Keynote by NVIDIA CEO Jensen Huang at 2024 SIEPR Economic Summit

welcome back everyone after the short break I know that many of you are looking forward to hearing from our next speaker Jensen Wong Jensen is at The Cutting Edge of artificial intelligence and all of the innovation technology and human capital that is needed to support it my good friend and Seer colleague John Chauvin is going to introduce Jensen and I hope he's here somewhere so I'm just going to keep talking and then the two of them will have a conversation before taking some of your questions John chovin certainly requires very little introduction to many most In This Crowd as my predecessor as the Tron director of seer John is the one who started the SE economic Summit 20 years ago so I would just like right now for all of us to give John chovin a huge round of applause and appreciate the community that he had the foresight to build uh for those of you who haven't been touched by John's research his mentorship or his friendship here's what here's just a snippet of what you might like to know about him along with being the former Seer director and a Seer Senor senior fellow Meritus John is the Charles R Schwab professor of Economics he is also a senior fellow at the Hoover institution and a research associate of the National Bureau of economic research he specializes in public finance and corporate finance and has published many articles over the years on social security health economics corporate personal taxation mutual funds pension plans economic demography applied General equilibrium economics and much more uh John isn't one for long introductions but I just will say that if I can be on10th as helpful to my successor as John uh has been to me I'll feel like I've uh succeeded so I will let you read more about his Publications and accomplishments in the programs you've received uh today and so please join me in welcoming our good friend John Chauvin and I'm really looking forward to this thanks wow thank you so I have always thought that the more famous the speaker the shorter the appropriate introduction and if I was to follow that rule I would stop right now and say Jensen Wong but I'm not going to do that um so the Oxford English Dictionary defines the American dream believe it or not it does that and it says that it's a situation where everybody has an equal opportunity for Success Through hard work dedication and initiative and I would like to say that Jensen Wong is an example of the American dream Jensen uh was born in Taiwan came to the US at age nine with his brother not with his parents went to a rough tough School in Kentucky survived that his parents came two years later he moved to Oregon skipped two grades and graduated from high school and went to Oregon state electrical engineering major 150 men and two women he said he was 16 he looked like he was 12 he had no chance with the women well he sort of liked one of them and said why don't we work on homework together did that over and over and over again six months later he after out for a date well he's still married to her so another American Dream now to skip to age 30 he co-founds Nvidia he's the only CEO there's ever been of Nvidia it's had its ups and its down more UPS than Downs it's now the fourth largest company in the world third largest American uh company so that sounds to me like the American dream um I should add that he also got a degree from Stanford master's degree I think he did it mostly at night uh and he was always good with homework at worked with his wife at worked with Stanford uh too um now of course we were here last week Nvidia announced its earnings in the finance crowd this got more attention than the Super Bowl that occurred a couple weeks earlier it was pretty uh amazing uh his company is at the absolute center of the most exciting develop vment I'd say of the 21st century technology development and uh so he's to be congratulated on that let me just say uh he's received a lot of awards a lot of recognition Enid has received a lot of awards a lot of recognition but I should have a short introduction so I'm about to quit I'm just going to talk about one award last month he was elected as a member of the National Academy of engineering this is a pretty prestigious award there are only three that I know of I actually asked chat GPT I didn't get an absolute clear answer how many CEOs of S&P 500 companies are members of the National Academy of engineering but I think it's three and two are in this room anaru Devan of Cadence Design Systems was awarded it last year so the two of them have that in common but let me now just conclude and congratulate Jensen not only on this award but on the amazing success of your company and thank you for speaking to us today at Seer Jensen how it thank you thank you you're here I'm here I guess so okay so why don't you start off with maybe some opening remarks and then I'll ask you a few questions and then then you get the tough questions well I think that after your opening remarks uh it is smartest for me not to make any opening remarks to to uh uh avoid risking uh damaging all the good things you said you know but but um let's see it's it's always good to have a pickup line um and mine was was uh do you want to see my [Laughter] homework and you're right we're married still we have two beautiful kids I have a perfect life uh two great puppies and um I love my job and and uh she still enjoys my homework well if you want I can ask you a few questions then yes please so if in my lifetime I thought the biggest technical development technology breakthrough was the transistor now I'm older than you yeah uh and it was pretty fundamental deal but should I rethink is AI now the biggest change in technology that has occurred in the last 76 years to to hint at my age yeah um well first first of all the the transistor was obviously a great invention but what was the greatest capability that enabled was software the ability for humans to express our ideas algorithms uh in a repeatable way computationally repeatable way uh was a was is the Breakthrough um what have we done we dedicated our company in the last 30 years 31 years uh to a new form of computing called accelerated Computing the idea is that general purpose Computing is not ideal for every every field of work and we said why don't we in invent a new way of doing computation such that we can solve problems that general purpose Computing is ill equipped at at solving and and uh uh what we what we have effectively done in in a particular area of a domain of computation that is that's that is algorithmic in nature that can be paralyzed we've taken the computational cost of computers to approximately zero so what happens when you when you uh are able to take the marginal cost of something to approximately zero some we enabled a new way of doing software where it used to be written by humans we now can use computers to write the software because the computational cost is approximately zero and so you might as well uh let the computer go off and grind on just a massive amount of experience we call data digital experience human dig digital experience called data and grind on it to find the relationships and patterns that as a result represents human knowledge and that miracle happened about a decade and a half ago we saw it coming and and we took the whole company and we shaped our computer which was already which was already driving the marginal cost of computing down to zero and we pushed it into this whole domain and as a result in the last 10 years we reduced the cost of computing by 1 million times the cost of deep learning by 1 million times and a lot of people said said to me but Jensen if you if you reduce the cost of computing your your cost by a million times then people buy less of it and it's exactly the opposite we saw that if we could reduce the marginal cost of computing down to approximately zero we might use it to do something insanely amazing large language models to literally extract all of digital human knowledge from the internet and put it into to a computer and let it go figure out what the wisd what the knowledge is that idea of scraping the entire internet and putting it in one computer let the computer figure out what the program is is an insane concept but you wouldn't ever consider doing it unless the marginal cost of computing was zero and so so we made we made that breakthrough and now we've enabled this new way of doing software imagine you know for for all the people that are still new to artificial intelligence we figured out how to use a computer to understand the meaning not the pattern but the meaning of almost all digital knowledge and everything you can digit anything you can digitize we can understand the meaning so let me give you an example Gene sequencing is digitizing genes but now with large language models we can go understand go un go learn the meaning of that Gene amino acids we digitized you know through Mass Spec we digitized um Pro amino acids now we can understand from the amino acid sequence without a whole lot of work with cryms and things like that we can go figure out what is the structure of the protein and what it does what is this meaning we can also do that on a fairly large scale pretty soon we can understand what's the meaning of a cell a whole bunch of genes that are connected together and this is from a computer's perspective no different than there's a a a whole page of words and you asked it to what is the meaning of it summarize what did it say summarize it for me what's the meaning this is no different than a hard you know big huge long page of genes what's the meaning of that big long page of proteins what's the meaning of that and so we're on the cusp of all this this is just this is the miracle of of what happened and so I would it's a longwinded answer of saying John that you're absolutely right that that that that AI which was enabled by this form this new form of computing we call Accelerated Computing that took three decades to do uh is probably the single greatest invention of the computer of the in of the technology industry this will likely be the most important thing of the 21st century I agree with that 21st century but maybe not the the 20th century which was the transistor which it's got to be close we'll let history decide that's right we'll let history decide could you look ahead you I I I take it that the the GPU chip that is behind uh artificial intelligence right now is your h100 and I know you're introducing an h200 and I think I read that you plan to upgrade that each year and so could you think ahead five years March 2029 you're introducing the H700 right what will it allow us to do that we can't do now um I'll go backwards but but let me first say something about the chip that John just described um as we say a chip all of you in the audience probably because you've seen a chip before you you imagine there's a chip kind of like you know like this um the chip that John just described uh weighs 70 lbs it consists of 35,000 Parts eight of those parts came from tsmc it that one chip replaces um a data center of old CPUs like this into one computer the savings because we compute so fast the savings of that one computer is incredible and yet it's the most expensive computer the world's ever seen it's it's a quarter of a million dollar per chip we sell the world's first quar million dollar chip but the system that it replaced the cables alone cost more than the chip this h100 the cables of connecting all those old computers that's the that's the incredible thing that we did we reinvented Computing and as a result Computing marginal cost of computing went to zero that's what I just explained we took this entire data center We Shrunk it into this one chip well this one chip uh uh is really really great at trying to figure out um uh uh this form this form of computation that that without without without getting weird on you guys um call Deep learning it's really good at this thing called Ai and so so uh the way that this chip works it works not just at the chip level but it works at the chip level and the algorithm level and the data center level it works together it can't it doesn't do all of its work by itself it works as a team and so you connect a whole bunch of these things together and it works at you know networking as part of it and so when you look at one of our computers it it's a it's a magnificent thing you know only only computer Engineers would think it's magnificent but it's magnificent okay um it weighs a lot miles and miles of cables hundreds of miles of cables and and the next one's soon coming is liquid cooled and you know it's beautiful in a lot of ways okay and and um uh and it computes at data center scales and together what's going to happen in the next 10 years say John um we'll increase the computational capability for M for deep learning by another million times and what happens when you do that what happens when you do that um today we we kind of learn and then we apply it we go train inference we learn and we apply it in the future we'll have continuous learning We could decide whether that whatever that continuous learning um result it will be uh uh deployed into you know the world's applications or not but the computer will will watch videos and and new text and uh from all the interactions that it's just continuously improving itself the learning process and the Train the the training process and the inference process the training process and the deployment process application process will just become one well that's exactly what we do you know we don't have like between now and o' in the morning I'm going to be doing my learning and then after that I'll just be doing inference you're learning and inferencing all the time and that reinforcement learning Loop will be continuous and that reinforcement learning will be grounded with real world data that is been um uh through interaction as well as synthetically generated data that we're creating in real time so this computer will be imagining all the time does that make sense just like just as when you're learning you you take take pieces of information and you go from first principles it should work like this and then we we do the the simulation the imagination in our brain and that that future imaginate imag imagin state in a lot of ways manifests itself to us as reality and so your AI computer in the future will kind of do the same it'll do synthetic data generation it'll do reinforcement learning it'll continue to be grounded by real world experiences um it'll imagine some things it'll test it with real world experience I'll be grounded by that and that entire Loop is just one giant Loop that's what happens when you can compute for a million times cheaper than today and so as I as I'm saying this notice what's what's at the core of it when you can drive the marginal cost of computing down to zero then there are many new ways of doing something you're willing to do this is no different than I'm willing to go further places because the marginal cost of Transportation has gone to zero I can fly from here to New York relatively cheap cheaply if it would if it would have taken a month you know probably never go and so it's exactly the same in transportation and all just about everything that we do and so we're we're going to take the marginal cost of computing down to approximately zero as a result we'll do a lot more computation that causes me as you probably know there have been some recent stories that Nvidia will face more competition in the inference Market than it has in the training Market but what you're saying is it's actually going to be one market I think can you comment about um you know is there going to be a separate training chip market and inference chip Market or it sounds like you're going to be continuously uh training and switching to inference maybe within one chip I I don't I don't know why don't you explain more well today today whenever you uh prompt uh an AI it could be chat GPT or it could be co-pilot or it could be uh if you're using a surface nail platform you using mid Journey um using Firefly from Adobe whenever you're prompting it's doing inference you know inference is right so it's it's generating information for you whenever you do that what's behind it 100% of them is NVIDIA gpus and so Nvidia most of the time you engage our our our platforms are when you're inferencing and so we are 100% of the world's inferencing today is NVIDIA now is inferencing hard or Easy A lot of people the the reason why people are picking on inferences when you look at training and you look at Nvidia system doing training when you just look at it you go that looks too hard I'm not going to go do that I'm a chip company that doesn't look like a chip and so there's a natural and you have to in order for you to even prove that something works or not you're $2 billion doll into it yeah and you turn it on to realize it's not very effective you're $2 billion in two years into it the risk the risk of exploring something new is too high for the for the customers and and so a lot of a lot of competitors tend to say you know we're not into we're not into training we're into inference inference is incredibly hard let's think about it for a second the the the the response time of inference has to be really high but this is the this is the easy part that's the computer science part the the E the hard part of inference is the goal of somebody who's doing inference is to engage a lot more users to to apply that software to a large install base inference is an install base problem this is no different than somebody who's writing a an application on on on an iPhone um the reason why they do so is because iPhone has such an large install base almost everyone has one and so if you wrote an application for that phone it's going to have the benefit of it it's going to be able to benefit everybody well in the case of Nvidia our accelerated Computing platform is the only accelerated Computing platform that's literally everywhere and because we we've been working on it for so long if you wrote an application for inference and you take that model and you Deploy on invidious architecture it literally runs everywhere and so you could touch everybody you can enable have greater impact and so the problem with inference is is actually install base and that takes enormous patience and years and years of success and dedication to architecture compatibility you know so on so forth you make completely State of-the-art chips is it possible though that you'll face competition that is claims to be good enough not as good as Nvidia but good enough and and much cheaper is that a is that a threat well first of all competition um we we have more competition than anyone on the planet has competition uh not only do we have competition from competitors we have competition from our customers and um and and I'm the only competitor to a customer um fully knowing they're about to design a chip to replace ours and I show them not only what my current chip is I show them what my next chip is and I'll show them what my chip after that is and so and the reason for that is because because look if you don't if you don't make an attempt at uh uh explaining why you're good at something they'll never get a chance to to buy your your products and so so we're we're completely open book in working with just about everybody in the industry um and and the reason the reason for that our our advantage is several our advantage what we're about is several things whereas you could build a chip to to be good at one particular algorithm remember Computing is more than even Transformers there's this idea called a Transformers there's a whole bunch of species of Transformers and their new Transformers being invented as we speak and the number of different types of software is really quite quite rich and the reason for that is because software Engineers love to create new things Innovation and we want that what Nvidia is good at is that our our architecture not only does it accelerate algorithms it's programmable meaning that that you can use it for SE we're the only accelerator for SQL SQL was came about in the 1960s IBM 1970s in storage Computing I mean sqls structured data is as important as it gets uh 300 zettabytes of data being created you know every couple of years Mo most of it is in sqls structured databases and so so we're we can accelerate that we can Accel accelerate quantum physics we can accelerate shortes equations we can accelerate just about you know every fluids particles um you know lots and lots of code and so what Nvidia is good at is the General field of accelerated Computing one of them is generative Ai and so for a data center that wants to have a lot of customers some of it in financial services some of it you know some of it in in manufacturing so on so forth in the world of computing we're you know we're we're a great standard we're in every single Cloud we're in every single computer company and so our company's architecture has become a standard if you will after some 30 somewhat years and and so that's that's really our advantage if a customer can can um do something specifically that's more cost effective quite frankly I'm even surprised by that and the reason for that is this remember artchip is only part think of when you see a when you see computers these days it's not a computer like a laptop it's a computer it's a Data Center and you have to operate it and so people who buy and sell chips think about the price of chips people who operate data centers think about the cost of operations our time to deployment our performance performance our utilization our flexibility across all these different applications in total allows our operations cost they call total cost of operations TCO our TCO is so good that even when the competitor's chips are free it's not cheap enough and that that is our goal to add so much value that the alternative um is not about cost and and so so we of course of course that takes a lot of a lot of hard work and we have to keep innovating and things like that and we don't take anything for granted but we have a lot of competitors as you know but maybe not everybody in the audience knows there's this term artificial general intelligence which basically I was hoping not to sound competitive but John asked a question that kind of triggered a competitive Gene and I came AC I I want to say I want to apologize I came across you know if if you will a little [Laughter] competitive I apologize for that I could have probably done that more artfully I will next time but he surprised me with a competitive I I I I thought I was on an economic Forum you know just walking in here I asked him I'd sent some questions to his team and I said did you look at the questions he says no I didn't look at the questions cuz I wanted to be spontaneous besides I might start thinking about it and then uh that that would be bad so we're just kind of winging it here um both of us um so I was asking when when do you think and of course it when do you think we will achieve artificial general intelligence the sort of human level intelligence is that is that 50 years away is it five years away what's your opinion um I'll give you a very specific answer but but first let me let me just tell you a couple things about what's happening that's super exciting first uh of course of course um uh we're training these models to be multimodality meaning uh that we will learn from sounds we will learn from uh words we'll learn from uh vision and we'll just watch TV and learn uh so on so forth okay just like all of us and the reason why that's so important is because we want AI to be grounded grounded not just by human value use which is what chat GPT um really innovated I remember we had large language models before but if it wasn't until reinforcement learning human feedback that human feedback that grounds the AI to something that that we feel good about human values okay um and now could you imagine now you have to generate images and videos and things like that how does it the AI know that hands don't penetrate through you know podiums uh that feet stand above the ground that when you step on water you all fall into it so you have to ground it on physics and so so now ai has to learn um by watching a lot of different examples and ideally mostly video uh that certain be certain properties um uh are are obeyed in in in the world okay it has to create what is called a world model and so so one we have to we have to understand multimodality there's a whole bunch of other modalities like as I mentioned before genes and amino acids and proteins and cells which leads to organs and you know so on so forth and so we would like to uh multim modality second is um uh greater and greater reasoning capabilities a lot of a lot of the things that we already do uh reasoning skills are encoded in common sense you know Common Sense is reasoning that we all kind of take for granted and so there are a lot of things in our knowledge in the internet that already encodes reasoning and and and models can learn that um but there's higher level reasoning uh capabilities for example example there's some questions that you ask me right now when we're talking I'm mostly doing generative AI I'm not spending a whole lot of time reasoning about the question however there are certain problems like for example planning problems where I'm going to that's interesting let me think about that and I'm cycling it in the back and I'm coming up with the multiple plans I've got I'm traversing a tree maybe I'm going through my graph and you know I'm I'm I'm pruning my tree and saying this doesn't make sense but this I'm going to put and I simulate it in my head and maybe I do some calculations and so on so forth that long thinking that long thinking AI is not good at today everything that you prompt into chat gbt it responds instantaneously we would like to prompt something into chat gbt give it a mission statement give it a problem and for it to think a while isn't that right and so so that kind of system you know what computer science call system 2 thinking or long thinking or planning those kind of things reasoning reasoning and planning those kind of problems I think we're going to we're working on those things and I think that you're going to see some breakthroughs and so in the future the way you're interact with AI will be very different some of it will be just just give me a question question and answer some of it say here's a problem go work on it for a while okay tell me tomorrow and it it it does the the largest amount of computation it can do U by tomorrow you you could also say I'm going to give you this problem U you know spend $1,000 on it but don't spend more than more than that and it comes back with the best answer within the Thousand or you you know so on so forth okay so so that's now AGI the question on AGI is what's the definition yeah in fact that's kind of the Supreme question now if you ask me uh if you say Jensen uh AGI is a list of a list of tests and remember an engineer can only know an engineer knows that we've you know anybody in the in in that you know prestigious organization that I'm now powered of it knows for sure about engineers is that you need to have a specification and you need to know what the definition of successes you need to have a test now if I if I gave uh an AI a lot of math tests and reasoning tests and a history test and biology tests and medical exams and bar exams and you name it SATs and mcats and every single test that you can possibly imagine you make that list of tests and you put it in front of put it in front of the computer science Industry I'm guessing in 5 years time we'll do well on every single one of them and so if your definition of AG is that it passes human tests yep then I will tell you five years if you tell me but is it if you asked it to me a little bit differently the way you asked it that AGI is going to be have human intelligence well I'm not exactly sure how to specify all of your intelligence yet and nobody does really and therefore it's hard to achieve as an engineer does that make sense okay and so so the answer is we're not sure and and um uh but we're we're all endeavoring to make it you know better and better so I'm going to ask two more questions and I'm going to turn it over because I think there's lots of uh good questions out there the first one I was going to ask about is could you just dive a little deeper into what you see as ai's role in drug discovery the first role is to understand understand the meaning of the digital information that we have right now we have we have all as you know we have U uh we have a whole lot of amino acids we can now uh because of alpha fold um understand the protein structure in many of them but the question is now what is the meaning of that protein what is the meaning of this protein what is this function uh it would be great just as you can chat with GPT uh as you guys know uh there's you can chat with a PDF you take a PDF file doesn't matter what it is my favorites are you take a PDF file of a of a research paper and you load it into chat G and you start at just talking to it it's like talking to the researchers is you know just ask what what inspired this this research what problem does it solve you know what was the Breakthrough what what was the what was the state- of art before then what were the what were the novel ideas just talk to it like a human okay in the future want to take a protein put it into chat GPT just like PDF what are you for what what enzymes activate you you know what makes you happy for example there'll be a whole whole sequence of genes and you're going to take the and represents a cell you you going to put that cell in what are you for what do you do what are you good for you know what do you hopes and dreams and so so that that's that's one of the most profound things we can do is to understand the meaning of biology does that make sense if we can understand the meaning of biology as you guys know once we understand the meaning of almost any information that it's in the world the computer science in the world of computing amazing engineers and amazing scientists know exactly what to do with it but that's the Breakthrough the multiomic multi multi-omic um understanding of biology and so that's if I could you know deep and shallow answer to your I think that's probably the single most profound thing that we can do boy Oregon State and Stanford are really proud of you so if I could switch gears just a little bit and just say Stanford has a lot of aspiring entrepreneurs students that are entrepreneurs and maybe they're computer science Majors or or engineering majors of some sort please don't build gpus what what advice would you give them uh to improve their chances of success um you know one one of my one of I think one of my my great advantages is that I have very low expectations um and and and I mean that um most of most of the Stanford graduates have very high expectations you you and you deserve to have have expectations because you came from a great school um uh you were very successful you're on top of your top of your class uh obviously you were able to pay for tuition um and and uh uh and then you're graduating from one of the finest institutions on the planet you're surrounded by other kids that are just incredible you should have very you you naturally have very high expectations um people with very high expectations have very low resilience and unfortunately resilience matters in success I don't know how to teach it to you except for I hope suffering happens to you and and uh I I was fortunate that I grew up with a with a with you know with my parents um uh uh providing a condition for us to be successful on the one hand um but there were plenty of plenty of opportunities for setbacks and suffering and um you know and and to to this day I use the word the phrase pain and suffering inside our company with great Glee and the reason and I mean that you know boy this is going to cause a lot of pain and suffering and I mean that in a happy way um because because you want to train you want to refine the character of your company you want want that you want greatness out of them and greatness is not intelligence as you know greatness comes from character and character isn't isn't formed out of smart people it's formed out of people who suffered and and so so that's that's kind of and so if I could if I could wish upon you I don't know how to do it but you know for all of you Stanford students I I wish upon you you know ample doses of pain and suffering I'm going to back out of my promise and ask you one more question how do you you seem incredibly motivated and energetic but how do you keep your employees motivated and energetic when they probably become richer than they ever expected to I'm surrounded I'm surrounded by 55 people my management team so you know my I I have a man my management team my director reports is 55 people um uh I write no reviews for any of them I give them constant reviews uh and they provide the same to me uh my compensation for them uh is the the bottom right corner of excel I just drag it down literally many of our executives are paid the same exactly to the dollar I know it's weird it works and and uh I don't do one-on ones with any of them unless they need me then I'll drop everything for them uh I never have meetings with them just alone and they never hear me say something to them uh that is only for them to know there's not one piece of information that I that I somehow secretly tell eaff that I don't tell the rest of the company um uh and so in in that in that way our company was designed for agility for information to be to flow as quickly as possible uh for people to be empowered by what they are able to do not what they know um and uh I and so that that's the architecture of our company um I don't remember your question but but oh oh oh oh oh oh oh I got it I got it I got it I got it uh and the the answer the answer for that is my behavior yeah the it's uh how do I celebrate success how do I celebrate failure how do I talk about success how do I talk about setbacks um every single thing that I'm looking for opportunities to instill every single day I'm looking for opportunities to to keep on uh instilling the culture of the company and what is important what's not important what's the definition of good how do you compare yourself to good how do you think about good um uh how do you think about a journey how do you think about results uh all of that all day long Mark dougen can you help us okay good so let's open it up uh for some questions let me start with Winston and I'll come to you oh we need a microphone can you just Ben you got this yeah board member Winston I have a couple question what's a story about your leather jacket and the second the second is according to your projection and calculation in 5 to 10 years how much more semiconductor manufacturing capacity is needed to support the growth of AI okay uh I appreciate two questions um uh the the uh the first question is this is what my wife bought for me and this is what I'm [Laughter] wearing and and because I do I do 0% of my own shopping uh as soon as something doesn't as soon as she finds something that doesn't make me itch because she knows she's known me since I was 17 years old and she thinks that everything makes me itch and the way I say I don't like something is it makes me itch and so as soon as she finds me something that doesn't make me itch if you look at my closet the whole closet is a shirt because she doesn't want to shot for me again and so so that's why uh this is all she bought me and this is all I'm wearing and if I if I don't like the answer I can go shopping otherwise I could wear it and it's good enough for me we second question on this the forecast is actually very this is very I'm horrible at forecasting but I'm very good at first principled reasoning of the size of the opportunity and so let me first reason for you um uh I have no idea how many f ABS but here's here's the thing that I do know the way that we do Computing today the the the information was was written by someone created by someone it's basically pre-recorded all the words all the videos all the sound everything that we do is retrieval based it was pre-recorded does that make sense as I say that every time you touch on a phone remember somebody wrote that and stored it somewhere it was pre-recorded okay every modality that you know in the future because we're going to have AIS it understands the current circumstance and because it can it's tapped into all of the world's you know latest news and things like it's called retrieval based okay and it understand your context meaning it understood why you asked what you're asking about when you and I ask about the economy we probably are meeting very different things and for very different context and based on that it can generate at exactly the right information for you so in the future it already understands context and most of computing will be generative in the today 100% of content is pre-recorded if in the future 100% of content will be generative the question is how many how does that change the shape of computing and so without torturing you anymore um I'll that's how I reason through things how much more networking do we need more less of that do we need memory of this and and the answer is we're going to need more Fabs however uh remember that we're also improving the algorithms and the processing of it um tremendously over time it's not as if the efficiency of computing is what it is today and therefore the demand is this much in the meantime I'm improving Computing by a million times every 10 years while demand is going up by a trillion times and that has to offset each other does that make sense and then there's technology diffusion and so on so forth that's just a matter of time but it doesn't change the fact that one day all of the computers in the world will be changed 100% every single data center will be all of those general purpose Computing data centers 100% of the trillion dollars worth of infrastructure will be completely changed and then there'll be new infrastructure built on even on top of that okay next question right here Ben and then over here to Rand so yeah thanks for coming today so recently you said that you encourage students not to learn how to code yeah um and that's the case it means one of maybe a few things but do you think the world starts to look like from a company formation an entrepreneurship perspective that it goes towards many many more companies that are created or do you think it's consolidation to just a number of the big big players so so first of all um I I I said it so poorly that you repeat it back poorly I I didn't if you would like to code for God's sakes code okay if if you want to make omelets make omelets I'm not not you coding has coding is a reasoning process it's good does is it going to guarantee you a job no not even a little bit uh the the number of coders in the world uh surely uh will continue to to uh uh be important and we Nvidia needs coders however in the future the way you interact with the computer is not going to be C++ mostly for some of us that's true for some of us that's but for you you know why why programming python so weird in the future you'll tell the computer what you want and the computer will will you you say hi I would like you to come up with a uh a build plan with all of the suppliers and build a material for a forecast that we have for you and based on all of the equip all the necessary components necessary coming up with a bill plan okay and then if you if you don't like that you write me a Python program that I can modify of that bill plan and so remember the first time I talk to the computer I'm just speaking in plain English the second time so English by the way human is the best programming language of the future how you talk to a computer how do you prompt it how do you prompt it it's called prompt engineering how you interact with people how do you interact with computers how do you make a computer do what you want it to do um how do you fine-tune uh the instructions with that computer that's called prompt engineering there's an there's an Artistry to that okay so for example most people are surprised by this but it's it's not surprising to me but but it's surprising for example you ask mour to generate a pcture an image of a puppy on a on a surfboard um uh uh in Hawaii uh at Sunset okay and then and then and it generates one and go and you say oh more cute make it more cute and it comes back it's more cute and you go no no cuter than that and it comes back why is it that software would do that there's a there's a structural reason why it does that but for example you need to know that that that capability exists in a computer in the future isn't that right that you if you don't like the answer first time you could you can find tuna and get it to within the context that you you know you can make it give you better and better results and once you you can even ask it to write the program Al together to generate that result in the future and so my point is that programming has has changed in a way that is probably less valuable on the other hand let me I will tell you this that because of artificial intelligence we have closed the technology divide of humanity today about about 10 million people are gainfully employed because we know how to program computers which leaves the other 8 billion behind that's not true in the future we all can program computers does that make sense you all know how to prompt a computer to make it do things and look at all you to do is look at YouTube and look at all the people who are using prompt engineering all the kids and you know who are making a do amazing things they don't know how to program they're just talking to chat GPT they just know that if I tell it to do this if do that you know and so it's no different than interacting with people in the future that's that's the great contribution we've the computer science Industry has made to the world we've closed the technology divide so that's that's inspiring okay over here we've got that sounds very we've got Randy with a question right over here oh um thank you very much I'm just wondering um about do you think very much about geopolitical risk and um how do you see it impacting your industry if you do uh geopolitical risk you know we we are almost a poster child of geopolitical risk and the reason for that is because uh we make a very important instrument for artificial intelligence and artificial intelligence as John and I were talking about earlier is the defining technology of this of this of this time and and um and so the United States uh has every right to determine that this instrument should be limited to uh to uh countries that that it determines that uh it should be limited limited uh with and so so the United States have has that right and they they exercise that right um and your question has to do with what is the implication to us I uh we first of all we we just have to understand these policies and we have to stay agile so that we can comply with the policies uh number one on the one hand it limits our opportunity and in some places and it it opens up opportunities in others one of the things that has happened in the last I would say maybe even 6 to n months is the Awakening of every single country every single Society The Awakening that they have to control their own digital intelligence that India can't Outsource its data so that some country transforms that Digital Data into India's intelligence and imports that intelligence back to India that Awakening that Sovereign AI that you have to you have to dedicate yourself to control your Sovereign AI your Sovereign intelligence protect your language protect your culture for your own industries that Awakening I think happened in the last 6 nine months the first part was we have to be we have to be mindful about safety then the second part was hold on a second we we all have to do this and so every single country from from India um uh Canada's doing this uh the UK France um Japan uh Singapore Malaysia uh the list goes on uh just about every single country now realize that they have to invest in their own Sovereign AI so geopolitics in the one hand limited opportunities but it created just enormous opportunities elsewhere and so hard hard to say okay so I think we I have multiple hands but I have time for one more question I am going to go right here you had to you were further on the now remember the last question has all big pressure you guys agree with that do you can we all agree right here the the person who La asked the last question don't don't leave us all depressed I'm going to don't trigger me please I'm I'm that's all I'm saying I'm just kidding I'm going to invoke your commandment to have low expectations at this juncture um you you mentioned your competing with your customers and I'm wondering you know given the advantages that you have why they're doing that and I'm wondering if in the future you see yourself building more customized solutions for customers of a certain scale um as opposed to you know uh the solutions that you have now which are more horizontal uh the the so so are we willing to customize the answerers yes now why is it that the bar is relatively High the the reason why the bar is high is because each generation of our our platform first of all there's a GPU there's a CPU there's a networking processor there's a SW there two types of switches I just build five chips for one generation people thinks it's one chip but it's five different chips each one of those chips are hundreds and hundreds of millions of dollars to do just hitting launch which is tape out for us launching a rocket is several hundred million dollar each time okay I I got five of them per generation then you've got to put them into into a system and then you got to put you know you got networking stuff you got C transceiver stuff you got optic stuff you got a mountain of software to do it takes a lot of software to run a computer as big as this room and so so all of that is complicated if I if if the customization is so different then then you have to repeat the entire R&D however if the customization leverages everything and adds something to it then it makes it's makes a great deal of sense maybe it's a it's a proprietary security system maybe it's a confidential Computing system maybe it's a a a new way of doing uh numerical processing um that that could be extended we're very open-minded to that and the custo our our customers know that I'm willing to do all that and recognizes the the the if you change it too far you've basically reset and you've squandered you know the the nearly hundred billion dollars that's taken us to get here um uh to to redo it from from scratch and so they want to leverage our ecosystem to the extent that that that that will be done I'm very open to it yeah and they know and they know that yeah okay so with that I think we need to wrap up thank you so much to John and Jensen

Channel: Stanford Institute for Economic Policy Research (SIEPR)

Views: 239,457

Keywords:

Id: cEg8cOx7UZk

Length: 55min 21sec (3321 seconds)

Published: Thu Mar 07 2024