【彭博专访·中英字幕】黄仁勋卸任CEO前最后一次正式采访
字幕摘录
| 时间 | 英文 | 中文 |
|---|---|---|
| 0:10 | Jensen likes to say AI is like a five layer cake. | Jensen喜欢说AI就像一个五层蛋糕. |
| 0:14 | From the bottom up, energy, chips, infrastructure, | 从下而上,能源,芯片,基础设施, |
| 0:18 | the models, and the apps on top. | 模型和上面的应用程序 |
| 0:22 | Now, NVIDIA started out squarely in layer two making the chips, | NVIDIA从第2层开始 制造芯片 |
| 0:26 | but that has since changed and now NVIDIA is investing up and | 但后来情况发生了变化 现在NVIDIA正在投资 |
| 0:30 | down all the layers of the cake in power, | 下方所有层的蛋糕 在权力, |
| 0:33 | in data centers, in the AI labs, | 在数据中心,在AI实验室, |
| 0:35 | in startups, which means no matter how you slice it, | 在初创阶段, 这意味着无论你如何切片, |
| 0:38 | more and more of the AI industry runs through NVIDIA. | 越来越多的AI产业通过NVIDIA. |
| 0:42 | 27年前,NVIDIA公开. |
展开字幕全文(478 条)
| 序号 | 英文 | 中文 |
|---|---|---|
| 1 | Jensen likes to say AI is like a five layer cake. | Jensen喜欢说AI就像一个五层蛋糕. |
| 2 | From the bottom up, energy, chips, infrastructure, | 从下而上,能源,芯片,基础设施, |
| 3 | the models, and the apps on top. | 模型和上面的应用程序 |
| 4 | Now, NVIDIA started out squarely in layer two making the chips, | NVIDIA从第2层开始 制造芯片 |
| 5 | but that has since changed and now NVIDIA is investing up and | 但后来情况发生了变化 现在NVIDIA正在投资 |
| 6 | down all the layers of the cake in power, | 下方所有层的蛋糕 在权力, |
| 7 | in data centers, in the AI labs, | 在数据中心,在AI实验室, |
| 8 | in startups, which means no matter how you slice it, | 在初创阶段, 这意味着无论你如何切片, |
| 9 | more and more of the AI industry runs through NVIDIA. | 越来越多的AI产业通过NVIDIA. |
| 10 | 27年前,NVIDIA公开. | |
| 11 | And they put me in this hotel. | 他们把我关在这酒店里 |
| 12 | Oh, really? | 是吗? |
| 13 | Yes. | 对 |
| 14 | It was the first time I ever had a butler. | 这是我第一次有管家。 |
| 15 | There was this 2000 and you'll be blown away $300 million. | 有2000年,你会被吹走3亿美元。 |
| 16 | It's, things have changed. | 事情变了 |
| 17 | $300 million. | 3亿美元。 |
| 18 | That was a lot back then though. | 那是很多的时候,虽然。 |
| 19 | No, no, no, it was fine. | 不,不,不,这是罚款。 |
| 20 | I was just happy to make it. | 我很高兴能来 |
| 21 | You were such an unlikely winner back then. | 那时候你是这么不可能的赢家 |
| 22 | Yeah, I was just happy to stay in business. | 是啊,我只是很高兴 留在商业。 |
| 23 | You know what I was doing. | 你知道我在做什么吗? |
| 24 | Didn't know how to write a business plan. | 不知道怎么写商业计划 |
| 25 | And now? | 现在呢? |
| 26 | No, still don't know how to write a business plan. | 不,还是不知道 如何写一个商业计划。 |
| 27 | What makes this trip a success? | 是什么使这次旅行成功? |
| 28 | Is it like government commitments, | 这就像政府的承诺, |
| 29 | selling chips, data centers? | 卖芯片,数据中心? |
| 30 | You know, I don't have a end game. | 你知道,我没有结束游戏。 |
| 31 | I don't have a roadmap. | 我没有地图 |
| 32 | You know, I don't have a plan. | 你知道,我没有计划。 |
| 33 | I'm not trying to get anywhere. | 我可没想走 |
| 34 | You know, there's no completion. | 你知道,没有完成。 |
| 35 | I'm kind of one of those people that | 我是那种那种人 那些 |
| 36 | the journey, the journey is the destination. | 旅程,旅程是目的地。 |
| 37 | Huang and his two business partners | 阿黄和他的两个生意伙伴 |
| 38 | wanted to create chips, specifically GPUs, | 想要创建芯片,特别是GPU, |
| 39 | that would make for better video game graphics. | 这样可以改进电子游戏图形。 |
| 40 | But it was a rocky start. | 但它是一个摇滚的开始。 |
| 41 | 2006年,NVIDIA揭幕CUDA, | |
| 42 | software that allowed its chips | 允许其芯片的软件 |
| 43 | to do more than run video games, | 做更多的 运行电子游戏, |
| 44 | from weather forecasting to gene sequencing. | 从天气预报到基因测序 |
| 45 | AI researchers soon discovered those same chips | AI研究人员很快发现了同样的芯片 |
| 46 | could be used to train their models. | 可以用来训练他们的模型。 |
| 47 | When word got back to Huang, he went all in, | 传回黄帝时,他全进去了 |
| 48 | betting big on tech built for AI | 为AI打造的科技大赌注 |
| 49 | before most of the industry caught on. | 之前,大多数行业 抓住。 |
| 50 | NVIDIA powered the early breakthroughs | NVIDIA推动了早期的突破 |
| 51 | that brought AI mainstream, | 这带来了AI的主流, |
| 52 | and it's been one of the most lucrative bets | 这是最赚钱的赌注之一 |
| 53 | in corporate history. | 在公司历史。 |
| 54 | 今天你会找到NVIDIA | |
| 55 | at the core of most major tech companies. | 在大多数主要技术公司的核心。 |
| 56 | OpenAI,Meta,谷歌,Anthropic,微软, | |
| 57 | SpaceX, Amazon, 和列表继续. | |
| 58 | AI is deeply unpopular, and people are scared. | AI非常不受欢迎,人们很害怕. |
| 59 | Sam Altman, Dario Amadei, Elon Musk | 萨姆·阿尔特曼、达里奥·阿马迪、埃隆·穆斯克 |
| 60 | all say we should slow down. | 我们都说我们应该慢下来。 |
| 61 | You have been very consistent in your optimism. | 你一直很乐观 |
| 62 | What do you see that they don't? | 你看到什么他们没有? |
| 63 | You don't advance capability, leaving safety behind. | 你不提高能力, 留下安全。 |
| 64 | You don't advance safety, leaving capability behind. | 你不推进安全, 留下能力。 |
| 65 | So let me give you an example. | 让我举个例子。 |
| 66 | When AI first came out, | 当AI第一次出来的时候, |
| 67 | the first criticism of the original chat GBT | 对最初聊天的首次批评 GBT |
| 68 | was, of course, hallucination. | 当然是幻觉 |
| 69 | It was just spewing nonsense. | 只是胡说八道 |
| 70 | And so in order for us to make these chatbots | 所以为了让我们做这些聊天机 |
| 71 | grounded in truth, we had to invent something like | 基于真理,我们不得不发明一些 |
| 72 | retrieval augmented generation. | 检索增强生成。 |
| 73 | We had to expand the context window. | 我们必须扩大上下文窗口。 |
| 74 | We connected to tools so that it could do a search | 我们把工具连接起来 以便它能够进行搜索 |
| 75 | before it answers the question, | 在它回答问题之前 |
| 76 | knows when to do a search, | 知道何时该进行搜索 |
| 77 | because maybe the information requires | 因为也许信息需要 |
| 78 | contextual information or contemporary information, | 背景资料或当代信息, |
| 79 | and so it has to go on and do a search. | 所以它必须继续搜索 |
| 80 | And all of that is part of its conditioning | 这一切都是它的一部分条件 |
| 81 | before it generates the answer. | 在它产生答案之前。 |
| 82 | It also reasons afterwards. | 这也是后来的原因。 |
| 83 | It reflects on the answer it gave you, | 它反映了它给你的答案, |
| 84 | and you go, you know, | 你走,你知道吗, |
| 85 | is this the best answer I can provide? | 这是最好的答案吗? |
| 86 | Or it might even generate two or three different versions | 也可能生成两三个不同的版本 |
| 87 | of the answer and reflects on which one's the best one. | 答案和反思 谁是最好的。 |
| 88 | All of that require technology. | 所有这些都需要技术。 |
| 89 | So safety and capability to me is one idea. | 所以对我来说安全和能力是一个想法。 |
| 90 | Microsoft is calling for an AI emergency break | 微软要求AI紧急休息 |
| 91 | and says AI is less regulated than washing machines. | 并说AI比洗衣机的规范更少. |
| 92 | Anthropic is saying their models could evade shutdown | Anthropic说他们的模型可以逃避关闭 |
| 93 | in a matter of months. | 几个月后 |
| 94 | What would you have to see that might convince you | 你要看到什么才能说服你 |
| 95 | that the industry should slow down or stop? | 行业应该放慢或停止? |
| 96 | Several things. | 几件东西 |
| 97 | One, if we're developing agenting systems | 第一,如果我们正在开发代理系统 |
| 98 | or autonomous systems, | 或自动系统, |
| 99 | the agents have to be in rigorous containment systems. | 毒剂必须处于严格的控制系统 |
| 100 | And those agents must be monitored. | 这些特工必须受到监视。 |
| 101 | Just job one. | 就做一个 |
| 102 | If you're not able to do that, you should not test it. | 若无此法不应验. |
| 103 | If you're not able to do that, you should not deploy it. | 若不能达者,不应部署. |
| 104 | Well, these models that went rogue are unreleased models. | 这些无赖的模型是无法释放的 |
| 105 | So is that reason that they should shut down, | 因此,他们应该关闭, |
| 106 | shut the labs down? | 关闭实验室? |
| 107 | No, what they should do is that they should contain | 不,他们应该做的是 他们应该包含 |
| 108 | and monitor those experiments. | 并监测这些实验。 |
| 109 | But they say they can't contain them. | 但他们说不能控制他们 |
| 110 | No, that's just made up. | 不,那只是编造。 |
| 111 | Nonsense. | 胡说 |
| 112 | I cannot be proven. | 我无法证明 |
| 113 | You can't prove that to me. | 你无法证明这一点给我。 |
| 114 | Yeah, you can absolutely contain, | 是的,你可以绝对地控制, |
| 115 | you can absolutely monitor. | 你可以完全监视。 |
| 116 | Until you try to do it, don't tell me you can't do it. | 在你试图做到这一点之前,不要告诉我你不能做到这一点。 |
| 117 | And so I'm just not gonna listen to that. | 所以我不会听你的 |
| 118 | So are you saying that you don't believe | 你是说你不相信 |
| 119 | what they're saying? | 他们说什么? |
| 120 | I don't know all that they're saying. | 我不知道他们说什么。 |
| 121 | If you're developing agentic systems | 如果你正在开发代理系统 |
| 122 | and you can't monitor it, you should not do so. | 你不能监视它,你不应该这样做。 |
| 123 | That's just the bottom line. | 这只是底线。 |
| 124 | Even the biggest AI optimists see risks. | 即使最大的AI乐观主义者也看到了风险. |
| 125 | They also see solutions. | 他们也看到了解决办法。 |
| 126 | What is the biggest risk you see? | 你看到的最大风险是什么? |
| 127 | Developing agentic systems without containment, | 开发无限制的代理系统, |
| 128 | developing agentic systems without monitoring, | 开发无监测的代理系统, |
| 129 | it is extremely high risk. | 风险极高。 |
| 130 | You should not do that. | 你不应该这样做。 |
| 131 | And until you have the ability to contain and monitor, | 直到你有能力控制和监测 |
| 132 | you should not run large scale experiments. | 你不应该进行大规模实验 |
| 133 | It just sounds like that's already happening. | 这听起来就像已经发生了。 |
| 134 | Then you just, you gotta stop. | 那你必须停下来 |
| 135 | That's it, it's really simple. | 就这样,它真的很简单。 |
| 136 | 所以这就是给Anthropic、OpenAI的信息 | |
| 137 | where we've seen the- | 我们在那里见过... |
| 138 | You should contain your technology, | 你应该控制你的技术, |
| 139 | contain your experiments, you should monitor your experiments. | 包含你的实验, 你应该监视你的实验。 |
| 140 | You should give these agents, when you deploy them, | 你应该给这些探员,当你部署他们时, |
| 141 | minimum rights. | 最低权利。 |
| 142 | They should be treated with zero trust. | 应当以零信任的态度对待他们。 |
| 143 | They should be monitored all the time. | 他们应该一直受到监视。 |
| 144 | If one of the labs was doing something | 如果其中一个实验室在做一些事情 |
| 145 | that was too dangerous, that you felt was too dangerous, | 太危险了,你觉得太危险了, |
| 146 | would NVIDIA stop selling to them? | NVIDIA会停止出售给他们吗? |
| 147 | Well, it's a hypothetical, but of course. | 当然是假设 |
| 148 | Yeah, yeah, it's a hypothetical, but of course. | 是啊,是啊,这是一个假设, 但当然。 |
| 149 | We don't run other people's companies. | 我们不经营别人的公司。 |
| 150 | And there's a limit to the judgment by which we apply | 我们的判断是有限度的 |
| 151 | to other companies' use of our technology, | 其它公司使用我们的技术, |
| 152 | our technology's general purpose. | 我们的科技总目的 |
| 153 | But if we know for certain that an application | 但如果我们知道申请 |
| 154 | of our technology is bringing harm to people, of course. | 当然,我们的技术正在给人们带来伤害。 |
| 155 | More than 100 data center projects have been blocked | 100多个数据中心项目受阻 |
| 156 | or delayed in the U.S. just so far this year. | 或在美国延迟 仅仅今年至今。 |
| 157 | There's pushback in Europe too. | 欧洲也有后退 |
| 158 | You have AIIPOs getting delayed, getting scrapped. | 你有AIIPO被拖延, 被废掉。 |
| 159 | Are we seeing cracks in the AI boom? | 我们在AI爆炸中看到裂缝吗? |
| 160 | And could you be left holding the bag? | 你能把包放一边吗? |
| 161 | It's too fundamental. | 这太根本了 |
| 162 | I think this is the future of computing. | 我认为这就是计算的未来。 |
| 163 | At the fundamental level, | 在基本层面, |
| 164 | computing will never go back. | 计算永远不会回头。 |
| 165 | We used to have computers | 我们以前有电脑 |
| 166 | that the software's written by hand, | 软件是手写的, |
| 167 | and the computer's waiting for us to use it. | 电脑在等我们使用 |
| 168 | In the future, and where we are now, | 将来,以及我们现在的位置, |
| 169 | computers are able to learn environments | 计算机能够学习环境 |
| 170 | and problems that are too complicated | 和过于复杂的问题 |
| 171 | for any human to program. | 为了任何人类的节目 |
| 172 | The capabilities of AI, now that we know what it can do, | AI的能力,现在我们知道它能做什么, |
| 173 | that's never gonna go back. | 永远也回不去了 |
| 174 | So I think in a lot of many ways, | 所以我想,在许多方面, |
| 175 | it's fundamental, irrespective of near-term challenges. | 这是根本的, 无论近期的挑战。 |
| 176 | And so I think the narrative of the industry | 所以我觉得这个行业的叙述 |
| 177 | is really out of control and ought to be stopped. | 已经失控了,应该停止。 |
| 178 | But that's just an industry growing up. | 但这只是一个行业成长。 |
| 179 | You probably remember the good old days of the internet, | 你可能记得互联网上的好时光 |
| 180 | and the industry grew up. | 和产业成长。 |
| 181 | And I think that our industry is, | 我认为我们的行业是, |
| 182 | many of the young companies | 许多年轻的公司 |
| 183 | are going through their growing pains, | 他们正在经历不断增长的痛苦, |
| 184 | and they're growing, of course, incredibly fast. | 他们正在成长,当然, 惊人的快速。 |
| 185 | There's an accountability, responsibility | 有问责制,责任 |
| 186 | that goes along with building companies | 与建筑公司并列 |
| 187 | that I think they're just growing into. | 我认为他们正在成长 |
| 188 | NVIDIA keeps adding new products, buying up companies. | NVIDIA不断添加新产品,收购公司. |
| 189 | In some cases, you're competing with your own customers. | 在某些情况下,你和自己的顾客竞争. |
| 190 | Why do you need to have such a broad range of offerings? | 为什么你需要如此广泛的祭品? |
| 191 | We do as little as we can. | 我们尽量少做点 |
| 192 | It seems like you're doing a lot. | 看来你做了很多 |
| 193 | Surprisingly little. | 令人惊讶的是小。 |
| 194 | I think we've done a good job | и谋眔и暗眔 |
| 195 | generalizing the work that we do | 概括我们所做的工作 |
| 196 | so that it could be applied to many different things. | 以便适用于许多不同的事情。 |
| 197 | Not because we're doing many different things, | 不是因为我们做了很多不同的事情, |
| 198 | it's because we generalized one thing. | 因为我们只概括了一件事 |
| 199 | And so if you look at our architecture, | 所以如果你看看我们的建筑 |
| 200 | we're the only architecture that's general purpose. | 我们是唯一具有普遍意义的建筑。 |
| 201 | Meaning we run every closed model, | 意味着我们运行每一个封闭的模型, |
| 202 | we run every open model, | 我们运行每一个开放的模型, |
| 203 | we run biology models, the language models, | 我们运行生物学模型,语言模型, |
| 204 | autoregressive models, the transformer models, | 自动后退型号 变压器型号 |
| 205 | the diffusion models, | 传播模型, |
| 206 | and all the different various technologies. | 以及各种不同的技术 |
| 207 | The challenge of generalization is, of course, performance. | 当然,一般化的挑战在于业绩。 |
| 208 | How can you be the world's fastest F1 race car | 你怎么会是世界上最快的F1赛车 |
| 209 | and also be a really great sedan? | 也是辆很棒的轿车? |
| 210 | And so that's the miracle of the technology | 这就是科技的奇迹 |
| 211 | that we've created. | 我们创造了。 |
| 212 | On the one hand, we're the world's fastest F1 racing cars. | 一方面,我们是世界上最快的F1赛车. |
| 213 | And we're the fastest computers. | 我们是最快的计算机。 |
| 214 | We're always at the top of the leaderboard. | 我们总是在领导板的顶端。 |
| 215 | You benchmark us, we're always the best. | 你以我们为基准 我们永远是最好的 |
| 216 | On the other hand, because we're fungible, | 另一方面,因为我们是可替换的, |
| 217 | we're also the lowest cost. | 我们也是最低的成本。 |
| 218 | NVIDIA最近承诺支出 | |
| 219 | about half a trillion dollars | 约5万亿美元 |
| 220 | across the global AI industry, | 整个全球AI产业, |
| 221 | and that number is quickly climbing. | 这个数字正在迅速攀升 |
| 222 | There's concern that in doing so, | 有人担心这样做, |
| 223 | NVIDIA is creating false demand, | NVIDIA正在制造虚假的需求, |
| 224 | giving companies the money they need | 给公司所需的钱 |
| 225 | to turn around and buy products from NVIDIA. | 转身从NVIDIA购买产品。 |
| 226 | It's no secret that critics have accused NVIDIA | 批评者指责NVIDIA不是秘密 |
| 227 | of circular financing. | 循环融资。 |
| 228 | How do you toe the line between boosting promising startups | 你如何在有前途的初创企业之间 |
| 229 | and blowing up a bubble? | 吹泡泡吗? |
| 230 | Like if AI is such a goldmine, | 如果AI是金矿 |
| 231 | why do you have to underwrite so much of it? | 为什么要买这么多? |
| 232 | Yeah, this technology requires high capex. | 是啊,这个技术需要高额的盖帽。 |
| 233 | If you look at this generation of AI startups, | 如果你看看这代人工智能的创始人 |
| 234 | it's very, very rare that you find a company | 找到公司是很罕见的 |
| 235 | that can be funded on $10 million. | 这笔资金可达1000万美元。 |
| 236 | And it's not unusual to find a company | 找到公司也没什么大不了的 |
| 237 | that requires $10 billion. | 这需要100亿美元。 |
| 238 | And the reason for that is because it needs | 原因就是因为它需要 |
| 239 | this computation resource in order | 此计算资源依次 |
| 240 | to deliver the intelligence. | 来传递情报 |
| 241 | Some of the highest quality companies | 一些质量最高的公司 |
| 242 | in the AI ecosystem, across each one of the layers | 在人工智能生态系统中,每个层 |
| 243 | of the five layer cake, it connects us | 五层蛋糕,它连接我们 |
| 244 | to the best and the most talented companies. | 给最优秀最有才华的公司 |
| 245 | It helps us continue to advance NVIDIA's platform, | 它帮助我们继续推进NVIDIA的平台, |
| 246 | if you will, because these companies | 如果你愿意的话 因为这些公司 |
| 247 | are gonna be built on NVIDIA's technology. | 将建立在NVIDIA的技术之上。 |
| 248 | And it helps us capitalize and be one | 它帮助我们利用 成为一体 |
| 249 | of the early investors in many of them. | 他们中的许多人是早期投资者。 |
| 250 | We have a highly disciplined approach | 我们的规矩很严谨 |
| 251 | of making sure that we are, | 确保我们是, |
| 252 | although one of the earliest investors, | 虽然是最早的投资者之一, |
| 253 | and many times even the anchor investor, | 许多次甚至主投资人, |
| 254 | we're not the lead investor. | 我们不是主要的投资者。 |
| 255 | We're not the largest investor. | 我们不是最大的投资者。 |
| 256 | In all cases, we own less than 20% of the company. | 无论如何,我们拥有的公司不到20%。 |
| 257 | We've invested some, I guess in the last five years, | 我们投资了一些, 我想在过去五年里, |
| 258 | maybe close to $100 billion or something like that. | 也许接近一千亿美元 或类似的东西。 |
| 259 | That $100 billion resulted in, well, | 这1000亿美元的结果,嗯, |
| 260 | at least five times as much other institutional investors | 至少是其他机构投资者的五倍 |
| 261 | into this ecosystem. | 进入这个生态系统。 |
| 262 | Most of the world's leading AI companies depend on NVIDIA. | 世界领先的AI公司大多依赖于NVIDIA. |
| 263 | Not just your hardware, but your software as well. | 不只是你的硬件,还有你的软件 |
| 264 | We're probably the only AI company in the world | 我们可能是世界上唯一的AI公司 |
| 265 | that works with every AI company in the world. | 和世界上每个AI公司合作 |
| 266 | So at what point does that dependence | 因此,在什么时间 依赖 |
| 267 | on one single company become a systemic risk? | 一个公司成为系统风险? |
| 268 | Well, they don't depend on us, they choose us. | 他们不依赖我们 他们选择我们 |
| 269 | I don't worry about that. | 我不担心这一点。 |
| 270 | And the reason for that is because | 原因就是因为 |
| 271 | we don't take our position for granted. | 我们不把我们的立场视为理所当然。 |
| 272 | We came from nothing. | 我们一无所获 |
| 273 | And we have to earn our keep every day. | 我们必须每天挣钱 |
| 274 | We have competitors in every company, | 每个公司都有竞争者 |
| 275 | every industry, every country. | 每一个行业,每一个国家。 |
| 276 | With respect to the size of the market, | 关于市场的规模, |
| 277 | just remember, it's just a number. | 记住,这只是一个数字。 |
| 278 | NVIDIA can only be as large as the, if you will, | NVIDIA只能和它一样大 如果你愿意的话 |
| 279 | the pond that we're in, the market that we're in. | 我们所在的池塘, 我们所在的市场。 |
| 280 | You said AI supply and demand will balance out | 你说AI供求平衡 |
| 281 | in maybe five to 10 years. | 也许五到十年。 |
| 282 | What happens when the world doesn't need as many new chips? | 当世界不需要那么多新筹码时会怎样? |
| 283 | New chips will replace the old chips, | 新的芯片将取代旧芯片, |
| 284 | just kind of like the personal computer industry. | 就像个人电脑工业一样 |
| 285 | You know, we've got, what, | 你知道,我们得到了,什么, |
| 286 | a billion users in the world today. | 今日世界十亿用户. |
| 287 | 1993年,NVIDIA成立之初, | |
| 288 | the personal computer industry | 个人计算机行业 |
| 289 | was doubling every single year. | 年复一年地翻一番。 |
| 290 | And of course, it doesn't double anymore. | 当然,它不再加倍了。 |
| 291 | At the moment, we're kind of doubling every year. | 目前,我们每年都要翻一番。 |
| 292 | And my sense is that we're gonna keep doing that | 我的感觉是,我们会继续这样做 |
| 293 | as we build out the infrastructure of the world. | 当我们建设世界的基础设施时 |
| 294 | And how far out can you see that doubling? | 你还能看多远? |
| 295 | You know, we're gonna do that for several years. | 你知道,我们要做到这一点 几年。 |
| 296 | And then we'll reevaluate. | 然后重审. |
| 297 | And so we're kind of two years | 我们差不多两年了 |
| 298 | into an infrastructure buildout. | 进入一个基础设施建设。 |
| 299 | And if you look at previous technology cycles, | 如果你看看以前的技术周期 |
| 300 | it takes about 10 years or so | 大约需要十年左右 |
| 301 | to reach some kind of maturity. | 达到某种成熟。 |
| 302 | And then after that, you're in replacement cycles. | 然后,你进入了替换周期。 |
| 303 | And so did that with phones, did that with the internet, | 和电话,和互联网, |
| 304 | did that with PCs. | 用电脑做的 |
| 305 | I think it'll happen here too. | 我想这也会发生在这里。 |
| 306 | Could there be a breakthrough | 会不会有突破 |
| 307 | that means AI is powered by a fraction of the energy? | 也就是说人工智能的动力是能量的一小部分? |
| 308 | Oh yeah. | 对 |
| 309 | Chips. | 芯片 芯片 芯片 |
| 310 | Oh yeah, yeah, yeah. | 哦,对,对,对。 |
| 311 | And then what does that mean for NVIDIA? | 那对NVIDIA意味着什么? |
| 312 | No, we're just gonna keep making things more efficient. | 不,我们只是要继续 使事情更有效率。 |
| 313 | There's a weird dynamic. | 有一个奇怪的动态。 |
| 314 | You know, people used to tell me, | 你知道,人们常告诉我, |
| 315 | hey, Jensen, if you make your chips | 嘿,詹森,如果你做的筹码 |
| 316 | that much better every generation, | 每一代人都会好得多 |
| 317 | doesn't it say that customers just buy less? | 上面不是说顾客少买一点吗? |
| 318 | You know? | 你知道吗? |
| 319 | And that same exact logic goes, | 同样的逻辑是, |
| 320 | if we make companies more productive, | 如果我们让公司更有生产力 |
| 321 | don't we hire less people? | 我们不雇更少的人吗? |
| 322 | And there's this weird thing called ambition. | 有个怪怪的叫野心 |
| 323 | If you gave me 36 hours a day, I'm not gonna sleep more. | 如果你每天给我36小时 我不会再睡觉了 |
| 324 | You know? | 你知道吗? |
| 325 | Because I've got ideas, I have things to do. | 因为我有主意,我有事情要做 |
| 326 | Well, I end up doing more. | 好吧,我最后会做更多。 |
| 327 | And so if AI makes us all more intelligent, | 所以如果AI让我们更聪明的话 |
| 328 | I become more ambitious. | 我变得更有野心了 |
| 329 | If the cost of doing something | 如果做一些事情的代价 |
| 330 | becomes marginally close to zero, | 变成零, |
| 331 | it turns out that resource becomes widely abundant. | 结果资源变得非常丰富。 |
| 332 | I think the economy's gonna get larger. | 我觉得经济会变大 |
| 333 | I think the industries will get larger. | 我认为这些行业会变得更大。 |
| 334 | I think the semiconductor industry will be, | 我认为半导体工业是, |
| 335 | at this point, very clearly | 此时此刻,非常清楚 |
| 336 | one of the largest industries in the world. | 世界上最大的工业之一。 |
| 337 | Obviously you've had this big week. | 显然你有一个大星期。 |
| 338 | You've gone to so many countries, so much exciting stuff. | 你去过这么多国家, 这么多令人兴奋的东西。 |
| 339 | But at the same time, there's like constant headlines. | 但与此同时,也有类似恒定的头条新闻. |
| 340 | This massive market sell-off, chip stocks under pressure. | 这种巨大的市场销售,芯片股票在压力下。 |
| 341 | As this trip has been going on, | 随着这次旅行的进行 |
| 342 | what have you been watching in the markets? | 你在市场上看什么? |
| 343 | What's your take on what's happening with CoreWeed? | 你对Corewed的事有什么看法? |
| 344 | All of this is gonna pass. | 所有这一切都会过去的。 |
| 345 | And the reason for that is very simple. | 原因很简单。 |
| 346 | We're going through a new platform transition. | 我们正在经历一个新的平台过渡. |
| 347 | News, as you know, happens all the time, | 新闻,你知道, 经常发生, |
| 348 | but it won't change the fundamentals. | 但它不会改变基础。 |
| 349 | You've said AI will create more jobs than industries. | 你说AI会创造比工业更多的工作 |
| 350 | AI does create more jobs. | 大赦国际确实创造了更多的就业机会。 |
| 351 | You've said people worried about losing their jobs | 你说有人担心失去工作 |
| 352 | should learn AI. | 应该学习AI。 |
| 353 | What do you say to the person who has learned AI, | 你对学会AI的人怎么说? |
| 354 | who is doing exactly that, | 到底是谁做的, |
| 355 | but still watching their job slip away? | 却还在看着他们的工作消失? |
| 356 | Well, first of all, I think you're asking | 首先,我想你问 |
| 357 | a theoretical question. | 一个理论问题。 |
| 358 | But let's apply the practical answer first. | 但先应用实际答案. |
| 359 | Now remember, AI is unlike previous technologies | 现在记住,AI不像以前的技术 |
| 360 | in the sense that this requires manufacturing. | 这需要制造 |
| 361 | It's creating construction jobs, | 它创造了建筑工作, |
| 362 | electrician jobs, plumbing jobs. | 电工 管道工 |
| 363 | In the last six months, | 过去六个月 |
| 364 | something like four to $600 billion, | 差不多四到六千亿美元 |
| 365 | four to $600 billion were invested into AI startups. | 40至6000亿美元投资给AI创业公司. |
| 366 | And the reason for that is people see AI is productive, | 原因是人们看到AI是生产性的, |
| 367 | it's effective, and people wanna invest in it. | 有效,人们想投资。 |
| 368 | Those four to $600 billion | 那40亿到6000亿美元 |
| 369 | has just created how many jobs? | 刚刚创造了多少工作? |
| 370 | And so, of course, this technology is a new tool, | 当然,这种技术是一种新工具, |
| 371 | and this new tool is therefore gonna cause | 因此这个新工具将导致 |
| 372 | everybody's jobs to change. | 每个人都需要改变工作 |
| 373 | Resisting the change won't make it go away. | 抗拒改变不会让它消失. |
| 374 | The superpowers that the tech companies have | 科技公司拥有的超能力 |
| 375 | are now made available to literally everybody. | 现在,每个人都可以使用。 |
| 376 | It democratized the technology. | 它使技术民主化。 |
| 377 | It's elevated everybody's capabilities. | 这是提高每个人的能力。 |
| 378 | And so, I see the benefits of AI. | 所以,我看到了AI的好处。 |
| 379 | I'm obviously very aware of the risks | 我显然非常清楚这些风险 |
| 380 | and the technology risks. | 技术风险。 |
| 381 | But in the final analysis, if I don't see the optimistic, | 但归根结底, 如果我不看到乐观, |
| 382 | if we don't see the optimistic, | 如果我们不看到乐观, |
| 383 | incredibly wonderful future that we're gonna create, | 令人难以置信的美好未来,我们要创造, |
| 384 | for what reason are we building it? | 为什么我们要建它? |
| 385 | Many people appreciate your optimism and get excited by it. | 很多人都欣赏你的乐观,并对此感到兴奋. |
| 386 | But what if your optimism proves wrong in the end? | 但如果你的乐观态度最终被证明是错误的呢? |
| 387 | Then I can say that I did everything I could. | 然后我可以说,我尽我所能。 |
| 388 | And the reason for that is because not only am I optimistic, | 原因 是因为我不仅乐观 |
| 389 | I'm doing everything I can to create the technology | 我在尽我所能创造出这个技术 |
| 390 | necessary to keep the industry safe, secure. | 要保证行业的安全 |
| 391 | And so, let's not talk about the future | 所以,我们不要谈论未来 |
| 392 | in a pessimistic way while we do nothing about it | 以悲观的方式,而我们却袖手旁观 |
| 393 | to prevent that from happening. | 防止发生这种情况。 |
| 394 | And so, maybe I'm older, I'm more parental. | 所以,也许我年纪大了,我更是父母。 |
| 395 | Don't just deliver the bad news. | 别只是传达坏消息 |
| 396 | Tell me what you're gonna do about it. | 告诉我你会怎么做 |
| 397 | And so, so long as we can prove to ourselves | 所以,只要我们能证明自己 |
| 398 | that we're doing both sides of that coin, | 我们做硬币的两边, |
| 399 | the optimistic future while safeguarding our way in doing so. | 乐观的未来,同时保障我们这样做的方式。 |
| 400 | So long as we do both sides of that, I'm perfectly fine. | 只要我们做两边,我完全没事。 |
| 401 | You famously run NVIDIA in a very unconventional way, | 你以非常不寻常的方式管理NVIDIA, |
| 402 | lots of direct reports, no meetings. | 很多直接报告,不开会。 |
| 403 | As the company has exploded in size | 随着公司规模的爆炸 |
| 404 | and the problems have become more complex, | 问题变得更加复杂, |
| 405 | how has your management style and your mentality changed? | 你的管理风格和心态如何改变? |
| 406 | Well, my management style is what made it possible | 嗯,我的管理风格 使得它成为可能 |
| 407 | for the company to grow. | 让公司成长 |
| 408 | Our organization is what made the company | 我们的组织就是公司 |
| 409 | to thrive and innovate. | 繁荣和创新。 |
| 410 | I wanted to create an architecture of a company | 我想建立一个公司的建筑 |
| 411 | with as few silos as possible. | 尽可能少的筒仓。 |
| 412 | The problem with silos, if the organizations are too rigid, | 仓子的问题,如果组织太僵硬, |
| 413 | then the company has very difficult time adapting to change. | 公司很难适应变化 |
| 414 | So I think a modern company really wants to be | 所以我觉得一个现代的公司真的想成为 |
| 415 | more and more agile and less and less structured. | 越来越敏捷 结构越来越小 |
| 416 | What about the CEO who runs that company? | 那经营那家公司的首席执行官呢? |
| 417 | You should be as mentally flexible as possible. | 你应该尽量灵活 |
| 418 | You know, I think you have to hold yourself responsible. | 你知道,我想你必须为自己负责。 |
| 419 | You have agency, you're the CEO of the company, | 你有代理,你是公司的首席执行官, |
| 420 | for God's sakes. | 看在上帝的份上 说吧 |
| 421 | You have to have high standards and expectations. | 你必须有很高的标准和期望。 |
| 422 | Your son and your daughter are working at NVIDIA. | 你儿子和你女儿在NVIDIA工作. |
| 423 | I think your son-in-law also works at NVIDIA. | 我想你的女婿也在NVIDIA工作 |
| 424 | Robotics, yeah. | 机器人,对。 |
| 425 | There is talk that your daughter is in line | 有人说你女儿排队 |
| 426 | to be your successor, is that true? | 成为你的继任者 是吗? |
| 427 | Well, first of all, I don't hope to job on her. | 好吧,首先,我不希望 工作给她。 |
| 428 | And the reason for that is because | 原因就是因为 |
| 429 | that's their aspiration, Spencer and Madison's. | 这是他们的愿望,斯宾塞和麦迪逊的。 |
| 430 | They have to go earn it. | 他们必须去挣钱。 |
| 431 | And it's really, really hard work. | 这真的很辛苦 |
| 432 | It takes a lot of sacrifice. | 这需要大量牺牲。 |
| 433 | You know, for many people, | 对很多人来说, |
| 434 | they can focus on their job and their families. | 他们可以专注于工作和家庭。 |
| 435 | When you're the CEO of the company, | 当你是公司的首席执行官时 |
| 436 | you have to focus on, of course, your family. | 当然,你必须专注于你的家人。 |
| 437 | But I've got a whole bunch of other families | 但我还有一堆其他家庭 |
| 438 | I'm caring for. | 我关心的。 |
| 439 | I'm responsible for an entire industry. | 我负责整个行业 |
| 440 | And so I think the pressure of being the CEO is high | 所以我觉得当首席执行官的压力很大 |
| 441 | and it's supposed to be high. | 它应该是高的。 |
| 442 | It should not be an easy job. | 这不应该是一个容易的工作。 |
| 443 | It shouldn't be a job that you take for granted. | 它不应该是一个你认为理所当然的工作。 |
| 444 | When I decided to be CEO, | 当我决定做首席执行官时 |
| 445 | I didn't realize the magnitude of the pressure | 我不知道压力有多大 |
| 446 | and responsibilities, but look, | 和责任,但你看, |
| 447 | that's what I signed up for. | 这就是我报名的 |
| 448 | And so now that my family and all of the people | 现在我的家人和所有人 |
| 449 | {\fn黑体\fs22\bord1\shad0\3aHBE\4aH00\fscx67\fscy66\2cHFFFFFF\3cH808080}我亲眼看到工作的责任 | |
| 450 | and what it takes, then they just gotta ask themselves | 然后他们只好自问 |
| 451 | whether this is the level of sacrifice | 这是否是牺牲的程度 |
| 452 | and responsibility you want. | 还有你想要的责任 |
| 453 | And I'm really hoping that the next CEO of the company | 我真的很希望公司下一任首席执行官 |
| 454 | is already at the company. | 已经到公司了 |
| 455 | But 20 years from now, I hope I stay healthy | 但二十年后 我希望我保持健康 |
| 456 | and I hope I continue to earn my keep. | 我希望我继续挣钱 |
| 457 | But so long as I can do that 20 years from now, | 但只要我20年后能做到这一点 |
| 458 | I hope they're already here, but they have to earn it. | 我希望他们已经在这里,但他们必须挣钱。 |
| 459 | For the next generation of leaders at NVIDIA, | 为NVIDIA的下一代领导人, |
| 460 | what will their challenge look like? | 他们的挑战会是什么样子? |
| 461 | I don't know. | 师曰. |
| 462 | And that's the, in fact, that's exactly the right answer. | 事实上,这就是正确的答案。 |
| 463 | We should create the conditions | 我们应该创造条件 |
| 464 | by which the company continues to thrive. | 公司继续蓬勃发展 |
| 465 | And we should create the conditions | 我们应该创造条件 |
| 466 | by which the leaders of our company continues to grow | 我们公司的领导人继续成长 |
| 467 | and that the next leader of the company | 而公司的下一位领导人 |
| 468 | 已经在NVIDIA。 | |
| 469 | That's my job, is to create that condition. | 这是我的工作,就是创造这个条件。 |
| 470 | And just as I can be fired, the next CEO has to be hired. | 就像我被炒鱿鱼一样 下一任首席执行官必须被雇用 |
| 471 | And so, I like it that way. | 所以,我喜欢这样。 |
| 472 | I like that we can't be overly complacent | 我喜欢这样,我们不能太自满 |
| 473 | about our position inside the company | 关于我们在公司中的地位 |
| 474 | and that we can't take it for granted. | 而我们不能把它视为理所当然。 |
| 475 | And you're always vulnerable. | 你总是很脆弱 |
| 476 | You're always earning it. | 你总是在赚钱 |
| 477 | You're always striving to be better. | 你总是努力变得更好 |
| 478 | That's a condition that's good. | 这条件很好 |
该视频共有字幕 478 条。解锁更多字幕为会员功能,请移动到 价格
