Transcription
Good afternoon, good evening. Um, I am thrilled to be here with all of you, but I am especially thrilled to be here with Jensen Wong. If there's anybody I want to talk to these days, it's this man because he knows how to look around corners when it comes to AI. Nvidia is in a quiet period right now. But what we want to hear from him, I think, is bigger picture. So, let's start Jensen just by taking a few steps back again. Everybody wants to know what's coming with AI. This is the fastest technological revolution we have ever seen. Things happen not on in years, not in decades, but probably, you know, weekly and monthly at this point. Things are changing pretty drastically. You're Wayne Gretzky. Tell us where the puck is going.
>> Well, first of all, uh, let me tell you where the puck came from.
>> Okay.
>> And in just the last couple of years, what happened? Um, I feel like I'm sitting on something that's causing all of this to happen. Is it me?
>> It might be me. Hold on. Better.
>> And so, so what happened was two years ago, Chad GPT came out. And what made Chad GPT revolutionary was the ability to generate generative AI. You give it a prompt, it can write you a story. You give it a prompt, it can make you a picture. You could give it a prompt, generate a video for you. You give it a video, it could generate a story. You can give it an image, it can right so on so forth. You give it an image in 2D, you could and it generates an image in 3D. And so generative AI um the ability to generate has two profound uh capabilities. One, in order to think, you have to generate tokens in your mind inside.
>> What do you mean mo tokens?
>> Tokens. Oh uh you have to generate thoughts. And so the fact that our ability to think and reason requires us to generate thoughts. So the moment we got AI to generate, we realized you can now think. You can now reason. The second thing is in order to use external tools, you have to generate commands. And so when you use the browser, you have to generate words, control something else. those two ideas. The moment that happened, the entire industry raced off to go figure out how to use generative AI to enable reasoning, which came out last year, and then now agentic AI, which is the ability for AI to understand, reason, plan, use tools to do something useful. So in the last several months what happened is the industry realized took the cloud code anthropics cloud code came out and uh it was the first agenting a a agentic system that was able to do really productive work like software coding and software coding is a a a good first place to go but remember what coding is. Coding is the codification of something you want to automate. And how many companies in the world, how many people in the world don't want to codify into a program something you want to automate over and over and over again. So it turns out coding is important for software engineers, but coding is super important for all companies. And this all happened in the last several months. And so AI in the last several months became useful. That's the big idea. The second big idea is that in order for AI to go through understanding, reasoning, planning, using tools to to take action, the amount of computation necessary compared to generative AI is like a thousand times more.
>> Yeah.
>> In two years time, just just think in two years time, the number of cars you need in the world grew by a thousand times. In two years time, the number of airplanes you need in the world grew by 2,000 times. Whatever it is that you want to use as a as a metric,
>> the thousand times difference is incredible. But then you multiply that by the number of people who now want to use it by a hundred times, which is the reason why, you know, GPU consumption is going through the roof and over, you know, even GPUs we sold four or five years ago now are rising in price faster than, you know, good wine. And so It it has defied what every
>> buying Nvidia GPU is like investing in art.
>> It has defied what everybody said was going to happen. Um, if it's a thousand times over the last year, the demand for compute, where do you see it next year? Um, and how far out do you try and figure these things? Well, you know, the the way to reason about this is you come back and you ask yourself um one uh what is the usefulness of automation of intelligence and who could benefit from it and how would you benefit from it? That's number one. Number two, why what kind of infrastructure is necessary to produce intelligence? And this is one of the big ideas in computer science and this is something I reason about. You know, right now I'm talking to you guys. I'm reasoning step by step and and 15 years ago I reasoned about what was going to happen with deep learning which led Nvidia to invest in all that stuff and here we are and so so what's happening is the difference between the way computers are going to work in the future and the way computers work for you right now you just have to pick up your phone and you just when you when you click on the news remember that news article or that video of Becky was pre-recorded she re she you recorded in advance she stored it in the cloud somewhere and then when I click it, I retrieve it. Now, in the future, you're going to still do a lot of that. However, most of the time that you're interacting with your computers, you're going to be giving the computer your intention. Are you asking a question? You would you like to do something? Are you thinking about something? You like somebody to debate with you? You know, is there a big planning thing that you want to do for a vacation or or a wedding or whatever it is? And then you just tell the computer what you want like you talk to a person and it understands your intention, reasons about how to solve it, comes up with a plan, uses whatever tools it needs to use, goes to all these different web browsers and use whatever Excel and you know maybe even use Photoshop and creates wonderful things for you, generates a bunch of images and then comes back with a brochure.
>> Yeah.
>> Right. And so you got to ask yourself therefore if that's the way you use if if the way the first way you use computer is retrievalbased and now everything is generative and it's contextually relevant meaning you can't pre-record every anything that I just said and so therefore the number of computers in the world is going to grow incredibly and this is the reason why I say AI is not just an application AI actually reinvented the computer industry. AI invented a whole new industry. And for many of many of you in the audience, you're you're working with us at the energy level. Um, of course, at the chip level, at the infrastructure level, land power and shell, cloud services, neo cloud, so on so forth. Then it's the model layer. Most most of the time we're talking about model models, but the fact of the matter is without this underlying part there is no models that's useful. And then most importantly is the application layer. The application layer in healthcare in well transportation, retail, you know, you name financial services, all these different industries are now going to get revolutionized by artificial intelligence. And so when you break it up from that perspective and you just kind of reason about it, it is very certain now that like the internet, which is everywhere, you're going to have computers computing like Nvidia's GPUs literally everywhere. so that every time you use the computer, it can generate the proper response to you.
>> When you say, you know, we're going to have all these computers, I think locusts, um, just how much demand there's going to be and how do we possibly keep up with that demand? And you you just mentioned this stacking. It's the five layer cake that you've talked about for a while. Where are the weaknesses? Where are the limiting factors with meeting that demand?
>> It changes all the time.
>> And two years ago we had enough energy but we were really short on chips and it depends on what part of the chips we're talking about. Um, most people think that Nvidia is a GPU company because we invented the GPU. Uh, but if you look at the systems we're creating today, there are seven different types of chips. The computer that I'm talking about is probably twice the width of this stage. When we say Vera Rubin, it's, you know, twice the width of this stage. Each one of the racks is about four or five million dollars. Three tons, one and a half million parts inside one of these racks. And and inside a data center is a football field of these racks. And so these these systems have silicon photonics inside. It's got the most advanced memories. Three threedimensional packaging, liquid cooling, extremely, you know, extremely sensitive uh electronics all over. And so it's really complicated stuff. We we work with just about every chip company in the world, every systems maker in the world. We have the largest supply chain in the world. And so there's there's a bottleneck somewhere. And so it takes a lot of work just to work through all that. And then and then of course um, you know, these days one of the biggest challenges and this is the part that people don't realize about AI. The first thing that AI is doing right now is creating an enormous number of jobs. It is AI creates jobs. AI is United States's best opportunity to re-industrialize ourselves because it takes three types of plants. Chip plants, computer plants, AI factories that we're in. And so three types of manufacturing plants probably several trillion dollars of re-industrialization we can do as a result of this. Nothing is more powerful than using market forces to drive re-industrialization. Like for example back in in the last administration there was this thing called chips act and everybody was reluctant to build the United States. When President Trump came into office, um, we talked and I said, "Hey, listen. Tell you what, I'm gonna give half a trillion dollars of orders. I'm gonna give half a trillion dollars of orders to these suppliers." And I bet you they come to United States. Boom. They all came to United States to build it here. And so using market forces to re-industrialize the United States, it created hundreds of thousands of jobs for the next four or five years. And then and then lastly, um, AI is the world's best opportunity to modernize the power grid. Our the United States power grid is is if anybody's in the audience related to this is a little antiquated and and and you know that. And so we have an opportunity now for the first time to use market forces to invest in sustainable energy. If you want to invest in nuclear, if you want to whatever whatever version of sustainable energy you choose, you now have, you know, plenty of customers who want to pay for it.
>> Are is Nvidia going to be investing in energy because you all have been making sure you're investing in whatever layer, whatever bottlenecks you find along the way. You recently said that this investment that you're making now in Open AI will probably be the last one because Open AI anthropic, they're going to go public. They're not going to need you kind of propping them up. They'll have their own money other places. But what you've done is find where the bottlenecks are and try and invest NVIDIA money there. Is energy a place you would do that or is that such a big investment hole that it's got to come from other places?
>> Uh, if there were good ideas and we could make a unique contribution, I'd be more than happy to. and and um uh but most most of the people who are investing in energy their their um their cycle times are you know their time horizon is fairly long especially the ones that we're really interested in and um uh where we need to invest with respect to energy is probably closer closer to home here in the United States and probably closer in on the horizon to make sure that the land power and shell are s sufficiently funded and maybe we maybe we backs stop some of that just so that they can get the financing going and get power, you know, inserted and um uh but that that's where we're focused. But you're right, we invest in the entire five layer cake and and we're looking at strategic points in there that if we invested $1, it activates AI maybe by $100. And so if we can make that kind of an amplification for the entire e ecosystem, it would be tremendous.
>> So where are you focusing right now in that five layer cake? if if the time to focus on the large language models may be nearing an end, where do you see the biggest choke holds and the places that you would be affected?
>> Great great really great great great question. So um, you you noticed that we invested in the infrastructure layer.
>> Yeah.
>> And at first people were a little you know wondered why we invested in companies like Coreweave and NBS.
>> Yeah. They said it was circular deals. Why are you doing this? It doesn't make any sense.
>> That's right. and we invested say a dollar, they still had to go go raise another nine.
>> Yeah.
>> And so, so we invest some amount and our anchor investment gave all of the investors confidence that we're behind this company. Well, all of that, everybody who invested along with me on Coreweave clearly extremely happy. Everybody who invested along with me on Nibbius incredibly happy. Anybody who invested along with me on NScale incredibly happy. And so, and the reason for that is because we could see the demand and we see the pipeline of opportunities that's coming their way. So, in a lot of ways, you know, we're highly informed investor.
>> That's why I want to know where you're investing next.
>> like I said, you know, and so see Becky is so alert. I when she asks me a question I'm not sure I want to answer, I tell you some history and she wants to see hear the future. Well, here it comes. No, I'm just kidding. And so and so that was my that was the first thing. The second thing and and I'm going to tell you about something that's really big. And so so of course we invested in open AAI, we invested in anthropic and um, but the big thing that's happening in the last three to six months is that both of these companies and most of the AI native companies have turned their gross margins have gone extremely positive.
>> Right. So they've turned the corner.
>> That's right. Now when you're when you're making something and your gross margins are highly highly profitable your goal is to make more of it which is the reason why both open AAI and anthropic are just racing for capacity because their their tokens these numbers these a intelligence that they produce the margins are excellent and so they're excellent cursors is excellent you look across the entire you know AI natives um ecosystem everybody body. And the reason for that is because finally AI has become useful.
>> That's the big idea.
>> And so so I I I'm hoping that that um the AI ecosystem is now on their own.
>> Every level of the cake.
>> Yeah. Yeah. Maybe. We'll see. You know, if I see some good investments, I'm going to not tell any of you first.
>> Okay. Let's get to this idea of AI is amazing. It does a lot of great things. it's going to do even greater things down the road. But then there is a lot of fear-mongering too and there's a lot of worry and you can break this into two camps. It's the AI doomers and the AI boomers and you are probably the leading boomer of what the promise is to come with this. Um
>> I'm the pragmatist.
>> Okay.
>> I'm a pragmatist and and um first of all, Becky, it's our responsibility as the industry to make AI safe. And the reason for that is because only we know how to do that.
>> How do you how do you do that?
>> Well, there's a lot of technology yet to invent, you know, and so this is no different than making making a an airplane safe. And so you have redundant systems. Uh, you have uh um a fair amount of diverse uh diverse sensor systems. And
>> but it's a little different because the airplane can't decide to fly itself into the ground. I mean, you have to do you think about putting maternalistic instincts into these large language models or what are the ways that we make it like us?
>> Uh, well, I was going to say um they're also guard rails.
>> Yeah.
>> And remember the difference between today's chatbots and two year old two-year-old previous chatbots, the guard rails have gotten so much better. And of course, you can't think of, you know, as an engineer, you can't think of every possible way something could not could malfunction. And so, you have to kind of try it out in the marketplace. There's no better way to become a better company or become build better products than having people use it. And that's unfortunate, but true. You know, today's planes and today's cars, today's health care systems can't be as secure or as safe as it is unless people actually use it. And so the guardrail systems are incredibly good. They still have places where where uh uh they, you know, people could could uh cause it to do things that you didn't want it to do. But every time somebody does that, then, you know, companies go go and fix it. And so I think that my my bigger point is that it is the technology industry's job to make it safe. and and of course we as we have to make sure that people understand of the capabilities of the technology and that we use it in a safe safe way. We also need to have relationships with people um so that with other countries so that we agree among ourselves that this technology is really powerful and capable and we ought not deploy these things against each other. And so we do that in a whole bunch of other places in chemicals and nuclear. And so of course we could surely do that do that in this case. But the the big point that I want to make is
>> is it's our job as the industry not to scare everybody
>> but to let everybody know this is important work and we're dedicated to it and we're serious about it and we have to hold ourselves account. The one thing that I am worried about is the worst outcome for AI for our nation is not that another country gets AI. Everybody should have AI. The global south should have AI. Every single company, every single company, well, every single company country should, everybody should have AI. It empowers them. It lifts them. It elevates them. It gives them superpowers. Of course, everybody should have it. My greatest concern is that we scare United States people, all the people that, you know, we're telling these science fiction stories to to the point where AI is so unpopular in the United States or people are so afraid of it they don't actually engage it
>> that we lose our lead as a nation.
>> That's right. Ultimately, you remember United States benefited from the last industrial revolution for a good reason. Not because we invented it, but because we applied it. There there's so many things to dig into on that. The first would be I understand your point that we need to be engaging with other nations and I I believe you're probably referring to China specifically that we need to be able to engage with them and I know your position has been that we should be giving them um the H200 chips, you know, not not necessarily the latest and greatest, but we should be giving them chips so that they are reliant on US companies for some of these things. That makes sense. Um,
>> we should compete globally. America should always have head start.
>> Should they have the latest and greatest chips?
>> No. Um, we have the right, the United States, we're an American company. United States has the right to make sure that and and we're delighted by that and we're huge supporters of it. That the United States has the first, the most, and the best. But simultaneously all American companies should compete globally because remember in the final analysis we're trying to maximize exports. We're trying to maximize American exports. We're trying to increase our revenues and by increasing our revenues, tax revenues. We improve our national security and economic security contributes to national security. Tax dollars helps us with defense. All of that increases national security. American technology has to win across the world at every single layer. Um, if we could export energy, we should. If we can export chips, we should. If we can export infrastructure, we should. If we can export models, we should. And if we can export applications, we should.
>> We we don't export our our best defense products to a lot of countries, particularly if they are are not our allies. Um, where does AI fall in that mix? It is it a weapon?
>> No. And the way you can test it is this a simple test. A 100% of everybody I'm looking at in the audience, and I could I can't see most of you, but 100% of you, even without seeing you, I can tell you you need AI.
>> None of you should have a nuke.
>> I just I just did the test.
>> That is a good line.
>> It's the test. It's the simple test. I I don't think any of you need an F-35.
>> All right. Are there weaponized AI versions? Let's just say mythos. Should we give that to everybody right now? Because the, you know, administration is convinced we should probably keep it in this smaller group for the moment. There has been some talk about rolling it out more widely and there's some hesitation about giving that to everybody because we want to make sure our companies are able to protect themselves before it falls into the hands of bad actors. First take a step back and ask yourself what's mythos? Mythos is a really great great model and but what's really important is mythos is a model that was that was designed for coding. Now remember what cyber security is code.
>> And if mythos can debug software, can test software, can write software, for what reason can Mythos debug cyber security, test for vulnerabilities, so on so forth, because it's just code.
>> But if a hacker gets access to that before the guys in the white hats do, what happens? They're going to find our weaknesses and maybe exploit them. The answer to that, as it turns out, is not another mythos. The way you defend against a super force is not with another super force. It's with an abundance of cheap force. And so the best answer for mythos is actually open source.
>> Okay?
>> Open source. So that so that we have swarms swarms of white blood cells. We have swarms of white blood cells. And these white blood cells are trained to detect and alert us of threats. And the moment that it detects threats, it figures out where the threat is coming from and closes the door. And so it we can't, you know, you can't count on the fact that your AI is better than their AI, but you can count on the fact that you got more AI than they got. that you can count on. And the reason for that is because there are more companies and more front doors than any threat. That thread has to decide which front door to focus on. And so the number of defenders we can have so long as they're open source because open source is cheap. Open source models are very good now.
>> And we can run all these open source models trained on defending ourselves. And that's the swarm, the the dome, if you will, the cyber security dome. And so that's the answer. But the the the doomers want to scare you.
>> I've got the world's biggest weapon. So what's your answer? And you're kind of think my only answer is another big weapon. Well, turns out asymmetry is what you're looking for.
>> Let's go take a step back just in terms of who should be deciding this. I know the industry knows best what's happening. Is there a role for the government to be involved with that? Self-regula, not just self-regulation, but government regulation, too.
>> Oh, absolutely. Uh, where where um every single application, the application of AI in medical imaging systems. No, no question about it. Every single medical imaging system in the future will essentially have a doctor embedded inside, an AI assistant embedded inside. And so it's going to know how to scan you exactly, right? And it's going to be, you know, while it's scanning you is be looking for disease and diagnosing it in real time. And so that instrument needs to be regulated and that AI has to be regulated in exactly the way that medical in instruments are cars. I'm surprised to this point that a self-driving car shouldn't have to get a license. That's a that's a Right. Don't you think?
>> Yeah.
>> Right. If your daughter needs to get a license, you don't think your self-driving car should get a license. Just put it on the road and see if it, you know, drives. makes it go through all the tests and make the teacher sit on the other side and you know yell at it while you're
>> so you don't believe in the move fast and break things sort of way of Silicon Valley because that seems how a lot of this got rolled out.
>> No, I think you should move fast but you shouldn't break things.
>> Yeah. The benefit of moving fast is because technology that is better is safer. I prefer today's car driving to I I prefer being driven in a car today than I prefer being a driven in a car 100 years ago because it's safer. It's got a lot more technology.
>> being driven by a self-driving car. I mean being driven by some
>> self-driving car or even a or even a real, you know, somebody a a human driver, you know. I I think the
>> I don't know why that was so hard to find.
>> It's a tell. Um,
>> it's so yesterday. That's adorable.
>> Okay, let's just talk about the doomers. I I'll get out of this in a minute, but
>> no, I'm happy to stay.
>> But, but I think it's important to address these things because you've got the boomers, you've got the doomers, and most Americans probably fall in the middle. So, they are listening to both camps and trying to figure out Yeah. where they go with this. Hey, hey, we the pragmatists need some air time. As you know, people who are pragmatic, the unless you're extreme, you don't nobody cares what you say.
>> True.
>> Yeah. And yet that's really where the world is,
>> right?
>> This thing I just wanted you to know it. It's not alive. It does has no consciousness. I know exactly what it is. And it's computers and software. Jeffrey,
>> we know how it's built. If we don't know how it's built, how do we keep making it better? The fact that that we don't know what it is, we don't know what it how it works. It's about to be conscious. We don't know how anything happened. Those kind of words just scare people and it's not true.
>> and it, you know, makes our work sound mysterious. Look, I I know you you look at other CEOs and say we probably shouldn't listen to what they're saying on some counts on this, but what about Jeffrey Hinton, who was the godfather of AI and who says that there's a 20 to 30% that it's ends human existence? Is is he completely wrong? And is he completely wrong that there's no chance or is he just wrong in his percentages?
>> Um, he he's he's uh he's completely wrong that that uh um that a whole bunch of smart people aren't working to prevent that from happening. Um, he because you could apply everything that he he said to all the circumstances in history. It's because there are so many good people working so hard to prevent these things. There are so many people working on making cars go faster, but there's 10 times more people trying to make the car safer. There are so many people trying to make make AI smarter. there are 10 times more people working on keeping it guardrailed and safe and you know not hallucinating and and producing useful work and and so so I think the part that is that's kind of missing is they they kind of project themselves that they are the only person who's worried about this they they forget that there's a whole bunch of people in the world whole bunch of computer scientists in the world who are trying to make the world a saf safer place and better place and so with respect respect to with respect to um uh, you know, there's some other things that were said like for example and we have to be careful they they in all of their their good intentions they think they're warning us um, but we have to be careful that if we scare people we're actually hurting us. So let me give you a tangible example.
>> Okay.
>> A tangible example and this is the one that it was the first prediction. Um, a very very well-known very important computer scientist said the first job that's going to be wiped out is radiology.
>> And the reason for that is because computer vision does an incredible incredibly good job studying scans and looking at images and detecting things that we can't detect. Obviously it can. And so so computer vision is now completely superhuman at that one narrow task. No human in the world can do a better job. Stay concentrated for as long anomalies as small. And so today they're the computer scientist is absolutely right. A decade later 100% of radiology is now infiltrated by AI. It is completely integrated into radiology. And so that was completely right. However, what was completely wrong is that radiologists that job as predicted was not wiped out. And the reason for that, surprisingly, it's the opposite to them. To me, it's completely obvious. So, what happened was radiologists could now study more scans. They can take more patients. They could do more scans on the patients, diagnose disease better. They could accept more patients. The hospital is making more money. The radiology department is the best, one of the biggest profit generating centers. Now, as a result, they want to hire more radiologists. Now, if it turned out that everybody listened to him and no, the world has no radiologists, we would be short of this incredible critical resource. We should be telling radiologists, your purpose in life is not to sit in a dark room to look at a workstation to study a scan. Your purpose in life is to work with doctors, help treat patients, to diagnose the disease, make people well. That's your purpose in life. Studying the scan is just a task you do. And so the fundamental thing that everybody is missing, all these computer scientists are going, "Yeah, that job is done. That job is done." is that they misunderstand that the purpose of a job and the task of the job
>> are related, not the same.
>> Yeah.
>> If you were to apply that to me, the task that I do 100% of the time is typing and talking. And talking and typing are both completely automated and completely superhuman. I should be out of a job. And so, yet you and I, you know, we're observing that two of us work harder than ever. I I agree with you 100% on this. I I I think that's right. This is the goal of capitalism to make us more productive, give us more free time, let us find newer, better ways to spend our brain power. I I I think that's the success of a capitalist society
>> to be more ambitious, to be more bold, to strive for more. I think the moment we lose ambition like for example if if the the work that we want to do today is all the work that humanity ever wants to do if this is it then I I would I'll concede okay automation is going to make more and more people unemployed however, as you know, we have so much human suffering we want to go solve we have so many hopes and dreams we want to go after there's so many things we still want to create and so if I just had more time well I'm finally going to have more time
>> I agree with you 100% but because this is happening so quickly is there a b big bigger dislocation than we've seen in the past that leads to greater inequality and what do we do about that
>> so let me give you the the pragmatic answer so the first the first discontinuity as we talked about earlier AI is not the model AI is the five layer cake the first thing that's happened is it's created tons of jobs as you know software
>> data centers
>> data yeah building data centers chip plants computer plants AI factories. Um, all the AI companies are hiring like crazy. Last year, a hundred billion dollars were invested into these into startup companies. The largest investment in human history. Those all went to jobs. AI natives, AI companies start number of software engineering jobs is rising, not declining. While we're sitting here that saying on the one hand AI, the first job, the first thing that AI has done well is software coding. Meanwhile, we're hiring more software engineers than ever.
>> Imagine the conflict. And so, the reason for that is because we now can use AI to do even more. We have so much ambition. And so, we're hiring more people. And so, I I think that that people have people have to reason about these things with a little bit more life experience, with a little bit more wisdom instead of looking at it strictly from the technical viewpoint. You know, the fact that now I see an AI write a complete a program completely by itself, we go, "Oh, that's it. Software engineering jobs are gone." That doesn't make any sense. The purpose of a software engineer is not to code. The purpose of a software engineer is to solve problems, innovate new things. That's their purpose. I never said when I when I was growing up, I said, "You know what I want to do more than anything? I want to type." You know, I was I was I can't I landed in America when I was nine years old. When I landed here, I go, "You know what? My dad, he sent us here so we can type. Type our butts off. We're just going to sit at a desk and we're going to hunch over this little tiny display and we're just going to type every single day. We're going to type. We're going to type from the moment we wake up to the moment we go to bed."
>> That doesn't make any sense.
>> And so,
>> but there I know I'm just having too much fun with
>> there could be some dislocations along like what I I agree with you that there's going to be greater job creation. There's going to be a lot of places where you see these.
>> There's probably going to be some dislocation just like there was when we left agricultural society and went to an industrialized one.
>> Every everybody's job will be impacted.
>> Yeah.
>> If I let me just give you one example. Um, if you're a college student graduating now,
>> yeah,
>> if you graduate and you're not an expert AI user, you're not going to take a job from another kid who's graduating who is an AI expert user. That's a dislocation,
>> a skill that was not necessarily yesterday, today essential.
>> And yet, if you were a young college graduate and you're an AI expert, tell me you're not going to get hired. We're gonna hire. Yeah. Right. Isn't that right? And so all of a sudden the demand for someone who uses AI versus someone who doesn't. That's a classic job dislocation. Now of course there are some jobs that are just basically the task.
>> You know maybe you're just answering phone calls. Now you could use AI to answer that phone call. And you know like when you call a restaurant these days I don't know you know I think there's all AIs isn't it? And so yeah, so the person who's at the reception who picks up the phone in the past to take a reservation no longer has to do that. So they can take care of a customer instead of having that customer wait at a restaurant. And so so I you know I don't know that that every job will be impacted. Many jobs will be created. Some jobs will be will be eliminated, but every job will be impacted.
>> Because we're in California. I want to bring this up and I've been trying to get at this idea. Um, most really wealthy people that I know are freaked out about the wealth tax proposal here in other places. Um, you are not, even though it could cost you about 8 billion.
>> Hold on. Say, what' you say?
>> Say that again.
>> I'm afraid it could cost you eight billion.
>> Let me let me do some factchecking before I answer that question.
>> Why are you not worried about it? What what um is it that you're so worried about so many other things or is it you think it's fair to have that redistribution that comes back or explain.
>> First of all, I prefer lower taxes and higher.
>> Um, however, I also don't mind paying taxes. You know, I love this country.
>> Um, we don't exercise that many tax loopholes. I think once a year we get a a bill, we pay it and it's big and I don't mind it. And um, we Lori and I never one time think about it. We love this country. Um, in a way that's our way of giving back. Uh, I would love California to be better. I would love United States to be better. Um, I would I would love that they would apply $10,000 of the taxes that I paid to fix that one pothole on a 101, but I might if you give me the chance, if they let me, I'll do it myself. and and uh, but but I I want you know, it's a it's fine. It I never once thought about it. And and then another thing is you know when we came out of school we didn't say okay all right states show me all your taxes and I looked at all of them and I go that's the one you won. We came to California because this was the state we chose because
>> we I loved the school that was here. I came to school Stanford here. I love the companies that were here. Um, we love the culture that's here. And and so we chose to come here. And it wasn't because, you know, we traded off among states with lower taxes. But but now I've got a lot of friends who who pay so little taxes that I'm just kidding.
>> Can you name them all?
>> No.
>> No. I love California and and I prefer lower taxes, but whatever they decide to to uh ask me to pay, I'll pay.
>> Let me ask you a question about anthropic. Um, because there has been a debate about, you know, they're caught up with the the Pentagon at this point. It seems like the White House is trying to maybe mend fences to find ways that anthropic can be used uh within the government and within contractors to the government. Is it key to you, do you think, to make sure that that happens for American competitiveness?
>> Um, absolutely. I I hope I hope that the the US government and Anthropic work it out. Um, Anthropic is an incredible company. They have an incredible culture. Um, they have a belief system that's really really deep rooted. uh and uh uh their contributions to AI um cloud code uh agent AI and all the work that they're doing incredible. Uh, we work with them on a technical basis and we work with them on a business basis and I'm delighted by all that. I don't agree with all of their posture and Dario and I are very clear that we don't have to agree on any of that and still be uh civil working with each other and so we are able to keep all of that separate. Here's my belief. Um, my belief is that that uh uh American companies who create technology are uh if the United States government decides to use it uh uh in defense of our nation and defense of my family that um to the extent that they use it uh in a constitutional way uh in a legal way and um uh in defense of our nation they're the fine men's and and women of of the military. Uh, I have every belief uh that they're going to apply it in the right way and that CEOs who are not elected officials. I'm not an elected official. And when we when we go when the United States go to war, I would really appreciate not getting a phone call asking whether my technology ought to be used. And the reason for that is because I would defer to their judgment and if I don't agree with them, uh, I can apply and and um exercise my rights as a citizen to vote next time. And that's how I can protest. Uh, I can speak out loud. I can um uh uh vote as a citizen. I could encourage other people to speak up. But the one thing that we will not do is get in the way of the United States defending our f our families. And so we are not elected officials. And so that's kind of my belief system and I believe that's how democracy works and how the country should work.
>> and um, but otherwise this is one thanks
>> this time
>> it is one extraordinary company though just to put it out there is you know if you if you think about it in the in the in the in the arc of history there's never been a company like this before to have grown from what are they 10 years old or something like that I think is to go from 10 a zero to nearly a trillion dollars in value. Um, at the the the business rate that they have, I mean, they're currently probably 4050 billion dollars annualized run rate. Uh, for a software company to generate the these kind of revenues, the this is historic in many many ways and their contributions to computer science to society is incredible. And so
>> tell us one thing before we go that you're kind of ruminating on right now um that we don't know about but something that's maybe surprised you in the last few months. Um, you you have you have um every reason to be optimistic and and the reason for that is because I every day I mean this morning I woke up I was talking to a professor and then later I was talking to a scientist and and um uh and then I flew over here to hang out with you and and so that's my day and so and in in that time frame uh we we spoke about about um uh AI for open science and the work that that uh AI can finally do. For example, what used to take months for a researcher to explore a new idea, they can now use AI to help them do that research in a day. What used to be months is now a day and you get the same and you know scientists is really a discovery process exploration pushing the frontiers of human knowledge. And so scientists, whether it's in in energy science, in climate science, of course, in biology, and all the places in healthcare and drug discovery, the physical sciences, the breakthroughs are incredible. If you could just see all of the things that I see every day, you would be so fired up, so excited about the future and realize that whatever ambition that you had in the past, the one thing that you have to say to yourself is whatever level ambition you have, it's just not high enough. That's the only change. The fundamental change that we have to make and I have to make is whatever expectations I have for the company, you got to increase it by about a 100x. So if my if people tell me they can do something, I've got a 100x in my head now. And so I've been I've been, you know, if you will, um, really transformed by what I now see that AI can do. And I can't wait for all of you to to enjoy that. It's going to be coming very soon and and in each one of the different fields in science and industry. It's going to be completely revolutionary. It's going to be great. Jensen, when you say it, I believe you because even though the things you're talking about are out there, it feels to me like you generally underpromise and overd deliver. Not the
>> Yeah. And also, Becky, as you know, if you go back in history, uh, most of my predictions have been right.
>> Yeah. Thank you everybody. Thank you, Jensen. Really appreciate it. We hope you enjoyed the discussion. Be sure to utilize the mobile app to stay uptodate on the latest programming changes. As you exit the room, please remember to bring your belongings with you. your best.