16:18sustainable energy, it's great for us.
16:21If China were to uh uh lead in many
16:24areas of science and math and we were
16:27able to learn from them and benefit from
16:29it, it's great for us. But listen, we're
16:32packing heat as a nation. This is a
16:35nation with extraordinary capability,
16:37extraordinary innovators. I'm proud to
16:40be part of the American technology
16:43industry that I would consider
16:45one of the three pillars of our national
16:48treasures. Of course, our financial
16:50system is one of them. Of course, our
16:52military is one of them. You know, the
16:54the the the industry that I serve, that
16:56I'm part of, the technology industry is
16:59absolutely one of our national
17:00treasures. We are going to go as fast as
17:02we can, but
17:04>> how important is it? We're going to go
17:06as fast as we can, but we would never
17:09ever and never should
17:12ship products before they're ready,
17:14deliver products that are unsafe, and
17:16nobody's expecting us to do that. But
17:18everybody's expecting us to succeed and
17:21help America be as prosperous as
17:23possible. David Saxs told me earlier
17:25this week that he believes that if we
17:28slow down the rate of AI development,
17:30then that could do huge damage to the
17:33American economy, the American military
17:36and put the United States at a
17:37disadvantage. So I just want to clearly
17:40understand like
17:42do you think that
17:45we should be pacing the development of
17:47AI right now given where we are visav
17:50China? First of all, David is incredibly
17:53smart. Um,
17:56this is thoughtful. He's deep. He's
17:59pragmatic. Um, he is he is one of the
18:03one of the great American patriots.
18:07Um, it's been a privilege to to know him
18:09and um I met him when he was an adviser
18:12at the White House and um uh incredible
18:16incredible human. Um, I'm not exactly
18:19what what the context of what he said
18:21was, uh, but I'm fairly certain he wants
18:24America to go as fast as we can, but not
18:27faster than we should. We should go as
18:30fast as we can, but not faster than we
18:31should. And if we're if we're
18:33compromising safety, that can't happen.
18:37Um, if you're compromising safety and
18:39you deliver products that are unsafe,
18:41that put people in harm's way, we have
18:44laws of all kinds to come after you. And
18:48so the companies that are developing
18:49these technologies are perfectly
18:52incentivized to do the right thing. And
18:55I have every confidence. I have every
18:57confidence because they're incentivized
18:59to do so. They have the capability to do
19:01so. that we will deliver, we will create
19:05AI, advance AI in a way that's safe to
19:08society.
19:09>> Dario Amade says we should stop selling
19:12chips to China. The quote is do not sell
19:15powerful AI chips or semiconductor
19:17manufacturing equipment to China. What's
19:20your response to that?
19:21>> Well, um we have to we have to make sure
19:24that that uh technology that's created
19:26by America advantage America first.
19:30Every country would do that. Every
19:32company does that. Every parent does
19:34that. Every company does that. Every
19:36industry does that. Every nation does
19:38that. And in the case of in the case of
19:40the American America has decided, the
19:43United States has decided and and I'm
19:45very supportive that we provide Nvidia's
19:48best technology to American companies
19:50first. And so we're in the generation of
19:53what we call Vera Rubin. Vera Rubin is
19:55shipping in volume production. Um we are
19:57several years ahead of China with this
20:00technology and um it's being made
20:03available to uh to Dario and uh Sam and
20:07you know the American labs and uh
20:09American industries and that's exactly
20:11what we're doing and we're we're
20:12delighted to do that. Um on the other
20:15hand remember that AI is a five layer
20:18cake. It is not just that model. The AI
20:21industry represent is represented by the
20:24energy sector, the chips sector, the
20:27model sector where where they are, the
20:28model layer where they are. Um, you
20:30could argue that there's six layers or
20:32five, but above that is basically the
20:34data and the application layer. And we
20:37need every single one of these layers,
20:40everybody in America to win. We want us
20:43to win in the energy sector. We'd like
20:45us to lead the world in the chip sector.
20:48Uh, we like to make sure that every
20:50single application company
20:52I love the fact that FedEx is going to
20:54be an AI company. I love the fact that
20:56Walmart's going to be an AI company. We
20:58need every single company in America to
21:01be an AI company. Not just one company,
21:04not just two companies. Remember, you
21:06got to take a step back and remember
21:08what what AI is. AI is intell digital
21:12intelligence.
21:14Why is it how is it possible that every
21:17single company in America is not
21:20supported and have capabilities of
21:22digital intelligence? And so of course
21:24every company is going to be a company
21:27of the future with digital intelligence
21:29or artificial intelligence as part of
21:31the company. And so I want everybody to
21:33win including the chips industry. And
21:35so,
21:36>> so you disagree with Daario's position
21:38that we should not sell?
21:40>> I completely disagree that uh we should
21:43ban China of all American products.
21:47The American company should have
21:48advanced capabilities. But then after
21:51that, every single layer of the American
21:54AI tech stack, including the traditional
21:58companies and the AI model companies of
22:01every single field and every single chip
22:02company should go and serve the world,
22:06compete for the world. Serving the
22:08world, having market have capturing
22:11market share, capturing economic
22:13prosperity, that's all about competing.
22:16America is about competition and we
22:19ought to go and compete and not seed
22:22entire markets.
22:24Our our market is not his to seed. Our
22:28market is not his to seed. We ought to
22:31go compete for the world's market.
22:34>> I want to ask you about President Trump
22:35and your relationship with him. Um
22:38>> do you agree with President Trump?
22:40You're one of his closest adviserss in
22:42tech and AI.
22:43>> He has a lot of adviserss. He has a lot
22:45of advice [clears throat]
22:46>> and he's he's an excellent listener.
22:48>> Um he has a broad range of people he
22:49talks to.
22:50>> I'm curious
22:51>> and I I that's that's one of the reasons
22:52why he sees through these complicated
22:56situations.
22:57>> Well, speaking of that, yeah,
22:58>> you know, he he said this week that he
23:00believes that the concerns, the fears
23:04about AI's dangers are a hoax. The
23:07potential to destroy humanity with AI is
23:09a hoax. Do you agree with President
23:11Trump on this?
23:12>> Oh, yeah. Now you and I started the
23:13conversation just now with exactly this.
23:16I completely disagree with that. I comp
23:18I completely disagree that AI is going
23:21to destroy is going to be the end of the
23:22world um in 2030.
23:24>> Do you agree? The president is calling
23:26this fear a hoax. Do you agree with his
23:29characterization of it?
23:33>> Um
23:34>> I just I was at the all-in summit
23:35listening to you talk to him on the
23:37phone.
23:37>> Well, I don't want to speak for him. I
23:40>> I'm asking you about you. I I believe I
23:43believe the claims of the end of the
23:47world um stirring fear across America
23:54um
23:56and doing it by the people who who are
23:59doing it makes no sense to me. So they
24:01must be doing it for ulterior reasons.
24:04Um that's that's uh uh maybe it's
24:08political, maybe it's otherwise, maybe
24:10just attention grabbing, maybe so that
24:12they could get a great interview with
24:14you. Um I don't I don't know what is the
24:17reason for it, but uh however it's
24:20characterized, uh 2030 is not going to
24:23be the end of the world. There is 0%
24:25chance that's going to be the end of the
24:27world. And so I I think that scaring
24:31people is unnecessary.
24:34it is irresponsible and I don't know
24:36what their their motives are. Um,
24:38however, uh, what he did say and some of
24:41the other things. So, so that's one
24:43thing that he said. The other thing that
24:45he said was was that um, uh, we don't
24:48need more regulations. Uh, we need to
24:51apply the current regulations we have.
24:53That's a very practical, very sensible
24:55thing and and we just spoke about that
24:57as well and I agree with that. He also
24:59said something interesting um about his
25:02ability to make sure that AI does not
25:05have this impact on Americans. Uh do you
25:09agree that President Trump has the
25:11ability to fend off and prevent an AI
25:13crisis while he's in office? You know
25:15him, you talk to him. Does he have that
25:18ability?
25:20uh as the leader of of of uh of America
25:25with all of the laws and regulations we
25:27currently have um and recognizing the
25:30incidents that we now have in front of
25:33us uh I believe we have the necessary
25:36laws and regulations to handle those
25:39things and
25:40>> so he can fend it off in your view
25:42>> I yeah because America has our our
25:46system already has a lot of laws and
25:48regulations that applies to many
25:50different sectors of products and
25:52services and so so I think he's saying
25:55practical things um and uh I and and
26:01like like we've already spoken earlier I
26:04I think we have lots of laws and
26:06regulations and they should be applied.
26:08>> I have two just quick questions for you
26:10here. Um you will be attending the state
26:13dinner with President Xiinping of China.
26:16Is that correct? Next week.
26:17>> Next week. Yes. if you get a chance to
26:19speak to him about this situation that
26:23we're in right now, the one that
26:24Americans are so concerned about, what
26:26will you say to the president of China?
26:28>> Um, I would I would say that that um I'm
26:32delighted to see uh AI's technology
26:35transitioning from from laboratories to
26:39now really useful capabilities and
26:41products and services. And because it's
26:44making that transition, it would be good
26:46for um all countries that are developing
26:49and advancing AI to do so in a way where
26:52we share best practices. Um maybe maybe
26:56set um uh global standards about uh
26:59product expectations.
27:01>> Work closer together.
27:02>> Work Yeah, of course. Yeah, of course.
27:04Work closer together. Um I if we want
27:09look
27:11they want China to be safe. They want
27:14China to have the benefit of AI. They
27:16want China to prosper just as we want
27:19United States to prosper. I'm glad they
27:21have ambitious
27:23expectations and aspirations. I want I
27:26wish every country Don't you wish every
27:29country has aspirations and ambitions
27:32for a better tomorrow? And I love that
27:35about them. Um, as they advance the
27:37technology, uh, we ought to make sure
27:39that we work with each other as much as
27:41we can, as much as we can. There are, of
27:44course, areas where we we can't work
27:46together for whatever reason. Um, maybe
27:48it's militarily, well, it could be
27:50military reasons. It could be um other
27:53national security reasons that that that
27:55um could arise. And uh, and those are
27:58those are for for the governments to
28:00sort out. Other than that, I think that
28:02it would be great for us to work in
28:04every possible way.
28:04>> Um, can I ask you about something about
28:06your your personal um, status here? I
28:08mean, I I would be remiss to we all know
28:10that Jensen Wong is the very successful
28:12CEO of Nvidia. You are a billionaire.
28:15You have said some of your fellow
28:17billionaires here in California are
28:19spending millions to fight Proposition
28:2140, the billionaire tax as it were. You
28:24have said that if California passes it,
28:27you'd be perfectly fine paying it. Why
28:30do you see it differently than so many
28:32of your peers?
28:34>> I don't mind paying taxes.
28:35>> Why not?
28:36>> Um, look, I don't like t I don't like
28:39tax dollars wasted. Um, I like that our
28:42country is efficient and runs
28:44efficiently. U, but I also support uh
28:48all that this country represents. Um, I
28:51I support uh uh all of our government
28:54officials um uh in their civil duties to
28:57uh to make America great and and better.
29:00Uh I love I love the infrastructure that
29:03we've provided for all of our our
29:05citizens. I love what it's created. Um
29:08for me uh you know this is I am the
29:11ultimate American dream and and um if
29:14not for this country the education uh
29:17the ability to create a new company
29:19pursue your dreams uh and and let merits
29:22be uh the reason why you succeed and
29:25your your your endeavor your grit your
29:28innovation let that be the reason why
29:30you succeed. if not for, you know, why
29:33I'm here. I I wouldn't have met all the
29:35people that made Nvidia possible. I
29:37wouldn't have met my wife in school. I
29:39mean, there's the list goes on. I, you
29:41know, I I love I love this country in
29:44every possible way. And it's not
29:46perfect, but it's incredible. And so I,
29:50you know, I've had the benefit of
29:51traveling around the world and I, you
29:53know, sometimes looking back in on your
29:56own country
29:57>> and and um our all of our trials and
30:01tribulations and discourse, you know,
30:04the messiness of of our country in a lot
30:06of ways, that's what makes it beautiful
30:08and and it's what makes it work and and
30:12um so I'm I'm delighted to to uh be part
30:15of it.
30:16>> Even if you might end up paying about
30:18perhaps eight billion dollars over five
30:22years still worth it?
30:24>> Well, I think that that first of all,
30:26I'm glad you said that. Um the fact that
30:29I can afford to pay $8 billion in taxes
30:32over five years is a privilege. It's not
30:36that it it's not the $8 billion. It's
30:38the fact that I can afford to pay$8
30:40billion. And and so so um uh I'm not
30:45afraid of paying taxes. I'm just afraid
30:47of being poor and and [laughter]
30:52I have no trouble contributing to
30:54others. I love paying all I I'm famed
30:57for for saying that I want to pay all of
30:59our employees as much as I can and I do
31:02instead of paying people as little as
31:04possible to get their work. I try to pay
31:06people as much as possible. That's a
31:08that's a privilege of being successful.
31:11and and um I that's you know frankly I'm
31:16delighted to do it. Uh I I I feel it's a
31:19privilege. It's a responsibility and um
31:23nothing would give me more joy than to
31:26pay be able to pay even more taxes.
31:28>> Wait, are you asking to be paying more
31:30in taxes?
31:31>> No, the more you make, the more taxes
31:32you pay. So I'm okay with that.
31:34>> That's true. That's true.
31:35>> I'm okay with that.
31:36>> Well, Jensen, thank you. Um, do we have
31:38time for a couple more questions or I
31:41want to be respectful of your time.
31:43>> I have all the time in the world.
31:44>> Oh, you do?
31:45>> It's my team.
31:46>> Excellent. Well, I think I want to ask
31:49you about something that's really
31:51bothering a lot of Americans right now.
31:53>> There's a lot of people who don't want
31:55data centers.
31:55>> Yeah.
31:56>> In their backyards.
31:57>> Yeah.
31:58>> This is rare bipartisan
32:01unity. The backlash is bipartisan.
32:04NVIDIA powers a lot of data centers.
32:08What do you say to the people out there
32:10who are speaking out and saying we do
32:13not want these in our neighborhoods?
32:16>> Well, first of all, I would say um we're
32:18sorry we didn't come talk to you sooner.
32:22We should have come talk to them sooner.
32:24um before before you even stand up a
32:27giant data center in some community,
32:29some rural rural area, some small town,
32:32you know, I grew up I grew up in a town
32:34uh of 600 people and I went to a school
32:38with 300 kids and those 300 kids with
32:41this that was Onita Onita in Kentucky,
32:44that was it. And if you were uh if all
32:47of out of nowhere a data center were to
32:50crop up right next to the one grocery
32:52store, um I think I I think I would be
32:54alarmed. And and so I I think that that
32:58first of all, we need to go engage the
33:00communities long before we uh set up the
33:04data centers. And I think uh all of my
33:07colleagues uh recognize this today. You
33:10know, it's just all of this happened so
33:11quickly and it happened all over America
33:13so quickly. uh uh sometimes we just we
33:17just didn't have we we didn't realize um
33:20and we didn't get a chance to be polite
33:23to be to be a visitor in some some new
33:26town. Um however I will also say uh um
33:31the data centers are also improving
33:32tremendously and tremendously quickly.
33:35Uh the water the the the cooling system
33:37is now is recirculating water. We use
33:39hot water to cool the data centers and
33:42so the amount of energy necessary the
33:44amount of cooling necessary is
33:46dramatically reduced. Uh the energy that
33:49we we consume uh we should be providing
33:52for the power generation plants. uh we
33:55should contribute in fact even to the
33:56local power grid enhancing it
33:58reliability and one of the things that
34:00United States uh is increasingly uh
34:03understanding using the market dynamics
34:07and the market forces of AI and just
34:09this incredible investment phase the
34:12first time in a 100 years we have an
34:14opportunity to invest in all kinds of
34:16sustainable energies where if if a
34:19particular town would like solar if you
34:21know if it like nuclear whether it's
34:23fusion or fision
34:25um hydro, every single form of power
34:27generation is on the table today and
34:29everybody is being invested in. And so
34:32that's a fantastic moment. If we want a
34:34sustainable future, the next five years
34:37recognize no doubt we'll use much more
34:40fossil fuel. But the the decades after
34:43that, we're going to see all kinds of
34:46applications of sustainable energy
34:48because of artificial intelligence. And
34:50so so I would say I would also say that
34:54there are so many great stories of towns
34:57that have been uh at the benefit of of
34:59investments from the data center
35:01builders. Uh you know their community
35:03centers improved, new schools, uh new
35:06baseball fields, new roads, uh uh lower
35:10property taxes, all of that is on the
35:12table. And so I think that that um uh
35:15data center builders should go to go to
35:18the communities, let them know that
35:20they're considering building in the
35:21community, let them know that the
35:24conversation includes an opportunity to
35:28upgrade their their town, uh lower their
35:31taxes, lower their energy uh bill and um
35:36uh and there should be some minimums.
35:38There should be some expectation about
35:40you know the the aesthetics of these
35:42data centers. I think they ought to be
35:44they ought to be designed better. They
35:45could be designed better.
35:46>> New a new approach to designing the data
35:49center. Although the main concern really
35:51is about
35:51>> making making something beautiful is is
35:55um just part of of uh the equation. I
35:59think that today nobody considers the
36:02aesthetics of a data center to be part
36:04of its part of its functionality. Well,
36:07people think a lot about the energy
36:08bills and the the water and the
36:11resources that go into it. That seems to
36:13be the primary concern from people who
36:14are protesting these data centers.
36:16>> And so those things um
36:20uh water consumption, that's a myth now.
36:23Uh these data centers use less water
36:27evaporative. They don't use evaporation
36:29anymore. It uses less water than a
36:31swimming swimming pool evaporating all
36:33year long. And so so I think all of that
36:36is kind of historic understandable but
36:40it's a myth now. Uh energy energy bills
36:43going up is actually the o the opposite.
36:46Most of these data center all these data
36:48centers uh should contribute to the grid
36:51and and and um if they contribute when
36:54they contribute to the grid it makes the
36:56grid more resilient. They reduce the
36:59power cost. Um,
37:01>> it sounds like you're trying to say that
37:03people should reconsider having these in
37:05their neighborhood,
37:06>> but that we should you should still do
37:07it nicely. You should still make it more
37:10aesthetic. The setback should be uh
37:12further away from the towns. Uh because
37:15you know the these data centers, there's
37:16a lot of traffic in and out. There's
37:18noise and so on so forth. And so you
37:20just got to put it you have to be
37:21sensible about these things. We have,
37:23you know, we should we should have we
37:25should have um uh very early on gone and
37:31and had dialogue with the communities
37:33before we showed up. It just happened so
37:35quickly and so uh for that for that um
37:39it's unfortunate. Uh but going forward
37:42we can do a much better job.
37:44>> You started this company you started
37:46Nvidia as a gaming chip company. M
37:49>> you were sitting in a Denny's
37:52>> 30 years ago.
37:54>> I was a dishwasher at the Denny's.
37:56>> Yeah. That's where I I love Denny's. I
37:59love Denny's. America's Diner. Who?
38:02Yeah.
38:02>> Did you ever think when you were a
38:03dishwasher at Denny's that you would end
38:06up on the front lines about a debate
38:08regarding AI and the end of humanity?
38:13>> Uh,
38:14no. I was just really really uh focused
38:17on doing doing dishes uh like no
38:20dishwasher ever did dishes and um I was
38:23the best bus boy of any bus boy that
38:26ever bust tables. Um you know I I
38:29brought enthusiasm to almost every job
38:31that I had. It it didn't matter what it
38:33was.
38:34>> It seems like you're bringing that same
38:35level of enthusiasm to your to your job
38:38today.
38:39>> This is consequential. This is
38:40important. It's an important
38:41conversation. I'm happy we're having it
38:43and and um I want to make sure that that
38:46the American public understands that
38:48this industry will it is our
38:50responsibility to build the technology
38:53safely just as it is the drug industry
38:56to build the drug safely the automotive
38:58the car industry to build the cars
39:00safely the plane industry to build the
39:02plane safely this industry will build AI
39:04safely we have to hold ourselves
39:06accountable to that number one number
39:08two uh don't let don't let the the myths
39:12and the the ridiculous dramatic stories
39:16uh cause us not to be optimistic about
39:18this incredible capability. We do not
39:21want to be the country that invented the
39:25technologies of this industrial
39:26revolution but have this country have
39:29our country be left behind. Don't let
39:31that happen. I don't want that happening
39:33for my kids. I don't want happening for
39:35your kids. Can you explain what data
39:38centers actually do
39:41>> and why
39:43AGI
39:44could be a good thing?
39:46>> Okay. So, so really great question. Um,
39:49first of all, so, so, um, uh, as you
39:52know, several hundred years ago, about
39:54350 years ago, uh, the modern industrial
39:56revolution started. Electricity was
39:59being manufactured. It was a mis it was
40:01a magical thing a almost a mythical
40:04thing where you could have uh water come
40:07into this machinery and this invisible
40:11technology comes out of it called
40:12electricity and we can transport energy
40:15all over the world replacing candles and
40:18and uh and stoves and and wood burning
40:22ovens and so on so forth, steam engines
40:25and and so we we made it possible uh for
40:28this new industry to happened and a
40:30whole industry was built on top of it.
40:32Then the second indust the next the last
40:34industrial revolution uh was really
40:36about the information age and we made it
40:39possible by building these data centers
40:42uh for the internet to be connected
40:44together so that we can uh find anything
40:47>> and so the fir the the previous
40:49industrial order we could power
40:51anything. Now we can find anything and
40:53where we are today we created this
40:55technology called artificial
40:56intelligence where given information
40:59given context it can reason about it can
41:02reason about the information that you
41:04gave it so that it can answer and do
41:08anything. You can even use tools like
41:10for example instead of you and I using a
41:12web browser or you and I using uh
41:14PowerPoint Excel or you know a 3D design
41:18tool called you know AutoCAD or Blender
41:20or something like that. This software
41:23can reason about what you asked it to do
41:27take the context the information that
41:30you need it to do your intentions and it
41:32does the work for you. And the latest
41:34generation from OpenAI called Astra is
41:36incredible. You just tell Astra, "Here's
41:39a picture of my backyard. Um, here's
41:42some other uh here's some examples that
41:44I would like, maybe a pool or a garden I
41:46like to place in it. I want you to now
41:49turn it into a design file for my
41:52gardeners and my carpenters so that they
41:55could do that backyard for me or design
41:58that house for me, um, build that house
42:00for me." And so Astra uses all the
42:03necessary tools and the images that you
42:05gave it and it creates the CAD files for
42:08the house or the kitchen or the
42:09backyard. It's incredible. And so that's
42:12what AGI does. Now what what is
42:16necessary, however,
42:18is we now need a new type of
42:21infrastructure. Instead of a power
42:22generator,
42:24instead of a data center for the
42:25internet, we now need a artificial
42:28intelligence AI factory. It's an AI
42:31factory. Energy comes into it, data
42:33comes into it, and these numbers we call
42:36tokens, but basically reformulated into
42:39intelligence. I know that sounds
42:41complicated, but basically
42:44energy comes in, intelligence comes out
42:46and it it basically is an AI factory.
42:49>> Input output.
42:50>> Input output. It's as simple as that.
42:52And of course, these factories have a
42:54lot of computers inside that Nvidia
42:56makes. And and so now the question is um
43:00uh how big would this industry be? Now
43:03it turns out that just about every part
43:05of the economy and every human uh we
43:08like to know things. We're like we like
43:11to do things. Maybe we like to, you
43:13know, automate our shopping. Maybe we
43:15want to u automate the planning of a of
43:17a vacation or a trip and find us the
43:20best fairs. Um uh so on so forth. I just
43:22gave you the example of maybe we want to
43:24improve our backyard or design a new
43:26kitchen. And so every single industry
43:30has information work to do and with the
43:35world's industries call it a hundred
43:36trillion dollars of economy and out of
43:38that some 15 trillion dollars of it um
43:41would benefit from more intelligence uh
43:45than it currently has. And so these
43:47factories is a layer that goes around
43:51every single country. Every company will
43:53use it. And um that's the reason why
43:55there's so many of these data centers.
43:57This is going to be a multi-t trillion
43:59dollar industry. Now the question the
44:01question then becomes how do we think
44:03about these data centers from America's
44:05perspective. And and so the thing to
44:08realize is that this is the first time
44:10in a 100 years well at least first time
44:12in 50 years that we have an opportunity
44:14to reindustrialize the United States
44:17meaning bring manufacturing back to
44:19United States. Now, of course, this is
44:21manufacturing of something that's
44:23invisible, just like power generation
44:25was.
44:25>> Not exactly your car manufacturing.
44:27That's right. Tradition.
44:28>> It's not manufacturing atoms. It's
44:30manufacturing electrons. Okay. If you
44:32will, it's invisible. And and um uh but
44:36it's manufacturing nonetheless. And it's
44:38creating all kinds of manufacturing jobs
44:40across the United States as you know. uh
44:42the the construction workers,
44:44electricians, the the fitters, the the
44:46designers, the the builders, everybody
44:49is so incredibly busy because they're
44:52building out the several trillion
44:53dollars worth of AI factories across
44:56America. And so this is the part that is
44:58really exciting to me is that this is
45:00the first time in several decades that
45:03the the bluecollar workers, the the
45:06builders, the crafts people are of our
45:08country, this is their great
45:10opportunity. and all of them are
45:12enjoying such prosperity now. And I I
45:14love seeing that we we have a real
45:16shortage of electricians and and and
45:18construction workers. But this is just a
45:21wonderful opportunity for United States
45:22to re-industrialize, be a manufacturing
45:24country again, for us to um uh uh create
45:28opportunities for the blue collar
45:31workers and and all the builders of our
45:33of our of our country and to bring back
45:36building and manufacturing again. And so
45:38I think it's incredible. So it's
45:40creating all kinds of jobs.
45:41>> A lot of people don't see data centers
45:43as a manufacturing story. So I think
45:45it's interesting.
45:46>> That's right. Yeah. It's just a new
45:47version of manufacturing just as I as I
45:50said earlier. Um power generation is a
45:52man it's industry. Power generation is
45:55industrial and you have people in there
45:58that's building these factories and and
46:00uh it has to be built. It's built by
46:02builders and data centers are just
46:04manufacturing the new version. Um, and
46:08this new version and of manufacturing
46:10products is it's invisible. It's
46:12numbers. I know it's it's weird to be
46:14manufacturing numbers. Um, but that's
46:16essentially what it is.

About this transcript

This page contains the full transcript of Extended interview: Nvidia CEO explains why he disagrees with AI warnings by CBS Sunday Morning, generated from the public captions YouTube serves with the video. The transcript has 7,570 words across 1,117 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.