黄仁勋：英伟达——市值四万亿美元的公司与 AI 革命
Jensen Huang: NVIDIA – The $4 Trillion Company & the AI Revolution
黄仁勋谈英伟达如何从图形芯片公司成长为四万亿美元巨头，以及 AI 革命的算力根基。
Lex Fridman 问黄仁勋，一个能够从零创办、并把一家科技公司运营到十亿美元估值的 AI 系统，还需要多久。
黄仁勋说，现在已经有了。
他紧接着补了一句解释：这个标准里没有“永远”两个字。一个 agent 可以做出某个网页服务，几亿人用上两个月，每人付五毛钱，然后公司又倒闭了。互联网时代满地都是这种公司，那些网站的技术含量，大多还不如今天的 Claude 随手生成的东西。
但他在同一段话里又拆了自己的台：十万个这样的 agent 里，能长出一个 NVIDIA 的概率是零。
这两句话放在一起看才有意思。这位做了 34 年 CEO、把公司带到四万亿美元市值的人，一边把 AGI 的门槛说得极低，一边把真正的难度说得极高。中间那段距离，才是这场两个半小时对话真正在谈的东西。
问题已经装不进一台计算机了
访谈从“极端协同设计”（extreme co-design）开始。黄仁勋给了一个非常物理的解释：要算的问题已经大到一台计算机装不下。
他用的框架是阿姆达尔定律。如果你把计算部分加速一百万倍，而计算只占整个工作量的 50%，那整体只快了两倍。真正的大模型训练和推理，问题的边界早就溢出单机，你必须把算法拆开、把流水线切分、把数据和模型分片。一旦开始分布式，所有东西都变成了瓶颈。
“你加了 10000 台计算机，但你想要快一百万倍。这时候 CPU 是问题，GPU 是问题，网络是问题，交换是问题，把工作负载分配到这一堆机器上也是问题。”（黄仁勋）
摩尔定律救不了这个局面。登纳德缩放（Dennard scaling）早就慢下来了，靠制程红利拿到的性能提升，撑不起这种量级的野心。剩下的路只有一条：把每一层技术同时推到极限，各层互相让路。
六十个直接下属，没有一对一会议
Lex 问，高带宽内存、NVLink、网卡、光模块和铜缆、供电、散热，每一块都有世界级专家，你怎么把他们弄到一个房间里。
黄仁勋的回答很直接：所以我们管理层这么大。
他的直接汇报人数是 60 人，而且还在增加。这个数字在管理学教科书里是灾难，因为一对一会议根本开不完。
他的解法是取消一对一。所有问题都摆在所有人面前解。讨论散热的时候，做供电的人在听；讨论内存的时候，做网络的人在听。谁该发言而没发言，会被点名。谁觉得跟自己无关，可以走神，但代价是错过自己该贡献的那一刻。
“公司存在的目的是成为一台产出东西的机器。公司的组织架构应该反映它所处的环境。我看过很多公司的组织架构图，全都长一个样，汉堡公司的、软件公司的、汽车公司的，看起来都一样，这对我来说毫无道理。”（黄仁勋）
这段话听起来像管理鸡汤，但放到 NVIDIA 的处境里就变了味。他们每年要出货一种由 130 万个零件、200 家供应商组成的机架系统，任何单点专家做决策都会撞墙。组织形态必须长得像产品形态，否则协同设计就是一句空话。
装机量定义架构，优雅的架构全都死了
NVIDIA 起步时是一家加速器公司。黄仁勋说得很坦白，加速器的问题是应用领域太窄。市场窄带来的不只是收入少，更致命的是研发能力受限，而研发能力最终决定了你能在计算领域施加多大影响。
所以他们必须往上走，成为一家计算公司。但“加速计算”这四个字里藏着根本矛盾：计算公司做得越好，专家属性就越弱；越专注，越做不了通用计算。他把这两个词硬拼在一起，就是要逼公司走一条极窄的钢丝。
路线图是一步步铺的。先做可编程像素着色器，这是通向可编程性的第一步。然后把 IEEE 兼容的 FP32 塞进着色器，这一步使所有做流处理器的人发现了他们：原来这个算力怪兽现在能跑我写的代码了。再往上加 C，就是 Cg，Cg 最终通向 CUDA。
CUDA 能不能赢，黄仁勋的判断标准只有一个：装机量。
“没有任何架构比 x86 挨过更多的批评，它被公认不够优雅，但它今天是定义性的架构。反过来有那么多 RISC 架构，设计得极其漂亮，出自全世界最聪明的计算机科学家之手，大部分都失败了。装机量定义架构，其他一切都是次要的。”（黄仁勋）
这个判断直接推出了那个后来被反复讲述的决定：把 CUDA 装进每一块 GeForce，不管用户用不用。
市值从八十亿掉到十五亿的那笔账
这笔账很难看。CUDA 使消费级 GPU 的成本暴涨，涨幅大概 50%。当时公司毛利率 35%，这一下几乎吃掉了全部毛利。
黄仁勋在董事会上把话说清楚了：我们可能会看到这样一个世界，GeForce 背着 CUDA 的成本，但没有一个玩家因此感激，也没有一个玩家愿意为此多付钱，因为玩家只认自己的心理价位，你的成本跟他们无关。
公司市值从大约七八十亿美元跌到十五亿。他们在谷底待了一段时间，慢慢爬回来，但 CUDA 一直留在 GeForce 上。
“我总说 NVIDIA 是 GeForce 建起来的房子，因为是 GeForce 把 CUDA 带给了所有人。”（黄仁勋）
为什么 GeForce 能做到这件事。因为那个年代的 PC 是主要计算载体，还没有云。一块装在学生机箱里的显卡，等于把一台超级计算机塞进了每个实验室、每个工程院系、每个自己攒机的玩家手里。这些人后来成了深度学习革命的第一批信徒。
黄仁勋说这个决定“花了十年”才真正兑现。
他不发宣言，他每天在改别人的信念
Lex 追问的是更软的东西：你怎么敢做这种赌注。
黄仁勋的回答拆成了两层。第一层是推理。他的好奇心很重，某些时点上会有一套推理系统把某个结论推得极其清晰：这件事一定会发生。一旦他心里认定了，那个未来就会变得不容置疑。
第二层才是真正的功夫，他称之为塑造信念系统。
他明确说自己讨厌那种管理风格：领导者憋着不说，某天突然发一份宣言，新财年、新使命、新 logo、大裁员、大重组，好像公司是一夜之间换了个灵魂。
他的做法是反过来的。当他开始被某个想法影响，他会立刻告诉身边所有人：这个东西有意思，这个会影响那个。他用外部信息、新发现、新的工程突破、新的里程碑当作砖，一块一块砌进别人的认知里。这个过程每天都在发生，对董事会、对管理层、对员工。
“等到我说‘我们把 Mellanox 买下来’的那天，所有人都会觉得这是明摆着的事。很多时候看起来像是你在后面推着大家走，其实你早就在塑造他们的信念，所以在宣布的那天，是百分之百的认同。”（黄仁勋）
他举了 GTC 的例子。整个 GTC 的功能之一，就是塑造产业链上下游合作方的信念系统，然后再反过来塑造自己员工的信念。他说 OpenClaw 这套东西他讲了两块半，两年半，回头一翻就发现，原来铺垫已经铺了这么久。
这也解释了 NVIDIA 为什么不能直接卖东西给终端用户。它是平台公司，垂直整合做设计和优化，然后每一层都开放出去，被别人的产品、云、超算、OEM 机器集成。它必须先说服别人，才有生意可做。
四个扩展定律，以及“推理很简单”这个错误
黄仁勋现在是四条扩展定律的信徒：预训练、后训练、测试时、agentic。
他回顾了每一轮“到头了”的恐慌。最早是数据不够。Ilya Sutskever 说过类似“数据用完了、预训练结束了”的话，全行业慌了，以为 AI 走到尽头。
他的反驳角度挺有意思：人类互相传授的知识，本来就是合成的。
“大部分人忘了，我们用来教彼此、互相告知的绝大部分数据都是合成的。因为它不是从自然界长出来的，是你创造出来的，我消费它、修改它、增强它、重新生成，再由别人消费。现在 AI 能够拿真实数据做增强，合成出巨量的数据，后训练这一段的扩展还在继续。”（黄仁勋）
结论是：人类产生的数据占比会越来越小，训练数据的瓶颈从“有多少数据”变成了“有多少算力”。
第二阶段是测试时扩展，也就是推理。他说当年一堆人跟他讲，推理很简单，预训练才难，推理芯片会是小芯片，会商品化，谁都能自己造。
他觉得这个逻辑从一开始就不成立。
“推理就是思考，而我认为思考很难。思考比阅读难得多。预训练只是记忆和泛化，找模式、找关系，是读书；测试时扩展是推理、规划、搜索、把没见过的经验拆成可解的小块，靠第一性原理或者过往案例去解。这怎么可能是算力轻的？”（黄仁勋）
第四条是 agentic 扩展。一个 agent 在测试时跑出去查数据库、调工具，最重要的一件事是不断派生出一堆子 agent。这相当于从一个人变成一个团队。他的类比很朴素：NVIDIA 要扩张，多招人比把自己变强容易得多。
四条定律构成一个闭环。agent 跑出大量经验和数据，好的部分拿回预训练做记忆和泛化，再精调进后训练，再通过测试时和 agent 系统放回产业界，循环往复。最后剩下的唯一变量就是算力。
模型架构半年一换，硬件架构三年一换
Lex 提出了这里最要命的错配：AI 模型架构的迭代周期大概是六个月，系统和硬件架构是三年。你没法在一周内掉头。
黄仁勋承认这是最难的预测问题，然后给了三个办法。
自己做基础研究和应用研究，自己训模型。NVIDIA 是全球唯一一家跟全世界所有 AI 公司都合作的 AI 公司，能听到各家的真实困难。最后是架构本身要足够灵活，能随风转向。CUDA 的价值就在这个平衡点上：一端是极致加速，一端是极致灵活。
他举了一个具体例子。混合专家模型（MoE）出现的时候，他们的回应是把 NVLink 从 8 扩到 72。原因很实在：只有当互连域足够大，你才能把一个四万亿到十万亿参数的模型放在一个计算域里，像跑在单块 GPU 上一样。
“如果你看 Grace Blackwell 机架的架构，它完全聚焦在一件事上：跑大语言模型推理。一年之后你再看 Vera Rubin 机架，它有存储加速器，有全新的 Vera CPU，有 Vera Rubin 加 NVLink 72 跑 LLM，还多了一个叫 Rock 的机架。整个系统的设计逻辑完全变了，因为上一代是给 MoE 推理用的，这一代是给 agent 用工具用的。”（黄仁勋）
Lex 追问，这套设计早于 Claude Code、Codex、OpenClaw 出现，你怎么提前知道。
黄仁勋说，不用知道，想就行了。
人形机器人会用你的微波炉，不会从手指发射微波
他的思想实验是这样的：假设十年后我们造出了最棒的人形机器人。它进你家干活，是倾向于用你家里现成的工具，还是手在某个瞬间变成十磅重的锤子、另一个瞬间变成手术刀，为了烧水就从手指射出微波？
显然是后者荒谬。它更可能走到微波炉前，第一次不知道怎么用，但它连着网，读一遍说明书就变成专家。
这一句话就推出了 agent 系统的全部特征：它要用工具，要访问文件，要能做研究，要有一套 IO 子系统。也推出了一个更大的判断：软件不会因为 AI 而消失，说“AI 会干掉软件，我们不再需要工具”是荒谬的。
“我们用 OpenClaw 的那张架构图，你在两年前的 GTC 上就能找到。我在两年前就在讲这样的 agent 系统。当然，要发生还需要很多事汇聚到一起。首先 Claude、GPT 这些模型的能力要到那个水平，然后得有人做出一个足够健壮、足够完整的开源项目。我认为 OpenClaw 对 agent 系统的意义，就像 ChatGPT 对生成式系统的意义。”（黄仁勋）
关于 OpenClaw 为什么比 Claude Code 和 Codex 更出圈，他的答案很短：因为普通消费者够得着。
安全问题他也正面回应了。NVIDIA 派了一批安全专家过去，做了 OpenShell，已经集成进 OpenClaw。规则是“三选二”：agent 可以访问敏感信息、可以执行代码、可以对外通信，任何时刻最多给两个，不给三个。再加上基于企业权限的访问控制和策略引擎对接。
一百三十万个零件，两百家供应商，没有一份合同
Lex 把供应链的问题摆出来：ASML 的 EUV、台积电的 CoWoS 先进封装、SK 海力士的高带宽内存，这些是不是让你睡不着。
黄仁勋说，天天都在想，也天天都在做。但他不焦虑，理由很直接：我告诉他们我需要什么，他们听懂了，他们告诉我他们要做什么，我相信他们。
这套信任是怎么建立的，他讲了 DRAM 行业的故事。三年前，数据中心里给 CPU 用的 DDR 是 DRAM 行业最大的量。他挨个说服几家 CEO：HBM 现在用量极少，几乎只有超算在用，但它会成为数据中心的主流内存。听起来很荒唐，但几家 CEO 信了，投钱建 HBM 产线。
更难说服的是另一件事：把手机用的低功耗内存 LPDDR5 改造成数据中心超算用的内存。对方的反应是“手机内存给超算用？”
结果是，LPDDR5、HBM4 和 DDR 三条线，背后都是 45 年历史的老公司，去年全部创下历史记录。
机架尺度带来的连锁反应，他讲得最具体。过去超算是在数据中心里组装的，现在是在供应链里组装好，一台两到三吨整机架运过来。这意味着供应链自己要先有电。
“如果你想要 50 吉瓦的超算同时在跑，而制造这批超算需要一周，那么供应链每周就需要一吉瓦的电力来建造和测试它，在我发货之前。所以我得飞去跟合作伙伴说，以前我们是这么造 DGX 的，以后要这么造，而且我会需要你们每人投几十亿美元的资本开支。”（黄仁勋）
电网有 40% 的容量在睡觉
谈能源的时候，黄仁勋讲了一段很少被提起的东西。
电网是按最坏情况加余量设计的。最坏情况是冬天几天、夏天几天、极端天气。剩下 99% 的时间，实际负载离峰值很远，大概只跑到峰值的 60% 左右。
“99% 的时间里，我们的电网都有富余电力，它们就闲置在那里。但它们必须闲置，因为万一那时候医院要供电、基础设施要供电、机场要运转。所以我的问题是，我们能不能帮他们理解，并设计出合同机制和机架、数据中心架构，在社会基础设施需要最大电力的时候，数据中心少拿一点。”（黄仁勋）
这种时刻极其罕见。真到了那天，可以启动备用发电机，可以把负载迁到别处，可以让计算机跑慢一点，延迟变长一点，服务质量降一点，但数据不能丢。
他把这件事卡住的原因拆成三方。第一是终端客户，合同里要求六个九的可用性，签合同的人跟 CEO 不是一个人，他打赌大部分 CEO 根本不知道自己公司签了什么。第二是数据中心要能优雅降级。第三是电力公司要愿意提供分等级的供电承诺，而不是一句“扩容要五年”。
“只要你能把它说清楚，你就能把它工程化。只要它遵循物理定律和第一性原理，我觉得就没问题。”（黄仁勋）
马斯克把“这事一定要这么久吗”问到了底
Lex 请他谈谈 xAI 在孟菲斯建 Colossus 那件事，四个月，现在 20 万块 GPU 还在扩。
黄仁勋的评价集中在三点。第一是系统思维，能跨多个学科思考。第二是极简，他不停追问：这是必要的吗，非得这么做吗，非得这么久吗，一直问到所有东西都被削到不能再削，而产品必需的能力还在。第三是他会亲自到场。
“有问题他就直接去现场，说‘把问题给我看’。当你把这些组合在一起，你就能推翻大量‘我们一直都是这么干的’‘我在等他们’。每个人都有很多借口。最后一件事是，当你亲自带着这么大的紧迫感做事，会逼得所有人都紧迫起来。每个供应商都有很多客户、很多项目，他就是要自己成为所有人项目里的第一优先级，而他靠亲自示范做到这一点。”（黄仁勋）
Lex 补充了一个细节：他见过马斯克跟一线工程师一起研究怎么把线缆插进机架，就为了弄明白这个动作为什么容易出错。从每一个具体任务里积累直觉，再叠加上“我们完全换个做法”的大锤，效率就是这么被压出来的。
光速思维：74 天是怎么来的
黄仁勋说他 30 年前就在用一套方法，叫“光速”。
光速跟速度无关，它是“物理极限在哪里”的简写。内存速度、数学速度、功耗、成本、时间、人力、制造周期，每一样都要拿光速去对标。延迟和吞吐、成本和产能，各自先算出物理极限，然后再放在一起做取舍，因为追求极低延迟的系统和追求极高吞吐的系统，架构根本不同。
他最不喜欢的是持续改进。
“有人跟我说这事现在要 74 天，我们能给你做到 72 天。我更愿意把它全部推倒，问：首先解释一下为什么是 74 天，今天什么是可能的，如果我从零开始造，要多久。答案经常让你吃惊，可能是 6 天。剩下那 68 天可能是各种合理的妥协和降本，但至少你现在知道它们是什么了。而一旦你知道 6 天是可能的，从 74 谈到 6 的对话会出乎意料地有效。”（黄仁勋）
Lex 报了一组 Vera Rubin pod 的数字：7 种芯片、5 种专用机架、40 个机架、1.2 千万亿个晶体管、近两万个 NVIDIA 芯片裸片、1100 多块 Rubin GPU、60 exaflops、每秒 10 PB 的扩展带宽。黄仁勋补充：这只是一个 pod，我们大概每周要造 200 个这样的 pod。
他的复杂度原则是一句话：该复杂的地方必须复杂，但尽可能简单。所有复杂性都要被质疑一遍是否必要，剩下的都是多余的。
中国的 AI 研究者占全球一半
聊到中国，黄仁勋先摆事实：全球大约一半的 AI 研究者是中国人，而且大部分还在中国。
他给的解释里有一条很少被提到：中国不是一个单一经济体，很多省市互相竞争，市长之间在比。这就是为什么会有那么多电动车公司、那么多 AI 公司，每个品类都能冒出好几家，内部竞争激烈到疯狂，最后活下来的是极强的公司。
第二条是文化。家庭第一，朋友第二，公司第三。工程师的兄弟在对手公司，朋友在对手公司，同学遍布全行业，同学这个概念在中国是一辈子的关系。知识流动得极快，藏着技术没有意义，不如放进开源。
“他们为什么对开源的贡献那么大，很好理解，他们在保护什么呢？工程师的兄弟在那家公司，朋友在那家公司，大家都是同学。分享知识非常快，所以开源社区反过来又放大和加速了创新。极强的人才，开源带来的快速创新，加上疯狂的竞争，最后涌现出惊人的东西。这是今天世界上创新最快的国家。”（黄仁勋）
他还说了一句对比：美国的领导者大多是律师，因为他们要维持秩序和法治；而中国是从贫困里建起来的，领导者大多是工程师，是这个国家最聪明的一批头脑。他用了一个词，建设者国家。
开源 Nemotron：三个理由
NVIDIA 开源了 Nemotron 3 Super，1200 亿参数的开放权重 MoE 模型，权重、数据、训练方法全部开放。
黄仁勋给了三个理由。
第一，要做一家伟大的 AI 计算公司，必须理解模型在怎么演进。Nemotron 3 是 transformer 加 SSM 的混合结构，他们在模型架构上做基础研究，才看得清未来什么样的计算系统能跑得动。这是极端协同设计的一部分。
第二，世界级的模型作为商品可以专有，但 AI 要扩散到每个行业、每个国家、每个研究者、每个学生，全靠专有是做不到的。开源是很多行业参与 AI 革命的前提。而 NVIDIA 有规模、有动机、也有能力一直造下去。
第三，AI 不止是语言。它要用别的模态训练出来的工具和子模型，可能是生物学、化学、物理定律、流体和热力学，这些东西不全在语言结构里。他们不造车，但要让每家车厂都有好模型可用；他们不做药物发现，但要让礼来有全世界最好的生物学 AI。
台积电最难复制的是信任
黄仁勋说，关于台积电最深的误解，是以为他们只有技术。
“他们的技术当然使他们与众不同，晶体管、金属化系统、封装、3D 封装、硅光子。但真正难的是他们能编排全世界几百家公司的动态需求，客户在往上走、在退出、在加量、在减量、在拉升、在收缩，wafer start、wafer stop、紧急插单。而他们同时做到高吞吐、高良率、好成本、优秀的客户服务。他们说好的交期，wafer 就会出现。这套制造系统我只能称之为奇迹。”（黄仁勋）
第二层是文化。同时做到技术最前沿和客户服务最好，这两件事很少出现在同一家公司身上。偏技术的不擅长服务，擅长服务的技术不够前沿。
第三层是信任，他用了一个数字：三十年的合作，几百亿上千亿美元的业务，没有合同。
聊到莫里斯·张 2013 年邀请他去当台积电 CEO 的往事，他确认是真的。他说自己非常荣幸，也深知台积电是历史上最有分量的公司之一，但他心里已经看到了 NVIDIA 会成为什么，那是他一个人的责任，所以只能拒绝。
护城河：开发者的账本
Lex 请他讲 NVIDIA 最大的护城河。
黄仁勋的答案没有任何犹豫：CUDA 的装机量。他补了一句，如果今天有人做出一个 GUDA 或者 TUDA，一点用都没有，因为从来就不只是技术问题。
“CUDA 成功靠的不是三个人，是四万三千人。还有几百万开发者相信我们会继续做 CUDA 1、2、3、13，他们决定把自己的软件、把自己堆积如山的代码搬上来。”（黄仁勋）
他站在开发者角度把这笔账算了一遍：支持 CUDA，半年后它就快十倍；支持 CUDA，你能触达几亿台机器，每一朵云、每一家电脑公司、每一个行业、每一个国家；而且你百分之百相信 NVIDIA 会一直维护它、优化它的库。三件事同时成立，所以如果我今天是开发者，我会先做 CUDA，投入最多。
第二条护城河是生态。垂直整合做极度复杂的系统，水平铺开进每一家公司的计算机：谷歌云、亚马逊、Azure、CoreWeave、Nscale、礼来的超算、企业机房、基站边缘、汽车、机器人、卫星，一个架构覆盖所有行业。
从仓库到工厂，从芯片到行星
计算的单位在变。黄仁勋说，过去是 GPU，然后是计算机，然后是集群，现在是一整个 AI 工厂。
他描述自己的心理模型变了。以前发布 Ampere，他会把芯片举起来。现在举起一块芯片仍然可爱，但那不是他在造的东西。他脑子里想的是几十上百兆瓦的庞然大物，接在电网上，配着 monstrous 的散热和网络系统，上万人在里面安装，几千名网络工程师，背后还有几千人做上电调试。上电从来不是谁按一下开关，是几千人一起把它点亮。
他说希望下一次思维跃迁，是行星级。
这个转变背后有一个更大的判断：计算机从仓库变成了工厂。
“过去的计算机本质上是文件检索系统，我们预先写好、预先录好、画好，放到网上，放进文件，再用推荐系统做过滤。现在 AI 计算机是上下文感知的，它必须实时处理和生成 token。我们从一个检索式计算系统，走到了一个生成式计算系统。”（黄仁勋）
仓库不赚钱，工厂的产出直接对应公司的营收。这不只是计算方式变了，计算机在世界上扮演的角色也变了：它不再是一台计算机，它是一座工厂，用来生产能卖钱的东西。
而且这些商品开始分层了。黄仁勋说，token 正在像 iPhone 一样分层：有免费 token，有 premium token，中间还有好几档。
“智能原来是一种可扩展的商品。有极高智能的 token，用在专业场景，人们愿意付钱。有人愿意为一百万 token 付一千美元这件事，就在拐角处，不是会不会，是什么时候。”（黄仁勋）
从这座工厂往下推，他给出的判断很硬：世界 GDP 会加速增长，其中用于计算的占比会比过去高一百倍，因为计算单元不再是存储单元，而是生产单元。
至于 NVIDIA 能不能成为一家三万亿美元营收的公司，他说当然能，因为看不到任何物理极限挡在那里。他顺便讲了个老故事：当年公司第一次突破十亿美元，有位 CEO 告诉他，无晶圆厂半导体公司理论上不可能超过十亿。后来又有人说你永远超不过 250 亿。“这些都不是第一性原理的思考。”
他也承认这里有个沟通难题。NVIDIA 不做市场份额生意，他讲的那些市场几乎都还不存在。如果一家百亿美元公司说要抢 NVIDIA 的份额，股东很容易算清楚；但要让人们想象一个没有对手可抢的市场有多大，很难。
“所以这个世界的一个挑战是想象未来。但我时间很多，我会一直推理下去，一直讲下去，每一次 GTC 都会变得更真实一点。”（黄仁勋）
token 的 iPhone 时刻到了
黄仁勋说，token 的 iPhone 时刻已经来了。
指的就是 agent，尤其是 OpenClaw。他给出的证据是增长速度：这是历史上增长最快的应用，直接垂直起飞。
Lex 承认自己在机场第一次对着笔记本电脑说话写代码，还觉得有点不好意思，假装在跟人类同事讲话。黄仁勋的回答把场景推得更远：更可能的情况是 AI 一直在烦你。
“它会不停地跟你汇报，我把这个做完了，接下来要我做什么。大多数人没意识到的是，将来跟你聊天、发消息最多的那个人，是你的 claw。”（黄仁勋）
压力大就拆问题，然后忘掉
Lex 把话题转向压力。NVIDIA 现在是被各国政府拿来规划 AI 基础设施、被无数普通人拿退休金押注的公司，CEO 怎么扛。
黄仁勋先讲了他清楚知道的事实：NVIDIA 的成功对美国很重要，巨额税收、技术领导力、国家安全的所有面向、再工业化带来的就业。他也知道自己身后是一整张生态网络，还有那些因为听信电视股评买了股票、如今成了百万富翁的教师和警察。
他的处理方式跟他处理工程问题完全一样：拆。
“我问自己，现在是什么情况，什么变了，难在哪里，我打算怎么办。把问题分解成可以动手的碎片。然后只剩下一个问题：你做了没有？你自己做了，还是让别人做了？如果你推理出该做但没做，也没让别人做，那就别再抱怨了。”（黄仁勋）
他承认自己对自己挺狠。但拆解使他保持冷静，能睡着。所有可能伤到公司、伤到伙伴、伤到产业的事，他都告诉了某个人，而且是能动手的那个人。说完之后，还能做什么呢。
至于心理上的低谷，他说有的是，经常有。
其中一部分靠遗忘。他提到 AI 学习里有个重要机制叫系统性遗忘，你不能什么都记住，什么包袱都背着。
“我会很快地拆解问题、推理问题，然后把它分出去。我说‘我告诉所有人’，本质上是在分摊这个负担。另一部分是忘记。很多时候你得对自己狠一点，行了，别哭了，走吧，然后起床。再往后就是被下一个亮点、下一个未来、下一个机会吸引过去。”（黄仁勋）
他说这跟顶级运动员一样，只关心下一分，上一分已经过去了，那些难堪和挫折也是。
Lex 提到那句著名的话：如果早知道建 NVIDIA 有这么难，比预想的难一百万倍，他根本不会开始。
黄仁勋说，这恰恰是他想解释的东西：拥有孩子般的心智是一种超能力。他看任何事情，第一反应都是“这能有多难”。
“你不能事先把所有的挫折、磨难、失望都模拟一遍，你不想知道那些。你要带着‘这会很棒、会很有趣’的心态进入一段新经历，然后在里面靠耐力和韧性撑住。因为挫折一定会来，而且会以让你意外的方式到来。”（黄仁勋）
他还补了一个判断依据：只要当初推导未来的那些前提没有实质改变，那么输出的结果也不该改变，那个模拟出来的未来仍然会发生，那就继续追。
他总结了几样东西的组合：新鲜心态进入、忘掉挫折、相信自己相信的东西、同时持续重新评估。他说自己很幸运，这几样都有一点。
公开做事的人，错的时候所有人都看得见
Lex 问了一个挺尖锐的问题：成为地球上最富有、最成功的人之一，钱、权力、名声会不会使你更难承认自己错了，更难听进反对意见。
黄仁勋说，恰恰相反。因为他大量工作是在公开场合做的，错了几乎所有人都看得见。
“我在内部会议上推理的东西，很多最后会不一样。但这从来不阻止我推理。我管理的方式，就是不断在人前推理。我跟你说一件事的时候，你能看到我是怎么走到这里的，然后你自己决定要不要相信最后那个结论。”（黄仁勋）
这套方法的好处在于，别人不必反驳你的结论，他们可以反驳你的推理步骤。反驳落在半路上，大家就能一起往前走。他把这叫集体的路径搜索。
Lex 接过这个词，说了一句更狠的：容忍尴尬是一种真实的能力。多年公开地使自己出丑，当众说过一个想法、后来被证明是错的，还能承认、还能从中长出来，这在人的层面上极其困难。
黄仁勋的回应很轻。他说，他们知道我第一份工作是刷厕所。
DLSS 5：我理解他们的担心，但那不是我们在做的事
聊到游戏，Lex 提了 DLSS 5 的争议：玩家担心它把游戏画面变成 AI 味的东西。
黄仁勋先站到了对面。
“他们的视角是有道理的，我能理解他们从哪来。我自己也不喜欢 AI slop。AI 生成的内容越来越像，都很漂亮，我对这种担心是有共情的。”（黄仁勋）
然后他解释 DLSS 5 到底在做什么：它是 3D 条件化的，受 3D 引导，受真实结构数据引导。几何是美术定的，每一帧都完全忠于这个几何；它由贴图和美术的工艺作为条件，每一帧只做增强，不改动任何东西。
而且系统是开放的，你可以在未来直接提示它：我要卡通渲染，我要这种风格。它会在忠于美术风格和
而且系统是开放的，你可以在未来直接提示它：我要卡通渲染，我要这种风格。它会在忠于美术风格和意图的前提下生成。
他点出了误解的来源：很多人以为游戏照原样做完发出来，然后我们在后期处理一遍。DLSS 不是这么设计的，它是跟美术一起工作的，是给美术的一件工具，美术也可以选择不用。
Lex 补了一段更宽的观察：人类对人脸极度敏感，我们正处在一个对 AI slop 敏感的时期，这其实是一面镜子，使人意识到自己真正想要的是不完美，是那些不完美的质感使世界变得动人。只要工具是用来帮人创造世界的，就很好。
黄仁勋同意，还顺手加了一句：这几年他们给开发者做了皮肤着色器，加入次表面散射，使皮肤看起来更像皮肤。游戏开发者一直在找更多表达工具，这只是又一件，用不用由他们决定。
《Doom》和《VR 战士》
Lex 问哪款游戏影响最大。黄仁勋的答案毫不犹豫：《Doom》。
理由是文化意义和产业意义的交叉点：它把 PC 从办公自动化工具变成了家庭和玩家的个人电脑。飞行模拟类游戏出现得更早，但没有《Doom》那样的普及度来完成这次转变。
如果只从游戏技术本身看，他选《VR 战士》。
Lex 提到《赛博朋克 2077》的全光线追踪，也提到自己反复重玩《上古卷轴 5：天际》，靠社区 mod 一遍遍重新体验一个已经爱上的世界。黄仁勋说他们做了 RTX Mod，使社区能把最新技术注入老游戏。
顺便还抖了个底：GeForce 到今天仍是 NVIDIA 最重要的营销策略。人在十几岁时通过《使命召唤》《堡垒之夜》认识 NVIDIA，上大学后开始用 CUDA，再后来用 Blender、达索、Autodesk。
“我们已经实现 AGI 了”
Lex 给出的 AGI 定义很具体：一个能创办、成长并运营一家十亿美元以上科技公司的 AI 系统。
黄仁勋说，现在已经实现了。
他的理由是抠字眼：你说十亿，你没说永远。一个 Claude 做出某个网页服务，几亿人用了两个月，每人付五毛，然后公司倒闭，这算不算。互联网时代满地都是这种公司，那些网站的技术复杂度，还不如今天 OpenClaw 能生成的东西。
但他紧接着把话说回来：十万个这样的 agent 里，跑出一个 NVIDIA 的概率是零。
然后他把话题拐到了他最想说的部分，工作焦虑。
“你的工作的目的，和你用来完成工作的任务和工具，是相关的，但不是一回事。我这份工作做了 33 年，是全世界在任最久的科技公司 CEO，34 年。这 34 年里我用的工具一直在变，有时候在两三年里剧变。但目的没变。”（黄仁勋）
他讲了那个被他反复讲的案例：放射科医生。
当年 AI 研究者预言，计算机视觉会达到超人水平，所以放射科医生这个职业会消失。2019 到 2020 年，计算机视觉真的达到了超人水平。今天每一个放射学平台和软件包都是 AI 驱动的。
但放射科医生的数量增长了，而且全世界现在缺放射科医生。
“那个危言耸听的警告走得太远了，它吓退了一批本可以进入这个对社会极其重要的职业的人。所以它造成了伤害。为什么错？因为放射科医生的目的是诊断疾病、帮助病人和医生诊断疾病。当你能更快地读片，你就可以读更多片、诊断得更好、病人更快入院、能见更多人，医院赚更多钱，住院病人更多，你就需要更多放射科医生。”（黄仁勋）
同样的逻辑放在 NVIDIA：软件工程师的人数会增长，不会减少。因为目的是解决问题，不是写多少行代码。
编程的定义变了，从三千万人变成十亿人
Lex 顺着问：那全世界的程序员数量会不会增加。
黄仁勋说会，理由是他重新定义了编程。今天的编程就是写规格说明，如果你想更强势一点，你甚至可以指定软件架构。问题是，有多少人能做这件事。
“我觉得我们刚从三千万人涨到了十亿人。未来每一个木匠都是一个程序员，只不过一个有 AI 的木匠同时还是建筑师。他们能交付给客户的价值被提升了，他们的工艺水平被大幅抬高了。每一个会计同时也成了你的财务分析师、你的财务顾问。如果我是木匠，看到 AI 我会彻底疯掉，我能给客户提供的服务太强了。如果我是水管工，也一样。”（黄仁勋）
他自己在工作中也是这样。给公司定战略、定方向的时候，他会把话说得足够具体，使人能行动，但故意欠规格，留出让 43000 个优秀的人把它做得比他想象的更好。
所以规格的工艺水平取决于你在解决什么问题、你在跟谁合作。你可以非常规定性，因为你要一个确定的结果；也可以故意欠规格，使 AI 反过来推你自己的创造力边界。这个谱系上的位置感，就是编程的未来。
他给大学生的建议很短：如果你今天招一个应届生，一个完全不懂 AI，一个是 AI 专家，我要 AI 专家。会计、市场、供应链、客服、销售、商务、律师，都一样。每个大学生毕业时都该是 AI 专家。如果你是木匠、电工、农民、药剂师，去用 AI，看看它能怎么改造你现在的活，把你自己变成这个行业的革新者。
Lex 补了一个更日常的角度：AI 最好用的地方之一是它能帮你拆解焦虑。“我担心我的工作”，然后它会给你一份逐条的行动计划。而且它把初学者的门槛拆掉了。你没法走到 Excel 面前说“我不会用 Excel”，那你就完了；但你可以跟 AI 说“我不会用 AI”，它会说“来，我教你”。
智能是商品，人性不是
Lex 问，有没有什么东西是人类意识里根本不可计算的，再强的芯片也复制不了。
黄仁勋的回答分两半。他不确定芯片会不会紧张。他相信 AI 能识别、理解紧张和焦虑，但芯片本身不会感到这些。而恰恰是这些感受在人类表现里显现出来：同样的处境，不同的人做出完全不同的成绩，那个谱系来自情绪，不来自计算。
“我不觉得我们造的任何东西会表明，两台计算机在完全相同的情境下会有根本不同的表现。当然统计上会有差异，但那不是因为它们感受不同。”（黄仁勋）
然后他把话锋转向了“智能”这个词本身。
“重要的是拆解什么是智能。这个词我们天天用，但它不神秘。智能是一个系统，包含感知、理解、推理、规划的能力，这个循环就是智能的根本。智能不等于人性，这是两个词。我不把智能过度浪漫化。我其实认为智能是一种商品。”（黄仁勋）
他拿自己举例。他身边全是比他聪明的人，学历比他好，学校比他好，在各自领域里钻得比他深。他的 60 个直接下属，对他来说都是超人。而他就坐在中间调度这些人。
“所以你不得不问自己，一个洗碗工凭什么坐在一群超人中间？这说明什么？我的意思是，智能是功能性的东西，人性不是用功能定义的，它是一个大得多的词。我们的生活经验、我们对痛苦的耐受、我们的决心，那些是跟智能完全不同的词。”（黄仁勋）
他说如果只能给听众留一样东西，那就是这个：我们社会把所有东西都压进了“智能”这一个词，而生活不止一个词。他自嘲说，他这一生恰好证明，在智能曲线上低于身边所有人，并不妨碍他成为最成功的那个。
所以别让智能的民主化、商品化带来焦虑，这件事值得为此感到鼓舞。
我不相信继任计划
Lex 问得很直接：你想过自己的死亡吗，你怕吗。
黄仁勋说，他真的很不想死。他有很好的生活、很好的家庭、很重要的工作。
“这不是一生一次的体验，这是人类史上一次性的体验，我正在经历的这个东西。NVIDIA 是历史上最有分量的科技公司之一，我们在做非常重要的事，我非常认真地对待它。”（黄仁勋）
然后他提到了继任计划，并且明确说自己不信这套。
他不信的理由不是觉得自己不会死。他的推理是：如果你为继任计划焦虑，那就把这个焦虑拆到底，今天最该做的事是什么。答案是：如果你关心你离开之后的公司，那就尽可能频繁、尽可能持续地把知识、信息、洞察、技能、经验传下去。
“这就是为什么我一直在团队面前推理。每一场会议都是推理会议。我在公司内外的每一刻，都是在尽可能快地把知识传给人们。我学到的任何东西，在我桌上停留的时间不超过零点几秒。我还没完全学明白，就已经在指给别人看了：快上，这个太酷了，你会想学的。”（黄仁勋）
所以他希望的结果是死在岗位上，最好是瞬间的那种，没有漫长的痛苦。
把智能请下神坛
最后 Lex 问，什么使你对人类和未来抱有希望。
黄仁勋说他一直对人的善良、慷慨、同情心和能力有极大的信心，有时候信心大过头，会被人占便宜，但这不会改变他的起点：人想做好事，人想帮别人。而绝大多数时候他都被证明是对的，而且常常超出预期。
使他真正兴奋的是现在能看到什么是可能的。
“有那么多问题我们想解决，有那么多东西我们想造，有那么多好事现在落在了我们够得着的地方，落在我这一辈子的范围内。你不可能对此不感到浪漫。”（黄仁勋）
他列举得很具体：疾病终结是一个合理的预期，污染大幅下降是一个合理的预期，以光速旅行在短距离内进入我们的未来，也是一个合理的预期。他还开了个玩笑：很快我会把一个机器人送上飞船，使它一路飞行一路自我改进，等到时候，我这辈子说过的话、写过的邮件早就上传成了我的 AI，用光速发过去追上那个机器人就行。
Lex 提到生物机器的破解就在眼前，黄仁勋说不是十年，大概是五年。意识如何解释，那个答案会很棒。
整场对话走到这里，其实有一条很清楚的线。他把智能拆成功能性能力，然后说它是商品；他把人性、品格、同情、耐力、决心拎出来，说那些才是超能力。他对 AGI 的定义给得极松，对做成一家 NVIDIA 的难度说得极紧。他一边用第一性原理把 74 天压到 6 天，一边承认如果早知道有多难，他根本不会开始。
这些说法看起来互相打架，其实指向同一件事：真正稀缺的从来不是算力，也不是智能，是在看不到尽头的时候还愿意每天往前挪一点的那种人。
给读完这篇的人，几条可以直接动手的事：
- 今天就找一个你工作里最耗时的任务，把它的完整流程讲给 AI 听，请它给出一份操作方案。不要问“AI 能做什么”，要问“我这个具体的事怎么做”。
- 练一件事：把问题拆到能动手为止。写下你现在最焦虑的那件事，拆成“我能做的”和“我做不了的”两栏，然后把第一栏的每一项都指给一个能动手的人，包括你自己。
- 如果你在带团队，别发宣言。把你在想的事情提前几个月、按周铺进日常对话里，等真正宣布的时候，使大家的反应是“你怎么才说”。
- 给自己设一个“光速”参照。下次有人告诉你某件事要 74 天，先别谈折扣，问一句：如果从零开始造，物理上最快是多少天。
- 学一点编程，哪怕你不打算当程序员。写规格说明就是新的编程，你描述得越清楚，AI 交付的东西越接近你脑子里那个版本。
- 如果你是学生、老师、木匠、会计、医生，任何职业，把“熟练使用 AI”当成和识字一样的基础能力，别等它变成门槛才开始。
The following is a conversation with Jensen Huang, CEO of NVIDIA, one of the most important and influential companies in the history of human civilization.
以下是与人类文明史上最重要、最具影响力的公司之一英伟达首席执行官黄仁勋的对话。
00:00 NVIDIA is the engine powering the AI revolution, and a lot of its success can be directly attributed to Jensen’s sheer force of will and his many brilliant bets and decisions as a leader, engineer, and innovator.
NVIDIA 是推动人工智能革命的引擎，它的成功在很大程度上可以直接归功于 Jensen 的坚定意志力以及他作为领导者、工程师和创新者的许多出色的赌注和决策。
This is Lex Fridman Podcast.
这是莱克斯·弗里德曼播客。
And now dear friends, here’s Jensen Huang.
亲爱的朋友们，这是黄仁勋。
00:33 You’ve propelled NVIDIA into a new era in AI, moving beyond his focus on chip scale design to now rack scale design.
你们推动 NVIDIA 进入了 AI 的新时代，从专注于芯片级设计转向了现在的机架级设计。
00:42 And I think it’s fair to say that winning for NVIDIA for a long time used to be about building the best GPU possible, and you still do, but now you’ve expanded that to extreme co-design of GPU, CPU memory, networking, storage, power cooling, software, the rack itself, the pod that you’ve announced, and even the data center.
我认为可以公平地说，长期以来 NVIDIA 的胜利在于构建尽可能最好的 GPU，现在仍然如此，但现在您已将其扩展到 GPU、CPU 内存、网络、存储、电源冷却、软件、机架本身、您宣布的 Pod 甚至数据中心的极端协同设计。
So let’s talk about extreme co-design.
那么我们来谈谈极限协同设计。
What is the hardest part of co-designing a system with that many complex components and design variables?
协同设计具有如此多复杂组件和设计变量的系统最困难的部分是什么？
Yeah, thanks for that question.
是的，谢谢你提出这个问题。
01:11 So first of all, the reason why extreme co-design is necessary is because the problem no longer fits inside one computer to be accelerated by one GPU.
首先，极端协同设计之所以必要，是因为这个问题不再适合在一台计算机内由一个 GPU 加速。
The problem that you’re trying to solve is you would like to go faster than the number of computers that you add.
您试图解决的问题是您希望运行速度比您添加的计算机数量更快。
So you added 10,000 computers, but you would like it to go a million times faster.
因此，您添加了 10,000 台计算机，但您希望它的运行速度快一百万倍。
Then all of a sudden you have to take the algorithm, you have to break up the algorithm, you have to refactor it, you have to shard the pipeline, you have to shard the data, you have to shard the model.
然后突然间你必须采用算法，你必须分解算法，你必须重构它，你必须对管道进行分片，你必须对数据进行分片，你必须对模型进行分片。
Now all of a sudden when you distribute the problem this way, not just scaling up the problem, but you’re distributing the problem, then everything gets in the way.
现在突然间，当你以这种方式分配问题时，不仅仅是扩大问题，而是分配问题，那么一切都会变得困难。
02:03 This is the Amdahl’s law problem where the amount of speed up you have for something depends on how much of the total workload it is.
这是阿姆达尔定律问题，其中某件事的加速量取决于它占总工作量的多少。
And so if computation represents 50% of the problem, and I sped up computation infinitely like a million times, you know, I only sped up the total workload by a factor of two.
因此，如果计算代表了问题的 50%，并且我将计算速度无限加快了一百万倍，你知道，我只将总工作负载速度加快了两倍。
Now all of a sudden, not only do you have to distribute a computation, you have to shard the pipeline somehow.
现在突然之间，您不仅必须分配计算，还必须以某种方式对管道进行分片。
You also have to solve the networking problem because you’ve got all of these computers are all connected together.
您还必须解决网络问题，因为所有这些计算机都连接在一起。
And so distributed computing at the scale that we do, the CPU is a problem, the GPU is a problem, the networking is a problem, the switching is a problem.
因此，我们所做的大规模分布式计算，CPU 是一个问题，GPU 是一个问题，网络是一个问题，交换也是一个问题。
And distributing the workload across all these computers is a problem.
在所有这些计算机上分配工作负载是一个问题。
02:57 It’s just a massively complex computer science problem.
这只是一个非常复杂的计算机科学问题。
And so we just gotta bring every technology to bear.
因此，我们必须运用所有技术。
Otherwise, we scale up linearly or we scale up based on the capabilities of Moore’s Law, which has largely slowed because Dennard scaling has slowed.
否则，我们会线性扩大规模，或者根据摩尔定律的能力扩大规模，摩尔定律的速度在很大程度上已经放缓，因为登纳德缩放速度已经放缓。
I’m sure there’s trade-offs there.
我确信这其中存在权衡。
03:16 Plus you have a complete disparate disciplines here.
另外，这里有完全不同的学科。
I’m sure you have specialists in each one of these high bandwidth memory, the network and the NVLink, the NICs, the optics and the copper that you’re doing, the power delivery, the cooling, all of that.
我确信你们在这些高带宽内存、网络和 NVLink、NIC、光学和铜缆、电力传输、冷却等方面都有专家。
I mean, there’s like world experts in each of those.
我的意思是，每一个领域都有世界专家。
How do you get ’em in a room together to figure out-
你如何让他们一起在一个房间里弄清楚-
That’s why my staff is so large.
这就是为什么我的员工如此庞大。
03:34 Yeah.
是的。
What’s the pro- can you take me through the process of the specialists and the generalists?
有什么问题-你能带我了解一下专家和通才的过程吗？
03:37 Like how do you put together the rack when you know the s- the set of things you have to shove into a rack together?
就像当您知道必须将一组东西一起推入架子时，如何将架子组装在一起？
Like what does that process look like of designing it all together?
一起设计的过程是什么样的？
Yeah.
是的。
03:51 There’s the first question, which is: what is extreme co-design?
第一个问题是：什么是极限协同设计？
We’re optimizing across the entire stack of software from architectures to chips, to systems, to system software, to the algorithms, to the applications.
我们正在优化整个软件堆栈，从架构到芯片、系统、系统软件、算法和应用程序。
That’s one layer.
这是一层。
The second thing that you and I just talked about goes beyond CPUs and GPUs and networking chips and scale up switches and scale out switches.
你和我刚才谈到的第二件事超出了 CPU、GPU 和网络芯片以及纵向扩展交换机和横向扩展交换机的范围。
And then of course, you gotta include power and cooling and all of that because all these computers are extremely power hungry.
当然，你必须包括电源和冷却以及所有这些，因为所有这些计算机都非常耗电。
They do a lot of work and they’re very energy efficient, but they in aggregate still consume a lot of power.
它们做大量的工作并且非常节能，但它们总的来说仍然消耗大量的电力。
And so that’s one.
这就是其中之一。
The first question is, what is it?
第一个问题是，它是什么？
04:34 The second question is, why is it, and we just spoke about the reason, you know you want to distribute the workload so that you can exceed the benefit of just increasing the number of computers.
第二个问题是，为什么会这样，我们刚才谈到了原因，你知道你想要分配工作负载，这样你就可以超越仅仅增加计算机数量所带来的好处。
And the, and then the third question is, how is it, how do you do it?
然后第三个问题是，怎么样，你是怎么做到的？
04:51 And, and that’s the, that’s kind of the miracle of this company.
这就是这家公司的奇迹。
You know, when you’re designing a computer, you have to have an operating system of computers.
你知道，当你设计一台计算机时，你必须有一个计算机操作系统。
When you’re designing a company, you should first think about what is it that you want the company to produce.
当你设计一家公司时，你应该首先考虑你希望公司生产什么。
You know, I see a lot of companies’ organization charts, and they all look the same.
你知道，我看过很多公司的组织结构图，它们看起来都一样。
Hamburger organization charts, soft organization charts, and car company organization charts.
汉堡组织结构图、软组织结构图、汽车公司组织结构图。
They all look the same.
他们看起来都一样。
And it doesn’t make any sense to me.
这对我来说没有任何意义。
You know, the goal of a comp- of a company is to be the machinery, the mechanism, the system that produces the output.
你知道，公司的目标是成为产生产出的机器、机制和系统。
And that output is the product that we like to create.
这个输出就是我们想要创造的产品。
It is also designed, the architecture of the company should reflect the environment by which it exists.
它也是设计的，公司的架构应该反映它存在的环境。
05:36 It almost directly says what you should do with the organization.
它几乎直接说明了你应该对组织做什么。
My direct staff is 60 people.
我的直接员工有 60 人。
You know, I don’t have one-on-ones with ’em because it’s impossible.
你知道，我不会与他们进行一对一的交流，因为这是不可能的。
You can’t have 60 people on your staff if you’re, you know, gonna get work done and-
如果你要完成工作，你就不可能拥有 60 名员工，而且——
So you still have 60 reports.
所以您还有 60 份报告。
05:51 You still have across-
你还有跨-
More, yeah.
更多，是的。
More.
更多的。
05:54 And most stars at least have a foot in engineering.
大多数明星至少都涉足工程学领域。
Almost all of them.
几乎全部。
05:59 There’s experts in memory, there’s experts in CPUs, there’s experts in optical.
有内存专家、CPU 专家、光学专家。
All-
所有-
That’s incredible.
这太不可思议了。
Yeah, GPUs and- Architecture, algorithms, design-
是的，GPU 和——架构、算法、设计——
So, you constantly have an eye on the entire stack, and you’re having to have, like, intense discussions about the design of the entire stack?
那么，您需要不断关注整个堆栈，并且必须对整个堆栈的设计进行激烈的讨论？
And no conversation is ever one person.
而且没有任何对话是一个人的。
06:18 That’s why I don’t do one-on-ones.
这就是为什么我不做一对一的。
We present a problem and all of us attack it.
我们提出一个问题，然后我们所有人都去攻击它。
You know, because we’re doing extreme co-design.
你知道，因为我们正在进行极端的协同设计。
And literally, the company is doing extreme co-design all the time.
从字面上看，该公司一直在进行极端的协同设计。
So, even if you’re talking about a particular component, like cooling, networking, everybody’s listening in?
那么，即使您正在谈论某个特定组件，例如冷却、网络，每个人都会倾听吗？
Yeah, exactly.
是的，完全正确。
And they can contribute, “Well, this doesn’t work for the power distribution.
他们可以贡献，“好吧，这对配电不起作用。
06:41 This doesn’t-“
这不——”
Exactly.
正是如此。
“…
“……
06:45 This doesn’t work for the memory.
这对记忆不起作用。
This doesn’t work for this.”
这对此不起作用。”
Exactly.
确切地。
06:49 And whoever wants to tune out, tune out.
谁想退出，就退出。
You know what I’m saying?
你知道我在说什么吗？
And the reason for that is because the people who are on the staff, they know when to pay attention.
原因是因为工作人员知道何时该注意。
There’s supposed…
据说有……
You know, it’s something they could have contributed to, they didn’t contribute to, “I’m going to call them out.” You know?
你知道，这是他们本可以做出贡献的事情，但他们却没有做出贡献，“我要叫他们出来。”你知道？
And so, “Hey, come on, let’s get in here.”
所以，“嘿，来吧，我们进去吧。”
So, as you mentioned, NVIDIA is this company that’s adapting to the environment.
所以，正如您提到的，NVIDIA 是一家正在适应环境的公司。
07:07 So, which point can you say, did the environment change and began adapting sort of secretly- …
那么，你能说哪一点是环境发生了变化并开始秘密适应——……
in the early days from GPU for gaming, maybe the early deep learning revolution to we’re now going to start thinking of it as an AI factory?
从早期用于游戏的 GPU，也许早期的深度学习革命到我们现在要开始将其视为人工智能工厂？
What does NVIDIA do?
英伟达是做什么的？
It produces AI; let’s build a factory that makes AI.
它产生人工智能；让我们建立一个制造人工智能的工厂。
I could reason through that systematically.
我可以系统地推理出来。
07:32 We started out as an accelerator company.
我们最初是一家加速器公司。
But the problem with accelerators is that the application domain’s too narrow.
但加速器的问题在于应用领域太窄。
It has the benefit of being incredibly optimized for the job.
它的优点是针对工作进行了令人难以置信的优化。
You know, any specialist has that benefit.
你知道，任何专家都有这种好处。
The problem with intense specialization is that, of course, your market reach is narrower, but that’s even fine.
当然，高度专业化的问题在于，你的市场范围会更窄，但那也没关系。
The problem is, the market size also dictates your R&D capacity.
问题是，市场规模也决定了你的研发能力。
And your R&D capacity ultimately dictates the influence and impact that you can possibly have in computing.
And your R&D capacity ultimately dictates the influence and impact that you can possibly have in computing.
And so, when we first started out as an accelerator, very specific accelerator, we always knew that was going to be our first step.
因此，当我们第一次开始作为加速器，非常具体的加速器时，我们一直知道这将是我们的第一步。
08:23 We had to find a way to become accelerated computing.
我们必须找到一种方法来加速计算。
But the problem is, when you become a computing company, it’s too general purpose and it takes away from your specialization.
但问题是，当你成为一家计算公司时，它的目的太通用了，它会削弱你的专业化。
The tur- I connected two words that actually have fundamental tension.
tur- I 连接了两个实际上具有基本张力的词。
The better computing company we become, the worse we became as a specialist.
我们的计算公司做得越好，我们的专家就越差。
The more of a specialist, the less capacity we have to do overall computing.
专家越多，我们进行整体计算的能力就越少。
And so, that…
所以，那……
And I connected those two words together on purpose, that the company has to find that really narrow path, step by step by step, to expand our aperture of computing, but not give up on the most important specialization that we had.
我故意将这两个词连接在一起，即公司必须一步一步地找到那条真正狭窄的道路，以扩大我们的计算范围，但不能放弃我们拥有的最重要的专业化。
Okay, so the first step that we took beyond acceleration was we invented a programmable pixel shader.
好的，除了加速之外，我们采取的第一步是发明了可编程像素着色器。
09:13 So, that was the first step towards programmability.
所以，这是迈向可编程性的第一步。
It was our first journey towards moving into the world of computing.
这是我们进入计算世界的第一次旅程。
The second thing that we did was we created, we put FP32 into our shaders.
我们做的第二件事是创建，将 FP32 放入着色器中。
That FP32 step, IEEE-compatible FP32, was a huge step in the direction of computing.
FP32 的进步，即 IEEE 兼容的 FP32，是计算方向的一大进步。
It was the reason why all of the people who were working on stream processors and, you know, other types of data flow processors discovered us.
这就是所有从事流处理器以及其他类型数据流处理器工作的人发现我们的原因。
And they said, “Hey, all of a sudden, you know, we might be able to use this GPU that’s incredibly computationally intensive, and it’s now, you know, compliant with IEEE.”
他们说，“嘿，突然之间，我们也许能够使用这种计算量极其巨大的 GPU，而且它现在符合 IEEE 的要求。”
09:55 I can take my software that I was writing, you know, previously on CPUs, and I can see about using the GPU for that.
I can take my software that I was writing, you know, previously on CPUs, and I can see about using the GPU for that.
And which led us to create, put C on top of FP32, what’s called, we call Cg.
这导致我们创建，将 C 放在 FP32 之上，我们称之为 Cg。
The Cg path took us to eventually CUDA.
Cg 路径最终将我们带到了 CUDA。
CUDA, step by step by step we…
CUDA，一步一步我们……
Well, the putting CUDA on GeForce, that was a strategic decision that was very, very hard to do, because it cost the company enormous amounts of our profits, and we couldn’t afford it at the time.
嗯，将 CUDA 放在 GeForce 上，这是一个非常非常难做的战略决策，因为这让公司损失了大量利润，而我们当时负担不起。
But we did it anyway because we wanted to be a computing company.
但我们还是这么做了，因为我们想成为一家计算公司。
A computing company has a computing architecture.
一家计算公司有一个计算架构。
A computing architecture has to be compatible across all of the chips that we build.
计算架构必须兼容我们制造的所有芯片。
Can you take me through that decision?
你能帮我完成这个决定吗？
10:42 So, putting CUDA on GeForce, could not afford to do?
那么，把CUDA放在GeForce上，没能力做什么？
Can you explain that decision?
你能解释一下这个决定吗？
Why boldly choose to do that anyway?
为什么还要大胆地选择这样做呢？
Can you explain that decision?
你能解释一下这个决定吗？
Yeah, excellent.
是的，非常好。
10:53 That was…
那是……
I would say that that was the first strategic decision that is as close to an existential threat.
我想说，这是第一个接近生存威胁的战略决策。
For people who don’t know, it turned out to be, spoiler alert, one of the most incredibly brilliant decisions ever made by a company.
对于那些不知道的人来说，事实证明，剧透警告，这是公司有史以来做出的最令人难以置信的明智决定之一。
11:06 So, CUDA turned out to be an incredible foundation for computation in this AI infrastructure world.
因此，CUDA 被证明是人工智能基础设施世界中令人难以置信的计算基础。
So, so- …
所以，所以——
just setting the context.
只是设置上下文。
It turned out to be a good decision.
事实证明这是一个很好的决定。
Yeah, it turned out to have been a good decision.
是的，事实证明这是一个很好的决定。
11:27 I think the…
我认为……
So, here’s the way it went.
所以，事情是这样的。
So, we invented this thing called CUDA, and it expanded the aperture of applications that we can accelerate with our accelerator.
因此，我们发明了这种称为 CUDA 的东西，它扩大了我们可以使用加速器加速的应用程序的范围。
The question is, how do we attract developers to CUDA?
问题是，我们如何吸引开发者使用 CUDA？
Because a computing platform is all about developers.
因为计算平台的核心就是开发人员。
And developers don’t come to a computing platform just because, you know, it could perform something interesting.
开发人员来到计算平台并不仅仅是因为它可以执行一些有趣的事情。
They come to a computing platform because the install base is large.
他们之所以选择计算平台，是因为安装基数很大。
Because a developer, like anybody else, wants to develop software that reaches a lot of people.
因为开发人员和其他人一样，希望开发出能够惠及很多人的软件。
So, the install base is, in fact, the single most important part of an architecture.
因此，安装基础实际上是架构中最重要的部分。
The architecture could attract enormous amounts of criticism.
该架构可能会招致大量批评。
12:18 For example, no architecture has ever attracted more criticism than the x86…
例如，没有任何架构比 x86 更招致批评……
you know, as a less than elegant architecture, but yet it is the defining architecture of today.
你知道，作为一个不太优雅的建筑，但它是当今的定义建筑。
It gives you an example that in fact so many RISC architectures which were beautifully architected, incredibly well-designed by some of the brightest computer scientists in the world, largely failed.
它给你举了一个例子，事实上，许多由世界上最聪明的计算机科学家精心设计、精心设计的 RISC 架构在很大程度上都失败了。
And so I’ve given you two examples where one is, you know, one is elegant, the other one’s barely aesthetic, and so yet x86 survived and the reason for-
所以我给了你们两个例子，一个是，你知道，一个是优雅的，另一个是几乎不美观的，但 x86 仍然幸存下来，原因是——
Install base is everything.
安装基础就是一切。
Install base defines an architecture.
安装基础定义了一个架构。
12:59 Not…
不……
Everything else is secondary, okay?
其他的都是次要的好吗？
And so there were other architectures at the time.
所以当时还有其他架构。
CUDA came out, OpenCL was here.
CUDA出来了，OpenCL来了。
There were…
有……
You know, there’s several other competing architectures.
您知道，还有其他几种相互竞争的架构。
But the thing that…
但那件事……
The decision that we made that was good was we said, “Hey, look, ultimately it’s about install base and what is the best way we could get a new computing architecture into the world?” By that timeframe, GeForce had become successful.
我们做出的很好的决定是，“嘿，看，最终是关于安装基础的，以及我们将新的计算架构推向世界的最佳方式是什么？”到那时，GeForce 已经取得了成功。
13:29 We were already selling millions and millions of GeForce GPUs a year, and we said, “You know, we, we ought to put CUDA on GeForce and put it into every single PC whether customers use it or not, and use it as a starting point of cultivating our install base.” Meanwhile, we’ll go and attract developers, and we went to universities and wrote books and taught classes and put CUDA everywhere.
我们每年已经售出数以百万计的 GeForce GPU，我们说，“你知道，我们应该将 CUDA 放在 GeForce 上，并将其放入每台 PC 中，无论客户是否使用它，并将其作为培养我们安装基础的起点。”与此同时，我们将去吸引开发人员，我们会去大学写书、教授课程，并将 CUDA 推广到各个地方。
And eventually people discover…
最终人们发现……
And at the time, the PC was the primary computing vehicle.
当时，个人电脑是主要的计算工具。
There was no cloud, and we could put a supercomputer in the hands of every researcher in school, every scientist, you know, every engineering school, every…
没有云，我们可以把一台超级计算机放在学校里的每个研究人员、每个科学家、每个工程学校、每个……手中。
or every student in school, and eventually something amazing will happen.
或者学校里的每个学生，最终会发生一些令人惊奇的事情。
14:15 Well, the problem was CUDA increased our cost of that GPU, which is a consumer product, so tremendously, it completely consumed all of the company’s gross profit dollars.
嗯，问题是 CUDA 增加了我们 GPU（消费产品）的成本，如此之大，以至于完全消耗了公司的所有毛利润。
And so at the time, the company was probably, you know, worth, I don’t know, at the time, eight…
所以当时，你知道，公司的价值，我不知道，当时可能是八……
Was it like $8 billion or something?
是不是80亿美元之类的？
Like six, $7 billion or something like that.
比如六、七十亿美元之类的。
After we launched CUDA, I recognized that it was going to add so much cost, but it was something we believed in.
After we launched CUDA, I recognized that it was going to add so much cost, but it was something we believed in.
You know, our market cap went down to like one and a half billion dollars.
You know, our market cap went down to like one and a half billion dollars.
And so we were down there for a while and we clawed our way back slowly, but we carried CUDA on GeForce.
所以我们在那里呆了一段时间，然后慢慢地回来，但我们在 GeForce 上搭载了 CUDA。
I always say that NVIDIA is the house that GeForce built, because it was GeForce that took CUDA out to everybody.
我总是说 NVIDIA 是 GeForce 建造的房子，因为正是 GeForce 将 CUDA 带给了每个人。
15:10 Researchers, scientists, they discovered CUDA on GeForce because they were all, you know…
研究人员、科学家，他们在 GeForce 上发现了 CUDA，因为他们都是，你知道……
Many of ’em were gamers.
他们中的许多人都是游戏玩家。
Many of them built their own PCs anyways.
不管怎样，他们中的许多人都建造了自己的电脑。
In a university lab, many of them built clusters themselves, you know, using PC components.
在大学实验室中，他们中的许多人使用 PC 组件自己构建集群。
And, and so that, you know, that’s kind of how we got going.
所以，你知道，我们就是这样开始的。
And then that became the platform and the foundation for the deep learning revolution.
然后这成为深度学习革命的平台和基础。
That was also another great, great observation.
这也是另一个非常非常伟大的观察。
15:35 Yeah.
是的。
That existential moment, do you remember…
那个存在的时刻，你还记得吗……
15:38 Like, what were those meetings like?
比如，那些会议是什么样的？
What were those discussions like, deciding as a company, risking everything?
作为一家公司，冒着一切风险做出决定，这些讨论是什么样的？
Well I had to make it clear to the board what we’re trying to do, and the management team knew our gross margins were gonna get crushed.
Well I had to make it clear to the board what we’re trying to do, and the management team knew our gross margins were gonna get crushed.
15:48 So you could imagine a world where GeForce would carry the burden of CUDA and none of the gamers would appreciate it and none of the gamers would pay for it.
因此，您可以想象这样一个世界：GeForce 将承担 CUDA 的负担，但没有一个游戏玩家会欣赏它，也没有一个游戏玩家会为此付费。
You know, they only pay certain price and it doesn’t matter what your cost is.
You know, they only pay certain price and it doesn’t matter what your cost is.
And so the…
所以……
You know, we increased our cost by 50% and that consumed…
You know, we increased our cost by 50% and that consumed…
And we were a 35% gross margin company, and so it was a…
And we were a 35% gross margin company, and so it was a…
It was quite a difficult decision to make.
这是一个相当困难的决定。
But you could imagine that someday this would go into workstations and it would go into supercomputers and in those segments, maybe we can capture more margin.
但你可以想象，有一天这将进入工作站，它会进入超级计算机，在这些领域，也许我们可以获得更多的利润。
16:36 So you could reason your way into being able to afford this, but it still took…
So you could reason your way into being able to afford this, but it still took…
It took a decade.
花了十年的时间。
But that, but that’s more of, like, conversation with the board convincing them, but you psychologically- …
但这，但这更像是与董事会的对话说服他们，但你在心理上——……
16:45 as NVIDIA’s continued to make bold bets that predict the future, and in part, especially now, define the future.
as NVIDIA’s continued to make bold bets that predict the future, and in part, especially now, define the future.
So I’m almost looking for wisdom about how you’re able to make those decisions, to make leaps- …
So I’m almost looking for wisdom about how you’re able to make those decisions, to make leaps- …
like that as a company.
作为一家公司就是这样。
Well, first of all, I’m informed by a lot of curiosity.
Well, first of all, I’m informed by a lot of curiosity.
17:14 At some point, there’s a reasoning system that convinces me so clearly this outcome will happen.
At some point, there’s a reasoning system that convinces me so clearly this outcome will happen.
That this will happen.
这将会发生。
And so I believe it in my mind, and when I believe it in my mind, you know how it is.
所以我心里相信它，当我心里相信它时，你就知道它是怎么回事。
You manifest a future and that future is so convincing, there’s no way it won’t happen.
You manifest a future and that future is so convincing, there’s no way it won’t happen.
There’s a lot of suffering in between, but you’ve gotta believe what you believe.
中间有很多痛苦，但你必须相信你所相信的。
So you, you, you envision the future- …
所以你，你，你憧憬未来——……
17:52 and you essentially, from a sort of engineering perspective, manifest it?
and you essentially, from a sort of engineering perspective, manifest it?
Yeah.
是的。
17:59 And you reason about how to get there.
然后你思考如何到达那里。
You reason about why it must exist.
你推理为什么它必须存在。
And you know, I reason…
你知道，我推理……
We all reason it here.
我们都在这里推理。
The management team would reason about it.
The management team would reason about it.
All the people that I…
所有我...的人
We spend a lot of time reasoning about it.
我们花了很多时间来思考它。
The thing that…
那件事……
The next part of it is probably a skill thing, which is, you know, oftentimes in leadership the leadership stays quiet or they learn about something, and then they do some manifesto, and it’s a brand-new year, and somehow at the end of the year, next year, we’re gonna have a brand-new plan.
下一部分可能是一个技能问题，你知道，在领导层中，领导层通常保持安静，或者他们了解一些东西，然后他们发表一些宣言，这是全新的一年，不知何故，在今年年底，明年，我们将有一个全新的计划。
Big huge layoff this way, big huge organization change this way, new mission statement…
巨大的裁员就这样，巨大的组织就这样改变，新的使命宣言……
brand new logos, you know, that kind of stuff.
全新的标志，你知道的，诸如此类的东西。
18:43 We’ve just never, I never do things that way.
我们从来没有，我从来没有这样做过。
When I learn about something and it’s starting to influence how I think, I’ll make it very clear to everybody near me that, you know, this is interesting.
当我了解到某件事并且它开始影响我的思维方式时，我会向我身边的每个人明确表示，你知道，这很有趣。
This is going to make a difference.
这将会带来改变。
This is going to impact that.
这将会影响到这一点。
And I reason about things step by step by step.
And I reason about things step by step by step.
Oftentimes, I’ve already made up my mind, but I’ll take every possible opportunity—external information, new insights, new discoveries, new engineering revelations, new milestones developed—I’ll take those opportunities and I’ll use it to shape everybody else’s belief system.
很多时候，我已经下定决心了，但我会抓住一切可能的机会——外部信息、新见解、新发现、新工程启示、新的里程碑——我会抓住这些机会，并用它来塑造其他人的信仰体系。
And I’m doing that literally every single day.
我几乎每天都在这样做。
I’m doing that with my board, I’m doing that with my management team, I’m doing that with my employees.
I’m doing that with my board, I’m doing that with my management team, I’m doing that with my employees.
19:33 I’m trying to shape their belief systems such that when I come the day I say, “Hey, let’s buy Mellanox,” it’s completely obvious to everybody that we absolutely should.
我正在努力塑造他们的信仰体系，以便当我来的那天我说，“嘿，让我们购买 Mellanox”，这对每个人来说都是完全显而易见的，我们绝对应该这样做。
On the day that I said, “Hey guys, let’s go all in on deep learning,” and let me tell you why.
那天我说：“嘿伙计们，让我们全力投入深度学习吧”，让我告诉你原因。
I’ve already been laying down the bricks to different organizations inside the company.
我已经为公司内部的不同组织奠定了基础。
Every organization and everybody, many of the people might have heard everything.
每个组织、每个人、很多人可能都听到了一切。
Most of the company hears, of course, pieces of it.
当然，公司的大多数人都听说过其中的一些内容。
And on the day that I announce it, everybody’s kind of bought in to many pieces of it.
在我宣布这一消息的那天，每个人都接受了其中的很多内容。
20:19 And in a lot of ways, I like to announce these things, and I imagine that the employees are kind of saying, “You know, Jensen, what took you so long?” And in fact, I’ve been shaping their belief system for some time, and therefore leadership.
在很多方面，我喜欢宣布这些事情，我想员工们会说，“你知道，詹森，你为什么花了这么长时间？”事实上，一段时间以来我一直在塑造他们的信仰体系，从而塑造他们的领导力。
Sometimes it looks like you’re leading from behind, but you’ve been shaping their, you know, to the point where on the day that I declared it, 100% buy-in.
有时看起来你是在后面领先，但你知道，你一直在塑造他们，直到我宣布这一点的那天，他们100%支持。
But that’s what you want.
但这就是你想要的。
You want to bring everybody along.
You want to bring everybody along.
Otherwise, we announce something about deep learning and everybody goes, “What are you talking about?” You know, you announce something about let’s go all in on this thing, and your management team, your board, your employees, your customers, they’re kind of like, “Where’s this coming from?”
否则，当我们宣布一些有关深度学习的内容时，每个人都会问：“你在说什么？”你知道，当你宣布让我们全力以赴做这件事时，你的管理团队、你的董事会、你的员工、你的客户，他们有点像，“这是从哪里来的？”
21:02 You know, this is insane.” And so, so GTC effect, if you go back in time, you look at, look at the keynotes, I’m also shaping the belief system of my partners in the industry and, and I’m using that to shape, you know, the belief system of my own employees.
你知道，这太疯狂了。”因此，GTC 效应，如果你回到过去，看看主题演讲，我也在塑造行业合作伙伴的信仰体系，并且我正在用它来塑造我自己员工的信仰体系。
And, and, and so by the time that I announce something, like for example, we just announ- we just announced Grok.
而且，并且，所以当我宣布某件事时，例如，我们刚刚宣布 - 我们刚刚宣布了 Grok。
We’ve been late…
我们迟到了……
I’ve been talking about the stepping stones for two and a half years.
我已经谈论踏脚石两年半了。
You just go back and go, “Oh my gosh, they’ve been talking about it for two and a half years.” And so I’ve been laying the foundation step by step by step, so when the time comes you announce it, everybody’s saying, “You know, what took you so long?”
你只要回去然后说：“天哪，他们已经谈论这件事两年半了。”所以我一直在一步一步地打基础，所以当你宣布这一点的时候，每个人都会说，“你知道吗，为什么花了这么长时间？”
But it’s not just inside the company.
But it’s not just inside the company.
21:44 You’re shaping the landscape, the broader global landscape of innovation.
你们正在塑造更广阔的全球创新格局。
Like, putting those ideas out there, you really are manifesting reality.
就像，把这些想法放在那里，你真的在体现现实。
We don’t build computers.
We don’t build computers.
21:53 We actually don’t build clouds.
我们实际上不构建云。
We don’t…
我们不……
As it turns out, we’re a computing platform company.
事实证明，我们是一家计算平台公司。
And so nobody can buy anything from us.
And so nobody can buy anything from us.
That’s the weird thing.
That’s the weird thing.
You know, we vertically design, vertically integrate to design and optimize, but then we open up the entire platform at every single layer to be integrated into other companies’ products and services and clouds and supercomputers and OEM computers, and so the amazing thing is, I can’t do what I do without having convinced them first.
你知道，我们垂直设计、垂直集成来设计和优化，然后我们在每一层开放整个平台，以便集成到其他公司的产品和服务、云、超级计算机和 OEM 计算机中，所以令人惊奇的是，如果不先说服他们，我就无法做我所做的事情。
And so most of GTC is about manifesting a future that by the time that we…
因此，GTC 的大部分内容都是为了展现一个未来，到那时我们……
My product is ready, they’re going, “What took you so long?”
我的产品已经准备好了，他们会问，“你怎么花了这么长时间？”
Yeah.
是的。
22:39 So one of the things you’ve been a believer for a long time is scaling laws, broadly defined.
所以你长期以来一直相信的一件事就是广义的缩放定律。
So are you still a believer in the scaling laws?
So are you still a believer in the scaling laws?
Yeah, yeah.
是啊，是啊。
22:49 Yeah, we have more scaling laws now.
Yeah, we have more scaling laws now.
So I think you’ve outlined four of them with pre-training, post-training, test time, and agentic scaling.
所以我认为你已经概述了其中的四个，包括训练前、训练后、测试时间和代理扩展。
22:51 What do you think, when you think about the future, deep future and the near-term future, what are the blockers that you’re most concerned about that keep you up at night that you have to overcome in order to keep scaling?
当您思考未来、长远的未来和近期的未来时，您认为哪些障碍是您最关心的、让您夜不能寐、为了继续扩展而必须克服的障碍？
Well, we can go back and reflect on what people thought were blockers.
好吧，我们可以回顾一下人们认为的阻碍因素。
23:12 So in the beginning, we were…
So in the beginning, we were…
The pre-training scaling law.
The pre-training scaling law.
You know, people thought, rightfully so, that the amount of data that we have, high-quality data that we have, will limit the intelligence that we achieve.
你知道，人们理所当然地认为，我们拥有的数据量、我们拥有的高质量数据将限制我们获得的智能。
And that scaling law was an important, very important scaling law.
这个比例定律是一个非常重要的比例定律。
The larger the model, the correspondingly more data results in a smarter AI.
模型越大，相应的数据就越多，人工智能就越智能。
And so that was pre-training.
And so that was pre-training.
And Ilya Sutskever said, “We’re out of data,” or something like that.
Ilya Sutskever 说，“我们没有数据了”，或者类似的话。
“Pre-training is over,” or something like that.
“预训练结束了”之类的话。
The industry panicked, you know, that this is the end of AI.
你知道，整个行业都惊慌失措，认为这就是人工智能的终结。
And of course, that’s obviously not true.
当然，这显然不是真的。
23:57 We’re gonna keep on scaling the amount of data that we have to train with.
我们将继续扩大训练所需的数据量。
A lot of that data is probably gonna be synthetic, and that also confused people, you know?
很多数据可能都是合成的，这也让人们感到困惑，你知道吗？
And what people don’t realize is they’ve kind of forgotten that most of the data that we are training, that we teach each other with, inform each other with, is synthetic.
人们没有意识到的是，他们有点忘记了，我们正在训练、互相教导、互相告知的大部分数据都是合成的。
You know, it’s synthetic because it didn’t come out of nature.
你知道，它是合成的，因为它不是来自自然界。
You created it.
你创造了它。
I’m consuming it.
我正在消耗它。
I modify it, augment it, I regenerate it, somebody else consumes it.
我修改它，增强它，我重新生成它，其他人消耗它。
And so we’ve now reached a level where AI is able to take ground truth, augment it…
所以我们现在已经达到了人工智能能够获取事实真相并对其进行增强的水平……
Enhance it, synthetically generate an enormous amount of data.
对其进行增强，综合生成海量数据。
24:47 And that part of post-training continues to scale, and so the amount of data that we could use that is human generated will be smaller, and smaller, and smaller.
训练后的这一部分继续扩大规模，因此我们可以使用的人类生成的数据量将越来越小。
The amount of data that we use to train models is going to continue to scale to the point where we’re no longer limited…
我们用于训练模型的数据量将继续扩大到我们不再受到限制的程度......
Training is no longer limited by…
Training is no longer limited by…
Data is now limited by compute.
Data is now limited by compute.
And the reason for that is most of the data is synthetic.
原因是大部分数据都是合成的。
Then the next phase is test time, and I still remember people telling me that, “Inference?
然后下一阶段是测试时间，我仍然记得人们告诉我，“推理？
Oh, yeah, that’s easy.
哦，是的，这很容易。
Pre-training, that’s hard.” These are giant systems that people are talking about.
预训练，这很难。”这些都是人们正在谈论的巨型系统。
Inference must be easy.
推理一定很容易。
And so inference chips are gonna be little tiny chips, and-
所以推理芯片将是很小的芯片，而且-
25:32 …
……
you know, they’re not, they’re not like NVIDIA’s chips.
你知道，它们不是，它们不像 NVIDIA 的芯片。
Oh, those are gonna be complicated and expensive, and, you know, we could make…
Oh, those are gonna be complicated and expensive, and, you know, we could make…
And this is- …
而这就是——……
in, in the future, inference is gonna be the biggest market, and it’s gonna be easy, and we’re gonna commoditize it.
未来，推理将成为最大的市场，而且会很容易，我们会将其商品化。
You know, everybody can build their own chips.
You know, everybody can build their own chips.
And, and that was always illogical to me because inference is thinking, and I think thinking is hard.
And, and that was always illogical to me because inference is thinking, and I think thinking is hard.
Thinking is way harder than reading.
思考比阅读困难得多。
25:59 You know, pre-training is just memorization and generalization, you know, and looking for patterns in relationships.
你知道，预训练只是记忆和概括，以及寻找关系中的模式。
You’re reading and reading, versus thinking, reasoning, solving problems, taking unexplored experiences, new experiences, and breaking it down into…
你不断地阅读，而不是思考、推理、解决问题、获取未探索的经验、新的经验，并将其分解为……
Decomposing it into, you know, solvable pieces that we then go off, either through first principle reasoning, or, you know, through previous examples, prior experiences.
将其分解为可解决的部分，然后我们可以通过第一原理推理，或者通过以前的例子和经验来解决这些问题。
You know, or just exploration and search and, you know, trying different things.
你知道，或者只是探索和搜索，你知道，尝试不同的事情。
And that whole process of test time scaling inference, is really about thinking.
测试时间缩放推理的整个过程实际上是关于思考的。
And it’s about reasoning, it’s about planning, it’s about search, it’s about…
这是关于推理、关于计划、关于搜索、关于……
26:50 And so how could that possibly be compute light?
And so how could that possibly be compute light?
And we were absolutely right about that.
我们对此是绝对正确的。
You know, so test time scaling is intensely compute intensive.
您知道，因此测试时间缩放是计算密集型的。
Then the question is, okay, now we’re at inference and we’re at test time scaling, what’s beyond that?
Then the question is, okay, now we’re at inference and we’re at test time scaling, what’s beyond that?
Well, obviously we have now created, you know, one agentic person, and that one agentic person has a large language model that we’ve now developed.
嗯，显然我们现在已经创建了，你知道，一个代理人，并且这个代理人拥有我们现在开发的大型语言模型。
But during test time, that agentic system goes off and does research and bangs on databases, and it goes out and, you know, uses tools, and one of the most important things it does is spins off and spawns off a whole bunch of sub-agents.
但在测试期间，该代理系统会启动并进行研究并敲击数据库，然后它会启动并使用工具，而它所做的最重要的事情之一就是剥离并产生一大堆子代理。
Which means we’re now creating large teams.
Which means we’re now creating large teams.
It’s so much easier to scale NVIDIA by hiring more employees than it is to scale myself.
It’s so much easier to scale NVIDIA by hiring more employees than it is to scale myself.
27:44 And so the next scaling law is the agentic scaling law.
因此，下一个标度法则是代理标度法则。
It’s kind of like multiplying AI.
It’s kind of like multiplying AI.
Multiplying AI, we could spin off agents as fast as you want to spin off agents.
通过人工智能的倍增，我们可以像您希望分拆代理一样快地分拆代理。
And so, you know, I…
所以，你知道，我……
You know, I have four scaling laws.
You know, I have four scaling laws.
And as we use the agentic systems, they’re gonna create a lot more data, they’re gonna create a lot of experiences.
And as we use the agentic systems, they’re gonna create a lot more data, they’re gonna create a lot of experiences.
Some of it we’re gonna say, “Wow, this is really good.
Some of it we’re gonna say, “Wow, this is really good.
We ought to memorize this.”
We ought to memorize this.”
28:12 That data set then comes all the way back to pre-training.
然后该数据集一路返回到预训练。
We memorize and generalize it.
We memorize and generalize it.
We then refine it and fine-tune it back into post-training.
然后我们对其进行完善并在训练后进行微调。
Then we enhance it even more with test time, you know, and the agentic systems, you know, put it out to the industry.
然后我们通过测试时间进一步增强它，你知道，代理系统，你知道，将其发布给行业。
And so this loop, this cycle, is gonna go on and on and on.
所以这个循环，这个循环，会一直持续下去。
It kinda comes down to basically intelligence is gonna scale by one thing, and that’s compute.
这有点归结为基本上智能将通过一件事来扩展，那就是计算。
But there’s a tricky thing there that you have to anticipate and predict, which is some of these components, it requires different kind of hardware to really do it optimally.
但有一件棘手的事情你必须预见和预测，其中一些组件需要不同类型的硬件才能真正以最佳方式完成。
28:41 So you have to anticipate where the AI innovation’s going to lead.
So you have to anticipate where the AI innovation’s going to lead.
For example, a mixture of-
For example, a mixture of-
Perfect.
完美。
…
……
28:58 experts with sparsity.
具有稀疏性的专家。
Perfect.
完美的。
With hardware, you can’t just pivot on a week’s notice.
对于硬件，你不能仅仅根据一周的通知进行调整。
29:00 You have to anticipate what that’s going to look like.
You have to anticipate what that’s going to look like.
It has some-
它有一些——
So good.
太好了。
…
……
29:07 that’s so scary and difficult to do, right?
that’s so scary and difficult to do, right?
For example, these AI model architectures are being invented about once every six months.
例如，这些人工智能模型架构大约每六个月发明一次。
29:09 Right?
正确的？
And system architectures and hardware architectures kind of every three years.
系统架构和硬件架构每三年更新一次。
And so you need to anticipate what likely is going to happen, you know, two, three years from now.
因此，你需要预测两三年后可能发生的情况。
And there’s a couple ways that you could do that.
有几种方法可以做到这一点。
First of all, we could do research internally ourselves, and that’s one of the reasons why we have basic research, we have applied research.
首先，我们可以自己内部做研究，这就是我们有基础研究、有应用研究的原因之一。
29:40 We create our own models.
We create our own models.
And so we have hands-on life experience right here.
所以我们在这里有实际的生活经验。
This is part of the co-design that I’m talking about.
这是我所说的协同设计的一部分。
We’re also the only AI company in the world that works with literally every AI company in the world.
我们也是世界上唯一一家与世界上几乎所有人工智能公司合作的人工智能公司。
And to the extent that we can, we try to get a sense of what are the challenges that people are experiencing.
我们会尽力了解人们正在经历的挑战。
So you’re listening to the whispers across the industry, the AI labs.
所以你正在倾听整个行业、人工智能实验室的窃窃私语。
That’s right.
这是正确的。
30:02 You got to listen and learn from everybody.
You got to listen and learn from everybody.
And have a…
并有一个……
And then the last part is to have an architecture that’s flexible, that can adapt and move with the wind.
最后一部分是拥有一个灵活的架构，能够适应并随风而动。
And one of the benefits of CUDA is that it’s, you know, on the one hand, an incredible accelerator.
CUDA 的好处之一是，一方面，它是一个令人难以置信的加速器。
On the other hand, it’s really flexible.
On the other hand, it’s really flexible.
And so that balance, incredible balance between specialization, otherwise we can’t accelerate the CPU, versus generalization, so that we can adapt with changing algorithms, that’s really, really important.
因此，专业化与泛化之间令人难以置信的平衡，否则我们无法加速 CPU，这样我们就可以适应不断变化的算法，这非常非常重要。
That’s the reason why CUDA has been so resilient on the one hand, and yet we continue to enhance it.
这就是为什么 CUDA 一方面如此具有弹性，但我们仍在不断增强它的原因。
30:44 We’re at CUDA 13.2, and so we’re evolving the architecture so fast that we can stay with the modern algorithms.
我们目前处于 CUDA 13.2 版本，因此我们正在快速发展架构，以便我们可以继续使用现代算法。
For example…
例如……
When mixtures of experts came out, that’s the reason why we had NVLink 72 instead of NVLink 8.
当专家混合出现时，这就是我们使用 NVLink 72 而不是 NVLink 8 的原因。
We could now take an entire 4 trillion, 10 trillion parameter model and put it in one computing domain as if it’s running on one GPU.
我们现在可以将整个 4 万亿、10 万亿参数模型放入一个计算域中，就好像它在一个 GPU 上运行一样。
People probably didn’t notice, I said it, but if you look at the architecture of the Grace Blackwell racks, it was completely focused on doing one thing, processing the LLM.
人们可能没有注意到，我说过，但是如果你看看 Grace Blackwell 机架的架构，你会发现它完全专注于做一件事，即处理法学硕士。
All of a sudden, one year later, you’re looking at a Vera Rubin rack.
突然间，一年后，你看到了维拉·鲁宾 (Vera Rubin) 的架子。
It has storage accelerators.
它有存储加速器。
It has this incredible new CPU called Vera.
It has this incredible new CPU called Vera.
It has Vera Rubin and NVLink 72 to run the LLMs.
它有 Vera Rubin 和 NVLink 72 来运行法学硕士。
31:46 It also has this new additional rack called Rock.
它还有一个名为 Rock 的新附加机架。
And so this entire rack system is completely different than the previous one, and it’s got all these new components in it.
因此，整个机架系统与以前的机架系统完全不同，并且其中包含所有这些新组件。
And the reason for that is because the last one was designed to run MoE large language models, inference.
原因是最后一个设计用于运行 MoE 大型语言模型，即推理。
And this one is to run agents and agents bang on tools, and-
这是运行代理和代理工具，然后-
Obviously, the design of the system had to have been done before Claude Code, Codex, OpenClaw.
显然，系统的设计必须在Claude Code、Codex、OpenClaw之前完成。
32:10 So you were anticipating the future, essentially.
所以你本质上是在预测未来。
And that, and that comes from what?
那，那来自什么？
From the whispers, from the understanding what all the state-
从耳语中，从对所有状态的理解中——
No
否
…
……
32:25 of the art is about?
艺术是关于什么的？
No, it’s easier than that.
不，比那更容易。
32:26 You just reason about it.
你只要推理一下就可以了。
First of all, you just reason.
First of all, you just reason.
No matter, no matter what happens, at some point in order for that large language model to be a digital worker…
无论如何，无论发生什么，在某个时刻，为了让大型语言模型成为数字工作者……
Let’s just use that metaphor.
Let’s just use that metaphor.
Let’s say that we want the LLM to be a digital worker.
假设我们希望法学硕士成为一名数字工作者。
What does that have to do?
这有什么关系？
It has to access ground truth.
It has to access ground truth.
That’s our file system.
That’s our file system.
It has to be able to do research.
它必须能够进行研究。
It doesn’t know everything.
它并不知道一切。
We don’t have…
我们没有……
And I don’t wanna wait until this AI becomes, you know, universally smart about everything, past, present, and future before I make it useful.
我不想等到这个人工智能对过去、现在和未来的一切都变得普遍智能，然后我才让它变得有用。
And so therefore, I might as well let it go do research.
所以，我还是放手去研究吧。
It’s obvious; if it wants to help me, it’s gotta use my tools.
这是显而易见的；如果它想帮助我，就必须使用我的工具。
33:13 You know, a lot of people would say, “You know AI is gonna completely destroy software.
你知道，很多人会说，“你知道人工智能将彻底摧毁软件。
We don’t need software anymore.
我们不再需要软件了。
We don’t even need tools anymore.” That’s ridiculous.
我们甚至不再需要工具了。”这太荒谬了。
Let’s use the…
让我们使用...
Let’s use a thought experiment.
让我们做一个思想实验。
And you could just sit there, enjoy a glass of whiskey, and think about all these things, and it would become completely obvious.
你可以坐在那里，享受一杯威士忌，并思考所有这些事情，它就会变得完全显而易见。
Like, if I were to create the most amazing agent that we can imagine in the next 10 years.
就像，如果我要创造出我们可以想象的未来 10 年内最令人惊叹的特工。
Let’s say it’d be a humanoid robot.
假设它是一个人形机器人。
If that humanoid robot were to be created, is it more likely that the humanoid robot comes into my house and uses the tools that I have to do the work that it needs to do?
如果要创建该人形机器人，该人形机器人是否更有可能进入我的房子并使用我必须完成其需要完成的工作的工具？
33:54 Or does this hand turn into a 10-pound hammer in one instance, turn into a scalpel in another instance, and in order to boil water, it beams, you know, microwaves out of its fingers?
或者这只手是否会在某种情况下变成一把 10 磅重的锤子，在另一种情况下变成一把手术刀，为了烧水，它会从手指中发出微波？
You know, or is it more likely just to use a microwave, you know?
你知道，或者更有可能只使用微波炉，你知道吗？
And the first time it goes up to the microwave, it probably doesn’t know how to use it.
而且第一次接触微波炉时，它可能不知道如何使用。
But that’s okay.
但没关系。
It’s connected to the internet.
它已连接到互联网。
It reads the manual of this microwave, reads it, instantly becomes an expert.
它看了这个微波炉的说明书，看了它，瞬间就成了专家。
And so it uses it.
所以它使用它。
And so I think the…
所以我认为……
I just described, in fact, almost all of the properties of OpenClaw.
事实上，我刚刚描述了OpenClaw的几乎所有属性。
34:35 You know, that it’s gonna use tools, that it’s gonna access files, it’s gonna be able to do research.
你知道，它将使用工具，它将访问文件，它将能够进行研究。
It has an IO subsystem.
它有一个IO子系统。
And when you’re done reasoning through it, reasoning about it in that way, then you say, “Oh, my gosh, the impact to the future of computing is deeply profound.” And the reason for that is, I think we’ve just reinvented the computer.
当你完成推理后，以这种方式推理，然后你会说，“哦，天哪，这对计算的未来影响是深远的。”原因是，我认为我们刚刚重新发明了计算机。
And then now you say, “Okay, when did we reason about that?
然后现在你说：“好吧，我们什么时候考虑过这个问题的？
When did we reason about OpenClaw?” If you take the OpenClaw schematic that I used at GTC, you’ll find it two years ago.
我们什么时候考虑过 OpenClaw 的？”如果你拿我在 GTC 上使用的 OpenClaw 原理图，你会发现它是两年前的。
Literally, two years ago at GTC, I was talking about agentic systems that exactly reflect OpenClaw today.
从字面上看，两年前在 GTC 上，我谈论的代理系统完全反映了当今的 OpenClaw。
And, of course, the confluence of many things had to happen.
当然，许多事情必须融合在一起。
35:26 First of all, we needed Claude and GPT and, you know, all of these models to reach a level of capability.
首先，我们需要 Claude 和 GPT，以及所有这些模型才能达到一定的能力水平。
So their innovation and their breakthroughs and their continued advances was really important.
因此，他们的创新、突破和持续进步非常重要。
And then, of course, somebody had to create an open source project that was sufficiently robust and sufficiently complete and that we can all put to work.
然后，当然，有人必须创建一个足够强大、足够完整并且我们都可以投入使用的开源项目。
And I think OpenClaw did for agentic systems what ChatGPT did for generative systems.
我认为 OpenClaw 对于代理系统的作用就像 ChatGPT 对于生成系统的作用一样。
And I just think it’s a very big deal.
我只是认为这是一件大事。
Yeah, it’s a really special moment.
是的，这真是一个特别的时刻。
36:02 I’m not exactly sure why it captured so much of the world’s attention, but it did, more than Claude Code and Codex and so on.
我不太清楚为什么它吸引了全世界如此多的关注，但它确实比 Claude Code 和 Codex 等更受关注。
Because consumers could reach it.
因为消费者可以接触到它。
Sure, yeah.
当然，是的。
36:13 But there’s also so much of this is vibes.
但其中也有很多共鸣。
And Peter, I had a podcast with him, he’s a wonderful human being.
彼得，我和他一起做了一个播客，他是一个很棒的人。
So part of it is also the humans that represent the thing.
所以它的一部分也是代表事物的人类。
Yeah, no doubt.
是的，毫无疑问。
Part of it is memes and the— ‘Cause we’re all trying to figure it out.
其中一部分是模因和——因为我们都在试图弄清楚它。
36:25 There’s really serious and complicated security concerns about when you have such powerful technology, how do you hand over your data so they can do useful stuff?
当你拥有如此强大的技术时，你如何移交你的数据以便他们可以做有用的事情，这确实是严重而复杂的安全问题？
But then there’s scary things associated with that.
但随之而来的是一些可怕的事情。
And we, as a civilization, as individual people and as a civilization, are figuring out how to find that right balance.
作为一个文明，作为个人和文明，我们正在寻找如何找到适当的平衡。
Yeah, we jumped on it right away and we sent a bunch of security experts this way.
是的，我们立即采取了行动，并派了一群安全专家来。
36:44 And we did this thing called OpenShell.
我们做了一个叫做 OpenShell 的事情。
It’s already been integrated into OpenClaw.
它已经集成到 OpenClaw 中。
And NVIDIA put forward NemoClaw.
而NVIDIA提出了NemoClaw。
Yep, exactly.
是的，完全正确。
They install super easy.
他们安装超级简单。
36:59 It makes sure that it’s secure.
它确保它是安全的。
We give you two out of three rights.
我们给予您三项权利中的两项。
37:03 Agentic systems can access sensitive information, it can execute code, and it can communicate externally.
代理系统可以访问敏感信息，可以执行代码，并且可以与外部通信。
We could keep things safe if we gave you two out of those three capabilities at any time, but not all three.
如果我们随时为您提供这三种功能中的两种，但不是全部三种，我们就可以保证事情的安全。
And out of those two out of three capabilities, we also give you access control based on whatever rights that you’re given by enterprise.
在这三分之二的功能中，我们还根据企业授予您的任何权限为您提供访问控制。
And then we connect it to a policy engine that all these enterprises already have.
然后我们将其连接到所有这些企业已经拥有的策略引擎。
And so we’re going to try to do our best to help OpenClaw become a better claw.
因此，我们将尽力帮助 OpenClaw 成为更好的爪子。
So you eloquently explained how we have a long history of blockers that we thought were going to be blockers, and we overcame them.
所以你雄辩地解释了我们如何拥有悠久的阻碍因素历史，我们认为这些阻碍因素将成为阻碍因素，并且我们克服了它们。
37:40 But now looking into the future, what do you think might be the blockers now that it’s clear that agents will be everywhere?
但现在展望未来，既然代理显然将无处不在，您认为可能的阻碍因素是什么？
So it’s obviously we’re going to need compute.
所以很明显我们需要计算。
So what is going to be the blocker for that scaling?
那么，阻碍这种扩展的因素是什么？
Power is a concern, but it’s not the only concern.
权力是一个问题，但不是唯一的问题。
37:59 But that’s the reason why we’re pushing so hard on extreme co-design, so that we can improve the tokens per second per watt orders of magnitude every single year.
但这就是我们如此努力推动极端协同设计的原因，以便我们每年都能将每秒每瓦的令牌数提高几个数量级。
And so in the last 10 years, Moore’s Law would have progressed computing about 100 times in the last 10 years.
因此，在过去 10 年中，摩尔定律将使计算在过去 10 年中进步约 100 倍。
We progressed and scaled up computing by a million times in the last 10 years.
在过去 10 年里，我们的计算能力取得了进步，规模扩大了一百万倍。
And so we’re gonna keep on doing that through extreme co-design.
因此，我们将继续通过极端的协同设计来做到这一点。
So energy efficiency, perf per watt, completely affects the revenues of a company.
因此，能源效率（每瓦性能）完全影响公司的收入。
It affects the revenues of a factory.
它影响工厂的收入。
And we’re just going to push that to the limit so that we can keep on driving token costs down as fast as we can.
我们只是要将其推向极限，以便我们能够继续尽可能快地降低代币成本。
38:51 You know, our computer price is going up, but our token generation effectiveness is going up so much faster that token cost is coming down.
您知道，我们的计算机价格正在上涨，但我们的代币生成效率提高得更快，以至于代币成本正在下降。
It’s just coming down an order of magnitude every year.
它只是每年下降一个数量级。
So power, that’s an interesting one.
所以权力，这是一个有趣的事情。
39:04 So the way to try to get around the power blocker is to try to, with the tokens per second per watt, try to make it more and more efficient.
因此，尝试绕过功率拦截器的方法是尝试使用每秒每瓦特的令牌，尝试使其变得越来越高效。
Of course, there’s the question of how do we get more power.
当然，还有一个问题是我们如何获得更多的权力。
We should also get more power.
我们还应该获得更多的权力。
That’s a really complicated one.
这是一个非常复杂的问题。
39:17 You’ve talked about small modular nuclear power plants.
您谈到了小型模块化核电站。
There’s all kinds of ideas for energy.
关于能源有各种各样的想法。
How much does it keep you up at night?
它让你彻夜难眠吗？
The bottlenecks in the supply chain of AI, like ASML with EUV lithography machines, TSMC with advanced packaging like CoWoS, and SK Hynix with the high bandwidth memory?
AI供应链的瓶颈，比如ASML的EUV光刻机、台积电的CoWoS等先进封装、SK海力士的高带宽内存？
All the time, and we’re working on it all the time.
一直以来，我们一直在努力。
39:38 No company in history has ever grown at a scale that we’re growing while accelerating that growth.
历史上还没有哪家公司能够达到我们在加速增长的同时增长的规模。
It’s incredible.
太不可思议了。
And it’s hard for people to even understand this.
人们甚至很难理解这一点。
In the overall world of AI computing, we’re increasing share.
在整个人工智能计算领域，我们的份额正在不断增加。
And so supply chain, upstream and downstream, are really important to us.
因此，上游和下游的供应链对我们来说非常重要。
I spend a lot of time informing all the CEOs that I work with: what are the dynamics that’s going to cause the growth to continue or even accelerate?
我花了很多时间告诉所有与我共事的首席执行官：哪些动力将导致增长持续甚至加速？
It’s part of the reasons why to the entire right-hand side of me were CEOs of practically the entire IT industry upstream and practically the entire infrastructure industry downstream.
这就是为什么我的整个右手边几乎都是上游 IT 行业和下游基础设施行业的首席执行官。
40:32 And they were all…
他们都是……
There were several hundred CEOs.
首席执行官有数百人。
And I don’t think there’s ever been keynotes where several hundred CEOs show up.
我认为从来没有过有数百名首席执行官出席的主题演讲。
And part of it is, I’m telling them about our business condition now.
其中一部分是，我现在告诉他们我们的业务状况。
I’m telling them about the growth drivers in the very near future and what’s happening.
我告诉他们不久的将来的增长动力以及正在发生的事情。
And I’m also describing where are we going to go next so that they could use all of this information and all of the dynamics that are here to inform how they want to invest.
我还描述了我们下一步要去哪里，以便他们可以利用所有这些信息和这里的所有动态来告知他们想要如何投资。
And so I inform them that way like I inform my own employees.
所以我会像通知我自己的员工一样通知他们。
41:06 And then of course, then I make trips out to them and make sure that, “Hey, listen, I want you to know this quarter, this coming year, this next year, these things are going to happen.” And if you look at the CEOs of the DRAM industry—the number one DRAM in the world was DDR memory for CPUs in data centers.
当然，然后我会去拜访他们，并确保，“嘿，听着，我想让你知道这个季度、明年、明年，这些事情都会发生。”如果你看看 DRAM 行业的首席执行官，就会发现世界上排名第一的 DRAM 是用于数据中心 CPU 的 DDR 内存。
About three years ago, I was able to convince several of the CEOs that even though at the time HBM memory was used quite scarcely, and barely by supercomputers, that this was going to be a mainstream memory for data centers in the future.
大约三年前，我说服了几位首席执行官，尽管当时 HBM 内存很少被使用，而且几乎没有被超级计算机使用，但这将成为未来数据中心的主流内存。
At first it sounded ridiculous, but several of the CEOs believed me and decided to invest in building HBM memories.
起初这听起来很荒谬，但几位首席执行官相信了我，并决定投资构建 HBM 存储器。
41:55 Another memory was rather odd to put into a data center: the low power memories that we use for cell phones.
另一种内存在数据中心中的使用相当奇怪：我们用于手机的低功耗内存。
And we wanted them to adapt them for supercomputers in the data center.
我们希望他们能够适应数据中心的超级计算机。
And they go, “Cell phone memory for supercomputers?” And I explained to them why.
他们问道：“超级计算机的手机内存？”我向他们解释了原因。
Well, look at these two memories, LPDDR5, HBM4.
好吧，看看这两种内存，LPDDR5、HBM4。
The volumes are so incredible.
数量是如此令人难以置信。
All three of them had record years in history, and these are 45-year companies.
这三家公司都拥有创纪录的历史年限，而且这些公司都有 45 年的历史。
And so, you know, I…
所以，你知道，我……
That’s part of my job, is to inform and shape, inspire, you know.
这是我工作的一部分，就是提供信息、塑造、激励，你知道。
So you’re not just manifesting the future and maybe inspiring NVIDIA, the different engineers of the company, you’re manifesting the supply chain of the future.
因此，您不仅体现了未来并可能激励 NVIDIA、公司的不同工程师，您还体现了未来的供应链。
42:36 So you’re having conversations with TSMC, with ASML.
所以你正在与台积电、ASML 进行对话。
Upstream, downstream.
上游，下游。
Upstream, downstream.
上游，下游。
42:51 So that’s the thing.
就是这样。
GEV, Caterpillar.
GEV，卡特彼勒。
42:53 Yeah, that’s downstream from us.
是的，那是我们的下游。
Yeah, yeah, there you go.
是啊，是啊，就这样吧。
Yeah, the whole thing.
是的，整个事情。
42:59 I mean, but that’s so…
我的意思是，但事实就是这样……
There’s so much incredibly difficult engineering that happens in the entire semiconductor industry, and it just feels scary how intricate the supply chain is, how many components there are, but it works somehow.
整个半导体行业发生了如此之多的极其困难的工程，供应链是多么复杂，有多少组件，但它以某种方式运作，让人感觉很可怕。
Exactly, the deep science.
准确地说，是深奥的科学。
43:18 The deep engineering, the incredible manufacturing, and so much of the manufacturing is already robotics, but we have a couple of hundred suppliers that contribute the technology that goes into our 1.3 million component rack.
深入的工程、令人难以置信的制造以及如此多的制造已经是机器人技术，但我们有数百家供应商为我们的 130 万个组件机架提供技术。
Each rack is 1.3, one and a half million components.
每个机架有 1.3，即五十万个组件。
There are 200 suppliers across the Vera Rubin rack.
Vera Rubin 货架上有 200 家供应商。
So it’s interesting that you don’t list that as the thing that keeps you up at night in the list of blockers.
所以有趣的是，你没有将其列为让你彻夜难眠的障碍列表中的事情。
But I’m doing, I’m doing all the things necessary to-
但我正在做，我正在做所有必要的事情——
Okay.
好的。
…
……
43:52 yeah, see?
是啊，看到了吗？
I can go to sleep because I checked it off.
我可以去睡觉了，因为我把它勾掉了。
I said, “Okay,” you know, I go, I can go to sleep and I go, “Well, let’s see, let’s reason about this.
我说，“好吧，”你知道，我走了，我可以去睡觉了，然后我说，“好吧，让我们看看，让我们来推理一下。
What’s important for us?” Because let’s reason about this.
对我们来说什么重要？”因为让我们来推理一下。
Because we changed the system architecture from the original DGX-1 that you remembered to NVLink-72 rack scale computing- …
因为我们把系统架构从大家记得的原来的DGX-1改成了NVLink-72机架规模计算——……
what’s gonna…
会发生什么……
What does that, what does that mean?
那是什么意思？
What does that mean to software?
这对软件意味着什么？
What does that mean to engineering?
这对工程意味着什么？
What does that mean to how we design and test?
这对我们的设计和测试意味着什么？
And what does that mean to the supply chain?
这对供应链意味着什么？
Well, one of the things that it meant was we moved supercomputer integration at the data center into supercomputer manufacturing in the supply chain.
嗯，这意味着我们将数据中心的超级计算机集成转移到供应链中的超级计算机制造中。
44:42 If you’re doing that, you also have to recognize you’re gonna move one…
如果你这样做，你还必须认识到你要移动一个......
And if your total footprint of whatever data center you’re gonna build, let’s say you would like to have, you know, 50 gigawatts of supercomputers that are running simultaneously, and it takes one week to manufacture that 50 gigawatts of supercomputers, then each week in the supply chain, the supercomputers are gonna need a gigawatt of power.
如果您要建造的任何数据中心的总占地面积，假设您希望拥有 50 吉瓦同时运行的超级计算机，并且需要一周的时间来制造这 50 吉瓦的超级计算机，那么在供应链中，每周超级计算机将需要 1 吉瓦的电力。
And so we’re gonna need the supply chain to increase the amount of power it has to build and test the supercomputers in the supply chain before I ship it.
因此，在发货之前，我们需要供应链增加其构建和测试供应链中的超级计算机的能力。
Oh.
哦。
Well, NVLink-72 literally builds supercomputers in the supply chain and ships ’em two, three tons at a time per rack.
嗯，NVLink-72 实际上是在供应链中构建超级计算机，并且每个机架一次运送两到三吨。
45:25 It used to be they used to come in parts and we used to assemble ’em inside the data center.
过去它们都是分成零件的，我们过去常常在数据中心内组装它们。
But that’s impossible now because NVLink-72 is so dense.
但现在这是不可能的，因为 NVLink-72 非常密集。
And so that’s an example.
这就是一个例子。
And I would have to go and fly into the supply chain, go meet my partners saying, “Hey,” I said, “guess what?
我必须飞进供应链，去见我的合作伙伴，说：“嘿，”我说，“你猜怎么着？
So here’s what I’m going to do with…
所以这就是我要做的......
This is the way we used to build our DGXs.
这就是我们用来构建 DGX 的方式。
We’re gonna build them this way.
我们将以这种方式构建它们。
This is gonna be so much better because we’re going to need ’em for inference.” The market for inference is, you know, coming.
这会好得多，因为我们将需要它们进行推理。”你知道，推理市场即将到来。
The inflection point for inference is coming.
推理的拐点即将到来。
It’s gonna be a big market.
这将是一个很大的市场。
46:05 And so I first explain to them what’s going on, why it’s gonna happen, and then I ask ’em to make several billion dollars of capital investments each.
所以我首先向他们解释发生了什么，为什么会发生，然后我要求他们每人进行数十亿美元的资本投资。
And because they trust me and I’m very respectful of ’em, and I give ’em every opportunity to question me and I spend time to explain things to people and I reason about it.
因为他们信任我，我非常尊重他们，我给他们一切机会提问我，我花时间向人们解释事情，并对此进行推理。
I draw pictures and I reason about it in first principles.
我画画并根据第一原理进行推理。
And by the time I’m done with them, they know what to do.
当我处理完他们的事情时，他们知道该怎么做。
So a lot of it is about relationships and building a shared view of the future.
所以很多都是关于关系和建立对未来的共同看法。
46:35 But do you worry about certain bottlenecks?
但您是否担心某些瓶颈？
I mean, what are the biggest bottlenecks in the supply chain?
我的意思是，供应链中最大的瓶颈是什么？
Are you worried about ASML’s EUV tooling?
您是否担心 ASML 的 EUV 工具？
Are you worried about the packaging, CoWoS packaging of TSMC, about how fast it could scale?
您是否担心台积电的封装，CoWoS 封装，担心它的扩展速度有多快？
Like you said, you’re not only growing incredibly fast, you’re accelerating your growth.
就像你说的，你不仅成长得非常快，而且还在加速成长。
So it feels like everybody in the supply chain, and those are certainly bottlenecks, would have to scale up.
因此，供应链中的每个人（这些肯定是瓶颈）都必须扩大规模。
Are you having conversations with them, like, how can you scale up faster?
您是否与他们进行对话，例如如何才能更快地扩大规模？
All the time.
一直以来。
Do you worry about it?
你担心吗？
No.
不。
Okay.
好的。
Because I told ’em what I needed.
因为我告诉他们我需要什么。
47:14 They understood what I need.
他们了解我的需要。
They told me what they’re gonna go do, and I believe them what they’re going to do.
他们告诉我他们要做什么，我相信他们要做什么。
Interesting.
有趣的。
47:22 That’s great to hear.
很高兴听到这个消息。
So maybe if we can just linger on the power for a little bit.
所以也许我们可以在电源上停留一会儿。
What are your hopes for how to solve the energy problem?
您对如何解决能源问题有什么希望？
One of the areas, Lex, that I would love us to talk about and just get the message out, you know, our power grid is designed for the worst case condition with some margin.
Lex，我希望我们能够讨论的领域之一，并传达出这样的信息，你知道，我们的电网是为最坏的情况而设计的，并有一定的余量。
47:30 Well, 99% of the time we’re nowhere near the worst case condition because the worst case condition is a few days in the winter, a few days in the summer, and extreme weather.
嗯，99% 的情况下，我们远未达到最坏的情况，因为最坏的情况是冬天的几天、夏天的几天和极端天气。
Most of the time we’re nowhere near the worst case condition and we’re probably running around, call it 60% of peak.
大多数时候，我们距离最坏情况还很远，我们可能会到处乱跑，称之为峰值的 60%。
48:08 And so 99% of the time, our power grid has excess power, and they’re just sitting idle, but they have to be there sitting idle because just in case, when the time comes, hospitals have to be powered and, you know, infrastructure has to be powered and airports have to run and so on and so forth.
所以99%的时候，我们的电网电力过剩，它们只是闲置，但它们必须闲置，因为以防万一，到时候，医院必须供电，你知道，基础设施必须供电，机场必须运行等等。
And so the question that I have is whether we could go and help them understand and create contractual agreements and design computer architecture systems, data centers, such that when they need the maximum power for infrastructure in society, that the data centers would get less.
因此，我的问题是，我们是否可以帮助他们理解和创建合同协议并设计计算机架构系统、数据中心，以便当他们需要社会基础设施的最大电力时，数据中心获得的电力就会更少。
48:49 But that’s in a very rare instance anyways.
但无论如何，这种情况都是非常罕见的。
And during that time, we either have a backup generator for that little part of it, or we just have our computers shift the workload somewhere else, or we have the computers just run slower.
在那段时间里，我们要么有一个备用发电机来处理那一小部分，要么我们只是让我们的计算机将工作负载转移到其他地方，或者我们让计算机运行得更慢。
You know, we could degrade our performance, reduce our power consumption and provide for a, you know, slightly longer latency response, you know, when somebody asks for, you know, asks for an answer.
你知道，我们可以降低我们的性能，减少功耗，并提供稍微更长的延迟响应，你知道，当有人要求时，你知道，要求答案。
And so I think that that, that way of using computers, of building data centers, instead of expecting 100% uptime—and these contracts that are really, really quite rigorous, it’s putting a lot of pressure on the grid to be able to…
所以我认为，使用计算机、建设数据中心的方式，而不是期望 100% 的正常运行时间——而且这些合同确实非常严格，它给电网带来了很大的压力，使其能够……
Now, they’re gonna have to increase from their maximum.
现在，他们必须从最大值开始增加。
I just wanna use their excess.
我只是想利用他们多余的东西。
It’s just sitting there.
它只是坐在那里。
Yeah, that’s not talked about enough.
是的，这还不够讨论。
49:36 So what’s stopping there?
那么是什么阻止了呢？
Is it regulation?
是监管吗？
Is it bureaucracy?
是官僚主义吗？
I think it’s a three-way problem.
我认为这是一个三向问题。
49:43 It starts with the end customer.
它从最终客户开始。
The end customer puts requirements on the data centers that they can never not be available, okay?
最终客户对数据中心提出了永远不可用的要求，好吗？
So that the end customer expects perfection.
让最终客户期望完美。
Now, in order to deliver that perfection, you need a combination of backup generators and your grid power supplier to deliver on perfection.
现在，为了实现完美，您需要结合备用发电机和电网电源来实现完美。
And so everybody’s gotta have six nines.
所以每个人都必须有六个 9。
Well, I think first of all, right now, we ought to have everybody understand that when the customer asks for these things, you have somebody in your data center operations team disconnected from the CEO.
嗯，我认为首先，现在我们应该让每个人都明白，当客户要求这些东西时，数据中心运营团队中的某个人就与首席执行官断开了联系。
I bet the CEO doesn’t know this.
我敢打赌首席执行官不知道这一点。
I’m gonna talk to all the CEOs.
我将与所有首席执行官交谈。
50:28 The CEOs are probably not paying any attention to the contracts that are being signed, and so everybody wants to sign the best contract, of course.
首席执行官们可能不会关注正在签署的合同，所以当然每个人都想签署最好的合同。
And they go down to cloud service providers, and the two contract negotiators that are…
他们归结为云服务提供商，以及两个合同谈判者……
You, I could just see them now.
你，我现在就可以看到他们了。
You know, negotiating these multi-year contracts.
你知道，谈判这些多年期合同。
Both sides want, you know, the best contract.
你知道，双方都想要最好的合同。
As a result, the CSPs then have to go down to the utilities, and they expect the nine, the six nines.
结果，CSP 必须向公用事业公司求助，他们期望得到 9 个、6 个 9。
And so I think, I think the first thing is just make sure that, that all of the customers, the CEOs and the customers realize what they’re asking for.
所以我认为，我认为第一件事就是确保所有客户、首席执行官和客户都意识到他们的要求。
Now, the second thing is we have to build data centers that gracefully degrade.
现在，第二件事是我们必须构建能够优雅降级的数据中心。
51:13 And so if the power, if the utility, if the grid tells us, “Listen, we’re gonna have to back you down to about 80%,” we’re gonna say, “That’s no problem at all.” We’re just gonna move our workload around.
因此，如果电力、公用事业、电网告诉我们，“听着，我们必须将电量降低到 80% 左右，”我们会说，“这根本没问题。”我们只是要转移我们的工作量。
We’re gonna make sure that data’s never lost, but we can reduce the computing rate and use less energy.
我们要确保数据永远不会丢失，但我们可以降低计算速率并使用更少的能源。
The quality of service degrades a little bit.
服务质量有点下降。
For the critical workloads, I shift that somewhere else right away so I don’t have that problem, and so, you know, whichever data center still has 100% uptime, and so…
对于关键工作负载，我会立即将其转移到其他地方，这样我就不会遇到这个问题，所以，你知道，无论哪个数据中心仍然具有 100% 的正常运行时间，等等......
How difficult of an engineering problem is that, that smart, dynamic allocation of power in a data center?
数据中心中智能、动态的电力分配是一个多么困难的工程问题？
As soon as you could specify, you could engineer it.
一旦您可以指定，您就可以对其进行设计。
51:49 Beautifully put.
摆得很漂亮。
So long as it obeys the laws of physics on first principles, I think we’re good.
只要它遵循第一原理的物理定律，我认为我们就很好。
What was the third thing you were mentioning?
你提到的第三件事是什么？
So the second thing is the, the data centers.
所以第二件事是数据中心。
52:00 And the third thing is we need the utilities to also recognize that this is an opportunity- …
第三件事是我们需要公用事业公司也认识到这是一个机会——……
and instead of saying, “Look, it’s gonna take me five years to increase my grid capability,” if you have, if you’re willing to take power of this level of guarantee, I can make them available for you next month and at this price.
而不是说，“看，我需要五年的时间来提高我的电网能力”，如果你愿意，如果你愿意获得这种级别的保证，我可以在下个月以这个价格为你提供它们。
And so if utilities also offered more segments of power delivery promises, then I think everybody will figure out what to do with it.
因此，如果公用事业公司也提供更多的电力输送承诺，那么我认为每个人都会弄清楚如何利用它。
Yeah, but there’s just way too much waste in the grid right now.
是的，但现在电网中的浪费太多了。
We should go after it.
我们应该去追寻它。
You’ve highly lauded Elon and xAI’s accomplishment in Memphis, in building Colossus supercomputer, probably in record time in just four months.
你们高度赞扬了 Elon 和 xAI 在孟菲斯建造 Colossus 超级计算机的成就，可能在短短四个月的时间里创下了纪录。
52:44 It’s now at 200,000 GPUs and growing very quickly.
现在 GPU 数量已达到 200,000 个，并且增长速度非常快。
Is there something that you could speak to the understanding about his approach that’s instructive to, broadly to all the data center creators that enabled that kind of accomplishment?
您是否可以谈谈对他的方法的理解，对所有实现这种成就的数据中心创建者具有指导意义？
His approach to engineering, his approach to the whole management of construction, everything?
他的工程方法，他的整个施工管理方法，一切？
First of all, Elon is deep in so many different topics.
首先，埃隆对许多不同的主题都有深入的了解。
53:15 Yet he’s also a really good systems thinker.
但他也是一位非常优秀的系统思考者。
And so he’s able to think through multiple disciplines, and he obviously pushes things, questions everything, where they’re, number one, is it necessary?
所以他能够思考多个学科，而且他显然会推动事情，质疑一切，它们在哪里，第一，有必要吗？
Number two, does it have to be done this way?
第二，一定要这样做吗？
And then number three, you know, does it have to take this long?
第三，你知道，需要花这么长时间吗？
And so he has the ability to question everything to the point where everything is down to its minimal amount that’s necessary, you can’t take anything else out.
所以他有能力质疑一切，直到一切都减少到必要的最低限度，你不能去掉任何其他东西。
And yet the necessary capabilities of the product remains, you know?
但产品的必要功能仍然存在，你知道吗？
And so he is as minimalist as you could possibly imagine, and he does it at a system scale.
所以他就像你想象的那样极简，而且他在系统规模上做到了这一点。
I think…
我认为……
I also love the fact that he is represented.
我也喜欢他有代表的事实。
He is present at the point of action.
他就在行动点。
54:25 You know, he’ll just go there.
你知道，他只会去那里。
If there’s a problem, he’ll just go there and then, “Show me the problem.” You know, when you do all of this in combination, you overcome a lot of previous, “This is just the way we do it.” “You know, I’m waiting for them.” You know, I mean, it’s just, everybody has a lot of excuses.
如果出现问题，他就会去那里，然后“告诉我问题。”你知道，当你把所有这些结合起来时，你就克服了很多以前的问题，“这就是我们做事的方式。” “你知道吗，我在等他们。”你知道，我的意思是，每个人都有很多借口。
And so, and then the last thing is when you act personally with so much urgency, it causes everybody else to act with urgency, you know?
所以，最后一件事是，当你个人采取如此紧迫的行动时，它会导致其他人采取紧急行动，你知道吗？
And every supplier has a lot of customers going on.
每个供应商都有很多客户。
Every supplier has a lot of projects going on, and he makes it his business that he’s the top priority of everybody else’s projects.
每个供应商都有很多项目正在进行，他把自己作为其他人项目的重中之重作为自己的事。
And so he does that by demonstrating it.
所以他通过展示来做到这一点。
Yeah, I’ve been in a bunch of those meetings.
是的，我参加过很多这样的会议。
55:09 It’s just, it’s fun to watch, ’cause really, not enough people ask the question like, “Okay, so can this be done a lot faster, and how?
只是，观看起来很有趣，因为实际上，没有足够多的人提出这样的问题：“好吧，这可以做得更快吗？如何做？
Why does it have to take this long?”
为什么要花这么长时间？”
Yeah, right.
是的，对。
And then in the…
然后在……
55:22 That becomes an engineering question often.
这常常成为一个工程问题。
And yes, I think when you get the ground truth of actually…
是的，我认为当你了解实际情况时……
I remember…
我记得……
One of the times I was hanging out with him, he literally is going through the entire process of how to plug in cables into a rack.
有一次我和他一起出去玩，他实际上正在经历如何将电缆插入机架的整个过程。
He’s working with an engineer on the ground that’s doing that task, and he’s just trying to understand what does that process look like so it can be less error-prone.
他正在与一位正在执行这项任务的工程师一起工作，他只是想了解这个过程是什么样的，这样就可以减少出错的可能性。
And just building up that intuition from every single task involved in putting together a data center-
只需从构建数据中心所涉及的每一项任务中建立起这种直觉-
55:52 …
……
you start to immediately get a sense at the detailed scale and at the broad systems scale of where the inefficiencies are, and so you can make it more and more and more efficient.
您开始立即了解详细范围和广泛系统范围中效率低下的地方，因此您可以使其变得越来越高效。
Plus you have the big hammer of being able to say, “Let’s do it totally different-“
另外，你还可以说，“让我们以完全不同的方式来做——”
Yeah.
是的。
56:08 That’s right.
这是正确的。
“…
“……
56:09 and remove all possible blockers.”
并消除所有可能的阻碍。”
That’s right.
这是正确的。
Is there parallels in the NVIDIA Extreme Systems co-design approach that you see in the way Elon approaches systems engineering?
您认为 NVIDIA Extreme Systems 协同设计方法与 Elon 的系统工程方法有相似之处吗？
Well, first of all, co-design is an ultimate systems engineering problem.
首先，协同设计是一个终极的系统工程问题。
56:18 And so we approach the work that we do from that first principle.
因此，我们从第一原则出发开展工作。
The other thing that we do and this is a philosophy that, a thought, a state of mind, I guess, a method that I started 30 years ago, and it’s called the speed of light.
我们做的另一件事是一种哲学，一种思想，一种心态，我想，是我 30 年前开始的一种方法，它被称为光速。
The speed of light is not just about the speed.
光速不仅仅是速度。
The speed of light is my shorthand for what’s the limit of what physics can do.
光速是我对物理学极限的简写。
And so every single thing that we do is compared against the speed of light.
因此，我们所做的每一件事都会与光速进行比较。
Memory speed, math speed, power, cost, time, effort, number of people, manufacturing cycle time.
内存速度、数学速度、功率、成本、时间、精力、人数、制造周期时间。
57:09 And when you think about latency versus throughput when you think about cost versus throughput, cost versus capacity, all of these things you test against the speed of light to achieve all of these different constraints separately.
当您考虑延迟与吞吐量时，当您考虑成本与吞吐量、成本与容量时，所有这些都需要根据光速进行测试，以分别实现所有这些不同的约束。
And then when you consider it together, you know you have to make compromises because a system that achieves extremely low latency versus a cheap, a system that achieves very high throughput are architected fundamentally differently.
然后，当您综合考虑时，您知道必须做出妥协，因为实现极低延迟的系统与实现非常高吞吐量的廉价系统的架构完全不同。
But you want to know what’s the speed of light of a system that achieves high throughput, what’s the speed of light of a system that achieves low latency?
但你想知道实现高吞吐量的系统的光速是多少，实现低延迟的系统的光速是多少？
And then when you think about the total system, you can make trade-offs.
然后当你考虑整个系统时，你可以做出权衡。
And so I force everybody to think about what’s the first principles, the limits-
所以我强迫每个人思考什么是首要原则，什么是限制——
58:01 …
……
the physical limits for everything before we do anything.
在我们做任何事情之前，先了解一切的物理限制。
And we test everything against that.
我们对此进行了测试。
And so that’s a good frame of mind.
所以这是一个很好的心态。
I don’t love the other methods, which is continuous improvement.
我不喜欢其他方法，那就是不断改进。
The problem with continuous improvement, it…
持续改进的问题，它......
First of all, you should engineer something from first principles at the speed, you know, with speed of light thinking.
首先，你应该以光速思维的速度从基本原理出发设计一些东西。
Limit it only by physical limits, and physics limits.
仅受物理极限和物理极限的限制。
And after that, of course you would improve it over time.
之后，随着时间的推移，你当然会改进它。
But I don’t like going into a problem and somebody says, “Hey, you know, it takes 74 days to do this today-” “…
但我不喜欢遇到问题时有人说，“嘿，你知道，今天需要 74 天才能完成这件事 -”“......
Right now.
现在。
And we can do it for you in 72 days.” You know, I’d rather strip it all back to zero-
我们可以在 72 天内为您完成。”你知道，我宁愿把一切都归零——
58:52 …
……
and say, “First of all, explain to me why 74 days in the first place.
并说：“首先，请向我解释一下为什么是 74 天。
And l- let’s note, let’s think about what’s possible today.
我-让我们注意，让我们想想今天可能发生什么。
And if I were to- to build it completely from scratch, you know, how long would it take?” Oftentimes, you’d be surprised.
如果我要——完全从头开始建造它，你知道，需要多长时间？”很多时候，你会感到惊讶。
It might come to six days.
可能需要六天。
Now, the rest of the six days, the 74, could be very well-reasoned and compromises, and, you know, cost reductions, and all kinds of different things.
现在，剩下的六天，即 74 天，可以进行非常充分的推理和妥协，而且，你知道，成本削减，以及各种不同的事情。
But at least you know what they are.
但至少你知道它们是什么。
And then now that you know that six days is possible, then the conversation from 74 to six, surprisingly much more effective.
现在您知道六天是可能的，然后将对话从 74 减到六天，出奇地有效得多。
In such incredibly complex systems that you’re working with, is simplicity sometimes a good heuristic to reach for?
在您正在使用的如此极其复杂的系统中，简单性有时是一个很好的启发式方法吗？
59:30 I mean, if I can just…
我的意思是，如果我可以……
I mean, the pod, the Vera Rubin pod that you announced is just incredible.
我的意思是，你宣布的 Vera Rubin 豆荚真是令人难以置信。
We’re talking about seven chips, seven chip types, five purpose-built rack types, 40 racks, 1.2 quadrillion transistors, nearly 20,000 NVIDIA dies, over 1,100 Rubin GPUs, 60 exaflops, 10 petabytes per second of scale bandwidth.
我们谈论的是七种芯片、七种芯片类型、五种专用机架类型、40 个机架、1.2 万亿个晶体管、近 20,000 个 NVIDIA 芯片、超过 1,100 个 Rubin GPU、60 exaflops、每秒 10 PB 的规模带宽。
That’s all just one…
这一切都只是一个……
That’s just one pod.
那只是一个吊舱。
That’s just one pod .
那只是一个吊舱。
Yeah, that’s just one pod.
是的，那只是一个吊舱。
I mean, in- …
我的意思是，在——……
1:00:07 so you have the…
所以你有……
And then even the NVL72 rack alone is 1.3 million components, 1300 chips, 4,000 pods crammed into a single 19-inch wide rack.
甚至仅 NVL72 机架就有 130 万个组件、1300 个芯片、4,000 个 pod 塞进一个 19 英寸宽的机架中。
And Lex, we’re probably gonna have to crank out about 200 of these pods a week, just to put it in perspective.
Lex，我们可能需要每周生产大约 200 个这样的 Pod，只是为了正确地看待它。
The amount of different components, I suppose simplicity is impossible, but is that a metric that you kind of reach for in trying to design things?
不同组件的数量，我认为简单是不可能的，但是这是您在尝试设计事物时所达到的指标吗？
You know, the phrase that I use most often is, we need things to be as complex as necessary, but as simple as possible.
你知道，我最常使用的一句话是，我们需要事情尽可能复杂，但又尽可能简单。
1:00:35 And so the question is, is all that complexity there necessary?
所以问题是，那么复杂有必要吗？
And we ought to test for that.
我们应该对此进行测试。
And we got to challenge that.
我们必须挑战这一点。
And then after that, everything else above it, you know, is gratuitous.
在那之后，你知道，上面的一切都是免费的。
But it’s still almost incredible.
但这仍然几乎令人难以置信。
1:00:56 Semiconductor industry broadly, but what NVIDIA is doing is some of the greatest engineering in history.
广义的半导体行业，但 NVIDIA 正在做的是一些历史上最伟大的工程。
So these systems are just truly, truly marvels of engineering.
所以这些系统是真正的工程奇迹。
It is the most complex computer the world has ever made.
它是世界上迄今为止制造的最复杂的计算机。
Yeah, the engineering teams, I mean- …
是的，工程团队，我的意思是——……
1:01:13 I don’t, it’s not a competition, but I don’t know.
我不知道，这不是比赛，但我不知道。
If it was like an Olympics of engineering teams, I mean, TSMC does incredible engineering.
如果这就像工程团队的奥运会，我的意思是，台积电的工程技术令人难以置信。
Like I said, ASML at every scale, but NVIDIA is gonna give them a run for their money.
就像我说的，ASML 在各个规模上都有，但 NVIDIA 会与他们展开竞争。
Just incredible, incredible teams.
令人难以置信，令人难以置信的团队。
Well, it’s gold medalists in every single, in every single sport, all assembled right here.
嗯，每一项运动的金牌得主都聚集在这里。
And have to work together.
并且必须共同努力。
1:01:33 And report directly to you.
并直接向您汇报。
This is wonderful.
这太棒了。
You recently traveled to China.
您最近去过中国。
So it’s interesting to ask you, China’s been incredibly successful in building up its technology sector.
所以很有趣地问你，中国在建设技术领域方面取得了令人难以置信的成功。
What do you understand about how China’s able to, over the past 10 years, build so many incredible world-class companies, world-class engineering teams, and just this technology ecosystem- …
您对中国如何在过去 10 年里建立如此多令人难以置信的世界级公司、世界级工程团队以及这个技术生态系统有何了解？
that produces so many incredible products?
生产出这么多令人难以置信的产品？
A whole bunch of reasons for…
一大堆理由……
1:02:05 Well, first of all, let’s start, let’s start with some facts.
好吧，首先，让我们开始，让我们从一些事实开始。
50% of the world’s AI researchers are Chinese, plus or minus, and they’re mostly in China still.
世界上 50% 的人工智能研究人员是中国人，无论是正数还是负数，而且他们大部分仍然在中国。
We have many of them here, but there’s amazing researchers still in China.
我们这里有很多这样的人，但中国仍然有很多令人惊叹的研究人员。
They—their tech industry showed up at precisely the right time.
他们——他们的科技产业恰好在正确的时间出现了。
At the time of the mobile cloud era, their way of contributing was software, and so this is a country’s incredible science and math really well-educated kids.
在移动云时代，他们的贡献方式是软件，所以这是一个国家令人难以置信的科学和数学受过良好教育的孩子。
Their tech industry was created during the era of software.
他们的科技产业是在软件时代创建的。
They’re very comfortable with modern software.
他们对现代软件非常满意。
China is not one giant economic country.
中国不是一个经济大国。
It’s got many provinces and cities with mayors all competing with each other.
很多省市的市长都是相互竞争的。
1:03:01 That’s the reason why there’s so many EV companies.
这就是为什么有这么多电动汽车公司的原因。
That’s the reason why there’s so many AI companies.
这就是为什么有这么多人工智能公司的原因。
That’s the reason why there’s so many—every company you could imagine, they all create some of them.
这就是为什么有这么多的原因——你能想象到的每家公司，他们都会创建其中的一些。
And, and as a result, they have insane competition internally.
结果，他们内部出现了疯狂的竞争。
And, you know, what remains is an incredible company.
而且，你知道，剩下的是一家令人难以置信的公司。
They also have a social culture where, where it’s family first, friends second, and company third.
他们还有一种社会文化，即家庭第一，朋友第二，公司第三。
And so the amount of conversation that goes back and forth between…
因此，双方之间来回的对话量......
They’re essentially open source all the time.
它们本质上一直都是开源的。
1:03:47 So the fact that they contribute more to open source is so sensible because they’re probably, “What are we protecting?” You know, my engineers, their brothers are in that company, their friends are in that company, and they’re all schoolmates.
因此，他们为开源做出更多贡献这一事实是如此明智，因为他们可能会问，“我们在保护什么？”你知道，我的工程师，他们的兄弟在那家公司，他们的朋友在那家公司，他们都是同学。
You know, the schoolmate concept.
你知道，同学的概念。
There’s a, you know, one schoolmate, you’re brother for life.
你知道，有一位同学，你们是终生的兄弟。
And and so they, they, they share knowledge very, very quickly.
所以他们，他们，他们非常非常快地分享知识。
And so there’s no sense keeping technology hidden.
因此，隐藏技术是没有意义的。
You might as well put it on open source.
您不妨将其开源。
And so the open source community then amplifies, accelerates the, the innovation process.
因此，开源社区放大、加速了创新过程。
So you get this rapid, incredible great talent, rapid innovation because of open source and just, you know, the nature of friends, and, and insane competition.
因此，由于开源以及朋友的本质，以及疯狂的竞争，你会获得这种快速、令人难以置信的优秀人才、快速创新。
1:04:35 Among the company, what emerges is incredible stuff.
在公司里，涌现出令人难以置信的东西。
And so this is the fastest innovating country in the world today, and this is something that has everything that, everything that I’ve just said is fundamental to just how the kids were grown, the fact that they have excellent education, the fact that they, parents want them to do well in school, the fact that they, their culture is that way.
所以这是当今世界上创新最快的国家，这里拥有我刚才所说的一切对于孩子们的成长至关重要的东西，他们接受了良好的教育，他们，父母希望他们在学校取得好成绩，他们的文化就是这样。
These are, you know, these are just the thing about their country, and they showed up at precisely the time when technology is going through that exponential.
你知道，这些都是他们国家的特色，而他们的出现正是在技术呈指数级增长的时候。
Plus culturally, it’s pretty cool to be an engineer.
另外，从文化角度来说，成为一名工程师非常酷。
1:05:09 It connects to all the components that you’re mentioning…
它连接到您提到的所有组件......
It’s a builder nation.
这是一个建设者国家。
It’s a builder nation.
这是一个建设者国家。
Yeah, it’s a builder nation.
是的，这是一个建设者国家。
1:05:19 Our country’s leaders, incredible, but they’re mostly lawyers.
我们国家的领导人，令人难以置信，但他们大多是律师。
Their country’s leaders—and because we’re, they’re trying to keep us safe, rule of law governing—their country was built out of poverty.
他们国家的领导人——因为我们是，他们正在努力保护我们的安全，实行法治——他们的国家是摆脱贫困而建立的。
And so most of their leaders are incredible engineers.
因此，他们的大多数领导者都是令人难以置信的工程师。
Some of the brightest minds.
一些最聪明的头脑。
To take a small tangent, because you mentioned open source, I have to go to Perplexity here, who you have been a fan of a long time.
稍微切题一下，因为你提到了开源，所以我必须在这里转到 Perplexity，你长期以来一直是它的粉丝。
Love it, yeah.
喜欢它，是的。
And thank you for releasing open source Nemotron 3 Super, which you can also use inside Perplexity to look stuff up.
感谢您发布开源 Nemotron 3 Super，您也可以在 Perplexity 中使用它来查找内容。
1:05:52 Now, which is 120 billion parameter open weight MoE model.
现在，这是1200亿参数的开放权重MoE模型。
What’s your vision with open source?
您对开源的愿景是什么？
So you mentioned China with, with DeepSeek and MiniMax, with all these companies really pushing forward the open source AI movement, and NVIDIA is really leading the way in close to state-of-the-art open source LLMs.
所以你提到了中国，DeepSeek 和 MiniMax，所有这些公司都在真正推动开源 AI 运动，而 NVIDIA 在接近最先进的开源 LLM 方面确实处于领先地位。
What’s your vision there?
您的愿景是什么？
First, if we’re gonna be a great AI computing company, we have to understand how AI models are evolving.
首先，如果我们想成为一家伟大的人工智能计算公司，我们就必须了解人工智能模型是如何演变的。
1:06:36 One of the things that I love about Nemotron 3 is it’s not just a pure transformer model, it’s transformer and SSMs.
我喜欢 Nemotron 3 的原因之一是它不仅仅是一个纯粹的变压器模型，它是变压器和 SSM。
And we were early in developing the, the conditional GANs, which, that progressive GANs, which led step-by-step to diffusion.
我们很早就开发了条件 GAN，渐进式 GAN，逐步导致扩散。
And so the fact that we’re doing basic research in model architecture and in different domains gives us visibility into, you know, what kind of computing systems would do a good job for future models.
因此，我们正在模型架构和不同领域进行基础研究，这一事实使我们能够了解什么样的计算系统可以为未来的模型做好工作。
And so it is part of our extreme co-design strategy.
因此，这是我们极端协同设计策略的一部分。
Second, I think we rightfully recognize that on the one hand, we want world-class models as products, and they should be proprietary.
其次，我认为我们正确地认识到，一方面，我们想要世界一流的模型作为产品，而且它们应该是专有的。
On the other hand, we also want AI to diffuse into every industry and every country, every researcher, every student.
另一方面，我们也希望人工智能扩散到每个行业、每个国家、每个研究人员、每个学生。
1:07:37 And if everything is proprietary, it’s hard to do research and it’s hard to innovate on top of, around, with.
如果一切都是专有的，那么就很难进行研究，也很难在其之上、周围进行创新。
And so…
所以……
Open source is fundamentally necessary for many industries to join the AI revolution.
开源对于许多行业加入人工智能革命来说是根本必要的。
NVIDIA has the scale and we have the motives—not only skills, scale, and motivation—to build and continue to build these AI models for as long as we shall live.
NVIDIA 拥有规模，我们也有动机（不仅仅是技能、规模和动力）来构建并继续构建这些 AI 模型，只要我们有生之年。
And so therefore, we ought to do that.
因此，我们应该这样做。
We can open up, we can activate every industry, every researcher, you know, every country to be able to join the AI revolution.
我们可以开放，我们可以激活每个行业、每个研究人员、每个国家都能够加入人工智能革命。
There’s the third reason, which is from that, to recognizing that AI is not just language.
第三个原因是认识到人工智能不仅仅是语言。
These AIs will likely use tools and models and sub-agents that were trained on other modalities of information.
这些人工智能可能会使用接受过其他信息模式培训的工具、模型和子代理。
1:08:39 Maybe it’s biology or chemistry or you know, laws of physics, or you know, fluids and thermodynamics, and not all of it is in language structure.
也许是生物学或化学，或者你知道，物理定律，或者你知道，流体和热力学，但并非所有这些都在语言结构中。
And so somebody has to go make sure that weather prediction, biology, AI, AI for biology, physical AI, all of that stuff stays, can be pushed to the limits and pushed to the frontier.
因此，必须有人去确保天气预报、生物学、人工智能、生物学人工智能、物理人工智能，所有这些东西都能保留下来，能够被推到极限并推到前沿。
We don’t build cars, but we wanna make sure every car company has access to great models.
我们不制造汽车，但我们希望确保每个汽车公司都能获得出色的车型。
We don’t discover drugs, but I wanna make sure that Lilly has the world’s best biology AI systems, so that they can go use it for discovering drugs.
我们不发现药物，但我想确保礼来公司拥有世界上最好的生物人工智能系统，以便他们可以用它来发现药物。
And so these three fundamental reasons, both in recognizing that AI is not just language, that AI is really broad, that we wanna engage everybody into the world of AI, and then also co-design of AI.
因此，这三个基本原因是认识到人工智能不仅仅是语言，人工智能确实很广泛，我们希望让每个人都进入人工智能的世界，然后也是人工智能的共同设计。
Well, I have to say, once again, thank you for open sourcing, really truly open sourcing Nemotron 3 and …
好吧，我不得不说，再次感谢你们的开源，真正的开源 Nemotron 3 和……
Yeah, I appreciate you were saying that.
是的，我很感激你这么说。
1:09:39 We open sourced the models, we open sourced the weights, we open sourced the data, we open sourced how we created it.
我们开源了模型，我们开源了权重，我们开源了数据，我们开源了我们如何创建它。
Yeah, it’s pretty amazing.
是的，这真是太神奇了。
It’s really incredible.
真是不可思议。
1:09:48 You’re originally from Taiwan and have a close relationship with TSMC.
您来自台湾，与台积电关系密切。
So I have to ask TSMC I think also is a legendary company in terms of the engineering teams, in terms of the incredible engineering work that they do.
所以我不得不问台积电，我认为就工程团队而言，就他们所做的令人难以置信的工程工作而言，它也是一家传奇公司。
What do you understand about TSMC culture and their approach that explains how they’re able to achieve this singular unmatched success in everything they’re doing with semiconductors?
您对台积电文化以及他们如何在半导体领域取得如此无与伦比的成功的方法有什么了解？
You know, first of all, the deepest misunderstanding about TSMC is that their technology is all they have.
要知道，首先，对台积电最深的误解是，他们的技术就是他们的全部。
1:10:19 That somehow they have a really great transistor, and if somebody shows up another transistor, game over.
不知何故，他们有一个非常出色的晶体管，如果有人展示另一个晶体管，游戏就结束了。
It’s the technology and, of course, you know, I don’t mean just the transistor, the metallization systems, the packaging, the 3D packaging, the silicon photonics, the, you know, all of the technology that they have.
当然，这是技术，我指的不仅仅是晶体管、金属化系统、封装、3D 封装、硅光子学，以及他们拥有的所有技术。
That technology is really what makes the company special.
这项技术确实使该公司与众不同。
Their technology makes the company special.
他们的技术使公司与众不同。
1:10:59 But their ability to orchestrate the demands, the dynamic demands of hundreds of companies in the world as they’re moving up, shifting out, you know, increasing, decreasing, pushing out, pulling in, changing from customer to customer, wafer starting, wafer stopping, emergency wafer starts, you know, all of this dynamics of the world’s complexity as the world is shape-shifting all the time, and somehow they’re running a factory with high throughput, high yields, really great costs, excellent customer service.
但他们有能力协调需求，世界上数百家公司的动态需求，因为它们正在向上移动，移出，你知道，增加，减少，推出，拉入，从客户到客户的变化，晶圆启动，晶圆停止，紧急晶圆启动，你知道，随着世界一直在发生变化，世界的复杂性的所有这些动态，不知何故，他们正在经营一个高吞吐量，高产量，真正巨大的成本，卓越的客户服务的工厂。
They take their promises seriously.
他们认真对待自己的承诺。
1:11:49 They, when your wafer—because they know that they’re helping you run your company—when the wafers were promised to show up, the wafers show up, you know, so that you could run your company appropriately.
他们，当你的晶圆——因为他们知道他们正在帮助你经营你的公司——当晶圆被承诺出现时，晶圆就会出现，你知道，这样你就可以适当地经营你的公司。
And so their system, their manufacturing system is completely miraculous, I would say.
所以我想说，他们的系统、他们的制造系统完全是奇迹。
Then the second thing is their culture.
那么第二件事就是他们的文化。
This culture is simultaneously technology focused on one hand, advancing technology; simultaneously customer service oriented on the other hand.
这种文化一方面关注技术，推动技术进步；另一方面关注技术。另一方面同时以客户服务为导向。
A lot of companies are very customer service oriented, but they’re not very technology excellent.
很多公司都非常注重客户服务，但他们的技术并不是很出色。
They’re not at the bleeding edge of technology.
他们并不处于技术的最前沿。
1:12:27 There are a lot of companies who are tech, at the bleeding edge of technology, but they’re not the best customer service oriented company.
有很多科技公司，处于技术前沿，但它们并不是最好的以客户服务为导向的公司。
And so it just depends on somehow they’ve, they’ve balanced these two and they’re world-class at both.
所以这只是取决于他们以某种方式平衡了这两者，而且他们在这两方面都是世界一流的。
And then probably the third thing is the technology that I most value in them that they created this, you know, this, this intangible called trust.
第三件事可能是我最看重他们的技术，他们创造了这个，你知道，这个，这个无形的东西，叫做信任。
I trust them to put my company on top of them.
我相信他们会把我的公司置于他们之上。
That’s a very big deal.
这是一件大事。
When they trust, I mean, there’s a really close relationship there that you’ve established, and that trust is established based on many years of performance, but there’s human relationships involved there as well.
我的意思是，当他们信任时，你已经建立了一种非常密切的关系，这种信任是基于多年的表现而建立的，但也涉及人际关系。
Three decades, I don’t know how many tens, hundreds of billions of dollars of business we’ve done through them, and we don’t have a contract.
三十年来，我不知道我们通过他们做了多少数百、数千亿美元的业务，而且我们没有合同。
1:13:05 That’s pretty great.
那真是太棒了。
Amazing.
惊人的。
1:13:15 Okay, there’s this story …
好吧，有这样一个故事……
…
……
That in 2013, the founders of TSMC, Morris Chang offered you the chance to become TSMC’s chief executive and you said you already had a job.
2013年，台积电创始人张忠谋给你提供了担任台积电首席执行官的机会，你说你已经找到工作了。
Is this story true?
这个故事是真的吗？
Story is true.
故事是真的。
1:13:30 I didn’t, I didn’t dismiss it.
我没有，我没有拒绝它。
But I was deeply honored and, and of course, I knew then as I know now, TSMC is one of the most consequential companies in history.
但我深感荣幸，当然，我当时就知道，正如我现在所知，台积电是历史上最有影响力的公司之一。
And Morris is one of the highest regarded executives and business and personal friend that I’ve had in my life.
莫里斯是我一生中最受尊敬的高管、商业和私人朋友之一。
And, for him to ask, I was humbled and really honored.
而且，对于他提出这样的问题，我感到很谦卑，也很荣幸。
But the work that I’m doing here is really important, and I’ve seen, you know, in my mind’s eye, what NVIDIA was going to be and what the impact that we could have.
但我在这里所做的工作非常重要，而且我已经在脑海中看到了 NVIDIA 的未来以及我们可能产生的影响。
And it was really important work.
这是非常重要的工作。
And it’s my responsibility, you know, my sole responsibility to make this happen.
你知道，这是我的责任，我唯一的责任就是让这一切发生。
And so I declined it, not because it wasn’t an incredible offer.
所以我拒绝了，并不是因为这不是一个令人难以置信的提议。
It’s an unbelievable offer, but I simply couldn’t take it.
这是一个令人难以置信的提议，但我就是无法接受。
I think NVIDIA, both NVIDIA and TSMC are two of the greatest companies in the history of human civilization.
我觉得NVIDIA，英伟达和台积电都是人类文明史上最伟大的两家公司。
1:14:38 And running either one, I’m sure, is an incredibly complicated effort and takes…
我确信，运行任何一个都是一项极其复杂的工作，并且需要……
You have to truly be all in.
你必须真正全力以赴。
Everybody at every scale, not just at the CEO level.
各个级别的每个人，而不仅仅是首席执行官级别的人。
Everybody is really truly all in-
每个人都真的全力以赴——
Yeah.
是的。
1:14:57 Yeah, no doubt.
是的，毫无疑问。
…
……
1:14:59 To, to accomplish this kind of complexity.
到，完成这种复杂性。
So now I can help both companies.
所以现在我可以帮助两家公司。
Exactly.
确切地。
1:15:02 So NVIDIA is now the most valuable company in the world.
所以NVIDIA现在是世界上最有价值的公司。
I have to ask, what is the NVIDIA’s biggest moat, as the folks in the tech sector say?
我不得不问，正如科技界人士所说，NVIDIA最大的护城河是什么？
The edge you have that protects you from the competition.
您拥有的优势可以保护您免受竞争。
Our single most important property as a company is the install base of our computing platform.
作为一家公司，我们最重要的资产是计算平台的安装基础。
1:15:20 Our single most important thing today is the install base of CUDA.
今天我们最重要的事情是 CUDA 的安装基础。
Now, the reason why 20 years ago, of course, there was no install base.
当然，20年前的原因是没有安装基础。
But what makes…
但是什么让……
And if somebody came up with a GUDA or TUDA, it wouldn’t make any difference at all.
如果有人提出 GUDA 或 TUDA，那根本没有任何区别。
And the reason for that is because it’s never been just about the technology.
其原因是因为它从来不仅仅是技术的问题。
The technology, of course, was incredible, visionary.
当然，这项技术是令人难以置信的、富有远见的。
But it’s the fact that the company was dedicated to it, stuck with it, expanded its reach.
但事实是，该公司致力于它，坚持下去，扩大了它的影响力。
It wasn’t three people that made CUDA successful.
CUDA 的成功并不是三个人的努力。
It was 43,000 people that made CUDA successful.
正是 43,000 人让 CUDA 取得了成功。
1:16:17 And the several million developers that believed in us that trusted that we were going to continue to make CUDA 1, 2, 3, 13, that they decided to port and dedicate their software on top of it, their mountain of software on top of it.
数百万相信我们的开发人员相信我们将继续开发 CUDA 1、2、3、13，他们决定将他们的软件移植并奉献给它，他们的软件山也建立在它之上。
And so the install base is the number one most important advantage.
因此，安装基础是第一个最重要的优势。
That install base, when you amplify it with the velocity of our execution at the scale that we’re talking about, no company in history had ever built systems of this complexity, period.
当你以我们所说的规模的执行速度来扩大安装基础时，历史上没有一家公司曾经构建过如此复杂的系统。
And then to build it once a year is impossible.
然后每年建造一次是不可能的。
And that velocity combined with the install base, in the developer’s mind, you just go now, take the developer’s mind.
在开发人员的心目中，这种速度与安装基础相结合，你现在就走，抓住开发人员的心。
From the developer’s perspective, if I support CUDA, tomorrow it’ll be 10 times better.
从开发者的角度来看，如果我支持CUDA，明天会好10倍。
I just have to wait six months on average.
我平均只需要等待六个月。
1:17:16 Not only that, if I develop it on CUDA, I reach a few hundred million people, computers.
不仅如此，如果我在 CUDA 上开发它，我会接触到数亿人和计算机。
I’m in every cloud, I’m in every computer company, I’m in every single industry, I’m in every single country.
我在每一个云中，我在每一个计算机公司，我在每一个行业，我在每一个国家。
So if I create an open source package and I put it on CUDA first, I get these both attributes simultaneously.
因此，如果我创建一个开源包并首先将其放在 CUDA 上，我就会同时获得这两个属性。
And not only that, I trust 100% that NVIDIA is going to keep CUDA around and maintain it and improve it and keep optimizing the libraries for as long as they shall live.
不仅如此，我 100% 相信 NVIDIA 将保留 CUDA、维护它、改进它，并在库存在期间不断优化它们。
You could take that to the bank, and that last part, trust.
你可以把它带到银行，最后一部分，信托。
You put all that stuff together, if I were a developer today, I would target CUDA first.
把所有这些东西放在一起，如果我今天是一名开发人员，我会首先瞄准 CUDA。
I would target CUDA most.
我最看重的是 CUDA。
And that’s the reason that I think in the final analysis is our first, that’s even our first-
这就是为什么我认为归根结底是我们的第一次，甚至是我们的第一次——
1:18:16 …
……
core advantage.
核心优势。
Our second one is our ecosystem.
我们的第二个是我们的生态系统。
The fact that we vertically integrated this incredibly complex system, but we integrate it horizontally into every single company’s computers.
事实上，我们垂直集成了这个极其复杂的系统，但我们将其水平集成到每家公司的计算机中。
We’re into Google Cloud, we’re into Amazon, we’re in Azure.
我们进入了谷歌云、亚马逊、Azure。
You know, we’re ramping up AWS like crazy right now.
您知道，我们现在正在疯狂地扩展 AWS。
We’re in new companies like CoreWeave and Nscale.
我们在 CoreWeave 和 Nscale 等新公司工作。
We’re in supercomputers at Lilly.
我们在礼来公司使用超级计算机。
We’re in enterprise computers.
我们在企业计算机领域。
We’re at the edge in radio base stations.
我们处于无线电基站的边缘。
You know, I mean, it’s just crazy.
你知道，我的意思是，这太疯狂了。
One architecture is in all these different systems.
所有这些不同的系统都采用一种架构。
We’re in cars, we’re in robots, we’re in satellites, we’re out in space.
我们在汽车里，我们在机器人里，我们在卫星里，我们在太空里。
And so the fact that you have this one architecture and the ecosystem is so broad, it basically covers every single industry in the world.
事实上，你拥有这样一个架构和生态系统是如此广泛，它基本上涵盖了世界上的每一个行业。
Well, how does the CUDA install base evolve into the future with AI factories as a moat?
那么，以 AI 工厂为护城河的 CUDA 安装基础将如何演进到未来呢？
1:19:03 What do you…
你做什么…
Do you think it’s possible that NVIDIA of the future is all about the AI factory?
您认为 NVIDIA 的未来有可能就是 AI 工厂吗？
Well, the unit of computing used to be GPU to us.
嗯，过去我们的计算单位是 GPU。
1:19:16 Then it became a computer, then it became a cluster.
然后它变成了一台计算机，然后它变成了一个集群。
Now it’s an entire AI factory.
现在它是一个完整的人工智能工厂。
When I see a computer, when I see what NVIDIA builds, in the old days, I would, you know, I visualize the chip.
过去，当我看到计算机时，当我看到 NVIDIA 构建的产品时，我会想象芯片。
And then when I announced the new product, new generation, like, “Ladies and gentlemen, we’re announcing Ampere today,” I’d pick up the chip.
然后，当我宣布新一代新产品时，就像“女士们先生们，我们今天宣布 Ampere”一样，我会拿起芯片。
That was my mental model- …
这就是我的思维模式——……
of what I was building.
我正在建造的东西。
Today, I wouldn’t…
今天，我不会……
Picking up the chip is kind of still adorable.
拿起芯片还是很可爱的。
1:19:47 But it’s adorable.
但它很可爱。
It’s not my mental model of what I’m doing.
这不是我正在做的事情的心理模型。
My mental model is this giant gigawatt thing that has power generations connected to the grid.
我的思维模型是这个巨大的千兆瓦级物体，其发电装置连接到电网。
It’s got cooling systems and networking of incredible monstrosity, you know.
你知道，它的冷却系统和网络令人难以置信。
10,000 people are in there trying to install it, hundreds of networking engineers in there, thousands of engineers behind it trying to power it up.
一万人在那里试图安装它，数百名网络工程师在那里，数千名工程师试图为其供电。
You know, powering up one of those factories, as you know, it’s not somebody going, “It’s on now.” It takes thousands of people to bring it up.
你知道，为其中一家工厂供电，正如你所知，并不是有人说，“现在就开始了。”需要成千上万的人才能提出来。
So mentally, you’re actually…
所以从心理上来说，你实际上……
1:20:22 When you’re thinking about a single unit of compute, you’re like literally, when you go to bed at night, you’re thinking now about a collection of racks, so pods, not individual chips.
当您考虑单个计算单元时，您就像字面意思一样，当您晚上睡觉时，您现在正在考虑一组机架，即 Pod，而不是单个芯片。
Entire infrastructure.
整个基础设施。
1:20:33 And I’m hoping my next click is when I’m thinking about building computers, it’s planetary scale.
我希望我的下一次点击是当我考虑建造计算机时，它是全球规模的。
That’ll be the next click.
这将是下一次点击。
Well, what do you think about the space angle that Elon has talked about, doing compute in space for solving some of the…
好吧，你对埃隆谈到的空间角度有什么看法，在空间中进行计算来解决一些......
1:20:42 It makes some of the energy issues in terms of scaling energy easier.
它使扩展能源方面的一些能源问题变得更加容易。
Cooling issues is not easy.
散热问题并不容易解决。
1:20:56 Yeah.
是的。
Cooling.
冷却。
1:20:58 Well, there’s a large number of engineering complexities involved with that.
嗯，这涉及到大量的工程复杂性。
So what…
那又怎样……
You know, NVIDIA has also announced that you’re already thinking about that.
你知道，NVIDIA 也宣布你已经在考虑这个问题了。
Yeah, we’re already there.
是的，我们已经在那里了。
1:21:09 NVIDIA GPUs are the first GPUs in space.
NVIDIA GPU 是第一批进入太空的 GPU。
And I didn’t realize it, it was so interesting to…
我没有意识到，这太有趣了……
I would have declared it maybe.
也许我会宣布这一点。
We’re in space.
我们在太空中。
You know, little, little astronaut suit on one of our GPUs.
您知道，我们的 GPU 上有一套小小的宇航员服。
But we’ve been in space.
但我们已经在太空了。
It’s the right place to do a lot of imaging.
这是进行大量成像的正确位置。
1:21:32 You know, because those satellites have really high resolution imaging systems, and they’re sweeping the Earth, you know, continuously now.
你知道，因为这些卫星拥有非常高分辨率的成像系统，而且它们现在正在连续地扫过地球。
And you want, you know, centimeter scale imaging that is done continuously for the world, so that, you know, you’ll basically have real time telemetry of everything.
你想要的是，你知道，对世界连续进行厘米级成像，这样，你知道，你基本上就能对一切进行实时遥测。
You don’t wanna beam that back down to Earth.
你不想将其传回地球。
It’s just, you know, petabytes and petabytes of data.
你知道，这只是数拍字节的数据。
You gotta just do AI right there at the edge, throw away everything you don’t need, you’ve seen before, didn’t change, and then just keep the stuff that you need.
你必须在边缘进行人工智能，扔掉所有你不需要的、你以前见过的、没有改变的东西，然后只保留你需要的东西。
And so AI had to be done at the edge.
因此人工智能必须在边缘完成。
Obviously we have 24/7 solar, if we put it at the polars.
显然，如果我们把它放在两极的话，我们就有 24/7 的太阳能。
And but, you know, there’s no conduction, no convection.
但是，你知道，没有传导，没有对流。
1:22:23 And so, you know, you’re pretty much just radiation.
所以，你知道，你几乎只是辐射。
And but, you know, space is big.
但是，你知道，空间很大。
I guess, you know, we’re just gonna put big, giant radiators out there.
我想，你知道，我们只是要把巨大的散热器放在那里。
How crazy of an idea do you think it is?
你认为这个想法有多疯狂？
1:22:32 Like is this five years out, 10 years out, 20 years out?
比如说五年后、十年后、二十年后吗？
So we’re talking about blockers for AI scaling.
所以我们正在讨论人工智能扩展的阻碍因素。
You know, I’m just so much more practical.
你知道，我只是比较实际而已。
1:22:41 I look for where my next, next bucket of opportunities are first.
我首先寻找我的下一个、下一个机会在哪里。
Meanwhile, I’m cultivating space.
与此同时，我正在培养空间。
And so I send, I send engineers to go work on the problem.
所以我派工程师去解决这个问题。
We’re starting to…
我们开始……
We’re learning a lot about it.
我们对此了解了很多。
How do we deal with radiation?
我们如何应对辐射？
How do we deal with degrading performance?
我们如何应对性能下降？
How do we deal with a continuous testing and attestation of defects?
我们如何处理持续的测试和缺陷证明？
And you know, how do we deal with redundancy?
你知道，我们如何处理冗余问题？
And how do we degrade gracefully and things like that?
我们如何优雅地降级之类的事情？
And so we could do a…
所以我们可以做一个……
What about software?
软件呢？
How do you think about software and redundancy and performance out in space?
您如何看待太空中的软件、冗余和性能？
1:23:24 Make it so that the computer never breaks, it just gets slower, you know.
让计算机永远不会崩溃，它只会变得更慢，你知道。
And I…
而我……
So we could start doing a lot of engineering exploration upfront.
因此我们可以预先开始进行大量的工程探索。
But in the meantime, my favorite answer is eliminate waste.
但与此同时，我最喜欢的答案是消除浪费。
You know, we’ve got all that idle power, I want to evacuate it as fast as possible.
你知道，我们有所有闲置电力，我想尽快将其疏散。
Yeah.
是的。
1:23:47 There, there…
那里，那里……
Yeah, there’s a lot of low-hanging fruit here on Earth- …
是的，地球上有很多唾手可得的果实——……
That we can utilize for the AI scaling.
我们可以利用它来进行人工智能扩展。
Quick pause.
快速暂停。
Quick 30-second thank you to our sponsors.
用 30 秒的时间快速感谢我们的赞助商。
Check them out in the description.
在描述中查看它们。
It really is the best way to support this podcast.
这确实是支持这个播客的最佳方式。
Go to lexfridman.com/sponsors.
请访问 lexfridman.com/sponsors。
We got Perplexity for curiosity-driven knowledge exploration, Shopify for selling stuff online, LMNT for electrolytes, Fin for customer service AI agents, and Quo for a phone system, like calls, texts, contacts, for your business.
我们有用于好奇心驱动的知识探索的 Perplexity、用于在线销售商品的 Shopify、用于电解质的 LMNT、用于客户服务人工智能代理的 Fin，以及用于电话系统（如电话、短信、联系人）的 Quo，用于您的业务。
Choose wisely, my friends.
明智地选择，我的朋友们。
And now, back to my conversation with Jensen Huang.
现在，回到我与黄仁勋的对话。
Do you think NVIDIA may be worth 10 trillion at some point?
您认为 NVIDIA 可能在某个时候价值 10 万亿美元吗？
Let’s, let’s ask it this way.
咱们这样问吧。
What does the future of the world look like where that’s true?
如果这是真的，世界的未来会是什么样子？
I think that NVIDIA’s growth is extremely likely, and in my mind, inevitable.
我认为 NVIDIA 的增长极有可能，而且在我看来，这是不可避免的。
1:24:45 And let me explain why.
让我解释一下原因。
We’re the largest computer company in history.
我们是历史上最大的计算机公司。
That alone should beg the question, why?
仅凭这一点就应该提出问题，为什么？
And the reason of course…
当然，原因是……
Two reasons.
有两个原因。
First, two foundational technical reasons.
首先，有两个基本的技术原因。
The first reason is that computing went from being a retrieval-based, file retrieval system.
第一个原因是计算不再是基于检索的文件检索系统。
Almost everything is a file…
几乎所有东西都是文件......
We pre-write something, we pre-record something.
我们预先写一些东西，我们预先录制一些东西。
You know, we draw something, we put it on the web, we put it in a file.
你知道，我们画一些东西，我们把它放在网上，我们把它放在一个文件中。
And we use a recommender system, some smart filter, to figure out what to retrieve for you.
我们使用推荐系统和一些智能过滤器来确定要为您检索什么。
And so we were a pre-recording, human pre-recording, and file retrieving system.
所以我们是一个预录音、人工预录音和文件检索系统。
That’s what a computer is, largely.
很大程度上，这就是计算机。
1:25:39 To now, AI computers are contextually aware, which means that it has to process and generate tokens in real time.
到目前为止，人工智能计算机具有上下文感知能力，这意味着它必须实时处理和生成令牌。
So we went from a retrieval-based computing system to a generative-based computing system.
因此，我们从基于检索的计算系统转变为基于生成的计算系统。
We’re gonna need a lot more processing in this new world than in the old world.
与旧世界相比，我们在这个新世界中需要更多的处理。
We need a lot of storage in the old world.
我们在旧世界需要大量的存储空间。
We need a lot of computation in this new world.
在这个新世界中我们需要大量的计算。
And so that’s the first part of it.
这就是第一部分。
We fundamentally changed computing and the way how computing is done.
我们从根本上改变了计算以及计算的完成方式。
The only thing that would cause it to go back……
唯一能让它回去的就是……
1:26:15 is if this way of computation, this way of computing generating information that’s contextually relevant, situationally aware, that is grounded on new insight before it generates information, this computation-intensive way of doing computing would only go back if it’s not effective.
如果这种计算方式，这种计算方式生成上下文相关的、情境感知的信息，在生成信息之前基于新的见解，那么这种计算密集型的计算方式只有在无效时才会返回。
So if…
所以如果...
For the last 10, 15 years while working on deep learning, if at any single moment I would have come to the conclusion that, “You know what?
在过去 10、15 年的深度学习研究中，如果有任何一个时刻我会得出这样的结论：“你知道吗？
This is not gonna work out.
这不会有结果。
I think this is a dead end.” Or, “It’s not gonna scale, it’s not gonna solve this modality, not gonna be used in this application.” Then, of course, I would feel very differently about it, but I think the last five years has given me more confidence than the previous ten years.
我认为这是一个死胡同。”或者，“它不会扩展，不会解决这种模式，也不会在这个应用程序中使用。”当然，我对此会有非常不同的感觉，但我认为过去五年比过去十年给了我更多信心。
1:27:04 The second idea is computers, because it was a storage system, it was largely a warehouse.
第二个想法是计算机，因为它是一个存储系统，它很大程度上是一个仓库。
We’re now building factories.
我们现在正在建设工厂。
Warehouses don’t make much money.
仓库赚不了多少钱。
Factories directly correlates with the company’s revenues.
工厂与公司的收入直接相关。
And so, the computer did two things.
因此，计算机做了两件事。
Not only did it change the way it did it, its purpose in the world changed.
它不仅改变了它的工作方式，也改变了它在世界上的目的。
It’s no longer a computer, it’s a factory.
它不再是一台计算机，而是一座工厂。
It’s a factory, it’s used for generation of revenues.
这是一家工厂，用于创收。
We’re now seeing not only is this factory generating products, commodities that people want to consume, we’re seeing that the commodities are so interesting, so valuable to so many different audiences that the tokens are starting to segment, like iPhones.
我们现在不仅看到这家工厂生产人们想要消费的产品和商品，我们还看到这些商品对许多不同的受众来说非常有趣、非常有价值，以至于代币开始细分，就像 iPhone 一样。
You have free tokens, you have premium tokens, and you have several tokens in the middle.
你有免费代币，你有高级代币，还有中间的几个代币。
1:28:10 And so intelligence, as it turns out, you know, it’s a scalable product.
事实证明，智能是一种可扩展的产品。
There’s extremely high intelligence products, tokens that you could…
有极高智能的产品、代币，你可以……
that are used for specialized things, people be willing to pay.
那些用于专门的东西，人们愿意付费。
You know, the idea that somebody’s willing to pay $1000 per million tokens is just around the corner.
你知道，有人愿意为每 100 万代币支付 1000 美元的想法即将到来。
It’s not if, it’s only when.
问题不在于是否，而在于何时。
And so, so now we’re seeing that the commodity that this factory makes is actually valuable, and is revenue generating and profit generating.
所以，现在我们看到这家工厂生产的商品实际上很有价值，并且可以产生收入和利润。
Now the question is how many of these factories does the world need?
现在的问题是世界需要多少这样的工厂？
How many tokens does the world need?
世界需要多少代币？
And how much is society willing to pay for these tokens?
社会愿意为这些代币支付多少钱？
And what would happen to the world’s economy if the productivity were to improve so substantially?
如果生产力大幅提高，世界经济会发生什么？
What would happen…
会发生什么……
1:29:08 Are we, are we gonna discover new drugs, new products, new services?
我们要发现新药、新产品、新服务吗？
And so when you take these things in combination, I am absolutely certain that the world’s GDP is going to accelerate in growth.
因此，当你综合考虑这些因素时，我绝对确信世界 GDP 将加速增长。
I’m absolutely certain the percentage of that GDP that will be used for computation will be 100 times more than the past—mm-hmm—because it’s no longer a storage unit.
我绝对确定用于计算的 GDP 百分比将是过去的 100 倍——嗯——因为它不再是一个存储单位。
It’s a product generation unit.
它是一个产品生成单元。
And so when you look at it in that context and then you back into what is NVIDIA’s, what does NVIDIA sh—what does NVIDIA do and how much of that new economics, new industry would we have to benefit t—to address, I think we’re gonna be a lot, lot bigger.
因此，当你在这种背景下审视这个问题，然后再回到 NVIDIA 是什么、NVIDIA 做什么——NVIDIA 做了什么以及我们必须从新经济、新行业中受益多少——为了解决这个问题，我认为我们会变得更大、更大。
1:29:58 And then the rest of it, to me, is: is it possible for NVIDIA to be a, you know, $3 trillion revenue company in the near future?
对我来说，剩下的问题是：NVIDIA 有可能在不久的将来成为一家营收 3 万亿美元的公司吗？
The answer is, of course, yes.
答案当然是肯定的。
And the reason for that is because it’s not limited by any physical limits.
原因是它不受任何物理限制的限制。
There’s nothing that I see that says, you know, gosh $3 trillion is not possible.
我没有看到任何内容表明，天哪，3 万亿美元是不可能的。
And as it turns out, NVIDIA’s supply chain is—the burden is shared by 200 companies.
事实证明，NVIDIA 的供应链——负担由 200 家公司分担。
And the fact that we scale out on the backs of, with the partnership of this ecosystem, the question is: do we have the energy to do so?
事实上，我们在这个生态系统的合作下进行扩展，问题是：我们有精力这样做吗？
And surely we will have the energy to do so.
我们肯定有能力这样做。
And so all of these things combined, that number is just a number, you know?
所以所有这些东西加在一起，这个数字只是一个数字，你知道吗？
1:30:51 And I still remember, NVIDIA was a…
我还记得，NVIDIA 是……
the first time we crossed a billion dollars, I was reminded of a CEO who told me, “You know, Jensen, it’s theoretically impossible for a fabless semiconductor company to exceed a billion dollars.” And I won’t bore you with why, but of course it’s illogical and there’s a lot of evidence we’re not.
当我们第一次突破 10 亿美元时，我想起一位首席执行官告诉我：“詹森，你知道，无晶圆厂半导体公司理论上不可能突破 10 亿美元。”我不会让你厌烦为什么，但这当然是不合逻辑的，而且有很多证据表明我们并非如此。
And then somebody told me, “You know, Jensen, you’ll never be more than $25 billion because of some other company.” Somebody told me that, “You’ll never be, you know, because…” And so those aren’t principled, first principled reason thinking.
然后有人告诉我，“你知道，Jensen，因为其他公司，你的市值永远不会超过 250 亿美元。”有人告诉我，“你知道，你永远不会成为这样的人，因为……”所以这些都不是有原则的、第一原则的理性思维。
And the simple way to think about that is what is it that we make and how large is the opportunity that we can create?
思考这个问题的简单方法是我们制造了什么以及我们可以创造多大的机会？
1:31:42 Now, NVIDIA is not in the market share business.
现在，NVIDIA已经不涉足市场份额业务了。
Almost everything that I just talked about don’t exist.
我刚才谈到的一切几乎都不存在。
That’s the part that’s hard.
这就是困难的部分。
You know, if NVIDIA was a $10 billion company trying to take NVIDIA’s share, then it’s easy to see for shareholders that, oh, yeah, if they could just take 10% share, they could be this much larger.
你知道，如果 NVIDIA 是一家价值 100 亿美元的公司，试图夺取 NVIDIA 的份额，那么股东很容易看到，哦，是的，如果他们能拿走 10% 的股份，他们的规模就会大得多。
But it’s hard for people to imagine how large we could be because there’s nobody I could take share from.
但人们很难想象我们能发展到多大，因为没有人可以让我分享。
You know?
你知道？
And so I think that that’s one of the challenges for the world is the imagination of the future.
所以我认为世界面临的挑战之一就是未来的想象力。
But I got plenty of time, and I’ll keep reasoning about it, and I’ll keep talking about it, and every single GTC will become more and more real.
但我有足够的时间，我会不断地推理它，我会不断地谈论它，每一个GTC都会变得越来越真实。
1:32:27 You know, and then more and more people will talk about it, and one of these days, you know, we’ll get there.
你知道，然后越来越多的人会谈论它，总有一天，你知道，我们会实现这一目标。
But I’m 100% we’ll get there.
但我百分百相信我们会到达那里。
Yeah, this view of you know, token factories essentially, this token per second per watt, and every token having value.
是的，你知道的这种观点，本质上是令牌工厂，每秒每瓦特的令牌，并且每个令牌都有价值。
1:32:34 Like it’s an actual thing that brings value, and it brings different kinds of value, different amounts of value to different people with value.
就好像它是一个真实存在的东西，能够带来价值，而且它给不同的人带来不同种类的价值、不同数量的价值。
That’s the actual product—it really could be loosely thought of as the token.
这就是实际的产品——它确实可以被宽松地视为代币。
And so you have a bunch of token factories.
所以你就有了一堆代币工厂。
And then it’s very easy, first principles, to imagine a future, given all the potential things that AI can solve, that you’re going to need an exponential number more of token factories.
然后很容易，根据第一原则，想象一个未来，考虑到人工智能可以解决的所有潜在问题，你将需要指数级数量的代币工厂。
Yeah.
是的。
1:33:05 And what’s really interesting, the reason why I was so excited about it, the iPhone of tokens arrived.
真正有趣的是，我如此兴奋的原因是代币 iPhone 的到来。
What do you call it?
你怎么称呼它？
1:33:11 Wait, are you saying OpenClaw’s iPhone?
等等，你是说 OpenClaw 的 iPhone 吗？
Yeah.
是的。
That’s interesting.
那很有意思。
Agents.
代理商。
Yeah, agents.
是啊，特工们。
1:33:16 True.
真的。
Agents in general.
代理一般。
1:33:18 The iPhone of tokens arrived.
代币 iPhone 到来了。
It is the fastest-growing application in history.
这是历史上增长最快的应用程序。
It went straight up.
它直线上升。
Went straight up.
径直走了上去。
That says something.
这说明了一些事情。
Yep, there’s no question OpenClaw is the iPhone of tokens.
是的，毫无疑问 OpenClaw 是代币界的 iPhone。
Is there something truly, as you know, something truly special happening from about December, where people have really woke up to the power of Claude Code of Codex, of OpenClaw?
正如你所知，从 12 月份开始，人们真的意识到了 Codex 的 Claude Code 和 OpenClaw 的力量吗？
1:33:31 I mean, I’m embarrassed to admit that on the way here in the airport, I’ve…
我的意思是，我很尴尬地承认，在去机场的路上，我……
It’s the first time I’ve done this in public.
这是我第一次在公共场合这样做。
I was programming, quote unquote, by talking to my laptop.
我通过与我的笔记本电脑交谈来编程。
Yeah, exactly.
是的，完全正确。
And I was embarrassed because I was pretending like I’m talking to a human colleague.
我很尴尬，因为我假装我正在和一个人类同事说话。
1:33:59 I’m not sure how I feel about the future where everybody- …
我不确定我对未来有何感想，每个人——……
is walking around talking to their AI, but it’s such an efficient way to get stuff done.
正在四处走动与他们的人工智能交谈，但这是一种完成工作的有效方式。
And it’s more likely that your AI is bothering you all the time.
而且更有可能的是，你的人工智能一直在困扰你。
1:34:13 And the reason for that is because it’s getting stuff done so fast.
原因是它完成工作的速度太快了。
It’s reporting back to you, “I got that done.” “You know, what do you want me to do next?” You know, it…
它会向你报告：“我已经完成了。” “你知道，接下来你想让我做什么吗？”你知道，它……
That’s the part that I think most people don’t realize is the person who’s gonna be chatting with them, texting them most, is their, is their claws or lobster.
我认为大多数人都没有意识到，那个会与他们聊天、给他们发短信最多的人是他们的，是他们的爪子或龙虾。
What an incredible future.
多么令人难以置信的未来。
1:34:37 I read that you attribute a lot of your success to your ability to work harder than anyone and withstand more suffering than anyone.
我读到，您将自己的成功很大程度上归功于您比任何人都更加努力工作并承受比任何人更多痛苦的能力。
So we can list many of the things that entails.
所以我们可以列出许多需要做的事情。
I mean, dealing with failure, the cost and engineering problems we’ve talked about.
我的意思是，处理我们讨论过的失败、成本和工程问题。
The human problems, uncertainty, responsibility, exhaustion, embarrassment, the near-death company moments that you’ve mentioned but also the pressure.
你提到的人类问题、不确定性、责任、疲惫、尴尬、濒临死亡的公司时刻，还有压力。
Now, as the CEO of this company that economies and nations strategize around, plan their financial allocations around, plan their AI infrastructure around, how do you deal with this much pressure?
现在，作为这家经济体和国家制定战略、规划财政分配、规划人工智能基础设施的公司的首席执行官，你如何应对如此大的压力？
What gives you strength, given how many nations and peoples depend on you?
鉴于有多少国家和人民依赖你，是什么给了你力量？
I’m conscious about the fact that NVIDIA’s success is very important to the United States.
我意识到 NVIDIA 的成功对美国非常重要。
1:35:38 We generate enormous amounts of tax revenues.
我们创造了大量的税收收入。
We established technology leadership for our nation.
我们为国家确立了技术领先地位。
Technology leadership is important for national security.
技术领先对于国家安全非常重要。
National security not just in one aspect of national security, all aspects of national security.
国家安全不仅仅是某一方面的国家安全，是各个方面的国家安全。
When our country’s more prosperous, we could do a better job with domestic policies and helping social benefits.
当我们的国家更加繁荣时，我们可以更好地做好国内政策和帮助社会福利。
Because we’re generating so much re-industrialization in the United States, we’re creating mountains of jobs.
因为我们在美国实现了如此多的再工业化，所以我们正在创造大量的就业机会。
We’re helping shift how we build things back to the United States in so many different plants, chips, computers, and of course, these AI factories.
我们正在帮助将我们在许多不同的工厂、芯片、计算机，当然还有这些人工智能工厂中的制造方式转变回美国。
I’m completely aware that, that…
我完全清楚，那个……
1:36:35 And I have the benefit, and this is a real gift with mainstream investors, teachers, policemen who have somehow, for whatever reason, invested in NVIDIA or because they watched Jim Cramer, bought some stock and now are millionaires.
我受益匪浅，这对主流投资者、教师、警察来说是一份真正的礼物，他们不知出于何种原因投资了 NVIDIA，或者因为观看了 Jim Cramer 的节目、购买了一些股票而现在成为了百万富翁。
1:36:57 And I am completely aware of that circumstance.
我完全了解这种情况。
I’m aware of the circumstance that NVIDIA is central to a very large network of ecosystem partners behind us and downstream from us.
我知道 NVIDIA 是我们背后和下游的一个庞大生态系统合作伙伴网络的核心。
And so the way I deal with that is exactly what I just did.
所以我处理这个问题的方式正是我刚才所做的。
I reason about what is…
我推理什么是……
what is it that we’re doing?
我们正在做什么？
What is it causing?
这是什么原因造成的呢？
What’s the impact that has on other people benefit, you know, positively or even through great burden, for example, to supply chain?
你知道，这对其他人的利益有什么影响，是积极的，还是给供应链带来巨大的负担？
And the question is therefore, what are you gonna do about it?
因此问题是，你会怎么做？
In almost everything that I feel, I break it down, I reason about, “Okay, what’s the circumstance?
几乎在我感受到的所有事情中，我都会将其分解，我会思考：“好吧，情况是什么？
What has changed?
发生了什么变化？
What’s hard?
有什么难的？
And what am I gonna do about it?” And I’m…
那我该怎么办呢？而我……
1:37:56 I break it down, decompose the problem, and the decomposition of these circumstances turns it into manageable things that I can do.
我把它分解，分解问题，这些情况的分解将它变成我可以做的可管理的事情。
And the only thing that after that I could do is, “Did you do it?
之后我唯一能做的就是：“你做到了吗？
Did you either do it or did you get somebody else to do it?
是你自己做的还是让别人做的？
And if you didn’t do it, you reasoned that you need to do it, and you didn’t do it, and you didn’t get anybody else to do it, then stop crying about it.”…
如果你没有这样做，你就会认为你需要这样做，但你没有这样做，也没有让其他人这样做，那就别再为此哭泣了。”
you know?
你知道吗？
And so, and so-
等等，等等——
1:38:27 so I’m fairly tough on myself.
所以我对自己相当严格。
And, but I also break things down so that I don’t panic.
而且，我也会把事情分解，这样我就不会惊慌。
I can go to sleep because I’ve made the list of things that needed to be done, and I’ve made sure that everything that could put our company in harm’s way, could put my partners in harm’s way, put our industry in harm’s way, I’ve told somebody.
我可以去睡觉了，因为我已经列出了需要做的事情，并且我已经确保一切可能使我们的公司、我的合作伙伴、我们的行业陷入危险的事情，我已经告诉过某人了。
Everything that I feel could put anybody in harm’s way, I’ve told someone.
我告诉过某人，我感觉到的一切都可能使任何人受到伤害。
And I’ve told that someone who could do something about it.
我已经告诉那个可以为此做点什么的人。
And so I’ve gotten it off my chest or I’m doing something about it.
所以我已经把这件事从心里吐出来了，或者说我正在为此做点什么。
And so after that, Lex, what else can you do?
那么在那之后，Lex，你还能做什么？
So given all the insane, intense amount of suffering on the journey of building up NVIDIA, have you hit low points psychologically?
那么，考虑到在创建 NVIDIA 的过程中所经历的所有疯狂、巨大的痛苦，您的心理是否已经陷入低谷？
Oh, yeah.
哦，是的。
1:39:22 Oh, yeah.
哦，是的。
Sure.
当然。
All the time.
一直以来。
All the time.
一直以来。
And there-
还有——
All the time
一直以来
…
……
1:39:27 you just break down the problem into pieces?
你只是把问题分解成碎片吗？
See what you could do about it?
看看你能做些什么呢？
And part of it, Lex, part of it is forgetting.
莱克斯，部分原因是遗忘。
1:39:33 One of the most important attributes of AI learning, as you know, is, right?
如您所知，人工智能学习最重要的属性之一是，对吧？
Systematic forgetting.
系统性遗忘。
You need to know when to forget some things.
你需要知道什么时候忘记一些事情。
You can’t memorize everything.
你不可能记住所有的事情。
You can’t keep everything and, you know, you don’t want to carry everything.
你无法保留所有东西，而且你知道，你也不想携带所有东西。
One of the things that I do very quickly is decompose the problem, I reason about the problem, and I share the load with it.
我很快做的事情之一就是分解问题，推理问题，并与其分担负载。
When I say I tell everybody, I’m essentially sharing that burden.
当我说我告诉所有人时，我本质上是在分担这个负担。
1:40:04 As quickly as possible.
尽快。
Whatever worries me, tell somebody else.
无论我担心什么，请告诉别人。
Don’t just keep it.
不要只是保留它。
You know, don’t freak them out.
你知道，别吓坏他们。
Decompose the problem into smaller parts and get people to, and inspire them to be able to go do something about it.
将问题分解成更小的部分，让人们去解决，并激励他们能够采取行动来解决这个问题。
But part of it is just forgetting.
但其中一部分只是遗忘。
You know, like, a lot of it is you gotta be tough on yourself.
你知道，很多事情都是你必须对自己严格要求。
You know, just come on, stop crying about it.
你知道，来吧，别再为此哭泣了。
Let’s get going.
让我们开始吧。
You know?
你知道？
And then you get out of bed.
然后你就起床了。
And then the other part is you’re attracted to the next shiny light, the next future, the next opportunity, the next, “Okay, that’s behind us.
另一方面，你会被下一个闪亮的光芒、下一个未来、下一个机会、下一个所吸引，“好吧，那已经过去了。
What’s next?” It’s a lot, I think, you know, you watch this with great athletes.
接下来怎么办？”我想，这很多，你知道，你和伟大的运动员一起观看这个比赛。
They just worry about the next point.
他们只是担心下一点。
The last point is behind them.
最后一点就在他们身后。
The embarrassment, the, you know- …
尴尬，你知道的——……
the setback.
挫折。
1:40:56 You know, and because I do so much of my job publicly, you know?
你知道，因为我的大部分工作都是公开的，你知道吗？
Lex, you do a fair amount of your job publicly too.
Lex，你也做了相当多的公开工作。
And so I do a lot of my job publicly.
所以我的很多工作都是公开进行的。
And so you know, I say a lot of things that seem sensible at the time or funny at the time, mostly it’s just because it’s funny to me at the time.
所以你知道，我说了很多当时看起来明智或有趣的话，大部分只是因为当时对我来说很有趣。
And then, you know, you reflect on it, it’s less funny, but…
然后，你知道，你反思一下，这并不那么有趣，但是......
Yeah.
是的。
1:41:20 No, trust me, I know.
不，相信我，我知道。
But you basically allow yourself to be pulled by the light of the future.
但你基本上让自己被未来之光所吸引。
Forget the past and just keep-
忘记过去，只保留——
That’s right.
没错。
…
……
1:41:28 keep working towards that.
继续为此努力。
I mean, you did say, there’s this kind of famous thing you said that if you knew how hard it would be to build NVIDIA it turned out to be—what is it?
我的意思是，您确实说过，您说过，如果您知道构建 NVIDIA 会有多困难，结果会是这样——那是什么呢？
A million times more hard than you anticipated—that you wouldn’t do it.
比你预期的困难一百万倍——你却做不到。
Yeah, right.
是的，对。
But isn’t…
但不是……
1:41:47 You know, when I hear that, that’s probably true about everything worth doing, right?
你知道，当我听到这句话时，这可能对所有值得做的事情都是正确的，对吧？
Exactly.
确切地。
1:41:53 That is, by the way, what I was trying to explain, is that there’s an incredible superpower of having the mind of a child.
顺便说一句，我想解释的是，拥有孩子般的思维有一种令人难以置信的超能力。
You know?
你知道？
And I say to myself oftentimes when I look at something, and almost everything my first thought is, “How hard can it be?” You know?
当我看到某样东西时，我经常对自己说，几乎所有事情我的第一个想法都是：“它能有多难？”你知道？
And so you get yourself into that mode, how hard could it be?
当你让自己进入这种模式时，这有多难？
And nobody’s ever done it.
从来没有人这样做过。
It looks gigantic.
它看起来巨大。
It’s gonna cost hundreds of billions of dollars.
这将花费数千亿美元。
It’s gonna take, you know, all this…
这将需要，你知道，这一切......
And you just go, “Yeah, but how hard could it be?” You know?
然后你就说：“是的，但这能有多难呢？”你知道？
How hard could it be?
这有多难？
1:42:37 And so, you gotta get yourself into that state of mind.
所以，你必须让自己进入那种心态。
You don’t wanna actually over-simulate everything and all the setbacks and all the trials and tribulations and all the disappointments.
你不想实际上过度模拟一切、所有的挫折、所有的考验、磨难和所有的失望。
You don’t wanna simulate all that in advance.
你不想提前模拟这一切。
You don’t wanna know that.
你不会想知道的。
You wanna go into a new experience thinking it’s gonna be perfect, it’s gonna be great, it’s gonna be incredibly fun.
你想要进入一种新的体验，认为它会是完美的、会很棒、会非常有趣。
And then while you’re there, you know, you need to have endurance, you need to have grit, so that when the setbacks actually happen, and those setbacks are gonna surprise you, the disappointments are gonna surprise you, the embarrassments are gonna surprise you, the humiliations are gonna surprise you.
然后，当你在那里时，你知道，你需要有耐力，你需要有勇气，这样当挫折真正发生时，这些挫折会让你感到惊讶，失望会让你惊讶，尴尬会让你惊讶，羞辱会让你惊讶。
1:43:17 You just can’t let…
你就是不能让……
Now you just gotta turn on the other bit, which is just forget about it.
现在你只需要打开另一位，就不用管它了。
Move on, keep moving.
继续前进，继续前进。
And to the extent that my assumptions about the future and why the future is gonna manifest, so long as those assumptions and that input doesn’t change or didn’t change materially, then I should expect that the output won’t change.
就我对未来的假设以及未来为何会显现而言，只要这些假设和输入没有改变或没有发生重大改变，那么我应该期望输出不会改变。
And so my simulated output of the future is still gonna happen.
所以我对未来的模拟输出仍然会发生。
And if it’s still gonna happen, I’m still gonna go after it.
如果它仍然会发生，我仍然会追求它。
1:43:54 I believe it’s gonna, you know, and so there’s a combination of two or three human characteristics: the ability to go into an experience fresh-minded, the ability to forget the setbacks, the ability to believe in yourself, you know, to believe what you believe and stay true to that belief.
我相信它会，你知道，所以这是两三个人类特征的结合：以新鲜的心态进入体验的能力，忘记挫折的能力，相信自己的能力，你知道，相信你所相信的并忠于这个信念。
But you’re constantly reevaluating.
但你不断地重新评估。
1:44:20 This combination of three, four, five things I think is really important for resilience.
我认为这三、四、五件事的结合对于恢复力来说非常重要。
And, you know, I’m fortunate that whatever life experiences led to this, I’ve got kind of those four, five things.
而且，你知道，我很幸运，无论生活经历如何导致了这一点，我都有这四、五件事。
You know, I’m always curious, always learning.
你知道，我总是充满好奇，总是在学习。
I’m always learning from everybody, you know?
我总是向每个人学习，你知道吗？
I’m always asking my…
我总是问我的……
And because I’m humble about everything, I’m always thinking, “Gosh, they did that so nicely.
因为我对一切都很谦虚，所以我总是在想，“天哪，他们做得这么好。
They did that so wonderfully.” You know, I wonder what they’re thinking through.
他们做得非常出色。”你知道，我想知道他们在想什么。
How do they…
他们如何……
So I’m simulating everybody.
所以我在模拟每个人。
In a lot of ways, you know, I’m emulating almost everybody I watch, right?
在很多方面，你知道，我几乎在模仿我所看过的每个人，对吧？
You’re empathetic towards everything that they do that you’re observing and respect.
您对您所观察和尊重的他们所做的一切表示同情。
And so you’re constantly learning and, you know.
所以你不断地学习，你知道。
You’re now one of the wealthiest people on Earth.
您现在是地球上最富有的人之一。
1:45:10 One of the most successful humans on Earth.
地球上最成功的人类之一。
Is it harder to be humble and to be able to…
谦虚并能够……是不是更难？
Do you feel the effect of money and power and fame in making it harder for you to sort of be wrong in your own head?
你是否感觉到金钱、权力和名誉的影响让你更难在自己的头脑中犯错？
Enough to hear out an opinion of somebody else when they disagree with you and learn from them?
当别人不同意你的观点时，足以听取他们的意见并向他们学习吗？
Those kinds of things.
诸如此类的事情。
Surprisingly, no.
令人惊讶的是，没有。
1:45:41 And I would actually go the other way.
我实际上会走另一条路。
Because I do so much of my work publicly, when I’m wrong, pretty much everybody sees it.
因为我的大部分工作都是公开进行的，所以当我犯错时，几乎每个人都会看到。
You get humbled.
你会变得谦卑。
1:45:53 Fair enough.
很公平。
And when I’m wrong—when I’m wrong or it didn’t turn out that way or, you know, I mean, most of the things that I say outside I’m fairly certain about.
当我错了——当我错了或者结果不是那样的时候，或者，你知道，我的意思是，我在外面说的大部分事情我都相当确定。
1:45:55 And the reason for that is because it’s gonna impact somebody else and I want to be quite concerned about that and quite circumspect about that.
这样做的原因是因为它会影响其他人，我想对此非常关心并非常谨慎。
For stuff that I’m reasoning about inside a meeting, you know, a lot of things could turn out differently.
你知道，对于我在会议中思考的事情，很多事情的结果可能会有所不同。
And so, but it doesn’t ever stop me from reasoning.
所以，但这并没有阻止我推理。
The way that I manage and lead, I’m constantly reasoning in front of people.
我的管理和领导方式是不断在人们面前推理。
And even when I’m talking to you, you can kind of see me reasoning through things.
即使当我和你说话时，你也能看到我在推理事情。
And I want to make sure that you understand what I’m saying not because I told you-
我想确保你明白我在说什么，而不是因为我告诉过你——
1:46:40 …
……
because I’m so humble about what I’m about to tell you.
因为我对我要告诉你的事情感到非常谦虚。
I kind of show you the steps that I got there.
我向您展示了我到达那里的步骤。
And then you can decide whether you believe what I said in the end.
然后你可以决定最后是否相信我说的话。
And so I’m doing that all day long in meetings.
所以我整天在会议中都这样做。
With all of my employees, I’m constantly reasoning through, “Let me tell you how I see it.” And then I reason through it.
我和我的所有员工一起不断地推理，“让我告诉你们我是如何看待它的。”然后我通过它进行推理。
It gives everybody the opportunity to intercept and say, “I disagree with that part.” The nice thing about reasoning through things and letting people interact with it is that they don’t have to disagree with your outcome.
它让每个人都有机会拦截并说：“我不同意那部分。”通过事物进行推理并让人们与之互动的好处是，他们不必不同意你的结果。
They can disagree with your reasoning steps.
他们可能不同意你的推理步骤。
And they could pull me in different directions, and then we can reason forward.
他们可以把我拉向不同的方向，然后我们就可以继续推理。
And so we’re kind of, you know, a collective path searching method.
所以我们是一种集体路径搜索方法。
And it’s really fantastic.
这真的很棒。
Yeah, you have this way about you of …
是的，你的性格是这样的……
1:47:29 When you’re explaining stuff, I can feel you actually reasoning on the spot about it with a constant open-mindedness where you could …
当你解释事情时，我能感觉到你实际上以一种持续开放的心态当场推理，你可以……
I could feel like I could steer your thinking.
我感觉我可以引导你的思维。
And that’s a—that’s really beautiful that you’ve been able to maintain that after so many years of success, and pain.
在经历了这么多年的成功和痛苦之后，你仍然能够保持这一点，这真的很美好。
I think sometimes pain closes you down a bit.
我想有时候疼痛会让你有点封闭。
And I think to maintain-
我认为要维持-
Yeah.
是的。
1:47:57 Tolerance for embarrassment, I think is…
对尴尬的容忍，我认为是……
Yes, that’s…
是的，那就是……
1:47:59 The tolerance…
耐受性…
I mean, that’s a real thing.
我的意思是，这是真实的事情。
Is many years of embarrassing yourself.
是多年的让自己难堪。
Even those meetings knowing that there’s people around you where you declared one idea and it was shown that that idea was wrong- …
即使在那些会议上，你知道你周围有人提出了一个想法，但事实证明这个想法是错误的——……
and be able to admit that and to grow from that.
并能够承认这一点并从中成长。
That’s not—that’s very difficult on a human level.
这不是——这对人类来说是非常困难的。
Yeah.
是的。
1:48:17 Well, you know.
嗯，你知道。
They knew I was—they knew that recently my first job was cleaning toilets, so.
他们知道我是——他们知道最近我的第一份工作是打扫厕所，所以。
I’m glad you maintained that same spirit of Denny’s, the work.
我很高兴你保持了丹尼的工作精神。
1:48:25 I mean, that was beautiful.
我的意思是，那太美了。
Your whole journey starting from Denny’s is a beautiful one.
从丹尼出发的整个旅程都是美好的。
Let me ask you about video games.
让我问你一些关于电子游戏的问题。
So I’m a big gaming fan.
所以我是一个游戏迷。
So I have to say thank you to NVIDIA for many years of incredible graphics.
因此，我必须对 NVIDIA 多年来提供的令人难以置信的图形表示感谢。
By the way, GeForce is our still, to this day- …
顺便说一句，GeForce 是我们的静止器，直到今天 - ...
1:48:47 our number one marketing strategy.
我们的第一营销策略。
Right.
正确的。
People learn about NVIDIA while they’re in their teenage years.
人们在青少年时期就开始了解 NVIDIA。
And then they go to college and they know who NVIDIA is and in the beginning it’s just, you know, playing Call of Duty, Fortnite.
然后他们进入大学，知道 NVIDIA 是谁，一开始只是玩《使命召唤》和《堡垒之夜》。
And then later they’re using CUDA, and then later they’re using NVIDIA and, you know, Blender and Dassault and Autodesk.
后来他们使用 CUDA，再后来他们使用 NVIDIA，还有 Blender、Dassault 和 Autodesk。
Yeah.
是的。
1:49:16 I mean, I should say I mentioned to a friend that I’m talking with you.
我的意思是，我应该说我向一位朋友提到我正在和你说话。
He said, “Oh, they make great gaming GPUs.”
他说：“哦，他们制造了很棒的游戏 GPU。”
Yeah, exactly.
是的，完全正确。
It’s like-
就像-
Exactly.
正是如此。
You know, there’s more to it, but, yeah, people really love it.
你知道，它还有更多的内容，但是，是的，人们真的很喜欢它。
1:49:27 It really brought a lot of joy to a lot of people.
确实给很多人带来了很多欢乐。
The hardware really brings these worlds to life.
硬件确实让这些世界变得栩栩如生。
There was some controversy around this with DLSS 5.
DLSS 5 对此存在一些争议。
Can you explain to me the drama around this?
你能给我解释一下这件事的剧情吗？
I guess people, the gamers online were concerned that it makes games look like AI slop.
我想人们，在线游戏玩家担心这会让游戏看起来像人工智能的废话。
What do you think of this drama?
你觉得这部剧怎么样？
Yeah.
是的。
1:49:56 I think their perspective makes sense and I could see where they’re coming from, because I don’t love AI slop myself.
我认为他们的观点是有道理的，我可以看到他们的观点来自哪里，因为我自己不喜欢人工智能。
You know, all of the AI-generated content increasingly looks similar and they’re all beautiful, and so I’m empathetic towards what they’re thinking.
你知道，所有人工智能生成的内容看起来越来越相似，而且都很漂亮，所以我很理解他们的想法。
That’s just not what DLSS 5 is trying to do.
这并不是 DLSS 5 想要做的事情。
I showed several examples of it.
我展示了几个例子。
But DLSS 5 is 3D-conditioned, 3D-guided.
但 DLSS 5 是 3D 调节、3D 引导的。
It’s ground truth structure data guided.
它以真实结构数据为指导。
And so the artist determined the geometry.
因此艺术家确定了几何形状。
We are completely truthful to the geometry maintained in every single frame.
我们完全忠实于每一帧中维护的几何形状。
It’s conditioned by the textures, the artistry of the artist.
它受到纹理和艺术家艺术性的影响。
And so every single frame, it enhances but it doesn’t change anything.
所以每一帧，它都会增强，但不会改变任何东西。
1:50:55 Now, the question is about enhancing.
现在的问题是关于增强。
DLSS 5 also lets, because the system is open, you could train your own models to determine, and you could even in the future prompt it.
由于系统是开放的，DLSS 5 还允许您训练自己的模型来确定，甚至可以在将来提示它。
You know, I want it to be a toon shader.
你知道，我希望它成为一个卡通着色器。
I want it to look like this kind of, so you can give it even an example.
我希望它看起来像这样，所以你甚至可以举个例子。
And it would generate in the style of that, all consistent with the artistry, the style, the intent of the artist.
它会以那种风格产生，一切都与艺术家的艺术性、风格和意图一致。
And so all of that is done for the artist, so that they can create something that is more beautiful but still in the style that they want.
因此，所有这一切都是为艺术家完成的，以便他们能够创造出更美丽但仍符合他们想要的风格的东西。
I think that they got the impression that the games are gonna come out the way the games are, shipped the way they do, and then we’re gonna post-process it.
我认为他们的印象是游戏将按游戏本身的方式推出，按其方式交付，然后我们将对其进行后期处理。
That’s not what DLSS is intended to do.
这不是 DLSS 的初衷。
1:51:50 DLSS is integrated with the artist, and so it’s about giving the artist the tool of AI, the tool of generative AI.
DLSS 与艺术家融为一体，所以它是为艺术家提供人工智能的工具，生成人工智能的工具。
They could decide not to use it, you know?
他们可以决定不使用它，你知道吗？
I think people are very sensitive to human faces.
我认为人们对人脸非常敏感。
1:52:01 And we’re now living in this moment, which I think is a, is a beautiful one, which is people are sensitive to AI slop.
我们现在生活在这一刻，我认为这是一个美好的时刻，人们对人工智能的错误很敏感。
It puts a mirror to ourselves to help us realize that what we seek is imperfections.
它给我们自己树立了一面镜子，帮助我们认识到我们所追求的是不完美。
What we seek is sometimes not perfect graphics.
我们追求的有时并不是完美的图形。
It helps us understand what we find compelling in the worlds we create.
它帮助我们了解我们所创造的世界中什么是引人注目的。
And that’s beautiful.
那很漂亮。
And as long as it’s tools that help us create those worlds-
只要它是帮助我们创造这些世界的工具——
Yeah, that’s right.
是的，没错。
…
……
1:52:28 it’s wonderful.
太棒了。
That’s right.
这是正确的。
1:52:29 Yet, yet another tool, and they want the generative models to generate the opposite of photo real.
然而，还有另一种工具，他们希望生成模型能够生成与照片真实相反的效果。
Yeah, it’ll do that too.
是的，它也会这样做。
And so it’s just yet another tool.
所以它只是另一个工具。
I think the gamers might also appreciate that in the last couple of years, we introduced skin shaders to the game developers.
我认为游戏玩家可能也会欣赏在过去几年中我们向游戏开发人员引入了皮肤着色器。
And many of those games have skin shaders that include subsurface scattering that make skin look more skin-like.
其中许多游戏都有皮肤着色器，其中包括次表面散射，使皮肤看起来更像皮肤。
And so the industries, you know, game developers are looking for more and more tools to express their art.
因此，游戏开发者行业正在寻找越来越多的工具来表达他们的艺术。
And so this is just yet one more tool, and they get to decide what to use.
所以这只是又一个工具，他们可以决定使用什么。
Ridiculous question.
可笑的问题。
1:53:16 What do you think is the greatest or most influential game ever made?
您认为有史以来最伟大或最有影响力的游戏是什么？
Maybe from NVIDIA’s perspective?
也许从 NVIDIA 的角度来看？
Doom.
厄运。
Doom, unquestionably.
厄运，毫无疑问。
1:53:25 That was the start of the 3D.
这就是 3D 的开始。
I would say Doom, from an art, the intersection of the cultural implication as well as the industry, turning a PC into a gaming device.
我想说《毁灭战士》是一门艺术，是文化内涵和行业的交集，将 PC 变成了游戏设备。
1:53:28 That was a very important moment.
那是一个非常重要的时刻。
Now, of course, flight simulation companies were before it.
当然，现在飞行模拟公司已经走在了前面。
And but they just didn’t have the popularity that Doom did to have made the industry turn the PC from an office automation tool into a personal computer for families and gamers and things like that.
但它们只是没有像《毁灭战士》那样受欢迎，无法让业界将个人电脑从办公自动化工具转变为家庭和游戏玩家等的个人电脑。
And so Doom was really impactful there.
所以《毁灭战士》在那里确实具有影响力。
From an actual game technology perspective, I would say Virtua Fighter.
从实际游戏技术的角度来看，我会说《Virtua Fighter》。
And so we’re great friends with both of them, you know?
所以我们和他们都是好朋友，你知道吗？
And then there’s games more recently—I mean, Cyberpunk 2077, really nice GPU-accelerated graphics.
还有最近的游戏——我的意思是《赛博朋克 2077》，非常好的 GPU 加速图形。
1:54:07 Like-
就像-
Fully ray traced.
完全光线追踪。
Fully ray traced.
完全光线追踪。
1:54:17 Also, I like, I personally, I’m a huge fan of Skyrim, Elder Scrolls, and the, you know, it’s, it’s been released a long, long time ago, but people release mods and-
另外，我个人喜欢，我是《天际》、《上古卷轴》的超级粉丝，而且，你知道，它已经发布了很久很久以前，但人们发布了模组，然后——
We love mods.
我们喜欢模组。
…
……
1:54:30 they create these inc- I mean, it’s like a different game and it just allows me to replay the game over and over.
他们创造了这些——我的意思是，这就像一个不同的游戏，它只是让我一遍又一遍地重玩游戏。
It makes you realize that you can re-experience in a totally new way the world you already love.
它让你意识到你可以以一种全新的方式重新体验你已经热爱的世界。
So-
所以——
That’s right.
没错。
…
……
1:54:45 I do that all the time.
我一直这样做。
One of my favorite things is just walk across Skyrim.
我最喜欢的事情之一就是步行穿过天际。
We created this thing called RTX Mod.
我们创建了这个叫做 RTX Mod 的东西。
1:54:48 Yeah, it’s a modding tool.
是的，这是一个改装工具。
Awesome.
惊人的。
It allows the community to inject the latest technology into an old game.
它允许社区将最新技术注入旧游戏中。
Of course, like what makes a great video game is not just graphics, it’s also story and character development, but-
当然，就像一款出色的视频游戏不仅仅是图像，还有故事和角色发展，但是——
That’s right
没错
…
……
1:55:06 beautiful graphics can add to the immersion.
漂亮的图形可以增加沉浸感。
The feeling like it’s another place you’re transported to.
感觉就像你被传送到了另一个地方。
Ah, what you said, I think accurately, that the AGI timeline question rests on your definition of AGI.
啊，你所说的，我认为准确的是，AGI 时间线问题取决于你对 AGI 的定义。
So let’s, let me ask you about possible timelines here.
那么，让我来询问一下可能的时间表。
Let’s, this ridiculous definition perhaps of what AGI is, but an AI system that’s able to essentially do your job.
让我们来看看这个荒谬的定义，也许 AGI 是什么，但它是一个基本上能够完成你的工作的人工智能系统。
So, run, no, start, grow, and run a successful technology company that’s worth-
所以，经营，不，创办、发展和经营一家成功的科技公司，它的价值是——
A good one or a one?
一个好还是一个好？
No.
不。
1:55:54 It has to be worth more than a billion, more than a billion dollars.
它的价值必须超过十亿，超过十亿美元。
So, you know, you know how hard it is to do all those components.
所以，你知道，你知道完成所有这些组件有多难。
So, how far are we away from that?
那么，我们离这个目标还有多远？
So, we’re talking about Open-Claude that does all the incredibly complex stuff that are required to, first of all, innovate, to find customers, to sell to them, to manage, to build a team of some agents, some humans, all that kind of stuff.
所以，我们谈论的是 Open-Claude，它可以完成所有极其复杂的事情，首先是创新、寻找客户、向他们销售、管理、建立一个由一些代理、一些人等组成的团队。
Is this five, 10, 15, 20 years away?
这是五年、十年、十五年、二十年后吗？
I think it’s now.
我想就是现在了。
1:56:31 I think we’ve achieved AGI.
我认为我们已经实现了通用人工智能。
Do you think you could have a company run by an AI system like this?
你认为你能拥有一家由这样的人工智能系统运营的公司吗？
Possible, and the reason for that is this.
有可能，原因就是这样。
1:56:37 You said a billion, and you didn’t say forever.
你说十亿，但没有说永远。
And so for example…
举例来说……
It is not out of the question that a Claude was able to create a web service, some interesting little app that all of a sudden, you know, a few billion people used for 50 cents, and then it went out of business again shortly after.
克劳德能够创建一个网络服务，一个有趣的小应用程序，突然之间，几十亿人花 50 美分使用它，然后不久之后它又停业了，这并不是不可能的。
Now, we saw a whole bunch of those type of companies during the internet era, and most of those websites were not anything more sophisticated than what Open-Claude could generate today.
现在，我们在互联网时代看到了大量此类公司，其中大多数网站并不比 Open-Claude 今天生成的网站更复杂。
Interesting.
有趣的。
1:57:20 Achieve virality and monetize that virality.
实现病毒式传播并通过这种病毒式传播货币化。
Yeah.
是的。
1:57:23 It’s just that I don’t know what it is, but I couldn’t have predicted any of those companies at the time either, you know?
只是我不知道它是什么，但我当时也无法预测这些公司，你知道吗？
And –
并且——
You’re gonna get a lot of people excited with that statement.
你的这句话会让很多人感到兴奋。
Yeah, no.
是的，不。
1:57:32 Yeah.
是的。
It’s like, what do you mean?
就像，你是什么意思？
1:57:33 I can just launch an agent and make a lot of money.
我只要开个代理就可以赚很多钱。
Well, by the way, it’s happening right now, right?
好吧，顺便说一句，这件事现在就发生了，对吧？
1:57:38 You know that when you go to China you’re gonna see, you’re gonna see a whole bunch of people teaching their, getting their Claudes to try to go out and look for jobs and, you know, do work, make money.
你知道，当你去中国时，你会看到，你会看到一大群人教导他们的克劳德，让他们的克劳德尝试出去寻找工作，然后，你知道，工作，赚钱。
And I’m not, I’m not actually…
而我不是，我实际上不是……
I wouldn’t be surprised if some social thing happened or somebody created a digital influencer, super, super cute, or some social application that, you know, feeds your little Tamagotchi or something like that, and it become out of the blue an instant success.
如果发生了一些社交事件，或者有人创建了一个超级、超级可爱的数字影响者，或者一些社交应用程序，你知道，可以喂养你的小电子宠物或类似的东西，并且它会突然获得成功，我不会感到惊讶。
A lot of people use it for a couple of months and it kind of dies away.
很多人使用它几个月后就消失了。
Now, the odds of 100,000 of those agents building NVIDIA is zero percent.
现在，100,000 名代理构建 NVIDIA 的可能性为零。
1:58:28 And then, and then the one part that I will, I won’t do and I wanna make sure we all do, is to recognize that people are really worried about their jobs.
然后，我会做、不会做但我想确保我们所有人都做的一件事就是认识到人们真的很担心他们的工作。
And I just want to remind them that the purpose of your job and the tasks and tools that you use to do your job are related, not the same.
我只是想提醒他们，你的工作目的和你用来完成工作的任务和工具是相关的，而不是相同的。
I’ve been doing my job for 33 years.
我已经从事我的工作 33 年了。
I’m the longest running tech CEO in the world, 34 years.
我是世界上任职时间最长的科技公司首席执行官，34 年。
And the tools that I’ve used to do my job has changed continuously in the last 34 years, and sometimes quite dramatically, you know, over the course of a couple, two, three years.
在过去的 34 年里，我用来完成工作的工具不断发生变化，有时甚至在几年、两年、三年的时间里发生了巨大的变化。
And the one story that I really wanna make sure that everybody hears is the story that the first job that computer scientists said, AI researchers said was gonna go away was radiology.
我真正想确保每个人都听到的一个故事是，计算机科学家和人工智能研究人员所说的第一个将消失的工作是放射学。
1:59:25 Because computer vision was going to achieve superhuman levels, and it did.
因为计算机视觉将达到超人的水平，而且它确实做到了。
CV…
简历…
Computer vision was superhuman in 2019, 20, maybe maybe a little bit later, 2020?
计算机视觉在 2019 年、20 年，也许晚一点，2020 年就已经是超人了？
1:59:39 Okay?
好的？
And so it’s been a long time since computer vision has been superhuman.
因此，计算机视觉成为超人类已经有很长一段时间了。
And so the prediction was radiologists would go away because studying radiology scans was a thing of the past.
因此，预测放射科医生将会离开，因为研究放射扫描已成为过去。
AI will do that.
人工智能会做到这一点。
Well, they were absolutely right.
嗯，他们是绝对正确的。
Computer vision is completely superhuman.
计算机视觉完全是超人的。
Every radiology platform and package today is driven by AI, and yet the number of radiologists grew.
如今，每个放射学平台和软件包都是由人工智能驱动的，但放射科医生的数量却在增加。
And so the question is why?
所以问题是为什么？
And we now have a shortage of radiologists in the world.
现在世界上放射科医生短缺。
And so, one, the alarmist warning went too far and it scared people from doing this profession that is so important to society.
因此，第一，危言耸听的警告太过分了，它吓坏了人们从事这个对社会如此重要的职业。
And so it did harm.
所以它确实有害。
Now, why was it wrong?
现在，为什么错了？
The reason why is because the purpose of a radiologist, the purpose is to diagnose disease and help patients and doctors diagnose disease.
之所以是因为放射科医生的目的，目的就是诊断疾病，帮助患者和医生诊断疾病。
2:00:38 And because we’re able to study scans so much faster now, you could study more scans, you could diagnose better, you could in-patient faster, you can see people more.
因为我们现在能够更快地研究扫描，你可以研究更多的扫描，你可以更好地诊断，你可以更快地住院，你可以看到更多的人。
The hospitals are making more money.
医院赚的钱越来越多了。
You have more patients in the hospital.
你们医院里还有更多的病人。
You need more radiologists.
您需要更多的放射科医生。
I mean, the amazing thing is, it’s so obvious this was gonna happen.
我的意思是，令人惊奇的是，很明显这会发生。
The number of software engineers at NVIDIA is gonna grow, not decline.
NVIDIA 的软件工程师数量将会增加，而不是减少。
And the reason for that is because the purpose of a software engineer and the task of a software engineer coding are related, not the same.
原因是软件工程师的目的和软件工程师编码的任务是相关的，而不是相同的。
I wanted my software engineers to solve problems.
我希望我的软件工程师能够解决问题。
I didn’t care how many lines of code they wrote, you know?
我不在乎他们写了多少行代码，你知道吗？
But their job, their purpose of their job didn’t change.
但他们的工作、工作目的并没有改变。
2:01:25 Solving problems, working as a team, diagnosing problems, evaluating the result, looking for new problems to solve, innovation, connecting dots.
解决问题、团队合作、诊断问题、评估结果、寻找新问题来解决、创新、连接点。
You know, none of that stuff is gonna go away.
你知道，这些东西都不会消失。
Do you think it’s possible that…
你认为有可能吗……
2:01:39 Let’s even take coding.
我们甚至可以进行编码。
Do you think the number of programmers in the world might increase, not decrease?
您认为世界上程序员的数量会增加而不是减少吗？
Yes.
是的。
2:01:45 And the reason for that is this.
原因是这样的。
What is the definition of coding?
编码的定义是什么？
I believe it is…
我相信这是……
The definition of coding, as of today, is simply specifying, specification, and maybe if you want to be rather directive, you could even give it an architecture of the software that you wanted to write.
截至今天，编码的定义只是简单地指定、规范，也许如果您想要更具指导性，您甚至可以给它一个您想要编写的软件的架构。
So the question is, how many people could do that?
那么问题来了，有多少人能做到呢？
Describe a specification for a computer to go…
描述计算机的规格...
telling the computer what to go build.
告诉计算机要构建什么。
How many people?
有多少人？
I think we just went from 30 million to probably 1 billion.
我认为我们刚刚从 3000 万增加到了 10 亿。
And so every carpenter in the future will be a coder, except a carpenter with AI is also an architect.
因此，未来的每个木匠都将是一名编码员，除了具有人工智能的木匠也是一名建筑师。
They’ve just increased the value that they could deliver to the customer.
他们刚刚增加了可以为客户提供的价值。
Their artistry just elevated tremendously.
他们的艺术水平得到了极大的提升。
2:02:43 I believe that every accountant is, you know, also your financial analyst, also your financial advisor.
我相信每个会计师都是您的财务分析师，也是您的财务顾问。
So, all of these professions have just been elevated…
所以，所有这些职业都刚刚得到提升……
and if I were a carpenter, I see AI, I would just completely go berserk.
如果我是一个木匠，我看到人工智能，我会彻底发疯。
You know, the services I can bring to my clients if I were a plumber, completely go berserk.
你知道，如果我是一名水管工，我能为客户提供的服务会完全疯狂。
And the, the people that are currently programmers and software engineers, I think they’re at the cutting edge of understanding intuitively how to communicate with the agents using natural language in order to design the best kind of software.
而且，目前的程序员和软件工程师，我认为他们处于直观地理解如何使用自然语言与代理进行通信以设计最好的软件的最前沿。
That’s right, exactly.
没错，正是如此。
So over time they’ll converge, but I think there’s still value in getting, I think learning how to program, like learning what programming languages are.
因此，随着时间的推移，它们会趋同，但我认为学习如何编程仍然有价值，就像学习什么是编程语言一样。
2:03:20 The old kind of programming, what are good practices for programming languages, what are design principles for programming-
旧式编程，编程语言的良好实践是什么，编程的设计原则是什么——
That’s right
没错
…
……
2:03:40 Languages for large software systems?
大型软件系统的语言？
And the reason for that, Lex, and you know, as you’re saying for the audience, I think the goal of, the goal of specification, the artistry of specification, the goal and the artistry of it is going to depend on what problem you’re trying to solve.
这样做的原因是，莱克斯，正如你对观众所说的那样，我认为目标，规范的目标，规范的艺术性，目标和艺术性将取决于你想要解决的问题。
2:03:43 When I’m thinking, when I’m thinking about giving the company strategies and formulating corporate directions and things that we should do, I describe it at a level that is sufficiently specific that people generally understand the direction and it’s actionable.
当我思考时，当我考虑为公司制定战略、制定公司方向以及我们应该做的事情时，我会在足够具体的水平上进行描述，以便人们普遍理解该方向并且它是可操作的。
It’s specific enough that they can take action on it, but I under-specify it on purpose, so that enables 43,000 amazing people to make it even better than I imagined.
它足够具体，他们可以对此采取行动，但我故意低估了它，这样 43,000 名出色的人就能把它做得比我想象的更好。
2:04:36 And so when I’m working with engineers and when I’m working with people, I think about who, what problem am I trying to solve?
因此，当我与工程师和人们一起工作时，我会思考我要解决谁的问题、什么问题？
Who am I working with?
我和谁一起工作？
And the level of specification, the level of architecture definition relates to that.
规范的级别、架构定义的级别都与此相关。
And so everybody’s going to have to learn how, where in the spectrum of coding they want to be.
因此，每个人都必须学习他们想要的编码方式、编码范围。
Writing a specification is coding.
编写规范就是编码。
And so you might decide to be quite prescriptive because there’s a very specific outcome you’re looking for.
因此，您可能会决定非常规范，因为您正在寻找一个非常具体的结果。
You might decide that, you know, this is an area you want to be much more exploratory, and so you might under-specify and enable you to go back and forth with the AI to even push your own boundaries of creativity.
你可能会决定，你知道，这是一个你想要更具探索性的领域，因此你可能会低估并允许你来回使用人工智能，甚至突破你自己的创造力界限。
And so this artistry of where you are in the spectrum, this is the future of coding.
因此，这种你在频谱中所处位置的艺术性，就是编码的未来。
But just to linger on it outside of coding, I think a lot of people, rightfully so, are worried about their jobs, have a lot of anxiety about their jobs, especially in the white-collar sector.
但只是在编码之外徘徊，我认为很多人理所当然地担心他们的工作，对他们的工作有很多焦虑，尤其是在白领领域。
2:05:31 I don’t think any of us know what to do with tumultuous times that always come when automations and new technology arrives.
我认为我们中的任何人都不知道如何应对自动化和新技术到来时总会到来的动荡时代。
And I just…
而我只是……
First of all, I think we all need to have compassion and the responsibility to feel sort of the burden of what the actual suffering feels like for individual people and families that lose their job.
首先，我认为我们都需要有同情心和责任感，感受到失业个人和家庭实际痛苦的负担。
I think whenever you have transformative technology like that’s coming with artificial intelligence, there’s going to be a lot of pain, and I don’t know what to do about that pain.
我认为，每当你拥有像人工智能这样的变革性技术时，就会有很多痛苦，而我不知道该如何应对这种痛苦。
2:06:21 Hopefully, it creates much more opportunities for those same people for the same kind of job as the tooling evolves and makes them more productive and makes them more fun, hopefully, as it does in the programming.
希望随着工具的发展，它能为同样的人创造更多从事同类工作的机会，使他们更有生产力，让他们更有趣，就像在编程中一样。
I have been having so much fun programming, I have to say.
我不得不说，我的编程非常有趣。
Like, I’ve never had this much fun.
就像，我从来没有玩过这么开心。
So hopefully it makes their job, automates the boring parts and makes the creative parts the ones that the human beings are responsible for.
因此，希望它能完成他们的工作，使无聊的部分自动化，并使创造性的部分成为人类负责的部分。
But still there’s going to be a lot of pain and suffering.
但仍然会有很多痛苦和磨难。
So my first recommendation before…
所以我之前的第一个推荐...
2:06:51 And this is now how I deal with anxiety.
这就是我现在处理焦虑的方式。
In fact, we just talked about it earlier.
其实我们刚才也谈到过。
Enormous anxiety about the future, enormous anxiety about the pressure, enormous anxiety about uncertainty, I first break it down, and then I’m gonna tell myself, “Okay, there are some things you can do something about, there’s some things you can’t do anything about.
对未来的巨大焦虑，对压力的巨大焦虑，对不确定性的巨大焦虑，我首先将其分解，然后我会告诉自己，“好吧，有些事情你可以做点什么，有些事情你什么也做不了。
But for the stuff that you can do something about, let’s reason, reason about it and let’s go do it.”
但对于那些你可以做点什么的事情，让我们推理，推理，然后我们就去做吧。”
2:07:20 If we were to hire a new college graduate today, and I have a choice between two, one that has no clue what AI is and one that is expert in using AI, I would hire the one who’s expert in using AI.
如果我们今天要雇用一名新的大学毕业生，我可以在两个人之间进行选择，一个不知道人工智能是什么，一个是使用人工智能的专家，我会雇用一个擅长使用人工智能的人。
If I had an accountant, a marketing person, the one that is expert in using AI, supply chain, customer service, a salesperson, business development, a lawyer, I would hire the one who is expert in using AI.
如果我有一名会计师、一名营销人员、一名擅长使用人工智能的人、供应链、客户服务、一名销售人员、一名业务开发人员、一名律师，我会聘请一名擅长使用人工智能的人。
And so I would advise that every college student, every teacher should encourage their student to go use AI.
所以我建议每个大学生、每个老师都应该鼓励他们的学生使用人工智能。
Every college student should graduate and be an expert in AI.
每个大学生毕业后都应该成为人工智能领域的专家。
And everybody, if you’re a carpenter, if you’re an electrician, go use AI.
大家，如果你是木匠，如果你是电工，就去使用人工智能。
Go see what it can do to transform your current job, elevate yourself.
去看看它能如何改变你目前的工作，提升你自己。
2:08:21 If I were a farmer, I would absolutely use AI.
如果我是农民，我绝对会使用人工智能。
If I were a pharmacist, I would use AI.
如果我是一名药剂师，我会使用人工智能。
I wanna see how, what it could do to elevate my job so that I could be the innovator to revolutionize this industry myself.
我想看看它如何、如何提升我的工作水平，以便我自己成为革新这个行业的创新者。
And so that would be the first thing that I would do.
所以这将是我要做的第一件事。
And then I would also help them…
然后我也会帮助他们……
It is the case that the technology will dislocate and will eliminate many tasks.
事实是，技术将会扰乱并消除许多任务。
And because it will automate it, if your job is the task—then you’re very highly going to be disrupted.
因为它将使其自动化，如果你的工作就是任务，那么你很可能会被打乱。
If your job’s purpose includes you, certain tasks- …
如果您的工作目的包括您，则某些任务 - ...
then it’s vital that you go learn how to use AI to automate those tasks.
那么学习如何使用人工智能来自动化这些任务就至关重要。
And then there’s the world of spectrum in between.
然后是介于两者之间的频谱世界。
And by the way, the beautiful thing about AI, so the chatbot versions, is you can break down…
顺便说一句，人工智能的美妙之处在于，聊天机器人版本，你可以崩溃……
2:09:14 You have anxiety and you can break down the problem by talking to it.
你有焦虑，你可以通过与它交谈来解决问题。
Like, I’ve recently…
比如，我最近...
It’s really just incredible how much you can think through your life’s problems, and through…
你能如此深入地思考生活中的问题，并通过……真是令人难以置信。
And I don’t mean, like, therapy problems.
我的意思不是治疗问题。
I mean, like, very practically, “Okay, I’m worried about my…” Literally, “I’m worried about my job.
我的意思是，实际上，“好吧，我担心我的……”字面意思是，“我担心我的工作。
What are the skills?
有什么技巧？
What are the steps I need to take?” How do I get better at AI?” Everything you just said, you could literally ask and it’s going to give you- …
我需要采取哪些步骤？” How do I get better at AI?”你刚才所说的一切，你都可以直接问，它会给你——……
a point-by-point plan.
逐点计划。
I mean, it’s just a great life coach, period.
我的意思是，这只是一个伟大的生活教练，就这样。
This-
这——
I don’t know how to use AI, and the AI goes, “Well, let me show you.”
我不知道如何使用人工智能，人工智能会说：“好吧，让我告诉你。”
Exactly.
正是如此。
2:09:54 It’s very meta, but it’s- It’s kind of incredible.
这是非常元的，但是它——有点令人难以置信。
So people definitely should-
所以人们绝对应该——
You can’t walk up to Excel and say, “I don’t know how to use Excel.”
您不能走到 Excel 面前说：“我不知道如何使用 Excel。”
Exactly.
正是如此。
You’re done.
你完成了。
I mean, that’s really what AI has done for me in all walks of life, is that initial friction of being a beginner of using a thing for the first time.
我的意思是，这确实是人工智能在各行各业为我所做的事情，那就是作为初学者第一次使用某样东西时的最初摩擦。
2:10:03 I can literally ask about any single thing, “What are the first steps I need to take?”
我确实可以问任何一件事：“我需要采取的第一步是什么？”
That’s right.
没错。
And that handholding that it does, removing the friction of all the experiences that the world offers is…
它所做的那种牵手，消除了世界提供的所有体验的摩擦……
2:10:17 You know, like I mentioned to you offline, you mentioned, “I’m going to China and Taiwan.”
你知道，就像我在线下向你提到的那样，你提到，“我要去中国和台湾。”
So awesome.
太棒了。
Just ask, “Where do I-“
只要问，“我在哪里——”
So excited for you.
为你感到兴奋。
“Where do I…
“我该去哪里……
2:10:32 What do…” “You know, where do I go?
What do…” “You know, where do I go?
How do I…” All of those questions- …
How do I…” All of those questions- …
immediately answered, and it’s beautiful.
immediately answered, and it’s beautiful.
Well, when you go to Taiwan, just ask AI…
Well, when you go to Taiwan, just ask AI…
2:10:37 “What are Jensen’s favorite restaurants in Taiwan?” And it’ll actually-
“Jensen 最喜欢的台湾餐厅是什么？”它实际上会——
You don’t know?
你不知道吗？
Oh, yeah.
哦，是的。
Is it accurate?
准确吗？
2:10:46 Okay.
好的。
All right.
好吧。
It’s all over Taiwan.
台湾各地都有。
Well, you’re a rockstar over there.
Well, you’re a rockstar over there.
2:10:50 And like we also mentioned offline, maybe our paths will cross, which would be really wonderful in computing.
就像我们在离线时也提到的那样，也许我们的路径会交叉，这在计算中真的很棒。
COMPUTEX.
电脑展。
2:10:58 NVIDIA GTC Taiwan.
NVIDIA GTC 台湾。
Do you think there’s some things about human nature, about human consciousness that is fundamentally non-computational?
你认为关于人性、关于人类意识的某些东西从根本上来说是非计算性的吗？
2:11:01 Maybe something a chip, no matter how powerful, can never replicate?
也许芯片无论多么强大，都永远无法复制？
I don’t know if the chip will ever get nervous.
我不知道芯片是否会紧张。
2:11:18 And that’s the, you know, of course, the conditions by which that causes anxiety or nervousness or whatever emotion.
当然，这就是导致焦虑、紧张或其他情绪的条件。
I believe that AI will be able to recognize those and understand those.
我相信人工智能将能够识别并理解这些。
I don’t think my chips will feel those.
I don’t think my chips will feel those.
And therefore, the…
因此，...
How that anxiety, how that feeling, how that excitement, how that, how that, you know…
怎样的焦虑，怎样的感觉，怎样的兴奋，怎样的，怎样的，你知道……
All of those feelings manifest in human performance.
所有这些感受都体现在人类的表现中。
For example, extremely amazing human performance, athletic performance, you know, average or lesser than average.
例如，极其惊人的人类表现，运动表现，你知道，平均或低于平均水平。
That entire spectrum of human performance that comes out of exactly the same circumstances for different people, manifesting a different outcome, manifesting a different performance.
人类的整个表现范围是在完全相同的环境下对不同的人产生的，表现出不同的结果，表现出不同的表现。
2:12:15 I don’t think there’s anything about anything that we’re building that would suggest that two different computers being presented with all of exactly the same context would perfo- Of course, it would produce statistically different outcomes, but it’s not because it felt different.
我不认为我们正在构建的任何东西都表明两台不同的计算机在完全相同的环境下会表现出来——当然，它会产生统计上不同的结果，但这并不是因为它感觉不同。
Yeah, the subjective…
是的，主观...
2:12:34 Boy, there’s something truly special about the subjective experience that we humans feel.
天哪，我们人类感受到的主观体验确实很特别。
Like I mentioned to you, I was pretty nervous talking to you.
就像我向你提到的那样，与你交谈我非常紧张。
Like I mentioned to you, that, the hope, the fear, the anxiety, and just life itself, the richness of life.
就像我向你提到的那样，希望、恐惧、焦虑，以及生活本身，生活的丰富性。
How amazing everything is.
一切是多么奇妙啊。
How deeply we fall in love, how deeply our hearts get broken, how afraid we are of death and how much pain we feel when our loved ones pass away.
我们爱得有多深，我们的心碎得有多深，我们对死亡有多恐惧，当我们所爱的人去世时，我们感到有多痛苦。
All of that, the whole thing.
所有这一切，整个事情。
I know it’s very hard to- …
我知道这很难——……
think AI being able to…
think AI being able to…
A computational device being able to do that.
A computational device being able to do that.
But there’s so many mysteries about this whole thing that we’re yet to uncover, that I am open to be surprised.
但这整件事还有很多谜团我们尚未揭开，所以我很惊讶。
I’ve been surprised a lot over the past-
I’ve been surprised a lot over the past-
2:13:23 …
……
few months and few years.
few months and few years.
Scaling can create some incredible miracles in the space of intelligence.
扩展可以在智能领域创造一些令人难以置信的奇迹。
It has been truly marvelous to watch, so I’m open to surprise.
观看起来真是太棒了，所以我愿意接受惊喜。
And it’s just really important to break down what is intelligence.
分解什么是智力非常重要。
2:13:34 You know, the word, that word we use all the time, it’s not a mysterious word.
你知道，这个词，这个我们一直使用的词，它并不是一个神秘的词。
Intelligence has a meaning, you know?
智力是有意义的，你知道吗？
2:13:46 And it’s a system that…
And it’s a system that…
You know, it’s something that we do that includes perception and understanding and reasoning and the ability to do plan.
你知道，我们所做的事情包括感知、理解、推理以及制定计划的能力。
And, you know, that loop, that loop, is the…
And, you know, that loop, that loop, is the…
Fundamentally what intelligence is.
Fundamentally what intelligence is.
Intelligence is not one word that is exactly equal to humanity.
智能这个词并不完全等同于人类。
And that’s, I think it’s really important to separate the two.
也就是说，我认为将两者分开非常重要。
We have two words for that.
We have two words for that.
I’m not…
我不是……
I don’t over-fantasize about, and I don’t over-romanticize about intelligence.
我不会对智力过度幻想，也不会过度浪漫化。
Intelligence is…
智力是……
And people have heard me say it before, I actually think intelligence is a commodity.
人们以前听我说过，我实际上认为智力是一种商品。
I’m surrounded by intelligent people.
我周围都是聪明人。
And I’m surrounded by intelligent people more intelligent than I am in each one of the spaces that they’re in.
我周围都是聪明人，他们所在的每一个空间都比我更聪明。
2:14:39 And yet, I have a role in that circle.
然而，我在那个圈子里扮演着一个角色。
It’s actually kind of interesting.
It’s actually kind of interesting.
They’re more educated than I am.
他们比我受过更多教育。
They went to better schools than I did.
他们去了比我更好的学校。
They’re deeper in any of the fields that they’re in.
他们在自己所在的任何领域都更深入。
All of them.
所有的人。
I have 60 of them.
我有 60 个。
They’re all superhuman to me.
他们对我来说都是超人。
And somehow, I’m sitting in the middle orchestrating all 60 of them.
不知何故，我坐在中间编排这 60 个人。
And so you gotta ask yourself…
所以你必须问自己……
What is it about a dishwasher that allows that dishwasher to sit in the middle of superhumans?
洗碗机到底是怎么回事，能让洗碗机坐在超人中间？
Does that make sense?
这有道理吗？
2:15:15 And so, but that’s my point.
所以，但这就是我的观点。
My point is intelligence is a functional thing.
我的观点是智力是一种功能性的东西。
Humanity is not specified functionally.
人性不是在功能上指定的。
It’s a much, much bigger word.
这是一个非常非常大的词。
And our life experience, our tolerance for pain, our determination, those are different words than intelligence.
我们的生活经历、我们对痛苦的忍耐、我们的决心，这些都是与智慧不同的词。
And so the thing that I wanna help the audience understand, if I could give them one thing, is intelligence is a word that we’ve elevated to a very high form over time.
因此，如果我能给观众一件事，我想帮助他们理解的是，随着时间的推移，智力这个词我们已经提升到了一个非常高的形式。
The word we should really elevate is humanity.
我们真正应该提升的词是人性。
Character, humanity.
性格，人性。
All those things.
所有这些事情。
All of those things.
所有这些事情。
2:15:55 Compassion, generosity, all of the things that you say just now, I believe those are superhuman powers.
慈悲、慷慨，你刚才所说的一切，我相信都是超人的力量。
And that now intelligence is gonna be commoditized.
现在情报将被商品化。
Because we’ve spoken about it, the most important thing is your education.
因为我们已经讨论过，最重要的是你的教育。
Now, even when they said the most important thing is your education, when you went to school, there’s more than just knowledge that you gained.
现在，即使他们说最重要的是你的教育，当你上学时，你获得的不仅仅是知识。
2:16:22 And so, but unfortunately, our society had put everything into one single word, and life is more than one word.
因此，不幸的是，我们的社会已将一切都集中在一个词中，而生活不仅仅是一个词。
And I’m just telling you, my life would suggest that being lower on the intelligence curve than everybody around me doesn’t change the fact I’m the most successful.
我只是告诉你，我的生活表明，在智力曲线上比我周围的每个人都低并不能改变我是最成功的事实。
And so, and I think that kind of is—I’m trying to hopefully to inspire everybody else—that don’t let this democratization of intelligence, this commoditization of intelligence, cause you anxiety.
所以，我认为，我希望能够激励其他人，不要让这种情报民主化、这种情报商品化引起你的焦虑。
You should be inspired by that.
你应该会受到启发。
Yeah.
是的。
2:17:00 I think AI will help us celebrate humans more.
我认为人工智能将帮助我们更多地庆祝人类。
And certainly humanity and human first, and I think what makes this world incredible is humans forever will be so, and just AI is this incredible tool that makes us-
当然，人性和人类优先，我认为让这个世界变得不可思议的是人类永远如此，而人工智能就是这个令人难以置信的工具，它使我们——
That’s exactly right.
完全正确。
…
……
2:17:18 humans more powerful.
人类更强大。
That’s exactly right.
完全正确。
So much of the success of NVIDIA and the lives of millions of people that I mentioned depend on you.
NVIDIA 的成功以及我提到的数百万人的生活很大程度上取决于您。
2:17:21 But you’re just one human, like we mentioned, a mortal like all of us.
但正如我们提到的，你只是一个人，和我们所有人一样都是凡人。
Do you think about your mortality?
你考虑过你的死亡率吗？
Are you afraid of death?
你害怕死亡吗？
I really don’t wanna die.
我真的不想死。
2:17:42 I have a great life.
我的生活很美好。
I have a great family.
我有一个美好的家庭。
I have really important work.
我有非常重要的工作。
This is not a once in a lifetime experience suggests that it has been experienced by many people, just not one person.
这不是一生一次的经历，表明很多人都经历过这种经历，但不是一个人经历过。
This is a once in a humanity experience, what I’m going through.
这是人类难得一见的经历，也是我正在经历的。
NVIDIA is one of the most consequential technology companies in history.
NVIDIA 是历史上最有影响力的科技公司之一。
We’re doing very important work.
我们正在做非常重要的工作。
I take it very seriously.
我非常认真地对待它。
And so some of the things that of course are practical things, like how do we think about succession planning?
当然，有些事情是实际的事情，比如我们如何考虑继任计划？
And I’m famous in saying that I don’t believe in succession planning.
我曾说过，我不相信继任计划。
Man.
伙计。
And the reason for that isn’t because I’m immortal.
其原因并不是因为我是不朽的。
2:18:36 The reason for that is because if you’re worried about succession planning, if you’re worried all that anxiety of succession planning, then what should you do about it?
原因是因为如果你担心继任计划，如果你担心继任计划的所有焦虑，那么你应该怎么做？
Then you break it all the way back down.
然后你把它全部打破。
The most important thing you should do today, if you care about the future of your company, post you, is to pass on knowledge, information, insight, skills, experience as often and continuously as you can, which is the reason why I continuously reason about everything in front of my team.
如果你关心公司的未来，你今天应该做的最重要的事情就是尽可能频繁、持续地传递知识、信息、洞察力、技能、经验，这就是我不断在团队面前推理一切的原因。
Every single meeting is a reasoning meeting.
每一次会议都是一次推理会议。
Every moment I spend inside a company, outside a company is about passing on knowledge to people as fast as I can.
我在公司内部和外部度过的每一刻都是为了尽快向人们传递知识。
2:19:23 Nothing I learn ever sits on my desk longer than, you know, a fraction of a second.
我学到的东西在我的桌子上停留的时间不会超过，你知道，不到一秒钟。
I’m passing that information, that knowledge—oh my gosh, this is cool.
我正在传递这些信息、知识——天哪，这太酷了。
Before I even finish learning all of it myself, I’m already pointing it to somebody else.
在我自己学完所有内容之前，我已经将其介绍给其他人了。
“Get on this.
“开始吧。
This is so cool.
这太酷了。
You’re gonna wanna learn this.” And so I’m constantly passing knowledge, empowering people, elevating the capability of everybody around me, so that the outcome that I seek, that I hope for, is that I die on the job, you know?
你一定会想学这个的。”所以我不断地传递知识，赋予人们权力，提升我周围每个人的能力，所以我所寻求的、我所希望的结果就是我死在工作中，你知道吗？
And hopefully I die on the job instantaneously, you know?
希望我能立即死在工作中，你知道吗？
And there’s no long periods of suffering, you know?
而且没有长期的痛苦，你知道吗？
It’s, uh –
这是，呃——
Well, from a fan perspective, given your extremely enormous positive impact on civilization, of course, I hope you keep going.
那么，从粉丝的角度来看，鉴于你对文明产生了极其巨大的积极影响，我当然希望你继续走下去。
2:20:06 But also it’s just fun to watch what NVIDIA is doing, you know.
但你知道，观看 NVIDIA 正在做的事情也很有趣。
It’s just the rate of innovation.
这只是创新率。
And I’m a huge fan of engineering.
我是工程学的忠实粉丝。
There’s so much incredible engineering continuously being done by NVIDIA.
NVIDIA 不断完成许多令人难以置信的工程。
It’s just fun to watch.
观看起来很有趣。
It’s a celebration of humanity, a celebration of great builders, a celebration of great engineering.
这是对人类的庆祝，对伟大建设者的庆祝，对伟大工程的庆祝。
So, it represents something special.
所以，它代表了一些特别的东西。
So I hope you and NVIDIA keep going.
所以我希望您和 NVIDIA 继续前进。
What gives you hope about this whole thing we got going on, about humanity, about the future of humanity?
是什么让你对我们正在发生的这一切、对人类、对人类的未来充满希望？
When you look out, when you think about the future quite a bit, when you look out 10, 20, 50, 100 years from now, what gives you hope?
当你展望未来，当你深入思考未来，当你展望 10、20、50、100 年后，什么给你带来希望？
I’ve always had a great confidence in the kindness, the generosity, the compassion, the human capacity.
我一直对善良、慷慨、同情心和人类的能力充满信心。
2:20:56 I’ve always been extremely confident of that.
我对此一直非常有信心。
Sometimes more so than I should.
有时比我应该做的还要多。
And I get taken advantage of, but it doesn’t ever cause me not to.
我被利用了，但这并不会让我不这么做。
I start with always that people want to do good.
我总是从人们想做善事开始。
People want to help others.
人们想要帮助别人。
And vastly, I am proven right.
在很大程度上，事实证明我是对的。
Constantly proven right.
不断被证明是正确的。
And often it exceeds my expectations.
而且常常超出我的预期。
And so I have complete confidence in the human capacity.
所以我对人类的能力充满信心。
I think the things that give me incredible hope is what I see now as possible, and as I extrapolate based on the things that we’re doing, what will very likely happen.
我认为给我难以置信的希望的事情是我现在看到的可能的事情，并且当我根据我们正在做的事情推断时，很可能会发生什么。
2:22:22 And that there’s so many things that we wanna solve.
我们想要解决的事情有很多。
There’s so many problems we wanna solve.
我们想要解决的问题太多了。
There’s so many things that we wanna build.
我们想要建造很多东西。
There’s so many good things that we wanna do that are now within our reach, and within the reach of my lifetime.
我们想做的好事有很多，现在就在我们的能力范围之内，在我的有生之年力所能及的范围内。
You just can’t possibly not be romantic about that.
你不可能对此不浪漫。
You know what I’m saying?
你知道我在说什么吗？
O-
欧-
What an exciting time to be alive.
活着是多么令人兴奋的时刻啊。
2:22:46 Like, truly-
就像，真的——
How can-
怎么可以——
…
……
2:22:49 truly so.
确实如此。
How can you not be romantic about that?
你怎么能不浪漫呢？
2:22:50 The fact that there is a—it’s a reasonable thing to expect the end of disease.
事实上，预期疾病的终结是合理的。
It’s a reasonable thing to expect.
这是一个合理的预期。
It’s a reasonable thing to expect that pollution will be drastically reduced.
预计污染将大幅减少是合理的。
It’s a reasonable thing to expect that traveling at the speed of light is actually in our future.
可以合理地预期，以光速旅行实际上就在我们的未来。
And then, you know, not for long distances, but short distances.
然后，你知道，不是长距离，而是短距离。
You know, and people ask me how.
你知道，人们问我怎么做。
Well, first of all, very soon, I’m gonna put a humanoid on a spaceship, and it’s gonna be, you know, my humanoid, and we’re gonna send it out as soon as possible, and it’s gonna keep improving and enhancing along the flight.
好吧，首先，很快，我将把一个人形机器人放在宇宙飞船上，这将是，你知道，我的人形机器人，我们将尽快将其发送出去，并且它将在飞行过程中不断改进和增强。
2:23:36 And then when it’s time, all of my consciousness has already been—you know, so much of my life has been uploaded in the internet.
然后到了时候，我所有的意识都已经——你知道，我生活的大部分内容都已经上传到互联网上。
Take all my inbox, take everything that I’ve done, everything I’ve said.
拿走我所有的收件箱，拿走我做过的一切，说过的一切。
You know, it’s been collected and becoming my AI.
你知道，它已被收集并成为我的人工智能。
And I’m just, when the time comes, we’ll just send that at the speed of light, catch up with my robot.
我只是，到时候，我们就会以光速发送，赶上我的机器人。
Oh, that’s brilliant.
哦，那太棒了。
2:24:00 I mean, but for me, that’s sorta application-focused.
我的意思是，但对我来说，这有点以应用程序为中心。
But also, for me, the curiosity-maxing perspective, I just, all of those mysteries.
但对我来说，好奇心最大化的视角，我只是，所有这些谜团。
There’s so much- …
有这么多——...
fascinating scientific questions there.
那里有令人着迷的科学问题。
Understanding the biological machine is right around the corner.
了解生物机器指日可待。
2:24:14 It’s, it’s not 10 years.
是的，这不是10年。
It’s five years probably.
大概是五年了。
And then your biological machine, the, the human mind and cracking physics, theoretical physics open.
然后你的生物机器，人类思维和破解物理，理论物理打开。
2:24:20 It’s so exciting.
太令人兴奋了。
Explaining consciousness, that one would be awesome.
解释意识，那就太棒了。
And it’s all within our reach.
而这一切都在我们力所能及的范围内。
2:24:29 Jensen, thank you so much for everything you’ve done over the years.
詹森，非常感谢你这些年来所做的一切。
Thank you for everything you’re doing for the world.
感谢您为世界所做的一切。
Thank you for being who you are.
感谢你做你自己。
I can tell you’re a great human being, and I wish you incredible success this year.
我可以看出你是一个伟大的人，我祝愿你今年取得令人难以置信的成功。
I can’t wait.
我等不及了。
As a fan, I can’t wait to see what you do next, and hopefully I’ll see you in Taiwan and thank you so much for talking today.
作为一名粉丝，我迫不及待地想看看你下一步会做什么，希望能在台湾见到你，非常感谢你今天的发言。
Thank you, Lex.
谢谢你，莱克斯。
2:24:52 I had a great time.
我玩得很开心。
And also, if I could just say one more thing.
另外，我还能再说一件事吗？
And thank you for all the interviews that you do, the depth, the respect that you go through with and the research that you do to reveal, you know, for all of us the amazing people that you’ve interviewed over the years.
感谢您所做的所有采访，感谢您所经历的深度、尊重以及您所做的研究，您知道，您为我们所有人揭示了您多年来采访过的那些了不起的人。
I’ve enjoyed them immensely.
我非常喜欢它们。
And as an innovator, to have created this long form, unbelievable, and yet, you know, it’s just captivating.
作为一名创新者，创造出如此长的形式，令人难以置信，但你知道，它只是令人着迷。
So anyways, thank you for everything you do.
所以无论如何，感谢你所做的一切。
It means the world.
它意味着世界。
2:25:25 Thank you, Jensen.
谢谢你，詹森。
Thank you, Lex.
谢谢你，莱克斯。
Thank you for listening to this conversation with Jensen Huang.
感谢您收听黄仁勋的对话。
2:25:29 To support this podcast, please check out our sponsors in the description, where you can also find links to contact me, ask questions, give feedback, and so on.
为了支持这个播客，请在描述中查看我们的赞助商，您还可以在其中找到联系我、提出问题、提供反馈等的链接。
And now, let me leave you with some words from Alan Kay.
现在，让我向你们传达艾伦·凯的话。
“The best way to predict the future is to invent it.” Thank you for listening, and hope to see you next time.
“预测未来的最好方法就是发明它。”感谢您的收听，希望下次再见。
本集资源与出处
- 视频在 YouTube 收看
- 音频官方 MP3 直链来自 lexfridman.com 官方播客 RSS
- 官网Lex Fridman 官网本集页面
- 文稿官方 Transcript（lexfridman.com）
