Shadowing Practice: Jensen Huang: The Mindset That Built NVIDIA - Learn English Speaking with Video
בחר מצב תרגול
כל מצב פותח חדר תרגול במסך מלא.
638 משפטים
אם המשפטים קצרים או ארוכים מדי, לחץ על ערוך כדי להתאים אותם.
1
Welcome to Startup School 2026.
2
Now let's get started.
3
Please join me in welcoming to the stage the founder and CEO of NVIDIA, Jensen Huang.
4
Oh my God.
5
This is a surreal moment for me.
6
Thank you.
7
Thank you for being here, Jensen.
8
I'm delighted to do it.
9
It's great to be here.
10
Apparently, if you're here, you are going to make it.
11
So, I'm happy I'm here.
12
Jensen, well, for the students who only know NVIDIA as a company at the center of AI,
13
what part of the early NVIDIA story do they most need to understand?
14
The thing that most people don't believe is
15
that the choice of our technology that we started the company with was absolutely wrong.
16
And so we had started with the idea that we would reinvent 3D graphics.
17
Well, the company's philosophy and perspective was that the general purpose computers,
18
the CPUs, were really useful, but if we could augment it with accelerators, we could solve problems that otherwise too hard to solve.
19
And one of the first problems we chose was 3D graphics.
20
And during that time, 1993, the PC was just rumored to be coming.
21
And our big idea was
22
that we would turn every single personal computer into a game console because we grew up in the era of game consoles.
23
And so we thought, you know, what if we could design a system that would fit into the personal computer
24
and it would turn it into a game console.
25
And so we thought we would reinvent the algorithm that would require these large supercomputers, and we would fit it into the PC.
26
And we came up with some new algorithms, and we were excited about it.
27
We believed in it.
28
We reasoned about it in a thoughtful way, and we went to start the company to go build it.
29
Well, it turns out the algorithm was exactly wrong.
30
And the technology that founded the company turns out to be exactly wrong.
31
And so in 1995, we realized that, and it was almost too late, because by then, there were some 35,
32
40 other companies that were building 3D graphics for PCs.
33
And so we realized that it didn't work, and I went back to the company, and we were at the company, I said, what are we going to do?
34
It doesn't work, and we're all talking about it.
35
And I said, look,
36
we won't have a company if we don't confront the fact that this doesn't work and start working towards the right algorithm.
37
And then somebody told me, it turns out none of us knew how to do it the right way.
38
And not only did we choose the wrong technology, we didn't know how to do it the right way.
39
And so that was a big day for me.
40
I had a couple of $60, a couple of $100 in my pocket.
41
And so I went down to Fry's and I bought three textbooks.
42
And the textbooks was about OpenGL and how to design OpenGL pipelines.
43
I brought it back to the company and gave it to the engineers.
44
And here we are.
45
We reinvented computer graphics.
46
We're the world leader in modern computer graphics.
47
We invented most of the major breakthroughs in the last 25 years.
48
everybody would have thought that nvidia is you know started out as word leaders in 3d graphics
49
and we learned it from a textbook and
50
so we actually started the company raised money and bought textbooks
51
when you think about it and so the the big lesson is
52
that for me is technology is changing all the time and
53
so long as you're able to confront the reality
54
so long as you are able to learn the technology itself actually doesn't matter and
55
so since then nvidia has been you know inventing all kinds
56
of technology since all kinds of technology we've never never really done before
57
and we approach everything with the same attitude you know this is
58
if it's important to do we're going to go learn it
59
and how hard can it be and it always turns out to be much much harder than we expect.
60
But you go into it with the attitude, how hard can it be?
61
I mean, backstage, we were talking about how, I mean, we were talking with some of the top YC companies, and you were saying that each one has an expertise in,
62
like, a domain that you and NVIDIA have an expertise in.
63
And they're all just, I forget what you said, it was like an algorithmic domain of a sort.
64
And so it sounds like 3D graphics was merely the first of an algorithmic domain.
65
That's right.
66
And it came from a textbook but then, you know, anyone could have read that textbook.
67
You created...
68
Particle physics, fluid dynamics, yeah.
69
But you created the thing that people want, like the end product that people want to pay a lot of money for.
70
The big idea of the company that was spot on is
71
that it is possible to augment the CPU to solve problems that otherwise are too difficult to solve.
72
And molecular dynamics is one of them.
73
Image processing is one of them.
74
Inverse physics is another one.
75
And so all kinds of different algorithms, of course, deep learning is one of the major ones.
76
And in order to create the company that we have today, we realized early on that it's not about building a great chip.
77
It's about accelerating an algorithm domain.
78
And so one of the things
79
that I've always believed in is what makes great companies is a unique perspective about the world that you deeply believe in.
80
It's not so much the technology.
81
It's not so much the market even.
82
Those things all matter.
83
And if you have the right technology for the right market at the right time, your life is going to be a lot easier.
84
A high -level vision about the future of some important thing, a perspective about it that's somehow unique, that you deeply believe in,
85
and ideally pursuing that vision is hard to do.
86
Those are kind of good combinations.
87
In our case, we realized that accelerated computing was going to be important.
88
And accelerated computing turns out to be very important.
89
And our realization is everything to do with the algorithm, not the chip, turns out to be exactly right.
90
So you've said a lot about, I guess, the hardships of a founder.
91
Are there a few stories that really jump out at you?
92
I mean, the people in this room would love to start a company, but, you know, are they really prepared for eating glass and,
93
you know, possibly having to shut down the company, like things going wrong?
94
Like, what are some of the pivotal moments that really jump out at you?
95
I think you were just in Japan, right?
96
And you were sort of honoring Sega, was it?
97
So I feel like that was a really powerful story.
98
The project that led us to realize the algorithm we chose was wrong was a partnership with Sega.
99
Sega had contracted us to build the game console after Saturn that turned out to have been Dreamcast.
100
I don't know if any, does anybody know what Dreamcast is?
101
Okay.
102
So, we did not build Dreamcast.
103
We were originally supposed to build Dreamcast.
104
But because our algorithm and our technology was fundamentally flawed, I went to Japan and I told Irimandri -san, the CEO at the time,
105
that the contract that they gave us was like $12 million contract fact,
106
we will not be able to fulfill because the technology doesn't work.
107
And I told him the reasons why, and then I advised that they choose somebody else to do it, but then I asked him,
108
I told him that I unfortunately still need the money, and he asked me, you know, you could just imagine the conversation, so what you're telling me is,
109
what I contract you to do, you can't do, but you would like all the money on the contract.
110
And I said, you got it.
111
That's exactly right.
112
But obviously, I was polite.
113
I was humble.
114
And he realized that I was honest, and everything made sense.
115
And if he didn't give us the money, we'd be out of business.
116
And I think that this happens in this room, you don't invest in companies, you invest in people.
117
And what Irmajuri recognized was here's somebody and a company that he trusted in the first place, the contract, and that he believed in,
118
and that he would love to see make it to the next day.
119
And so that $5 million kept us alive and gave me enough time to discover what to do.
120
And then I guess if they They sold it for $15 million, I heard.
121
Yeah, they sold it the moment we went public.
122
When NVIDIA went public, our valuation was $300 million.
123
$300 million in 1999.
124
That was real money.
125
I think it's north of a trillion dollars now or so.
126
It's more than true, yeah.
127
Yeah, that's wild.
128
So you're sort of the core, you know, we like to say that you're the man who controls the spice.
129
um You know, before that, you know, I don't think anyone could have really predicted, per se, how important GPUs and,
130
you know, the technology you built would be for this AI revolution.
131
You know, what did you see to, I mean, was it the accelerator and being in the right place at the right time?
132
Or surely there were a lot of things that led up to that that allowed you to sort of capture this position.
133
Yeah.
134
I saw AlexNet just like everybody else saw AlexNet.
135
But remember, our lens of the world, my view of the world, was always looking for algorithms.
136
And that algorithm, the algorithm could be NAMD, the algorithm could be VASC, the algorithm could be OpenGL, you know, it could be SQL,
137
some domain -specific language, some algorithm.
138
And so my lens of the world was always looking for some problem that we might be able to help solve.
139
When AlexNet came along, the algorithm was deep learning.
140
And so the question is, what is this algorithm?
141
And why does it matter?
142
Why was it so effective?
143
And what else can it do?
144
And if you were to scale algorithms and scale it beyond that, what could it solve that otherwise you can't solve today?
145
And the breakthrough for us was realizing that AlexNet was not Alex net
146
that Alex net was an approach with deep deep learning
147
that allows you to learn any function and
148
so 15 years ago I was telling everybody
149
that hey guess what we just learned a universal function approximator
150
we just discovered the universal function approximator we can give it
151
we could you know give it the the answer for almost any function and it could learn what the function is.
152
And for a lot of functions, you don't have to be precise.
153
And in fact, it's impossible to be precise.
154
And so most of the interesting problems are imprecise in this way.
155
And so the day that we realized we have a universal function approximator,
156
the question then is, what does that do to the computing stack?
157
What does that happen to software? what are the industries that this could impact, so on and so forth.
158
Almost right away, we started working on computer vision.
159
Almost right away, we started working on robotics,
160
self -driving cars, because that fundamental capability you could imagine solving some important problems in the area of computer vision and robotics.
161
And so I think the big breakthrough was simply
162
that this is much more foundational than AlexNet this is a way of doing software.
163
And the implications to the processor, the middleware, the algorithms, the applications, you know, what I now describe as the five -layer cake,
164
that entire industrial stack, I imagine reinventing all together about 15 years ago.
165
And this is simply about asking questions, reasoning about things to first principles, asking questions like, if this, then what?
166
If this can get better, then so what?
167
Asking all of the basic questions about something that you observe that's really impactful.
168
I mean, one of the things
169
that really jumps out at me is to what degree you go all the way into the weeds.
170
You read papers, you talk directly to the principal scientists who are sort of coming up with these things.
171
Do you have any advice for people in the audience?
172
I mean, that's like true founder mode
173
and then at the same time you probably you have an organization
174
and you have executives
175
and you have people who say like here's the graph we
176
want to stay on this graph you know sometimes it ruffles
177
feathers like do you have any advice for people about an organization
178
and how you navigate that really like how do you build an org
179
that allows you to think in first principles because
180
if the fortune 500 did
181
that like the fortune 500 will probably look a lot more like nvidia than not
182
and it doesn't like you you have built a very unique company
183
my state of mind
184
when i'm my state of mind is always starts with curiosity i have a whole bunch of questions myself
185
and and of course like anybody else i'll seek the shortest path to the answer
186
but oftentimes the the answers from the people that are near me might not be satisfying and I might have other questions,
187
and maybe they're busy doing something and they're pursuing something.
188
And so my first inclination is to go discover the answers to my own curiosity.
189
My second is if I find that the information is in
190
that the domain of information or in a particular field could be really important to somebody
191
and could be important to our company.
192
Then my next inclination is, how can I learn as much as possible so
193
that I could be of service to the company and share it with everybody else?
194
You know, this is no different than you when you're sharing knowledge.
195
I mean, I watch your podcasts, and I watch your videos, and I really enjoy them.
196
You're sharing ideas with everybody else.
197
In a lot of ways, I think a CEO is in service of the company, in service of all the people that are working there,
198
and you want to empower them with some insight.
199
and so that's really where it's coming from it's not
200
so much a management technique but a personality technique you know I want to empower you
201
and this is something really important
202
that I just observed let me tell you why it's
203
so important now part of part of having to be near the ground
204
and be in the weeds if you will is
205
because oftentimes the technology is complicated or it's changing really fast and especially
206
when it's changing fast like like our world unless you have
207
a tactile sensation of what is actually happening it could either
208
to you feel like it's just moving way too fast to understand but
209
if you understand the first principles of it over time then
210
everything kind of makes sense you know it's kind of like
211
surfing I would imagine I don't know how to surf
212
but I can imagine it's kind of like surfing you get on the wave to me it looks like chaos
213
but to a surfer you know somehow they get right they can read the waves, and they know how to stay on top of it.
214
And so I think being CEO is very similar to that.
215
You know, you have to learn how to serve.
216
In order to learn how to serve, you have to understand the waves, you have to be able to read the wind, and you have to have good timing.
217
And you can't have any of that unless you try, unless you actually do it.
218
And so partly is to inform myself, partly is to try to figure out, you know, what is, try to break down the problem so
219
that the company can learn it in a way that they can do something about.
220
Part of it is by inspiring other people
221
and you know it's all those basic traits of all the people in this room.
222
You don't have to change your personality or your behavior when you become CEO.
223
It is possible for you to continue to be yourself
224
and one of the things that I learned a long time ago
225
and I had no idea where i saw this uh but
226
but um you know the ceo
227
or the founders you are the you you're building a car
228
that you are going to race you're going to build an f1 racer
229
but you're going to build it in a way
230
that you can drive you should adapt the car to you you know somebody i think asked me
231
you know jensen if you if you don't use conventional management techniques
232
and organizational techniques you know what's going to happen when you leave the company
233
well you know
234
when i die on the job um someday you know i
235
told them they'll just have to reshape the company for the next ceo
236
and the reason that's wisdom is because we're the f1 drivers You know, we're the racers.
237
And the world is really competitive.
238
And we've got to stay, we've got to, you know, we've got to win.
239
And we've got to achieve our mission.
240
And so whatever it takes to fit the car to you, whatever it takes to fit the organization to you, that's what you ought to do.
241
And the next CEO, whatever the personality is, they can figure it out.
242
Amazing.
243
I mean, it does seem like any change you make to the car will just slow you down
244
and lose you races that, you know, isn't fit to you.
245
Yeah, or we're constantly tweaking the car to our needs, and that's really what I'm doing all the time.
246
I'm constantly tweaking the company, constantly reshaping business processes and the way things work, so that I can, you know, be more effective for the company.
247
True founder mode.
248
Yeah, founder mode.
249
Founder mode could scale for 34 years.
250
That's right.
251
From zero to five trillion.
252
No evidence.
253
No. I'd love to switch gears to like what,
254
you know, what are the frontier algorithms that you're most interested in now?
255
I mean, I love that you're all the way down into the material science, all the way up into the app level.
256
You know, you're the first to speak on stage about Open Claw and now Hermes Agent.
257
I wonder
258
if you can sort of like walk us through a day
259
in the life of like how you think about the different stages.
260
I mean, going from materials to chips to data centers to even like the app level, like how people are going to work,
261
like there's sort of this idea of a full stack AI factory.
262
Well, this is one of the things that is probably going to be the most useful skill in the future.
263
And in fact, just in listening to you talk about technology and your use of it, you know, one of the most important things is systems understanding.
264
Systems awareness, system design, system organization, but systems thinking.
265
And the reason for that is because most of the low -level things
266
that has to be done are going to be done agentically anyways.
267
They're going to be automated anyhow.
268
And so whether it's, you know, in my generation, it's about compiling chips and synthesizing transistors and gates and functional blocks.
269
And all of that is now synthesized.
270
And so most of our designers are systems designers.
271
In the case of software, most software is going to be done agentically anyhow.
272
So you have to be much more able to think abstractly about systems.
273
What are the problems you're trying to solve?
274
What are the constraints?
275
Where's the input?
276
Where's the output?
277
Where are information coming from?
278
What is the rate of information flowing in and out of the system?
279
What are the constraints?
280
It's, you know, and so is a processor, is a memory, is a networking, you know. And
281
so understanding these systems problems at a sufficiently technical level is
282
going to be very helpful to all of the people in this room.
283
And I don't think that that way of that fundamental knowledge is ever going to be useless.
284
I think it's going to be more and more useful.
285
And so I try to understand systems the best I can.
286
One of the things, speaking of agents, the fact of the matter is we kind of have course -level recursive self -improvement already.
287
And the fact that every time you use it, it improves the markdown files.
288
Every time you use it, it updates its long -term memory, and the long -term memory is being processed, either compacted or turned into knowledge graphs or,
289
you know, so on and so forth.
290
It's been improved all the time, you know, asynchronously.
291
And so the agent's getting smarter and smarter every time.
292
Still, the problem is, and this is one of the problems that I think would be helpful for everybody to solve, is how can we have very, very specific fine -grained control?
293
You know, if not for rags, if not for conditional inputs, if not for all of our prompts directly into output,
294
was was too coarse and so the fact that we can condition the fact that we can control the agents
295
all the way down to eventually
296
when it comes up with a plan i change one word in a plan file
297
and that one word makes a delta difference not complete difference but specific difference
298
maybe it's one pixel maybe it's one triangle maybe it's one component in a cat file maybe one layer,
299
one via, one connection, and then it regenerates everything else.
300
I think that that level of control and that level of collaboration with agents will be game -changing.
301
We don't need the agents to be 100 % accurate, 100 % high quality in order for us to use it.
302
It could literally be 80 % and then we help it the rest of the way, or it could be 99 % we help it the rest of the way.
303
And so I think controllability is probably the single biggest breakthrough that we need for agents at every single level.
304
Do you think people will like, I mean, with Hermes or OpenClaw, it feels like that might actually be somewhat existential.
305
Like people should control their own personal AGI.
306
They shouldn't outsource that and, you know, have it be just in the cloud and someone else's agent that like kind of tells you what to do.
307
Like you kind of want it to be your own.
308
Is that part of the thrust behind NVIDIA being so involved?
309
I think, well, first of all, I need to understand Agents because Agents is the new software.
310
And how is this new software processed matters a lot to computer architecture.
311
And the more intimate we are about the nature of Agents and how it's different than chatbots,
312
which is how different than maybe Inference in the very beginning, however we think about these processing layers,
313
the more intimate we are about the nature of the processing, the better we can design systems.
314
We kind of have to live in the future five to ten years
315
because it takes three or so years just to build a system.
316
It takes a couple of years to ramp it up
317
and you're dealing and you would like them to be able to use the computer for ten years after.
318
So you kind of have to live in the future for a while.
319
And so agentic systems for us at the first principles is just what is the workload?
320
What's the algorithm?
321
How is it going to evolve?
322
Where are the bottlenecks?
323
You know, where are the MDOS laws problems?
324
And how does it scale?
325
What happens to concurrency?
326
How do you deal with sandboxes?
327
How do you deal with MCP?
328
How do you deal with, you know, working memory, long -term memory?
329
How do you have all these autonomous system, asynchronous systems working all the time.
330
And so what kind of design architecture makes perfect sense for that?
331
And so we have to go and go discover that.
332
And then, of course, the second thing is I want to use agents ourselves to make NVIDIA go faster.
333
And so we have, you know, Boris is in the back, and we've got cloud code autonomously running in sandboxes all over NVIDIA, and that's really fantastic.
334
And some people use codex, some people use cloud code, some people use cursor, some people use cognition.
335
And we let kind of a thousand flowers bloom, let people select the tools they want to use.
336
And then we learn from all of that.
337
And so the second part is just helping the company move faster, use the tools.
338
And the more they use it, the more we're going to learn about how to make it work better in the future.
339
And the last part is discovering the future of solutions technology for the future.
340
And maybe, you know, when we saw the early versions of Chain of Thought come out of Stanford, this is probably a decade ago at this point,
341
maybe eight years ago, you know, the question is how effective is that going to be in reasoning and how scalable it's going to be?
342
And what is the implication, for example, in computer vision if we can reason from prior knowledge?
343
And then the big breakthrough, of course, just in thinking through that small little domain,
344
you come to realize that maybe we don't need as much data for cars to train a self -driving car, which led us to creating Alpamayo,
345
which is the world's first thinking self -driving car.
346
And with just a million miles or so, a couple million miles, it's an incredibly great self -driving car.
347
And the reason for that is it's kind of like us, right?
348
We don't need that many miles before we could drive fairly well most of our lives.
349
And the reason for that is because we have prior knowledge from our language model, and we can decompose a situation we've never seen before
350
and build it up out of things that we understood and know very well.
351
So that's an example of seeing something and then realizing the impact sometime later.
352
When the agentic systems came along, it's very, very clear that obviously a large language model needs memory,
353
it needs prior knowledge, it needs tools, it needs ways to network with other agents.
354
And so that kind of, you know, that once you see some early indicators and you're able to reason about the future, it helps you get a leap,
355
you know, into the future.
356
I feel like there's this pattern that I'm starting to see around NVIDIA.
357
You see a problem, there's a new algorithm, there's some new thing happening, and then actually you're right there with open source.
358
I mean, I remember when OpenClaw came out and people said it was unsafe,
359
but you guys came out with a sandboxing sort of toolkit that surrounds any harness and makes it safe.
360
When I saw OpenClaw, my first thought was, well, first of all, I learned about it.
361
And then, you know, without much imagination, you just realized we just designed a modern computer.
362
This is the operating system that's going to hold a large language model.
363
And in a lot of ways, OpenClaw to me was very Linux moment to me.
364
And now everybody can build their own AI.
365
And I was so excited about that.
366
And we contacted Peter.
367
And we said, hey, you know, all of NVIDIA's engineers are your engineers.
368
That's what I told Peter.
369
You've got this battleship outside your house.
370
You, you, you know, break down the problem as you desire and will contribute as you wish.
371
Same thing with the Hermes team, you know, and I'm so excited about the work that they're doing.
372
I do think that the world needs the ability for everybody to build their own AI.
373
and you could you could of course
374
and i encourage everybody to to uh use cloud services as much as possible everybody should use chat gpt
375
and clod and right everybody should use that and um but
376
if you if you need to build your own ai
377
because you're a company and
378
and um you need to build your own domain specific ais now you have hermes
379
and you have open claw you've got all kinds of you've
380
got lang chain uh deep agent you've got all these different ways right to build your own ai And it's,
381
quite frankly, relatively easy because the software is smart, you know?
382
And so AI is smart, and therefore AI must be so smart you could adapt it easily.
383
And so I think that we want to encourage everybody in every company to build their own AIs.
384
And who knows what innovation will come from the fact that it's open source.
385
I feel like all the alpha is in building your own AI.
386
I mean, if someone else is using whatever is off the shelf, but you have a thing that can.
387
recursively self -improve.
388
And it is, you know, I mean, people are very flippant about markdown files.
389
They say like, oh, haha, it's just text.
390
But like, text is intelligence.
391
And we're in a different way.
392
Words are thoughts.
393
Yeah.
394
Yeah, words are thoughts.
395
Yeah.
396
And it turns out you can try to think without words.
397
Yeah, that's right.
398
So switching gears again, I mean, a lot of people are, anytime you move the cheese, people get a little worried.
399
Intelligence is going to be on tap, which is really awesome.
400
I think it bodes well for everyone in this room.
401
What do you think changes about the economy?
402
What do you think, you know, happens in sort of a broader sense?
403
Obviously, what I'm going to say is uneven.
404
There are some, you know, we're going to automate tasks.
405
We're going to automate cognitive tasks.
406
If that task is somebody makes a phone call and sends a bunch of words, you know, across the phone to you,
407
and your job is to provide a response.
408
And if all the information is at your fingertip, because you have all the database here, and you should be able to answer that question completely,
409
in that case, that task will be automated away, okay?
410
Ignoring that for a second, not that we ignored this, but my point is I'm going to answer the question about really the great opportunity.
411
And so many tasks will be automated away.
412
Many jobs, every single job will change, and there'll be a whole bunch of new jobs.
413
And that I think we know.
414
The bottom line is this.
415
The evidence would show that, and it makes perfect sense, that AI and automation is creating jobs everywhere.
416
The narrative about AI destroying jobs is exactly backwards.
417
AI eliminates tasks.
418
AI automates tasks away, but it doesn't necessarily eliminate jobs.
419
And the reason for that is because the job of a person has a purpose, and that purpose has many tasks.
420
Some of those tasks could be automated away.
421
Many of those tasks cannot be.
422
And so the evidence suggests that here we are, we've automated coding, which is a task, but the job of a software engineer appears to be growing, right?
423
The number of software engineer jobs year over year has increased 10%.
424
The task of reading radiology scans has been automated,
425
but the number of radiology jobs has increased some 20 % in the last several years, even though AI has taken over the whole field.
426
And the reason for that is because the backlog of patients is incredibly high.
427
Now doctors and hospitals could admit a lot more patients.
428
In order to admit a lot more patients, you need more nurses, more radiologists.
429
And so the same thing with software.
430
The backlog of ideas, the backlog of ambition and aspiration is
431
so high that if we can automate away the task of programming, we could hire more software engineers to do more things we could be more ambitious.
432
Same thing, you know, just across the board.
433
They said Harvey is going to eliminate all of the paralegal jobs and the number of lawyers will be reduced.
434
Turns out paralegals are growing like crazy.
435
And the reason for that is because the backlog of lawsuits is really high.
436
And now these law firms could get a lot more cases through.
437
In order to do so, you got to hire more people and
438
so this is a classic classic example of productivity increasing growth
439
increasing growth drives more employment this is the reason why there's more employment today than there was
440
when i first came out of school
441
so we've been talking a lot about software
442
and agents um another really exciting thing
443
that nvidia is all the way out on the edge on
444
is actually physical robots um you know how far out i think in the past you might have even said
445
As soon as this year, what's the latest thinking on, you know, when can we expect practical robotics?
446
Yeah, the moment that I saw us generating video, that was a great moment for me.
447
The moment that I saw us generating video, I mean, we did the original work on auto -progressive GANs, okay?
448
And we did the original work on conditional GANs.
449
Long before the first videos were generated outside that people saw, a couple of years earlier inside our labs,
450
we were driving a simulator completely generated by video and completely generated by neural networks.
451
And so the moment I saw us generating articulation, if I can generate video of a finger moving,
452
if I could generate video of a hand picking up a glass, why can't I cause a robot to do the same?
453
And so the moment I saw that generative AI happening, I realized that robotics articulation was around the corner.
454
And so now the question is, you know, how is the robot going to understand, to generate motions that obey the laws of physics?
455
How does it understand causality?
456
How does it understand, you know, friction, tension?
457
How does it understand the laws of physics? And
458
so it started us down the journey of creating what we call physical AI now everybody calls it physical ai
459
and physical ai we started working on world foundation model
460
an ai that understands the laws of physics and how the world works
461
and we started down the journey of of working on robotics
462
i would say the chat gpt moment of robots happened a couple years ago already wow
463
and and the reason for that is remember when chat gpt first came out
464
it didn't do anything productive it didn't do anything useful
465
but it opened our imagination about what's possible
466
and i would say a couple of years ago you know robots walking around
467
that we could do reinforcement learning fine -tune it for and ground it in physics
468
really happened a couple years ago
469
so now what what do we need to do we need to do all the same things
470
that we're doing now for agentic systems we have to create environments for them to learn in to eval in, eval against.
471
And so we have to do real to sim to create environments.
472
We have to do, we have to generate simulators that are based on simulation, grounded physics simulation, as well as generative physics simulations.
473
And so Isaac Sim, Cosmos, and all the work that we do in that area is related to simulation.
474
And then the last part is sim to real.
475
And so that part is, has something to do with reinforcement learning,
476
grounding it on physics, grounding it on all the electromechanical systems that robots require.
477
And so, but these three basic systems, I think, builds up the eval, if you will, the post -training of robotics.
478
And I think we're going to see it right around the corner.
479
Amazing.
480
Where does physical AI show up first in a way that's really economically real?
481
Are you seeing that already?
482
We conjectured that robotics was going to come along and decided
483
that the first application of robotics that has both a large enough market,
484
relatively standardized technology so that we could scale and get the flywheel going and has real economic value was self -driving cars.
485
And so inside Waymo are chips from NVIDIA.
486
At Tesla, we were in the car.
487
Now we're in the data center.
488
Mercedes, we're in the data center.
489
We're in the car with a software stack.
490
We worked on Alpamayo, and we open -sourced it.
491
And the reason why we open -sourced the self -driving car stack is because you need it for agriculture.
492
You need it for mail delivery.
493
You need it for warehouse AMRs.
494
There's so many different ways that you could apply autonomous navigation
495
and none of those markets are big enough to be a self -driving car market
496
and we thought it was sufficiently diverse that we would create the whole stack for it.
497
And so we're working with autonomous vehicles in all kinds of different places.
498
Our robotics business, autonomous vehicle business, basically physical AI business is probably almost like ten billion dollars so it's really really big already.
499
Likely, this will be one of the largest industries in the world.
500
And it'll take longer than a couple, two, three years.
501
It'll take less than 10.
502
And so this will be our next $100 billion business.
503
Amazing.
504
I want to take a moment.
505
I think this is the exact right crowd to, you know, maybe as an arena, we can welcome Jensen to X.
506
Welcome to X.
507
I mean, you made your first post.
508
And thank you for your leadership.
509
you know that's it that just that shows you how introverted i am it took me until
510
2026 to have the first post on x you know it's i'm probably the last human on earth
511
that that did it but but what i posted was too important to me
512
and too important to the to the industry and too important to the world.
513
And so, so I, I over, overcame my, my shyness and put my first thing out on X.
514
No, thank you for your leadership.
515
I mean, open source, open weights, open source models are incredibly important for, I mean, what all of us in this room want to do.
516
We want to create setups.
517
If not for open source, the mobile cloud industry would have never happened.
518
If not for open if not for Linux if not for Kubernetes
519
if not for all of these you know platform if not for TensorFlow or more important PyTorch right?
520
And the early versions of a cafe, right?
521
Torch.
522
I mean, all of the, Theano.
523
Remember the early versions of all?
524
Those were all open source.
525
If not for all of that, how would we have modern AI?
526
Well, thank you for your leadership and your voice is incredibly important here.
527
Thank you.
528
Before we go, I feel like we, I just really resonate with your story.
529
I think that everyone here, I mean, would love the wisdom of, you know, your journey coming here.
530
I mean, what should a young person learn now, given all the things that you're seeing, all the algorithms that are going to take hold in society?
531
What should a young person learn now that will still matter based on what you're seeing?
532
Well, some of the things that I saw today and some of the starters I met today was really, really quite, quite encouraging.
533
And, and, and the thing that, that the big takeaway is, of course, the simple stuff is going to get automated away.
534
And when I say simple stuff, I mean, software, you know, coding, the idea that you would,
535
you would do a, you would solve a problem by sitting in front of a computer and you're, you're, you're actually writing, you know,
536
writing code, that concept is obviously going to get automated away.
537
In my generation, when I was growing up, we had to do long division.
538
I mean, for God's sakes, who has to learn long division?
539
And so that got coded away.
540
That got automated away.
541
And so I think the simple stuff is going to get automated away.
542
But the hard problems, the hard sciences, physics, chemistry, biology, computer science,
543
computer engineering, systems thinking, and particularly the domains that are intersecting, those hard problems will never go away.
544
And so AI is just an incredible tool that helps us become even more ambitious,
545
even more impatient about solving these extraordinarily large and incredibly hard problems than before.
546
And so if you look at my generation, when I first graduated, a chip designer would design a chip with maybe a thousand transistors and
547
that would be a very large chip you know now designing a trillion transistor chips is not even you know
548
if somebody would have told me jensen our next chip is
549
a trillion transistor i said okay you know it's not a thing
550
and the reason for that is because we are
551
so ambitious now the the the scale of the problem the scale of the task is no longer a matter
552
and so you don't have to worry about about how much coding, how many engineers.
553
You don't have to think about those things anymore.
554
You just have to think about what is the problem you have to solve.
555
And so I think that the deep tech stuff, the deep science stuff, understanding the intersection between technology and social issues,
556
understanding market gaps and holes, opportunities, I think all of that still exists.
557
And the better you are at systems thinking so that you could orchestrate millions of agents solving problems autonomously,
558
the better off you are.
559
And so that's why system thinking is going to be so important.
560
But otherwise, I think the world's going to continue to have a lot of great challenges for us to solve.
561
Go to school the same old way.
562
You know, stay in school.
563
Stay in school.
564
I guess I usually like to end with, you're looking out on the crowd there are a lot of people who I mean,
565
I started this the opener with like I honestly look in the crowd
566
and I see people who are not different than us per se,
567
you know, we actually just are technical and like love systems Thank you,
568
thank you What advice would you give to this room of, you know and do you see yourself in this room
569
and like I'm curious what you would say
570
if you could send a telegram a message to the 18 to 22 year old version of yourself what would
571
that be i could tell you exactly how i felt
572
when i first when nvidia first founded and
573
and uh the three of us started um the the thing i felt at the time is there was
574
so much for me to know and so much for me to learn
575
and i didn't know it and i was telling you earlier
576
at the time there was there were no YouTube there's you
577
know no YC nobody's teaching you how to start a company
578
and so I went to the bookstore and I bought a book and the book said how to start a company.
579
Unfortunately the book was like 500 pages long.
580
And so, you know, I figured by the time I read it, you know, I'd be out of business, and Lori and I'd be out of money, and so there's no sense reading it.
581
But the thing I remember very vividly is that how scared I was to go raise money,
582
because I felt that I was about to talk to a bunch of people, and I didn't know how to answer their questions.
583
And it's true, and I barely know how to answer their questions even today.
584
But the thing that I learned is none of that stuff matters, as it turns out.
585
And you're always going to have things that you don't know.
586
And every single day, the world is changing, technology is changing.
587
Obviously, this is the greatest time in the last 60 years to start a company.
588
The whole industry has changed.
589
It's a complete reset from a technology perspective.
590
The single most important technology in human history, the computer, has been completely reset.
591
And so this is absolutely the single greatest time to start a company.
592
And I'm jealous of all of you and the opportunities you have ahead.
593
I mean, it's going to be incredible.
594
So it's the perfect time on the one hand.
595
On the other hand, the technology is changing so fast.
596
And so the question is, what's the right feeling for you?
597
And eventually, and I told you the story of us, of me buying the other book, the textbook.
598
I think the psychology and the feeling that I have today
599
on all of the new experiences and the new technology and new markets and new dynamics, I look at it and I say, this is important.
600
I've got to go learn it.
601
And I've got to go do something about it.
602
And I better get to it as fast as I can.
603
And how hard can it be?
604
I always have this feeling, how hard can it be? and truth be told
605
it is way harder than you think and
606
but you don't want your mind to be to be there
607
you want your mind to be how hard can it be
608
and let the suffering come to you a little bit at
609
a time you know don't don't imagine how hard it's going to be
610
and let all of that turn into anxiety and not doing something about it.
611
You want to imagine in your head, how hard can it be?
612
You know, I've got a whole bunch of AI agents helping me anyways.
613
And so how hard can it be?
614
And then you get going on working on it.
615
And so that's probably the attitude of an entrepreneur.
616
You know you have to learn a bunch of stuff along the way.
617
You believe in your ability to learn, which is, you know, learning is the single greatest superpower.
618
And if you go into it with the attitude, how hard can it be? if anybody can do it, I can do it, and just realize that it will be hard.
619
And you just have to have the resilience to overcome it every single day.
620
You don't have to overcome life in one day.
621
You just have to overcome that morning, that morning.
622
You know, you have to overcome today, today.
623
And so it's not a big deal.
624
Just get through today.
625
Wait till, right?
626
Work towards tomorrow.
627
Keep following your dreams.
628
And the rest of everything, if you stick with it long enough, you know, NVIDIA happens.
629
And so, you know, I think that the wisdom that I can, if there's anything, is resilience is probably the single most important thing.
630
And if you believe in something, just get going on it and get your mind,
631
you know, out of keeping yourself from pursuing it because of, you know, fear or anxiety or lack of confidence or whatever it is.
632
And you're just going to tell yourself, I'm going to learn my way there.
633
Jensen Wong, everybody.
634
All right, guys.
635
Thank you.
636
Thank you so much.
637
Yes.
638
Thank you, guys.
הורד אפליקציה
כל מה שאתה צריך כדי לדבר בשטף
הגיית AIציון לכל משפט
תרגול IPAשלוט בכל צליל
אוצר מיליםבנה את מאגר המילים שלך
משחק אוצר מיליםלמד תוך כדי משחק
אותו ערוץ
C11
Garry Tan: Own Your Intelligence
Daily ConversationWorking Adults
B22
Dot Plots: How to Actually See What Your Users Are Doing
Daily ConversationWorking Adults
B25
How Pitching Investors is Different Than Pitching Customers - Michael Seibel
Business EnglishWorking Adults
C115
Alexandr Wang on what matters in an age of abundant intelligence
Daily ConversationUniversity StudentsWorking Adults
C1Alexandr Wang: “This is a Once-in-a-Civilization Opportunity”
Daily ConversationUniversity StudentsWorking Adults
B2How to Get Your First 10 Customers
Business EnglishWorking Adults
C1How to Build an AI-Native Services Company
Business EnglishUniversity StudentsWorking Adults
FOR YOU
המלץ על סרטון
C12
318 משפטים
實戰練英文｜黃仁勳：你們要加油，不然我會拿走所有機會，然後我會變很有錢｜德國沒有下一個機會了｜投資德國高峰會｜長文、註解、大字版、中英對照
Daily ConversationUniversity StudentsWorking Adults
B21
286 משפטים
Theranos – Silicon Valley’s Greatest Disaster
Daily ConversationUniversity StudentsWorking Adults
B21
331 משפטים
Dear Higgsfield, I reverse engineered you in 12 hours
Daily ConversationUniversity StudentsWorking Adults
B21
546 משפטים
Do Elite Degrees Still Matter In The AI Era? | Jamie Beaton from Crimson Education | EP 410
Daily ConversationUniversity StudentsWorking Adults
C11
532 משפטים
How an Ex-Googler Accidentally Founded a Database Company
Daily ConversationUniversity StudentsWorking Adults
C11
71 משפטים
Did the AI bubble just pop?
Daily ConversationUniversity StudentsWorking Adults
C11
40 משפטים
The Engineering that Runs the Digital World 🛠️⚙️💻 How do CPUs Work?
Daily ConversationUniversity StudentsWorking Adults
C11
60 משפטים
Quick Study: How Venture Capitalists Make Decisions
Daily ConversationUniversity StudentsWorking Adults
B21
351 משפטים
Before the Comeback: Steve Jobs’ Most Honest Interview (1994) #tech #apple #news
Business EnglishDaily ConversationUniversity Students
C14
188 משפטים
NVIDIA GTC 2025 - Keynote Highlights by NVIDIA CEO Jensen Huang
Daily ConversationUniversity StudentsWorking Adults
C11
274 משפטים
A 4-Step System to Process Information with Extreme Clarity
Daily ConversationUniversity StudentsWorking Adults
B21
25 משפטים
Elon was my only customer of our Al Supercomputer #jensenhuang #elonmusk #openai
Business EnglishDaily ConversationWorking Adults
Vocabulary and speaking notes for this lesson
This video has 638 sentences and 8027 words to shadow. The speech runs for 48:52. The speaker talks at a natural 164 words per minute, close to everyday conversation. 86% of the words are among the 3,000 most common in English; the rest is worth studying before you start.
Key vocabulary in this video
15 words from the video worth learning, with pronunciation and meaning:
| Word | Pronunciation | Meaning |
|---|---|---|
| algorithm noun | /ˈælɡəɹɪðm̩/ | A collection of ordered steps that solve a mathematical problem. A precise step-by-step plan for a computational procedure that possibly begins with an input… |
| solve verb | /sɒlv/ | To find an answer or solution to a problem or question; to work out. |
| physics noun | /ˈfɪz.ɪks/ | The branch of science concerned with the study of the properties and interactions of space, time, matter and energy. |
| chip noun | /t͡ʃɪp/ | A small piece broken from a larger piece of solid material. |
| generate verb | /ˈd͡ʒɛn.ə.ɹeɪt/ | To bring into being; give rise to. |
| robotics noun | /ɹoʊˈbɑ.tɪks/ | The science and technology of robots, their design, manufacture, and application |
| domain noun | /ˌdəʊˈmeɪ̯n/ | A geographic area owned or controlled by a single person or organization. |
| graphics noun | /ˈɡɹæfɪks/ | The making of architectural or design drawings. |
| textbook noun | /ˈtɛkst.bʊk/ | A coursebook, a formal manual of instruction in a specific subject, especially one for use in schools or colleges. |
| founder noun | /ˈfaʊ̯n.dəː/ | One who founds or establishes (a company, project, organisation, state, etc.). |
| robot noun | /ˈɹoʊ.bɑt/ | An intelligent mechanical being designed to look like a human or other creature, and usually made from metal. |
| stack noun | /stæk/ | A large pile of hay, grain, straw, or the like, larger at the bottom than the top, sometimes covered with thatch. |
| cloud noun | /ˈklaʊ̯d/ | A visible mass of water droplets suspended in the air. |
| incredibly adverb | /ɪŋˈkɹɛdɪbli/ | In an incredible manner; not to be believed. |
| principle noun | /ˈpɹɪn.sɪ.pəl/ | A fundamental assumption, fundamental law or guiding belief. |
Phrasal verbs you will hear
| Word | Pronunciation | Meaning |
|---|---|---|
| turn out verb | To end up; to result. | |
| come along verb | To accompany. | |
| come up with verb | To manage to produce, deliver, or present (something) by inventing, creating, thinking of, or obtaining it. | |
| break down verb | To stop functioning. | |
| come out verb | /ˌkʌm ˈaʊt/ | To be discovered; to be revealed. |
| build up verb | To erect; to construct. | |
| come out with verb | To say (something) unexpectedly. | |
| figure out verb | To come to understand; to discover or find a solution; to deduce. |
Pronunciation to watch
The speaker uses 128 contractions and reduced forms, such as you're, don't, we're. Say them the short way, as you hear them.
- The “th” sounds: algorithm /ˈælɡəɹɪðm̩/, breakthrough /ˈbɹeɪk.θɹuː/
- The “sh” and “zh” sounds: ambitious /æmˈbɪʃ.əs/, simulation /ˌsɪm.jəˈleɪ.ʃn̩/, imagination /ɪˌmæd͡ʒəˈneɪʃən/, sufficiently /səˈfɪʃəntli/, conditional /kənˈdɪʃ.ə.nəl/
- Long words — get the stress right: incredibly /ɪŋˈkɹɛdɪbli/, autonomous /ɔˈtɑnəməs/, universal /ˌjunəˈvɜɹs(ə)l/, anxiety /æŋˈzaɪ.əti/, architecture /ˈɑː.kɪˌtɛk.t͡ʃə/
How to practise with this video
- Listen to the whole video once without speaking and note the words you do not know.
- Start at 0.75× speed, shadow it sentence by sentence, then go back to normal speed once it feels easy.
- Record yourself and compare with the original, paying attention to words like algorithm, solve, physics.
דקדוק בסרטון הזה
המבנים שהדובר משתמש בהם הכי הרבה, עם המילים המדויקות מהסרטון:
| מבנה | בסרטון |
|---|---|
| הווה מושלם (Present perfect) have/has + past participle — פעולה שקרתה בעבר ועדיין יש לה חשיבות או השפעה עכשיו | has been · I've always believed · you've said |
| סביל (Passive voice) be + past participle — הדגש הוא על מה שקורה, לא על מי שעושה את הפעולה | was just rumored · be done · be automated |
| משפטי זיקה (Relative clauses) who / which + clause — מידע נוסף על אדם או דבר | students who only · man who controls · people who say |
מהי טכניקת השאדואינג?
שאדואינג היא טכניקת למידת שפה מבוססת מדע שפותחה במקור להכשרת מתורגמנים מקצועיים ופופולרית על ידי הפוליגלוט ד"ר אלכסנדר ארגואלס. השיטה פשוטה אך עוצמתית: אתה מקשיב לאודיו באנגלית ילידית וחוזר עליו בקול רם מיד — כמו צל העוקב אחרי הדובר עם עיכוב של 1–2 שניות בלבד. בניגוד להאזנה פסיבית או תרגילי דקדוק, שאדואינג מכריח את המוח והשרירים בפה לעבד ולשחזר דפוסי דיבור אמיתיים בו זמנית. מחקרים מראים שזה משפר משמעותית את דיוק ההגייה, האינטונציה, הקצב, הדיבור המחובר, הבנת השמיעה ושטף הדיבור — מה שהופך אותו לאחת השיטות היעילות ביותר להכנה לIELTS Speaking ולתקשורת באנגלית בעולם האמיתי.