來(lái)源：華爾街見(jiàn)聞

　　黃仁勛表示，沒(méi)有物理定律限制AI數(shù)據(jù)中心擴(kuò)展到百萬(wàn)芯片，我們現(xiàn)在可以將AI軟件擴(kuò)展到多個(gè)數(shù)據(jù)中心運(yùn)行。我們已經(jīng)為能夠在一個(gè)前所未有的水平上擴(kuò)展計(jì)算做好了準(zhǔn)備，而且我們現(xiàn)在才剛剛開(kāi)始。在未來(lái)十年，計(jì)算性能每年將翻倍或翻三倍，而能源需求每年將減少2-3倍，我稱之為超摩爾定律曲線。

　　本周，英偉達(dá)CEO黃仁勛接受了《No Priors》節(jié)目主持人的采訪，就英偉達(dá)的十年賭注、x.AI超級(jí)集群的快速發(fā)展、NVLink技術(shù)創(chuàng)新等AI相關(guān)話題進(jìn)行了一場(chǎng)深度對(duì)話。

　　黃仁勛表示，沒(méi)有任何物理定律可以阻止將AI數(shù)據(jù)中心擴(kuò)展到一百萬(wàn)個(gè)芯片，盡管這是一個(gè)難題，多家大公司包括OpenAI、Anthropic、谷歌、Meta和微軟等，都在爭(zhēng)奪AI領(lǐng)域的領(lǐng)導(dǎo)地位，競(jìng)相攀登技術(shù)的高峰，但重新創(chuàng)造智能的潛在回報(bào)是如此之大，以至于不能不去嘗試。

　　摩爾定律曾是半導(dǎo)體行業(yè)發(fā)展的核心法則，預(yù)測(cè)芯片的晶體管數(shù)目每?jī)赡陼?huì)翻倍，從而帶來(lái)性能的持續(xù)提升。然而，隨著物理極限的接近，摩爾定律的速度開(kāi)始放緩，芯片性能提升的瓶頸逐漸顯現(xiàn)。

　　為了解決這一問(wèn)題，英偉達(dá)將不同類型的處理器（如GPU、TPU等）結(jié)合起來(lái)，通過(guò)并行處理來(lái)突破傳統(tǒng)摩爾定律的限制。黃仁勛表示，未來(lái)10年，計(jì)算性能每年將翻一番或三倍，而能源需求每年將減少2-3倍，我稱之為“超摩爾定律曲線”。

　　黃仁勛還提到，我們現(xiàn)在可以將AI軟件擴(kuò)展到多個(gè)數(shù)據(jù)中心：“我們已經(jīng)做好準(zhǔn)備，能夠?qū)⒂?jì)算擴(kuò)展到前所未有的水平，而我們正處于這一領(lǐng)域的起步階段?！?/p>

　　以下是黃仁勛講話的亮點(diǎn)：

1．我們?cè)谖磥?lái)10年進(jìn)行了重大的投資。我們正在投資基礎(chǔ)設(shè)施，打造下一代AI計(jì)算平臺(tái)。我們?cè)谲浖?、架?gòu)、GPU以及所有實(shí)現(xiàn)AI開(kāi)發(fā)所需的組件上都進(jìn)行了投資。

2．摩爾定律，即晶體管數(shù)目每?jī)赡攴兜念A(yù)言，曾經(jīng)是半導(dǎo)體行業(yè)的增長(zhǎng)指南。然而，隨著物理極限的接近，摩爾定律已不再能夠單獨(dú)推動(dòng)芯片性能的提升。為了解決這一問(wèn)題，英偉達(dá)采用了類似于“異構(gòu)計(jì)算”的方式，即將不同類型的處理器（如GPU、TPU等）結(jié)合起來(lái)，通過(guò)并行處理來(lái)突破傳統(tǒng)摩爾定律的限制。英偉達(dá)的技術(shù)創(chuàng)新，如CUDA架構(gòu)和深度學(xué)習(xí)優(yōu)化，使得AI應(yīng)用得以在超越摩爾定律的環(huán)境中高速運(yùn)行。

3．我們推出了NVLink作為互連技術(shù)，它使得多個(gè)GPU能夠協(xié)同工作，每個(gè)GPU處理工作負(fù)載的不同部分。通過(guò)NVLink，GPU之間的帶寬和通信能力大幅提升，使得數(shù)據(jù)中心能夠擴(kuò)展并支持AI工作負(fù)載。

4．未來(lái)的AI應(yīng)用需要?jiǎng)討B(tài)和彈性強(qiáng)的基礎(chǔ)設(shè)施，能夠適應(yīng)各種規(guī)模和類型的AI任務(wù)。因此，英偉達(dá)致力于構(gòu)建可以靈活配置和高效運(yùn)營(yíng)的基礎(chǔ)設(shè)施，滿足從中小型AI項(xiàng)目到超大規(guī)模超級(jí)計(jì)算集群的需求。

5．構(gòu)建AI數(shù)據(jù)中心的關(guān)鍵是要同時(shí)優(yōu)化性能和效率。在AI工作負(fù)載中，你需要巨大的電力，而散熱成為一個(gè)巨大的問(wèn)題。所以我們花了大量時(shí)間優(yōu)化數(shù)據(jù)中心的設(shè)計(jì)和運(yùn)營(yíng)，包括冷卻系統(tǒng)和電力效率。

6．在硬件快速發(fā)展的背景下，保持軟件與硬件架構(gòu)的兼容性顯得尤為重要。黃仁勛提到，我們必須確保我們的軟件平臺(tái)，如CUDA，可以跨代硬件使用。開(kāi)發(fā)者不應(yīng)當(dāng)每次我們推出新芯片時(shí)都被迫重寫(xiě)代碼。因此，我們確保保持向后兼容，并讓軟件能夠在我們開(kāi)發(fā)的任何新硬件上高效運(yùn)行。

7．我們正在建設(shè)一個(gè)超級(jí)集群，叫做X.AI，它將成為世界上最大的AI超級(jí)計(jì)算平臺(tái)之一。這個(gè)超級(jí)集群將提供支持一些最雄心勃勃的AI項(xiàng)目所需的計(jì)算能力。這是我們推動(dòng)AI前進(jìn)的一大步。

8．?dāng)U展AI數(shù)據(jù)中心的一個(gè)大挑戰(zhàn)是管理它們消耗的巨大能源。問(wèn)題不僅僅是構(gòu)建更大、更快的系統(tǒng)。我們還必須處理運(yùn)行這些超大規(guī)模系統(tǒng)時(shí)面臨的熱量和電力挑戰(zhàn)。為了應(yīng)對(duì)這一切，需要?jiǎng)?chuàng)新的工程技術(shù)來(lái)確?；A(chǔ)設(shè)施能夠應(yīng)對(duì)。

9.AI在芯片設(shè)計(jì)中的作用日益重要，黃仁勛指出，AI已經(jīng)在芯片設(shè)計(jì)中發(fā)揮著重要作用。我們使用機(jī)器學(xué)習(xí)來(lái)幫助設(shè)計(jì)更高效的芯片，速度更快。這是我們?cè)O(shè)計(jì)下一代英偉達(dá)芯片的一個(gè)關(guān)鍵部分，并幫助我們構(gòu)建專為AI工作負(fù)載優(yōu)化的芯片。

10．英偉達(dá)市值的激增是因?yàn)槲覀兡軌驅(qū)⒐巨D(zhuǎn)型為AI公司。我們從一開(kāi)始是GPU公司，但我們已經(jīng)轉(zhuǎn)型成了AI計(jì)算公司，這一轉(zhuǎn)型是我們市值增長(zhǎng)的關(guān)鍵部分。AI技術(shù)的需求正在迅速增長(zhǎng)，我們處在一個(gè)能夠滿足這一需求的有利位置。

11．具象化AI是指將AI與物理世界進(jìn)行結(jié)合。通過(guò)這種方式，AI不僅可以在虛擬環(huán)境中進(jìn)行任務(wù)處理，還能在現(xiàn)實(shí)世界中進(jìn)行決策并執(zhí)行任務(wù)。具象化AI將推動(dòng)智能硬件、自動(dòng)駕駛等技術(shù)的快速發(fā)展。

12.AI不僅僅是工具，它也可以成為‘虛擬員工’，幫助提升工作效率。AI能夠在數(shù)據(jù)處理、編程、決策等領(lǐng)域替代或輔助人類工作，進(jìn)而改變整個(gè)勞動(dòng)市場(chǎng)和工作方式。

13.AI將在科學(xué)與工程領(lǐng)域產(chǎn)生巨大影響，特別是在藥物研發(fā)、氣候研究、物理實(shí)驗(yàn)等領(lǐng)域。AI將幫助科學(xué)家處理大量數(shù)據(jù)，揭示新的科學(xué)規(guī)律，并加速創(chuàng)新。它還將在工程領(lǐng)域優(yōu)化設(shè)計(jì)，提高效率，推動(dòng)更具創(chuàng)新性的技術(shù)發(fā)展。

14．我自己也在日常工作中使用AI工具，來(lái)提高效率和創(chuàng)造力。我認(rèn)為，AI不僅能夠幫助我們處理復(fù)雜的數(shù)據(jù)和決策任務(wù)，還能提升我們的創(chuàng)意思維和工作效率，成為每個(gè)人工作中不可或缺的一部分。

　　以下是采訪文字實(shí)錄全文，由AI翻譯：

主持人：Welcome back， Johnson， 30 years in to Nvidia and looking 10 years out， what are the big bets you think are still to make？ Is it all about scale up from here？ Are we running into limitations in terms of how we can squeeze more compute memory out of the architectures we have？ What are you focused on？ Well．

嗨，Johnson，歡迎回來(lái)！你在英偉達(dá)工作了30年，展望未來(lái)10年，你認(rèn)為還有哪些重要的投資機(jī)會(huì)？是不是說(shuō)我們只需要繼續(xù)擴(kuò)大規(guī)模？我們?cè)诂F(xiàn)有架構(gòu)中是否會(huì)遇到限制，無(wú)法再擠出更多的計(jì)算內(nèi)存？你目前關(guān)注的重點(diǎn)是什么？

黃仁勛：If we take a step back and think about what we‘ve done， we went from coding to machine learning， from writing software tools to creating AIs and all of that running on CPUs that was designed for human coding to now running on GPUs designed for AI coding， basically machine learning． And so the world has changed the way we do computing the whole stack has changed． And as a result， the scale of the problems we could address has changed a lot because we could， if you could paralyze your software on one GPU， you’ve set the foundations to paralyze across a whole cluster or maybe across multiple clusters or multiple data centers． And so I think we‘ve set ourselves up to be able to scale computing at a level and develop software at a level that nobody’s ever imagined before． And so we‘re at the beginning that over the next 10 years， our hope is that we could double or triple performance every year at scale， not at chip， at scale． And to be able to therefore drive the cost down by a factor of 2 or 3， drive the energy down by a factor of 2，3 every single year． When you do that every single year， when you double or triple every year in just a few years， it adds up． So it compounds really aggressively． And so I wouldn’t be surprised if， you know， the way people think about Moore‘s Law， which is 2 x every couple of years， you know， we’re gonna be on some kind of a hyper Moore‘s Law curve． And I fully hope that we continue to do that． Well， what．

以前我們編程都是靠自己寫(xiě)代碼，現(xiàn)在我們開(kāi)始讓機(jī)器自己學(xué)習(xí)，自己寫(xiě)代碼。以前我們用的那種電腦芯片（CPU）是給人寫(xiě)代碼用的，現(xiàn)在我們用的電腦芯片（GPU）是給機(jī)器學(xué)習(xí)用的。因?yàn)檫@些變化，我們現(xiàn)在處理問(wèn)題的方式和以前完全不一樣了。打個(gè)比方，如果你能讓一個(gè)機(jī)器學(xué)習(xí)程序在一個(gè)GPU上運(yùn)行，那你就可以讓它在整個(gè)電腦群里，甚至在很多電腦群或者數(shù)據(jù)中心里運(yùn)行。這意味著我們現(xiàn)在能處理的問(wèn)題比以前大多了。所以，我們相信自己已經(jīng)建立了能夠大規(guī)模擴(kuò)展計(jì)算能力和開(kāi)發(fā)軟件的基礎(chǔ)，這個(gè)規(guī)模是以前沒(méi)人想象過(guò)的。

我們希望在未來(lái)10年里，每年都能讓計(jì)算能力翻兩倍或者三倍，不是單個(gè)芯片的能力，而是整體的能力。這樣的話，我們就能每年把計(jì)算成本降低兩倍或三倍，把能耗也降低兩倍或三倍。這種增長(zhǎng)如果每年都能實(shí)現(xiàn)，那么幾年下來(lái)，這個(gè)增長(zhǎng)會(huì)非常驚人。因此，我認(rèn)為未來(lái)的計(jì)算將會(huì)超越傳統(tǒng)的“摩爾定律”（即每?jī)赡暧?jì)算能力翻倍），可能會(huì)走上一條更快的增長(zhǎng)曲線，我也非常希望能夠繼續(xù)沿著這個(gè)方向前進(jìn)。

主持人：Do you think is the driver of making that happen even faster than Morse law？ Cuz I know morezo was sort of self reflexive， right？ It was something that he said and then people kind of implemented it to me to happen．

你認(rèn)為是什么因素推動(dòng)了計(jì)算能力增長(zhǎng)速度超過(guò)摩爾定律的？因?yàn)槲抑溃柖杀旧砭褪且环N“自我實(shí)現(xiàn)”的規(guī)律，對(duì)吧？也就是說(shuō)，摩爾定律本身是摩爾提出的，然后大家就按照這個(gè)規(guī)律去做，結(jié)果它就實(shí)現(xiàn)了。

黃仁勛：Yep， too． Fundamental technical pillars． One of them was Denard scaling and the other one was Carver Mead‘s VLSI scaling． And both of those techniques were rigorous techniques， but those techniques have really run out of steam． And， and so now we need a new way of doing scaling． You know， obviously the new way of doing scaling are all kinds of things associated with co design． Unless you can modify or change the algorithm to reflect the architecture of the system or change and then change the system to reflect the architecture of the new software and go back and forth． Unless you can control both sides of it， you have no hope． But if you can control both sides of it， you can do things like

move from FP64 to FP32 to BF16 to FPA to， you know， FP4 to who knows what， right？ And so， and so I think that code design is a very big part of that． The second part of it， we call it full stack． The second part of it is data center scale． You know， unless you could treat the network as a compute fabric and push a lot of the work into the network， push a lot of the work into the fabric． And as a result， you‘re compressing， you know， doing compressing at very large scales． And so that’s the reason why we bought Melanox and started fusing infinite and MV Link in such an aggressive way．

過(guò)去推動(dòng)技術(shù)進(jìn)步的兩個(gè)關(guān)鍵技術(shù)柱子是Denard縮放（Denard Scaling）和Carver Mead的VLSI縮放。但是這兩種方法現(xiàn)在都不太管用了，我們需要新的方法來(lái)變得更快。

新方式就是“協(xié)同設(shè)計(jì)”（co-design），也就是軟件和硬件必須同時(shí)考慮和優(yōu)化。具體來(lái)說(shuō)，如果你不能修改或調(diào)整算法，使其與系統(tǒng)的架構(gòu)匹配，或者不能改變系統(tǒng)架構(gòu)，以適應(yīng)新軟件的需求，那么就沒(méi)有希望。但如果你能同時(shí)控制軟件和硬件，你就能做很多新的事情，比如：從高精度的FP64轉(zhuǎn)到低精度的FP32，再到BF16、FPA、甚至FP4等更低精度的計(jì)算。

這就是為什么“協(xié)同設(shè)計(jì)”這么重要的原因。另外，另一個(gè)重要的部分是全棧設(shè)計(jì)。這意味著，你不僅要考慮硬件，還要考慮數(shù)據(jù)中心級(jí)別的規(guī)模。比如，必須把網(wǎng)絡(luò)當(dāng)作一個(gè)計(jì)算平臺(tái)來(lái)使用，把大量的計(jì)算任務(wù)推到網(wǎng)絡(luò)里，利用網(wǎng)絡(luò)和硬件進(jìn)行大規(guī)模壓縮運(yùn)算。

因此，我們收購(gòu)了Mellanox，并開(kāi)始非常積極地推動(dòng)InfiniBand和NVLink這類高速連接技術(shù)，來(lái)支持這種全新的大規(guī)模計(jì)算架構(gòu)。

And now look where MV Link is gonna go． You know， the compute fabric is going to， I scale out what appears to be one incredible processor called a GPU． Now we get hundreds of GPUs that are gonna be working together.And now look where MV Link is gonna go． You know， the compute fabric is going to， I scale out what appears to be one incredible processor called a GPU． Now we get hundreds of GPUs that are gonna be working together.You know， most of these computing challenges that we‘re dealing with now， one of the most exciting ones， of course， is inference time scaling， has to do with essentially generating tokens at incredibly low latency because you’re self reflecting， as you just mentioned． I mean， you‘re gonna be doing tree surge， you’re gonna be doing chain of thought， you‘re gonna be doing probably some amount of simulation in your head． You’re gonna be reflecting on your own answers． Well， you‘re gonna be prompting yourself and generating text to your in， you know， silently and still respond hopefully in a second． Well， the only way to do that is if your latency is extremely low.Meanwhile， the data center is still about producing high throughput tokens because you know， you still wanna keep cost down， you wanna keep the throughput high， you wanna， right， you know， and generate a return． And so these two fundamental things about a factory， low latency and high throughput， they’re at odds with each other． And so in order for us to create something that is really great in both， we have to go invent something new， and Envy Link is really our way of doing that.We now you have a virtual GPU that has incredible amount of flops because you need it for context． You need a huge amount of memory， working memory， and still have incredible bandwidth for token generation all of the same time．

現(xiàn)在看NVLink（英偉達(dá)的高速連接技術(shù)）將走向哪里，未來(lái)的計(jì)算架構(gòu)將變得非常強(qiáng)大。你可以把它想象成一個(gè)超級(jí)強(qiáng)大的處理器，就是GPU（圖形處理單元）。而現(xiàn)在，英偉達(dá)的目標(biāo)是把數(shù)百個(gè)GPU集成到一起，協(xié)同工作，形成一個(gè)龐大的計(jì)算平臺(tái)。

現(xiàn)在我們面臨的計(jì)算挑戰(zhàn)中，有一個(gè)非常令人興奮的問(wèn)題就是推理時(shí)間的縮短。特別是在生成文本時(shí)，需要非常低的延遲。因?yàn)榫拖衲銊偛盘岬降?，我們的思維其實(shí)是一種自我反思的過(guò)程：你可能在腦海中進(jìn)行“樹(shù)形搜索”（tree search）、思考鏈條（chain of thought），甚至可能會(huì)進(jìn)行某種模擬，回顧自己的答案。你會(huì)自己給自己提問(wèn)，并生成答案，在大腦里“默默地”思考，然后希望能在幾秒鐘內(nèi)回應(yīng)出來(lái)。

為了做到這一點(diǎn)，計(jì)算的延遲必須非常低，因?yàn)槟悴豢赡艿忍貌拍艿玫浇Y(jié)果。

但與此同時(shí)，數(shù)據(jù)中心的任務(wù)是產(chǎn)生大量的高吞吐量的“token”（符號(hào)）。你需要控制成本，保持高吞吐量，并且確保能夠獲得回報(bào)。所以，低延遲和高吞吐量是兩個(gè)相互矛盾的目標(biāo)：低延遲要求快速響應(yīng)，而高吞吐量則需要處理更多的數(shù)據(jù)。這兩者之間存在沖突。

為了同時(shí)做到這兩點(diǎn)，必須創(chuàng)造一些全新的技術(shù)，而NVLink就是我們解決這個(gè)問(wèn)題的方法之一。通過(guò)NVLink，英偉達(dá)希望能夠在確保高吞吐量的同時(shí)，也能提供低延遲，從而解決這一計(jì)算上的矛盾，提升整體性能。

現(xiàn)在我們有了虛擬GPU，它的計(jì)算能力非常強(qiáng)大，因?yàn)槲覀冃枰@么強(qiáng)的計(jì)算能力來(lái)處理上下文。也就是說(shuō)，當(dāng)我們?cè)谔幚硪恍┤蝿?wù)時(shí)，需要非常大的內(nèi)存（特別是工作內(nèi)存），同時(shí)還要有極高的帶寬來(lái)生成token（即文本或數(shù)據(jù)符號(hào)）。

主持人：Building the models， actually also optimizing things pretty dramatically like David and my team pull data where over the last 18 months or so， the cost of 1 million tokens going into a GPT four equivalent model is basically dropped 240 x． Yeah， and so there‘s just massive optimization and compression happening on that side as．

構(gòu)建模型的過(guò)程其實(shí)也包括了很多優(yōu)化工作，比如David和他的團(tuán)隊(duì)，通過(guò)過(guò)去18個(gè)月的努力，成功地將每百萬(wàn)個(gè)token的成本（用于GPT-4類模型的成本）降低了240倍。

黃仁勛：Well． Just in our layer， just on the layer that we work on． You know， one of the things that we care a lot about， of course， is the ecosystem of our stack and the productivity of our software． You know， people forget that because you have Kuda Foundation and that‘s a solid foundation． Everything above it can change． If everything， if the foundation’s changing underneath you， it‘s hard to build a building on top． It’s hard to create anything and interesting on top． And so could have made it possible for us to iterate so quickly just in the last year． And then we just went back and benchmarked when Lama first came out， we‘ve improved the performance of Hopper by a factor of five without the algorithm， without the layer on top ever changing． Now， well， a factor of five in one year is impossible using traditional computing approaches． But it’s already computing and using this way of code design， we‘re able to explain all kinds of new things．

在我們的工作領(lǐng)域里，有一件非常重要的事情，就是技術(shù)棧的生態(tài)系統(tǒng)和軟件的生產(chǎn)力。我們特別重視的是Kuda Foundation這個(gè)基礎(chǔ)平臺(tái)，它是非常穩(wěn)定和堅(jiān)實(shí)的。因?yàn)槿绻A(chǔ)平臺(tái)不斷變化，想要在上面構(gòu)建出一個(gè)系統(tǒng)或者應(yīng)用就非常困難，根本無(wú)法在不穩(wěn)定的基礎(chǔ)上創(chuàng)造出有趣的東西。所以，Kuda Foundation的穩(wěn)定性讓我們能夠非常快速地進(jìn)行迭代和創(chuàng)新，尤其是在過(guò)去一年里。

然后，我們還做了一個(gè)對(duì)比測(cè)試：當(dāng)Lama首次推出時(shí)，我們通過(guò)優(yōu)化Hopper（一種計(jì)算平臺(tái)或架構(gòu)），在不改變算法和不改變上層架構(gòu)的情況下，提升了性能5倍。而且這種5倍的提升，在傳統(tǒng)的計(jì)算方法下是幾乎不可能實(shí)現(xiàn)的。但通過(guò)協(xié)同設(shè)計(jì)這種新的方法，我們能夠在現(xiàn)有的基礎(chǔ)上不斷創(chuàng)新和解釋更多新的技術(shù)可能性。

主持人：How much are， you know， your biggest customers thinking about the interchangeability of their infrastructure between large scale training and inference？

你的那些最大客戶有多關(guān)心他們?cè)诖笠?guī)模訓(xùn)練和推理之間基礎(chǔ)設(shè)施的互換性？

黃仁勛：Well， you know， infrastructure is disaggregated these days． Sam was just telling me that he had decommissioned Volta just recently． They have pascals， they have amperes， all different configurations of blackwall coming． Some of it is optimized for air cool， some of it‘s optimized liquid cool． Your services are gonna have to take advantage of all of this． The advantage that Nvidia has， of course， is that the infrastructure that you built today for training will just be wonderful for inference tomorrow． And most of Chat GBT， I believe， are inferenced on the same type of systems that we’re trained on just recently． And so you can train on， you can inference on it． And so you‘re leaving a trail of infrastructure that you know is going to be incredibly good at inference， and you have complete confidence that you can then take that return on it， on the investment that you’ve had and put it into a new infrastructure to go scale with， you know you‘re gonna leave behind something of use and you know that Nvidia and the rest of the ecosystem are gonna be working on improving the algorithm so that the rest of your infrastructure improves by a factor of five， you know， in just a year． And so that motion will never change．

現(xiàn)在的基礎(chǔ)設(shè)施不像以前那樣是一成不變的了。比如Sam剛告訴我，他們最近淘汰了Volta型號(hào)的設(shè)備。他們有Pascal型號(hào)的，有Ampere型號(hào)的，還有很多不同配置的Blackwall型號(hào)即將到來(lái)。有些設(shè)備是優(yōu)化了空氣冷卻的，有些則是優(yōu)化了液體冷卻的。你們的服務(wù)需要能夠利用所有這些不同的設(shè)備。

英偉達(dá)的優(yōu)勢(shì)在于，你今天為訓(xùn)練搭建的基礎(chǔ)設(shè)施，將來(lái)會(huì)非常適合用于推理。我相信大多數(shù)的Chat GBT（可能是指大型語(yǔ)言模型）都是在最近訓(xùn)練過(guò)的相同類型的系統(tǒng)上進(jìn)行推理的。所以你可以在這個(gè)系統(tǒng)上訓(xùn)練，也可以在這個(gè)系統(tǒng)上進(jìn)行推理。這樣，你就留下了一條基礎(chǔ)設(shè)施的軌跡，你知道這些基礎(chǔ)設(shè)施將來(lái)會(huì)非常適合進(jìn)行推理，你完全有信心可以把之前投資的回報(bào)，投入到新的基礎(chǔ)設(shè)施中去，擴(kuò)大規(guī)模。你知道你會(huì)留下一些有用的東西，而且你知道英偉達(dá)和整個(gè)生態(tài)系統(tǒng)都在努力改進(jìn)算法，這樣你的其他基礎(chǔ)設(shè)施在僅僅一年內(nèi)就能提高五倍的效率。所以這種趨勢(shì)是不會(huì)變的。

And so the way that people will think about the infrastructures， yeah， even though I built it for training today， it‘s gotta be great for training． We know it’s gonna be great for inference． Inference is gonna be multi scale． 說(shuō)話人 2 08：53 I mean， you‘re gonna take， first of all， in order to， the still smaller models could have a larger model that’s still from and so you‘re still gonna create these incredible a frontier models． They’re gonna be used for， of course， the groundbreaking work． You‘re gonna use it for synthetic data generation． You’re gonna use the models， the big models that teach smaller models and distill down to smaller models． And so there‘s a whole bunch of different things you can do， but in the end， you’re gonna have giant models all the way down to little tiny models． The little tiny models are gonna be quite effective， you know， not as generalizable， but quite effective． And so， you know， they‘re gonna perform very specific stunts incredibly well that one task． And we’re gonna see superhuman task in one little tiny domain from a little tiny model． Maybe you know， it‘s not a small language model， but you know， tiny language model， TLMs are， you know， whatever． Yeah， so I think we’re gonna see all kinds of sizes and we hope isn‘t right， just kind of like softwares today．

人們看待基礎(chǔ)設(shè)施的方式在變，就像我現(xiàn)在建的這個(gè)設(shè)施雖然是為了訓(xùn)練用的，但它也必須很適合訓(xùn)練。我們知道它將來(lái)也會(huì)非常適合做推理。推理會(huì)有很多不同的規(guī)模。

我是說(shuō)，你會(huì)有各種不同大小的模型。小模型可以從大模型那里學(xué)習(xí)，所以你還是會(huì)創(chuàng)造一些前沿的大模型。這些大模型會(huì)用來(lái)做開(kāi)創(chuàng)性的工作，用來(lái)生成合成數(shù)據(jù)，用來(lái)教小模型，然后把知識(shí)蒸餾給小模型。所以你可以做的事情有很多，但最后你會(huì)有從巨大的模型到非常小的模型。這些小模型將會(huì)非常有效，雖然它們不能通用，但在特定任務(wù)上會(huì)非常有效。它們會(huì)在某個(gè)特定任務(wù)上表現(xiàn)得非常好，我們將會(huì)看到在某個(gè)小小的領(lǐng)域里，小模型能完成超乎人類的任務(wù)。也許它不是一個(gè)小型的語(yǔ)言模型，但你知道，就是微型語(yǔ)言模型，TLMs，反正就是類似的東西。所以我覺(jué)得我們會(huì)看到各種大小的模型，就像現(xiàn)在的軟件一樣。

Yeah， I think in a lot of ways， artificial intelligence allows us to break new ground in how easy it is to create new applications． But everything about computing has largely remained the same． For example， the cost of maintaining software is extremely expensive． And once you build it， you would like it to run on a large of an install base as possible． You would like not to write the same software twice． I mean， you know， a lot of people still feel the same way． You like to take your engineering and move them forward． And so to the extent that， to the extent that the architecture allows you， on one hand， create software today that runs even better tomorrow with new hardware that‘s great or software that you create tomorrow， AI that you create tomorrow runs on a large install base． You think that’s great． That way of thinking about software is not gonna．

我覺(jué)得在很多方面，人工智能讓我們能夠更容易地創(chuàng)造新的應(yīng)用程序。但是在計(jì)算方面，大部分事情還是老樣子。比如說(shuō)，維護(hù)軟件的成本非常高。一旦你建好了軟件，你希望它能在盡可能多的設(shè)備上運(yùn)行。你不想重復(fù)寫(xiě)同樣的軟件。我的意思是，很多人還是這么想的。你喜歡把你的工程推向前進(jìn)。所以，如果架構(gòu)允許你，一方面，今天創(chuàng)建的軟件明天在新硬件上能運(yùn)行得更好，那就太好了；或者你明天創(chuàng)建的軟件，后天創(chuàng)建的人工智能能在很多設(shè)備上運(yùn)行。你認(rèn)為那很棒。這種考慮軟件的方式是不會(huì)變的。

主持人：Change． And video has moved into larger and larger， let‘s say， like a unit of support for customers． I think about it going from single chip to， you know， server to rack and real 72． How do you think about that progression？ Like what’s next？ Like should Nvidia do you full data center？ But

隨著技術(shù)的發(fā)展，英偉達(dá)的產(chǎn)品已經(jīng)不僅僅是單個(gè)的芯片了，而是擴(kuò)展到了支持整個(gè)數(shù)據(jù)中心的規(guī)模。你怎么看待這種發(fā)展？接下來(lái)會(huì)是什么？比如，英偉達(dá)是不是應(yīng)該做整個(gè)數(shù)據(jù)中心？

黃仁勛：In fact， we build full data centers the way that we build everything． Unless you‘re building， if you’re developing software， you need the computer in its full manifestation． We don‘t build Powerpoint slides and ship the chips and we build a whole data center． And until we get the whole data center built up， how do you know the software works until you get the whole data center built up， how do you know your， you know， your fabric works and all the things that you expected the efficiencies to be， how do you know it’s gonna really work at scale？ And that‘s the reason why it’s not unusual to see somebody‘s actual performance be dramatically lower than their peak performance， as shown in Powerpoint slides， and it is， computing is just not used to， is not what it used to be． You know， I say that the new unit of computing is the data center． That’s to us． So that‘s what you have to deliver． That’s what we build.Now we build a whole thing like that． And then we， for every single thing that every combination， air cold， x 86， liquid cold， Grace， Ethernet， infinite band， MV link， no NV link， you know what I‘m saying？ We build every single configuration． We have five supercomputers in our company today． Next year， we’re gonna build easily five more． So if you‘re serious about software， you build your own computers if you’re serious about software， then you‘re gonna build your whole computer． And we build it all at scale．

實(shí)際上，我們建造完整的數(shù)據(jù)中心就像我們建造其他所有東西一樣。如果你在開(kāi)發(fā)軟件，你需要電腦的完整形態(tài)來(lái)測(cè)試。我們不只是做PPT幻燈片然后發(fā)貨芯片，我們建造整個(gè)數(shù)據(jù)中心。只有當(dāng)我們把整個(gè)數(shù)據(jù)中心搭建起來(lái)后，你才能知道軟件是否正常工作，你的網(wǎng)絡(luò)布線是否有效，所有你期望的效率是否都能達(dá)到，你才知道它是否真的能在大規(guī)模上運(yùn)行。這就是為什么人們的實(shí)際性能通常遠(yuǎn)低于PPT幻燈片上展示的峰值性能，計(jì)算已經(jīng)不再是過(guò)去的樣子了。我說(shuō)現(xiàn)在的計(jì)算單元是數(shù)據(jù)中心，對(duì)我們來(lái)說(shuō)就是這樣。這就是你必須交付的東西，也是我們建造的東西。

我們現(xiàn)在就這樣建造整個(gè)系統(tǒng)。然后我們?yōu)槊恳环N可能的組合建造：空氣冷卻、x86架構(gòu)、液體冷卻、Grace芯片、以太網(wǎng)、無(wú)限帶寬、MVLink，沒(méi)有NVLink，你懂我的意思嗎？我們建造每一種配置。我們公司現(xiàn)在有五臺(tái)超級(jí)計(jì)算機(jī)，明年我們輕易就能再建造五臺(tái)。所以，如果你對(duì)軟件是認(rèn)真的，你就會(huì)自己建造計(jì)算機(jī)，如果你對(duì)軟件是認(rèn)真的，你就會(huì)建造整個(gè)計(jì)算機(jī)。我們都是大規(guī)模地建造。

This is the part that is really interesting． We build it at scale and we build it very vertically integrate． We optimize it full stack， and then we disagree everything and we sell lemon parts． That‘s the part that is completely， utterly remarkable about what we do． The complexity of that is just insane． And the reason for that is we want to be able to graft our infrastructure into GCP， AWS， Azure， OCI． All of their control planes， security planes are all different and all of the way they think about their cluster sizing all different． And， but yet we make it possible for them to all accommodate Nvidia’s architecture． So that could， it could be everywhere． That‘s really in the end the singular thought， you know， that we would like to have a computing platform that developers could use that’s largely consistent， modular， you know， 10% here and there because people‘s infrastructure are slightly optimized differently and modular 10% here and there， but everything they build will run everywhere． This is kind of the one of the principles of software that should never be given up． And it， and we protected quite dearly． Yeah， it makes it possible for our software engineers to build ones run everywhere． And that’s because we recognize that the investment of software is the most expensive investment， and it‘s easy to test．

這部分真的很有趣。我們不僅大規(guī)模建造，而且是垂直整合建造。我們從底層到頂層全程優(yōu)化，然后我們把各個(gè)部分分開(kāi)，單獨(dú)賣。我們做的事情復(fù)雜得讓人難以置信。為什么這么做呢？因?yàn)槲覀兿氚盐覀兊幕A(chǔ)設(shè)施融入到GCP、AWS、Azure、OCI這些不同的云服務(wù)提供商中。我們的控制平臺(tái)、安全平臺(tái)都不一樣，我們考慮集群大小的方式也各不相同。但是，我們還是想辦法讓他們都能適應(yīng)英偉達(dá)的架構(gòu)。這樣，我們的架構(gòu)就能無(wú)處不在。

最終，我們希望有一個(gè)計(jì)算平臺(tái)，開(kāi)發(fā)者可以用它來(lái)構(gòu)建軟件，這個(gè)平臺(tái)在大部分情況下是一致的，可以模塊化地調(diào)整，可能這里那里有10%的不同，因?yàn)槊總€(gè)人的基礎(chǔ)設(shè)施都略有優(yōu)化差異，但是無(wú)論在哪里， 我們構(gòu)建的東西都能運(yùn)行。這是軟件的一個(gè)原則，我們非常珍視這一點(diǎn)。這使得我們的軟件工程師可以構(gòu)建出到處都能運(yùn)行的軟件。這是因?yàn)槲覀冋J(rèn)識(shí)到，軟件的投資是最昂貴的投資，而且它很容易測(cè)試。

Look at the size of the whole hardware industry and then look at the size of the world‘s industries． It’s $100 trillion on top of this one trillion dollar industry． And that tells you something.The software that you build， you have to， you know， you basically maintain for as long as you shall live． We‘ve never given up on piece of software． The reason why Kuda is used is because， you know， I called everybody． We will maintain this for as long as we shall live． And we’re serious now． We still maintain． I just saw a review the other day， Nvidia Shield， our Android TV． It‘s the best Android TV in the world． We shifted seven years ago． It is still the number one Android TV that people， you know， anybody who enjoys TV． And we just updated the software just this last week and people wrote a new story about it． G Force， we have 300 million gamers around the world． We’ve never stranded a single one of them． And so the fact that our architecture is compatible across all of these different areas makes it possible for us to do it． Otherwise， we would be sub， we would be， we would have， you know， we would have software teams that are hundred times the size of our company is today if not for this architectural compatibility． So we‘re very serious about that， and that translates to benefits the developers．

看看整個(gè)硬件行業(yè)的規(guī)模，再比比全世界所有行業(yè)的規(guī)模。硬件行業(yè)只有一萬(wàn)億美元，而全世界的行業(yè)加起來(lái)有一百萬(wàn)億億美元。這個(gè)對(duì)比告訴你，軟件行業(yè)要比硬件行業(yè)大得多。

你們做的軟件，基本上要一直維護(hù)下去。我們從沒(méi)有放棄過(guò)任何一款軟件。Kuda之所以被大家用，是因?yàn)槲蚁蛩腥顺兄Z，我們會(huì)一直維護(hù)它，只要我們還在。我們現(xiàn)在還是很認(rèn)真的，我們還在維護(hù)它。我前幾天還看到一篇評(píng)論，說(shuō)我們的英偉達(dá)Shield，我們的安卓電視，是世界上最好的安卓電視。我們?cè)谄吣昵巴瞥龅?，它仍然是排名第一的安卓電視，任何喜歡看電視的人都愛(ài)它。我們上周才更新了軟件，然后人們就寫(xiě)了新的文章來(lái)評(píng)論它。我們的G Force，全世界有3億玩家。我們從沒(méi)有拋棄過(guò)他們中的任何一個(gè)。我們的架構(gòu)在所有這些不同領(lǐng)域都是兼容的，這使得我們能做到這一點(diǎn)。如果不是因?yàn)槲覀兊募軜?gòu)兼容性，否則我們今天的軟件團(tuán)隊(duì)的規(guī)模會(huì)比現(xiàn)在公司大一百倍。所以我們非常重視這一點(diǎn)，這也給開(kāi)發(fā)者帶來(lái)了好處。

主持人：One impressive substantiation of that recently was how quickly brought up a cluster for X dot AI． Yeah， and if you want to check about that， cuz that was striking in terms of both the scale and the speed with what you did． That

最近有一個(gè)讓人印象深刻的例子是我們?yōu)閄 dot AI迅速搭建了一個(gè)集群。如果你想了解這件事，因?yàn)樗谝?guī)模和速度上都讓人驚訝。我們很快就完成了這個(gè)任務(wù)。

黃仁勛：You know， a lot of that credit you gotta give to Elon． I think the， first of all， to decide to do something， select the site． I bring cooling to it． I power hum and then decide to build this hundred thousand GPU super cluster， which is， you know， the largest of its kind in one unit． And then working backwards， you know， we started planning together the date that he was gonna stand everything up． And the date that he was gonna stand everything up was determined， you know， quite， you know， a few months ago． And so all of the components， all the Oems， all the systems， all the software integration we did with their team， all the network simulation we simulate all the network configurations， we， we pre， I mean like we prestaged everything as a digital twin． We， we pres， we prestaged all of his supply chain． We prestaged all of the wiring of the networking． We even set up a small version of it． Kind of a， you know， just a first instance of it． You know， ground truth， if you reference 0， you know， system 0 before everything else showed up． So by the time that everything showed up， everything was staged， all the practicing was done， all the simulations were done．

這里得給埃隆·馬斯克很多功勞。首先，他決定要做這件事，選了地方，解決了冷卻和供電問(wèn)題，然后決定建造這個(gè)十萬(wàn)GPU的超級(jí)計(jì)算機(jī)群，這是迄今為止這種類型中最大的一個(gè)。然后，我們開(kāi)始倒推，就是說(shuō)，我們幾個(gè)月前就一起計(jì)劃了他要讓一切運(yùn)行起來(lái)的日期。所以，所有的組件、所有的原始設(shè)備制造商、所有的系統(tǒng)、所有的軟件集成，我們都是和他們的團(tuán)隊(duì)一起做的，所有的網(wǎng)絡(luò)配置我們都模擬了一遍，我們預(yù)先準(zhǔn)備，就像數(shù)字孿生一樣，我們預(yù)先準(zhǔn)備了所有的供應(yīng)鏈，所有的網(wǎng)絡(luò)布線。我們甚至搭建了一個(gè)小版本，就像是第一個(gè)實(shí)例，你懂的，就是所有東西到位之前的基準(zhǔn)，你參考的0號(hào)系統(tǒng)。所以，當(dāng)所有東西都到位的時(shí)候，一切都已經(jīng)安排好了，所有的練習(xí)都做完了，所有的模擬也都完成了。

And then， you know， the massive integration， even then the massive integration was a Monument of， you know， gargantuan teams of humanity crawling over each other， wiring everything up 247． And within a few weeks， the clusters were out． I mean， it‘s， it’s really， yeah， it‘s really a testament to his willpower and how he’s able to think through mechanical things， electrical things and overcome what is apparently， you know， extraordinary obstacles． I mean， what was done there is the first time that a computer of that large scale has ever been done at that speed． Unless our two teams are working from a networking team to compute team to software team to training team to， you know， and the infrastructure team， the people that the electrical engineers today， you know， to the software engineers all working together． Yeah， it‘s really quite a fit to watch． Was．

然后，你知道，大規(guī)模的集成工作，即使這個(gè)集成工作本身也是個(gè)巨大的工程，需要大量的團(tuán)隊(duì)成員像螞蟻一樣辛勤工作，幾乎是全天候不停地接線和設(shè)置。幾周之內(nèi)，這些計(jì)算機(jī)群就建成了。這真的是對(duì)他意志力的證明，也顯示了他如何在機(jī)械、電氣方面思考，并克服了顯然是非常巨大的障礙。我的意思是，這可是第一次在這么短的時(shí)間內(nèi)建成如此大規(guī)模的計(jì)算機(jī)系統(tǒng)。這需要我們的網(wǎng)絡(luò)團(tuán)隊(duì)、計(jì)算團(tuán)隊(duì)、軟件團(tuán)隊(duì)、訓(xùn)練團(tuán)隊(duì)，以及基礎(chǔ)設(shè)施團(tuán)隊(duì)，也就是那些電氣工程師、軟件工程師，所有人一起合作。這真的挺壯觀的。這就像是一場(chǎng)大型的團(tuán)隊(duì)協(xié)作，每個(gè)人都在努力確保一切順利運(yùn)行。

主持人：There a challenge that felt most likely to be blocking from an engineering perspective， active， just．

從工程角度來(lái)看，有沒(méi)有哪個(gè)挑戰(zhàn)最可能成為絆腳石，就是說(shuō)，有沒(méi)有哪個(gè)技術(shù)難題最可能讓整個(gè)項(xiàng)目卡住，動(dòng)彈不得？

黃仁勛：A tonnage of electronics that had to come together． I mean， it probably worth just to measure it． I mean， it‘s a， you know， it tons and tons of equipment． It’s just abnormal． You know， usually a supercomputer system like that， you plan it for a couple of years from the moment that the first systems come on， come delivered to the time that you‘ve probably submitted everything for some serious work． Don’t be surprised if it‘s a year， you know， I mean， I think that happens all the time． It’s not abnormal． Now we couldn‘t afford to do that． So we created， you know， a few years ago， there was an initiative in our company that’s called Data Center as a product． We don‘t sell it as a product， but we have to treat it like it’s a product． Everything about planning for it and then standing it up， optimizing it， tuning it， keep it operational， right？ The goal is that it should be， you know， kind of like opening up your beautiful new iPhone and you open it up and everything just kind of works．

我們需要把大量的電子設(shè)備整合在一起。我的意思是，這些設(shè)備的量多到值得去稱一稱。有數(shù)噸又?jǐn)?shù)噸的設(shè)備，這太不正常了。通常像這樣的超級(jí)計(jì)算機(jī)系統(tǒng)，從第一個(gè)系統(tǒng)開(kāi)始交付，到你把所有東西都準(zhǔn)備好進(jìn)行一些嚴(yán)肅的工作，你通常需要規(guī)劃幾年時(shí)間。如果這個(gè)過(guò)程需要一年，你要知道，這是常有的事，并不奇怪。

但現(xiàn)在我們沒(méi)有時(shí)間去這么做。所以幾年前，我們公司里有一個(gè)叫做“數(shù)據(jù)中心即產(chǎn)品”的計(jì)劃。我們不把它當(dāng)作產(chǎn)品來(lái)賣，但我們必須像對(duì)待產(chǎn)品一樣對(duì)待它。從規(guī)劃到建立，再到優(yōu)化、調(diào)整、保持運(yùn)行，所有的一切都是為了確保它能夠像打開(kāi)一部嶄新的iPhone一樣，一打開(kāi)，一切都能正常工作。我們的目標(biāo)就是這樣。

Now， of course， it‘s a miracle of technology making it that， like that， but we now have the skills to do that． And so if you’re interested in a data center and just have to give me a space and some power， some cooling， you know， and we‘ll help you set it up within， call it， 30 days． I mean， it’s pretty extraordinary．

當(dāng)然了，能這么快就把數(shù)據(jù)中心建好，這簡(jiǎn)直就是科技的奇跡。但現(xiàn)在我們已經(jīng)有了這樣的技術(shù)能力。所以如果你想要建一個(gè)數(shù)據(jù)中心，只需要給我一個(gè)地方，提供一些電力和制冷設(shè)備，我們就能在差不多30天內(nèi)幫你把一切都搭建好。我的意思是，這真的非常了不起。

主持人：That‘s wild． If you think， if you look ahead to 200，000，500，000， a million in a super cluster， whatever you call it． At that point， what do you think is the biggest blocker？ Capital energy supply in one area？

那真是厲害。如果你想想，要是將來(lái)有個(gè)超級(jí)大的計(jì)算機(jī)集群，里面有個(gè)二十萬(wàn)、五十萬(wàn)、甚至一百萬(wàn)的計(jì)算機(jī)，不管你叫它什么。到那個(gè)時(shí)候，你覺(jué)得最大的難題會(huì)是什么呢？是資金問(wèn)題、能源供應(yīng)問(wèn)題，還是別的什么？

黃仁勛：Everything． Nothing about what you， just the scales that you talked about， though， nothing is normal．

你說(shuō)的那些事情，不管是哪個(gè)方面，只要涉及到你提到的那些巨大規(guī)模，那就沒(méi)有一件事情是正常的。

主持人：But nothing is impossible． Nothing．

但是，也沒(méi)什么事是完全不可能的。啥事都有可能。

黃仁勛：Is， yeah， no laws of physics limits， but everything is gonna be hard． And of course， you know， I， is it worth it？ Like you can‘t believe， you know， to get to something that we would recognize as a computer that so easily and so able to do what we ask it to do， what， you know， otherwise general intelligence of some kind and even， you know， even if we could argue about is it really general intelligence， just getting close to it is going to be a miracle． We know that． And so I think the， there are five or six endeavors to try to get there． Right？ I think， of course， OpenAI and anthropic and X and， you know， of course， Google and meta and Microsoft and you know， there， this frontier， the next couple of clicks that mountain are just so vital． Who doesn’t wanna be the first on that mountain． I think that the prize for reinventing intelligence altogether． Right． It‘s just， it’s too consequential not to attempt it． And so I think there are no laws of physics． Everything is gonna be hard．

確實(shí)，沒(méi)有物理定律說(shuō)我們做不到，但每件事情都會(huì)非常難。你也知道，這值得嗎？你可能覺(jué)得難以置信，我們要達(dá)到的那種電腦，能夠輕松地做我們讓它做的事情，也就是某種通用智能，哪怕我們能爭(zhēng)論它是否真的是通用智能，接近它都將會(huì)是一個(gè)奇跡。我們知道這很難。所以我認(rèn)為，有五六個(gè)團(tuán)隊(duì)正在嘗試達(dá)到這個(gè)目標(biāo)。對(duì)吧？比如說(shuō)，OpenAI、Anthropic、X，還有谷歌、Meta和微軟等等，他們都在努力攀登這個(gè)前沿科技的山峰。誰(shuí)不想成為第一個(gè)登頂?shù)娜四?？我認(rèn)為，重新發(fā)明智能的獎(jiǎng)勵(lì)是如此之大，它的影響太大了，我們不能不去嘗試。所以，雖然物理定律上沒(méi)有限制，但每件事都會(huì)很難。

主持人：A year ago when we spoke together， you talked about， we asked like what applications you got most excited about that Nvidia would serve next in AI and otherwise， and you talked about how you led to， your most extreme customers sort of lead you there． Yeah， and about some of the scientific applications． So I think that‘s become like much more mainstream of you over the last year． Is it still like science and AI’s application of science that most excites you？

一年前我們聊天時(shí)，我問(wèn)你，你對(duì)英偉達(dá)接下來(lái)在AI和其他領(lǐng)域能服務(wù)的哪些應(yīng)用最興奮，你談到了你的一些最極端的客戶某種程度上引導(dǎo)了你。是的，還有關(guān)于一些科學(xué)應(yīng)用的討論。所以我覺(jué)得過(guò)去一年里，這些科學(xué)和AI的應(yīng)用變得更主流了?，F(xiàn)在，是不是仍然是科學(xué)以及AI在科學(xué)領(lǐng)域的應(yīng)用讓你最興奮？

黃仁勛：I love the fact that we have digital， we have AI chip designers here in video． Yeah， I love that． We have AI software engineers． How．

我就直說(shuō)了，咱們現(xiàn)在有數(shù)字版的，也就是用人工智能來(lái)設(shè)計(jì)芯片的設(shè)計(jì)師，就在視頻里。是的，我喜歡這個(gè)。我們還有AI軟件工程師。

主持人：Effective our AI chip designers today？ Super．

我們今天用人工智能來(lái)設(shè)計(jì)芯片的效果怎么樣？非常好。

黃仁勛：Good． We can‘t， we couldn’t build Hopper without it． And the reason for that is because they could explore a much larger space than we can and because they have infinite time． They‘re running on a supercomputer． We have so little time using human engineers that we don’t explore as much of the space as we should， and we also can explore commentary． I can‘t explore my space while including your exploration and your exploration． And so， you know， our chips are so large， it’s not like it‘s designed as one chip． It’s designed almost like 1，000 ships and we have to ex， we have to optimize each one of them． Kind of an isolation． You really wanna optimize a lot of them together and， you know， cross module code design and optimize across much larger space． But obviously we‘re gonna be able to find fine， you know， local maximums that are hidden behind local minimum somewhere． And so clearly we can find better answers． You can’t do that without AI． Engineers just simply can‘t do it． We just don’t have enough time．

我們的AI芯片設(shè)計(jì)師真的很厲害。如果沒(méi)有它們，我們根本造不出Hopper這款芯片。因?yàn)樗鼈兡芴剿鞯姆秶任覀內(nèi)祟悘V得多，而且它們好像有無(wú)窮無(wú)盡的時(shí)間。它們?cè)诔?jí)計(jì)算機(jī)上運(yùn)行，而我們?nèi)祟惞こ處煹臅r(shí)間有限，探索不了那么大的范圍。而且，我們也不能同時(shí)探索所有的可能，我探索我的領(lǐng)域的時(shí)候，就不能同時(shí)探索你的領(lǐng)域。

我們的芯片非常大，不像是設(shè)計(jì)一個(gè)單獨(dú)的芯片，更像是設(shè)計(jì)1000個(gè)芯片，每個(gè)都需要優(yōu)化。就像是一個(gè)個(gè)獨(dú)立的小島。但我們其實(shí)很想把它們放在一起優(yōu)化，跨模塊協(xié)同設(shè)計(jì)，在整個(gè)更大的空間里優(yōu)化。顯然，我們能找到更好的解決方案，那些隱藏在某個(gè)角落里的最好的選擇。沒(méi)有AI我們做不到這一點(diǎn)。工程師們就是時(shí)間不夠，做不到。

主持人：One other thing has changed since we last spoke collectively， and I looked it up at the time in videos， market cap was about 500 billion． It‘s now over 3 trillion． So the last 18 months， you’ve added two and a half trillion plus of market cap， which effectively is $100 billion plus a month or two and a half snowflakes or， you know， a stripe plus a little bit， or however you wanna think about.A country or two． Obviously， a lot of things are stayed consistent in terms of focus on what you‘re building and etc． And you know， walking through here earlier today， I felt the buzz like when I was at Google 15 years ago was kind of you felt the energy of the company and the vibe of excitement． What has changed during that period， if anything？ Or how， what is different in terms of either how Nvidia functions or how you think about the world or the size of bets you can take or．

自我們上次一起聊天以來(lái)，有一件事變了，我查了下，當(dāng)時(shí)英偉達(dá)的市值大概是5000億美元?，F(xiàn)在超過(guò)了3萬(wàn)億美元。所以在過(guò)去18個(gè)月里，你們?cè)黾恿藘扇f(wàn)五千億美元以上的市值，這相當(dāng)于每個(gè)月增加了1000億美元，或者說(shuō)增加了兩個(gè)半的Snowflake公司或者一個(gè)Stripe公司多一點(diǎn)的市值，無(wú)論你怎么想。

這相當(dāng)于增加了一兩個(gè)國(guó)家的市值。顯然，盡管市值增長(zhǎng)了這么多，你們?cè)诮ㄔ斓臇|西和專注的領(lǐng)域上還是保持了一致性。你知道，今天我在這里走了一圈，我感受到了一種活力，就像15年前我在谷歌時(shí)感受到的那樣，你能感覺(jué)到公司的能量和興奮的氛圍。在這段時(shí)間里，有什么變化了嗎？或者，英偉達(dá)的運(yùn)作方式、你對(duì)世界的看法、你能承擔(dān)的風(fēng)險(xiǎn)大小等方面有什么不同了嗎？

黃仁勛：Well， our company can‘t change as fast as a stock price． Let’s just be clear about． So in a lot of ways， we haven‘t changed that much． I think the thing to do is to take a step back and ask ourselves， what are we doing？ I think that’s really the big， you know， the big observation， realization， awakening for companies and countries is what‘s actually happening． I think what we’re talking about earlier， I‘m from our industry perspective， we reinvented computing． Now it hasn’t been reinvented for 60 years． That‘s how big of a deal it is that we’ve driven down the marginal cost of computing， down probably by a million x in the last 10 years to the point that we just， hey， let‘s just let the computer go exhaustively write the software． That’s the big realization． 說(shuō)話人 2 24：00 And that in a lot of ways， I was kind of， we were kind of saying the same thing about chip design． We would love for the computer to go discover something about our chips that we otherwise could have done ourselves， explore our chips and optimize it in a way that we couldn‘t do ourselves， right， in the way that we would love for digital biology or， you know， any other field of science．

我們公司的變化速度可沒(méi)有股價(jià)變化那么快。所以這么說(shuō)吧，我們?cè)诤芏喾矫娌](méi)有太大變化。我認(rèn)為重要的是要退一步來(lái)問(wèn)問(wèn)我們自己，我們到底在做什么。這真的是對(duì)公司和國(guó)家來(lái)說(shuō)一個(gè)很大的觀察、認(rèn)識(shí)和覺(jué)醒，那就是真正發(fā)生的事情。

就像我們之前討論的，從我們行業(yè)的角度來(lái)看，我們重新發(fā)明了計(jì)算。這可是60年來(lái)都沒(méi)有發(fā)生過(guò)的事情。我們把計(jì)算的邊際成本降低了，可能在過(guò)去10年里降低了一百萬(wàn)分之一，以至于我們現(xiàn)在可以讓計(jì)算機(jī)去詳盡地編寫(xiě)軟件。這是一個(gè)重大的領(lǐng)悟。

在很多方面，我們對(duì)芯片設(shè)計(jì)也是這么說(shuō)的。我們希望計(jì)算機(jī)能自己去發(fā)現(xiàn)我們芯片的一些東西，這些東西我們本來(lái)可以自己做，但計(jì)算機(jī)可以探索我們的芯片并以我們自己做不到的方式進(jìn)行優(yōu)化，就像我們希望在數(shù)字生物學(xué)或其他科學(xué)領(lǐng)域那樣。

And so I think people are starting to realize when we reinvented computing， but what does that mean even， and as we， all of a sudden， we created this thing called intelligence and what happened to computing？ Well， we went from data centers are multi tenant stores of files． These new data centers we‘re creating are not data centers． They don’t， they‘re not multi tenant． They tend to be single tenant． They’re not storing any of our files． They‘re just， they’re producing something． They‘re producing tokens． And these tokens are reconstituted into what appears to be intelligence． Isn’t that right？ And intelligence of all different kinds． You know， it could be articulation of robotic motion． It could be sequences of amino acids． It could be， you know， chemical chains． It could be all kinds of interesting things， right？ So what are we really doing？ We‘ve created a new instrument， a new machinery that in a lot of ways is that the noun of the adjective generative AI． You know， instead of generative AI， you know， it’s， it‘s an AI factory． It’s a factory that generates AI． And we‘re doing that at extremely large scale． And what people are starting to realize is， you know， maybe this is a new industry． It generates tokens， it generates numbers， but these numbers constitute in a way that is fairly valuable and what industry would benefit from it．

所以我覺(jué)得人們開(kāi)始意識(shí)到，當(dāng)我們重新發(fā)明計(jì)算時(shí)，這到底意味著什么。突然間，我們創(chuàng)造了這個(gè)叫做智能的東西，計(jì)算發(fā)生了什么變化？嗯，我們以前把數(shù)據(jù)中心看作是多租戶存儲(chǔ)文件的地方。我們現(xiàn)在創(chuàng)建的這些新數(shù)據(jù)中心，其實(shí)已經(jīng)不是傳統(tǒng)意義上的數(shù)據(jù)中心了。它們往往是單一租戶的，它們不存儲(chǔ)我們的文件，它們只是在生產(chǎn)一些東西。它們?cè)谏a(chǎn)數(shù)據(jù)令牌。然后這些數(shù)據(jù)令牌重新組合成看起來(lái)像智能的東西。對(duì)吧？而且智能有各種各樣的形式?？赡苁菣C(jī)器人動(dòng)作的表達(dá)，可能是氨基酸序列，可能是化學(xué)物質(zhì)鏈，可能是各種有趣的事情，對(duì)吧？所以我們到底在做什么？我們創(chuàng)造了一種新的工具，一種新的機(jī)械，從很多方面來(lái)說(shuō)，它就是生成性人工智能的名詞形式。你知道，不是生成性人工智能，而是人工智能工廠。它是一個(gè)生產(chǎn)人工智能的工廠。我們正在非常大規(guī)模地做這件事。人們開(kāi)始意識(shí)到，這可能是一個(gè)新行業(yè)。它生成數(shù)據(jù)令牌，它生成數(shù)字，但這些數(shù)字以一種相當(dāng)有價(jià)值的方式構(gòu)成，哪些行業(yè)會(huì)從中受益。

Then you take a step back and you ask yourself again， you know， what‘s going on？ Nvidia on the one hand， we reinvent a computing as we know it． And so there’s $1 trillion of infrastructure that needs to be modernized． That‘s just one layer of it． The big layer of it is that there’s， this instrument that we‘re building is not just for data centers， which we were modernizing， but you’re using it for producing some new commodity． And how big can this new commodity industry be？ Hard to say， but it‘s probably worth trillions． 說(shuō)話人 2 26：18 And so that I think is kind of the viewers to take a step back． You know， we don’t build computers anymore． We build factories． And every country is gonna need it， every company‘s gonna need it， you know， give me an example of a company who or industry as us， you know what， we don’t need to produce intelligence． We got plenty of it． And so that‘s the big idea． I think， you know， and that’s kind of an abstracted industrial view． And， you know， someday people realize that in a lot of ways， the semiconductor industry wasn‘t about building chips， it was building， it was about building the foundational fabric for society． And then all of a sudden， there we go． I get it． You know， this is a big deal． Isn’t not just about chips．

然后你退一步，再次問(wèn)自己，到底發(fā)生了什么？Nvidia一方面，我們重新發(fā)明了我們所知道的計(jì)算。所以有一萬(wàn)億美元的基礎(chǔ)設(shè)施需要現(xiàn)代化。這只是其中一層。更大的一層是，我們正在建造的這個(gè)工具不僅僅是為了數(shù)據(jù)中心，我們正在現(xiàn)代化數(shù)據(jù)中心，而是你用它來(lái)生產(chǎn)一些新的商品。這個(gè)新商品行業(yè)能有多大？很難說(shuō)，但可能價(jià)值數(shù)萬(wàn)億美元。

所以我認(rèn)為這是觀眾需要退一步的地方。你知道，我們不再制造電腦了。我們制造工廠。每個(gè)國(guó)家都會(huì)需要它，每個(gè)公司都會(huì)需要它，給我一個(gè)不需要生產(chǎn)智能的公司或行業(yè)的例子，你知道，我們有很多智能。所以這就是這個(gè)大主意。我認(rèn)為，你知道，這是一種抽象的工業(yè)觀點(diǎn)。然后，有一天人們意識(shí)到，在很多方面，半導(dǎo)體行業(yè)不是關(guān)于制造芯片，它是關(guān)于為社會(huì)建立基礎(chǔ)結(jié)構(gòu)。然后突然間，我們明白了。這不僅僅是關(guān)于芯片的大事。

主持人：How do you think about embodiment now？

你現(xiàn)在怎么看待“體現(xiàn)”或者“具體化”這個(gè)概念？就是說(shuō)，你怎么考慮把智能或者人工智能真正應(yīng)用到實(shí)際的物理世界中，比如機(jī)器人或者其他實(shí)體設(shè)備上？

黃仁勛：Well， the thing I‘m super excited about is in a lot of ways， we’ve， we‘re close to artificial general intelligence， but we’re also close to artificial general robotics． Tokens are tokens． I mean， the question is， can you tokenize it？ You know， of course， tokenis， tokenizing things is not easy， as you guys know． But if you‘re able to tokenize things， align it with large language models and other modalities， if I can generate a video that has Jensen reaching out to pick up the coffee cup， why can’t I prompt a robot to generate the token， still pick up the rule， you know？ And so intuitively， you would think that the problem statement is rather similar for computer． And， and so I think that we‘re that close． That’s incredibly exciting．

我現(xiàn)在非常興奮的一點(diǎn)是，我們?cè)诤芏喾矫娑伎煲獙?shí)現(xiàn)通用人工智能了，而且我們也快實(shí)現(xiàn)通用機(jī)器人技術(shù)了。數(shù)據(jù)令牌就是數(shù)據(jù)令牌。我的意思是，問(wèn)題是，你能把它變成數(shù)據(jù)令牌嗎？當(dāng)然，把東西變成數(shù)據(jù)令牌并不容易，你們知道這一點(diǎn)。但如果你能做到這一點(diǎn)，把它和大型語(yǔ)言模型和其他方式對(duì)齊，如果我能生成一個(gè)視頻，視頻里有Jensen伸手去拿咖啡杯，為什么我不能提示一個(gè)機(jī)器人去生成數(shù)據(jù)令牌，實(shí)際上去拿起那個(gè)規(guī)則，你知道嗎？所以直觀上，你會(huì)認(rèn)為這個(gè)問(wèn)題對(duì)計(jì)算機(jī)來(lái)說(shuō)相當(dāng)相似。所以我認(rèn)為我們已經(jīng)很接近了。這非常令人興奮。

Now the， the two brown field robotic systems． Brown field means that you don‘t have to change the environment for is self driving cars． And with digital chauffeurs and body robots right between the cars and the human robot， we could literally bring robotics to the world without changing the world because we built a world for those two things． Probably not a coincidence that Elon spoke is then those two forms． So robotics because it is likely to have the larger potential scale． And and so I think that’s exciting． But the digital version of it， I is equally exciting． You know， we‘re talking about digital or AI employees． There’s no question we‘re gonna have AI employees of all kinds， and our outlook will be some biologics and some artificial intelligence， and we will prompt them in the same way． Isn’t that right？ Mostly I prompt my employees， right？ You know， provide them context， ask him to perform a mission． They go and recruit other team members， they come back and work going back and forth． How‘s that gonna be any different with digital and AI employees of all kinds？ So we’re gonna have AI marketing people， AI chip designers， AI supply chain people， AIs， you know， and I‘m hoping that Nvidia is someday biologically bigger， but also from an artificial intelligence perspective， much bigger． That’s our future company． If．

現(xiàn)在有兩種“棕色地帶”機(jī)器人系統(tǒng)。“棕色地帶”意味著你不需要改變環(huán)境，比如自動(dòng)駕駛汽車。有了數(shù)字司機(jī)和機(jī)器人助手在汽車和人類機(jī)器人之間，我們可以在不改變世界的情況下把機(jī)器人技術(shù)帶到世界上，因?yàn)槲覀優(yōu)檫@兩樣?xùn)|西建造了世界。埃隆·馬斯克可能不是偶然提到這兩種形式的。所以機(jī)器人技術(shù)因?yàn)榭赡苡懈蟮臐撛谝?guī)模而令人興奮。而數(shù)字版的機(jī)器人也同樣令人興奮。你知道，我們談?wù)摰氖菙?shù)字或AI員工。毫無(wú)疑問(wèn)，我們將擁有各種AI員工，我們的前景將是一些生物和一些人工智能，我們將以相同的方式提示他們。不是嗎？大多數(shù)情況下，我提示我的員工，對(duì)吧？給他們提供上下文，讓他們執(zhí)行任務(wù)。他們?nèi)フ心计渌麍F(tuán)隊(duì)成員，他們回來(lái)工作，來(lái)回工作。這和各種數(shù)字和AI員工有什么不同呢？所以我們將有AI營(yíng)銷人員，AI芯片設(shè)計(jì)師，AI供應(yīng)鏈人員，AI，等等，我希望英偉達(dá)有一天在生物學(xué)上更大，同時(shí)從人工智能的角度來(lái)看，也更大。這是我們未來(lái)公司的樣子。

主持人：We came back and talked to you year from now， what part of the company do you think would be most artificially intelligent？

如果我們一年后回來(lái)再和你聊聊，你覺(jué)得公司里哪個(gè)部分會(huì)是最智能化的？

黃仁勛：I‘m hoping it should sign．

我希望公司里最重要的、最核心的部分能實(shí)現(xiàn)智能化。

主持人：Okay． And most．

好的，然后繼續(xù)詢問(wèn)。

黃仁勛：Important part． And the read． That‘s right． Because it because I should start where it moves the needle most also where we can make the biggest impact most． You know， it’s such an insanely hard problem． I work with Sasina at synopsis and rude at cadence． I totally imagine them having synopsis chip designers that I can rent． And they know something about a particular module， their tool， and they train an AI to be incredibly good at it． And we‘ll just hire a whole bunch of them whenever we need， we’re in that phase of that chip design． You know， I might rent a million synopsis engineers to come and help me out and then go rent a million Cadence engineers to help me out． And that， what an exciting future for them that they have all these agents that sit on top of their tools platform， that use the tools platform and other， and collaborate with other platforms． And you‘ll do that for， you know， Christian will do that at SAP and Bill will do that as service．

我認(rèn)為最重要的部分應(yīng)該是公司里最能產(chǎn)生影響的地方。他說(shuō)，這個(gè)問(wèn)題非常難，但他希望從最能推動(dòng)公司發(fā)展的地方開(kāi)始智能化。他和Synopsys的Sasina和Cadence的Rude一起工作，他想象著可以租用Synopsys的芯片設(shè)計(jì)師AI。這些AI對(duì)某個(gè)特定模塊、工具非常了解，并且已經(jīng)被訓(xùn)練得非常擅長(zhǎng)這方面的工作。當(dāng)他們需要進(jìn)行芯片設(shè)計(jì)的某個(gè)階段時(shí)，他們會(huì)租用一大批這樣的AI設(shè)計(jì)師。比如，他可能會(huì)租用一百萬(wàn)個(gè)Synopsys工程師AI來(lái)幫忙，然后再租用一百萬(wàn)個(gè)Cadence工程師AI來(lái)幫忙。我認(rèn)為，對(duì)于我們來(lái)說(shuō)，有一個(gè)激動(dòng)人心的未來(lái)，因?yàn)槲覀冇兴羞@些AI代理，它們位于我們工具平臺(tái)的頂部，使用這些工具平臺(tái)，并且與其他平臺(tái)協(xié)作。SAP的Christian會(huì)這樣做，Bill會(huì)作為服務(wù)來(lái)做這件事。

Now， you know， people say that these Saas platforms are gonna be disrupted． I actually think the opposite， that they‘re sitting on a gold mine， that they’re gonna be this flourishing of agents that are gonna be specialized in Salesforce， specialized in， you know， well， Salesforce， I think they call Lightning and SAP is about， and everybody‘s got their own language． Is that right？ And we got Kuda and we’ve got open USD for Omniverse． And who‘s gonna create an AI agent？ That’s awesome． At open USD， we‘re， you know， because nobody cares about it more than we do， right？ And so I think in a lot of ways， these platforms are gonna be flourishing with agents and we’re gonna introduce them to each other and they‘re gonna collaborate and solve problems．

現(xiàn)在，有些人說(shuō)這些基于網(wǎng)絡(luò)的軟件服務(wù)平臺(tái)（SaaS）將會(huì)被顛覆。但我實(shí)際上認(rèn)為恰恰相反，他們就像坐在金礦上一樣，將會(huì)有一個(gè)專業(yè)化的智能代理（AI）的繁榮時(shí)期。這些智能代理將會(huì)專門針對(duì)Salesforce、SAP等平臺(tái)進(jìn)行優(yōu)化。比如Salesforce有個(gè)叫做Lightning的平臺(tái)，每個(gè)平臺(tái)都有自己的語(yǔ)言和特點(diǎn)。我們有Kuda，還有為Omniverse準(zhǔn)備的開(kāi)放USD。誰(shuí)會(huì)來(lái)創(chuàng)造這些AI代理呢？那將會(huì)是非常酷的事情。在開(kāi)放USD方面，我們會(huì)來(lái)做，因?yàn)闆](méi)有人比我們更關(guān)心它，對(duì)吧？所以我認(rèn)為在很多方面，這些平臺(tái)將會(huì)因?yàn)檫@些智能代理而繁榮起來(lái)，我們會(huì)把它們相互介紹，它們將會(huì)協(xié)作并解決問(wèn)題。

主持人：You see a wealth of different people working in every domain in AI． What do you think is under notice or that people that you want more entrepreneurs or engineers or business people could work on？

你覺(jué)得在人工智能領(lǐng)域，有沒(méi)有什么被忽視的地方，或者你希望更多的創(chuàng)業(yè)者、工程師或商業(yè)人士能關(guān)注和投入工作的領(lǐng)域？

黃仁勛：Well， first of all， I think what is misunderstood， and I misunderstood， maybe it may be underestimated， is the， the under the water activity， under the surface activity of groundbreaking science， computer science to science and engineering that is being affected by AI and machinery． I think you just can‘t walk into a science department anywhere， theoretical math department anywhere， where AI and machine learning and the type of work that we’re talking about today is gonna transform tomorrow． If they are， if you take all of the engineers in the world， all of the scientists in the world and you say that the way they‘re working today is early indication of the future， because obviously it is． Then you’re gonna see a tidal wave of gender to AI， a tidal wave of AI， a tidal wave machine learning change everything that we do in some short period of time．

首先，我認(rèn)為可能被誤解或低估了的是，那些在水面下的、正在進(jìn)行的、突破性的科學(xué)、計(jì)算機(jī)科學(xué)以及科學(xué)與工程活動(dòng)，這些活動(dòng)正受到人工智能和機(jī)械的影響。如果你走進(jìn)任何一個(gè)科學(xué)系，任何一個(gè)理論數(shù)學(xué)系，你會(huì)發(fā)現(xiàn)今天的人工智能和機(jī)器學(xué)習(xí)的工作將改變明天。如果你把世界上所有的工程師、所有的科學(xué)家都看作是未來(lái)的早期跡象，因?yàn)轱@然他們是，那么你就會(huì)看到一股涌向人工智能的潮流，一股人工智能的潮流，一股機(jī)器學(xué)習(xí)改變我們所做的一切的潮流，這將在很短的時(shí)間內(nèi)發(fā)生。

in some short period of time.ion． And to work with Alex and Elian and Hinton at at at in Toronto and Yan Lekun and of course， Andrew Ang here in Stanford． And， you know， I saw the early indications of it and we were fortunate to have extrapolated from what was observed to be detecting cats into a profound change in computer science and computing altogether． And that extrapolation was fortunate for us． And now， of course， we， we were so excited by， so inspired by it that we changed everything about how we did things． But that took how long？ It took literally six years from observing that toy， Alex Net， which I think by today‘s standards will be considered a toy to superhuman levels of capabilities in object recognition． Well， that was only a few years． 說(shuō)話人 2 33：40 Now what is happening right now， the groundswell in all of the fields of science， not one field of science left behind． I mean， just to be very clear． Okay， everything from quantum computing， the quantum chemistry， you know， every field of science is involved in the approaches that we’re talking about． If we give ourselves， and they‘ve been added for a couple to three years， if we give ourselves in a couple， two， three years， the world’s gonna change． There‘s not gonna be one paper， there’s not gonna be one breakthrough in science， one breakthrough in engineering， where generative AI isn‘t at the foundation of it． I’m fairly certain of it． And， and so I， I think， you know， there‘s a lot of questions about， you know， every so often I hear about whether this is a fad computer． You just gotta go back to first principles and observe what is actually happening．

就在很短的時(shí)間內(nèi)，我們看到了科學(xué)領(lǐng)域的大浪潮，沒(méi)有一個(gè)科學(xué)領(lǐng)域被落下。我的意思是，每一件事都非常清楚。從量子計(jì)算到量子化學(xué)，你知道的，每個(gè)科學(xué)領(lǐng)域都涉及到我們正在討論的方法。如果我們給自己，比如說(shuō)，兩三年的時(shí)間，世界將會(huì)改變。不會(huì)有一篇科學(xué)論文，不會(huì)有一項(xiàng)科學(xué)突破，一項(xiàng)工程突破，不是以生成性人工智能為基礎(chǔ)的。我對(duì)此相當(dāng)確定。所以，我認(rèn)為，你知道，有很多問(wèn)題，時(shí)不時(shí)我聽(tīng)到關(guān)于這是否是計(jì)算機(jī)的一時(shí)風(fēng)尚。你只需要回到基本原則，觀察實(shí)際發(fā)生的事情。

人工智能和機(jī)器學(xué)習(xí)的發(fā)展非?？欤矣绊懮钸h(yuǎn)。我在人工智能領(lǐng)域有重大貢獻(xiàn)的科學(xué)家合作的經(jīng)歷，比如多倫多的Alex Krizhevsky、Eliasmith、Hinton和斯坦福的Yan LeCun以及Andrew Ng。、從識(shí)別貓咪的簡(jiǎn)單任務(wù)到物體識(shí)別能力的超人水平的發(fā)展，這個(gè)過(guò)程只用了幾年時(shí)間。我相信，在未來(lái)幾年內(nèi)，每個(gè)科學(xué)領(lǐng)域的每項(xiàng)科學(xué)和工程突破都將以生成性人工智能為基礎(chǔ)。鼓勵(lì)人們不要懷疑這是否只是一時(shí)的流行，而應(yīng)該觀察實(shí)際發(fā)生的事情，基于事實(shí)來(lái)判斷。

The computing stack， the way we do computing has changed if the way you write software has changed， I mean， that is pretty cool． Software is how humans encode knowledge． This is how we encode our， you know， our algorithms． We encode it in a very different way． Now that‘s gonna affect everything， nothing else， whatever， be the same． And so I， I think the， the， I think I’m talking to the converted here and we all see the same thing． And all the startups that， you know， you guys work with and the scientists I work with and the engineers I work with， nothing will be left behind． I mean， this， we‘re gonna take everybody with us again．

計(jì)算的整個(gè)體系，也就是我們進(jìn)行計(jì)算的方式，已經(jīng)改變了，連我們編寫(xiě)軟件的方式也改變了。這意味著我們編碼知識(shí)的方法也變了，這是一種全新的編碼方式。這將會(huì)改變一切，其他的事情都不會(huì)和以前一樣了。他認(rèn)為他在這里是對(duì)已經(jīng)認(rèn)同這一點(diǎn)的人說(shuō)話，大家都看到了同樣的趨勢(shì)。無(wú)論是他們合作的初創(chuàng)公司，還是他合作的科學(xué)家和工程師，所有人都將被這一變革所影響。他的意思是，這次變革將會(huì)帶領(lǐng)所有人一起前進(jìn)。

主持人：I think one of the most exciting things coming from like the computer science world and looking at all these other fields of science is like I can go to a robotics conference now． Yeah， material science conference． Oh yeah， biotech conference． And like， I‘m like， oh， I understand this， you know， not at every level of the science， but in the driving of discovery， it is all the algorithms that are．

計(jì)算機(jī)科學(xué)領(lǐng)域的一個(gè)最令人興奮的事情是，現(xiàn)在可以應(yīng)用于所有其他科學(xué)領(lǐng)域。比如，他可以去機(jī)器人會(huì)議、材料科學(xué)會(huì)議、生物技術(shù)會(huì)議，他會(huì)發(fā)現(xiàn)自己能理解那些內(nèi)容。雖然不是在每個(gè)科學(xué)領(lǐng)域的每個(gè)層面上都懂，但在推動(dòng)發(fā)現(xiàn)方面，都是算法在起作用。

黃仁勛：General and there‘s some universal unifying concepts．

對(duì)，有一些普遍統(tǒng)一的概念。

主持人：And I think that‘s like incredibly exciting when you see how effective it is in every domain．

我認(rèn)為這非常令人興奮，當(dāng)你看到算法在每個(gè)領(lǐng)域都如此有效時(shí)。

黃仁勛：Yep， absolutely． And eh， I‘m so excited that I’m using it myself every day． You know， I don‘t know about you guys， but it’s my tutor now． I mean， I， I， I don‘t do， I don’t learn anything without first going to an AI． You know？ Why？ Learn the hard way． Just go directly to an AI． I should go directly to ChatGPT． Or， you know， sometimes I do perplexity just depending on just the formulation of my questions． And I just start learning from there． And then you can always fork off and go deeper if you like． But holy cow， it‘s just incredible．

我絕對(duì)同意。我很興奮，因?yàn)槲易约好刻於荚谑褂肁I。不知你們?cè)趺礃?，但AI已經(jīng)成為我的導(dǎo)師。我現(xiàn)在學(xué)任何東西都會(huì)先去問(wèn)AI。為什么？何必要費(fèi)勁去學(xué)呢，直接去找AI就行了。比如他會(huì)直接去問(wèn)ChatGPT，或者根據(jù)問(wèn)題的不同，有時(shí)他會(huì)去問(wèn)Perplexity。他會(huì)從那里開(kāi)始學(xué)習(xí)，然后如果愿意，可以深入研究。天哪，這真是太不可思議了。

And almost everything I know， I check， I double check， even though I know it to be a fact， you know， what I consider to be ground truth． I‘m the expert． I’ll still go to AI and check， make double check． Yeah， so great． Almost everything I do， I involve it．

我現(xiàn)在幾乎做任何事情都會(huì)用到AI。哪怕是他知道的事實(shí)，就算是他是那個(gè)領(lǐng)域的專家，他也會(huì)用AI再檢查一遍。他覺(jué)得這樣很好，因?yàn)樗麕缀跛械氖虑槎紩?huì)讓AI參與。

主持人：I think it‘s a great note to stop on． Yeah， thanks so much that time today．

這是個(gè)很好的結(jié)束話題。感謝大家今天的參與，時(shí)間到了。

黃仁勛：Really enjoyed it． Nice to see you guys．

我今天很開(kāi)心見(jiàn)到大家。

　　風(fēng)險(xiǎn)提示及免責(zé)條款

　　市場(chǎng)有風(fēng)險(xiǎn)，投資需謹(jǐn)慎。本文不構(gòu)成個(gè)人投資建議，也未考慮到個(gè)別用戶特殊的投資目標(biāo)、財(cái)務(wù)狀況或需要。用戶應(yīng)考慮本文中的任何意見(jiàn)、觀點(diǎn)或結(jié)論是否符合其特定狀況。據(jù)此投資，責(zé)任自負(fù)。