We're about to get started.

我們馬上就要開始了。

My next guest, I am so impressed by this person, three reasons.

下一位嘉賓，我對這個人印象深刻，原因有三。

First reason is there are only a handful of entrepreneurs, founders that started a company that literally touched the lives of billions of people around the world as part of the social fabric, invented services, and a state-of-the-art computing company.

第一個原因是，只有極少數企業家和創始人創辦的公司，作為社會結構的一部分，實實在在地影響了全世界數十億人的生活，他們發明了服務，創辦了最先進的計算機公司。

Two, very few entrepreneurs, founders founded the company and led it to over a trillion dollars of value.

兩位極少數的企業家、創始人創立了這家公司，並帶領它創造了超過萬億美元的價值。

And three, a college dropout.

第三，大學輟學。

All three things simultaneously true.

這三件事同時都是真的。

Ladies and gentlemen, please help me welcome Mark Zuckerberg.

女士們，先生們，請允許我歡迎馬克-扎克伯格。

Mark, welcome to your first SIGGRAPH.

馬克，歡迎你第一次參加 SIGGRAPH。

Can you believe this?

你能相信嗎？

One of the pioneers of computing, a driver of modern computing, and I had to invite him to SIGGRAPH.

他是計算的先驅之一，是現代計算的推動者，我必須邀請他參加 SIGGRAPH。

So, anyways, Mark, sit down.

不管怎樣 馬克 坐下吧

It's great to have you here.

很高興你能來。

Welcome.

歡迎光臨

Thanks for flying down.

謝謝你飛過來。

Yeah, no, this will be fun.

是啊，不，這會很有趣的。

I hear you've been going for like five hours already or something.

我聽說你已經走了五個小時了

Well, yeah, sure.

嗯，是的，當然。

This is SIGGRAPH, you know, these 90% PhDs.

這是 SIGGRAPH，你知道，這些 90% 都是博士。

And so the thing that's really great about SIGGRAPH, as you know, this is the show of computer graphics, image processing, artificial intelligence, and robotics combined.

大家都知道，SIGGRAPH 是一個集計算機圖形學、圖像處理、人工智能和機器人技術於一體的展會。

And some of the companies that over the years have demonstrated and revealed amazing things here from Disney, Pixar, Adobe, Epic Games, and of course, you know, NVIDIA.

多年來，迪士尼、皮克斯、Adobe、Epic Games，當然還有英偉達（NVIDIA）等公司都在這裡展示和揭示了令人驚歎的東西。

We've done a lot of work here.

我們在這裡做了很多工作。

This year we introduced 20 papers at the intersection of artificial intelligence and simulation.

今年，我們推出了 20 篇有關人工智能與仿真交叉領域的論文。

So we're using artificial intelligence to help simulation be way larger scale, way faster.

是以，我們正在利用人工智能來幫助模擬更大規模、更快速度。

For example, differentiable physics.

例如，可微分物理學。

We're using simulation to create simulation environments for synthetic data generation for artificial intelligence.

我們正在利用仿真技術為人工智能的合成數據生成創建仿真環境。

And so these two areas are really coming together.

是以，這兩個領域真的走到了一起。

Really proud of the work that we've done here at Meta.

我為我們在 Meta 所做的工作感到自豪。

You guys have done amazing AI work.

你們的人工智能工作非常出色。

I mean, one of the things that I find amusing is when the press writes about how Meta's jumped into AI this last couple of years.

我的意思是，我覺得很有趣的一件事是，當媒體寫到 Meta 在過去幾年裡是如何躍入人工智能領域的。

As if, you know, the work that FAIR has done, remember, we all use PyTorch.

就好像，你知道，FAIR 所做的工作，請記住，我們都在使用 PyTorch。

That comes out of Meta.

這來自於 Meta。

The work that you do in computer vision, the work in language models, real-time translation, groundbreaking work.

你們在計算機視覺、語言模型、實時翻譯方面所做的工作，都是開創性的工作。

I guess my first question for you is, how do you see the advances of generative AI at

我想我要問您的第一個問題是，您如何看待生成式人工智能的進步？

Meta today, and how do you apply it to either enhance your operations or introduce new capabilities that you're offering?

今天的 Meta，以及您如何應用它來提升您的營運或推出您正在提供的新功能？

Yeah.

是啊

So, a lot to unpack there.

所以，這裡有很多東西需要解讀。

First of all, really happy to be here.

首先，很高興來到這裡。

Meta has done a lot of work and has been at SIGGRAPH for eight years.

Meta 做了大量工作，在 SIGGRAPH 已經工作了八年。

So, we're noobs compared to you guys.

所以，和你們相比，我們是新手。

But I think it was back in 2018...

但我想那是在 2018 年...

You're dressed right, but this is my hood.

你穿得沒錯，但這是我的兜帽。

I just wanted to...

我只是想...

I mean, well, thank you for welcoming me to your hood.

我是說，謝謝你歡迎我來到你的地盤。

I think it was back in 2018, we showed some of the early hand-tracking work for our VR and mixed reality headsets.

我想應該是在 2018 年，我們展示了 VR 和混合現實頭顯的一些早期手部追蹤工作。

I think we've talked a bunch about the progress that we're making on codec avatars, the photorealistic avatars that we want to be able to drive from consumer headsets, which we're getting closer and closer to.

我想我們已經談了很多我們在編解碼器頭像方面取得的進展，我們希望能夠通過消費級頭顯驅動逼真的頭像，我們離這個目標越來越近了。

So, pretty excited about that.

所以，我對此非常興奮。

And also, a lot of the display systems work that we've done.

此外，我們還做了很多顯示系統方面的工作。

So, some of the future prototypes and research for getting the mixed reality headsets to be able to be really thin, with just pretty advanced optical stacks and display systems, the integrated system.

是以，未來的一些原型和研究都是為了讓混合現實頭盔能夠變得非常輕薄，並配備相當先進的光學堆棧和顯示系統，即集成系統。

That's been stuff that we've typically shown here first.

我們通常會先在這裡展示這些內容。

So, excited to be here this year, not just talking about the metaverse stuff, but also all the AI pieces, which, as you said, we started FAIR, the AI research center.

所以，很高興今年能來到這裡，不僅是討論元宇宙的東西，還有所有人工智能的部分，正如你所說，我們成立了人工智能研究中心 FAIR。

Back then it was Facebook, now Meta, before we started Reality Labs.

當時是 Facebook，現在是 Meta，在我們創辦 Reality Labs 之前。

We've been at this for a while.

我們這樣做已經有一段時間了。

All the stuff around Gen AI, it's an interesting revolution.

所有與人工智能有關的東西，都是一場有趣的革命。

And I think that it's going to end up making all of the different products that we do different in interesting ways.

我認為，它最終會以有趣的方式使我們生產的所有不同產品變得與眾不同。

You can look at the big product lines that we have already, so things like the feed and recommendation systems and Instagram and Facebook.

你可以看看我們已經擁有的大型產品線，比如 feed、推薦系統、Instagram 和 Facebook。

We've kind of been on this journey where that's gone from just being about connecting with your friends.

我們已經走過了這樣一段旅程，從單純的與朋友建立聯繫。

And the ranking was always important, because even when you were just following friends, if someone did something really important, like your cousin had a baby or something, it's like, you want that at the top, you'd be pretty angry at us if it was buried somewhere down in your feed.

排名一直都很重要，因為即使你只是關注朋友，如果有人做了非常重要的事情，比如你的表弟生了個孩子什麼的，你就會希望把它放在最前面，如果它被埋在你的推送中的某個地方，你就會對我們非常生氣。

So, the ranking was important, but now, over the last few years, it's gotten to a point where more of that stuff is just different public content that's out there.

是以，排名是很重要的，但現在，在過去幾年裡，更多的內容都是不同的公開內容。

The recommendation systems are super important, because now, instead of just a few hundred or thousand potential candidate posts from friends, there's millions of pieces of content.

推薦系統非常重要，因為現在，潛在的候選帖子不再是幾百個或幾千個來自朋友的帖子，而是數以百萬計的內容。

And that turns into a really interesting recommendation problem.

這就變成了一個非常有趣的推薦問題。

And with generative AI, I think we're going to quickly move into the zone where not only is the majority of the content that you see today on Instagram just recommended to you from stuff that's out there in the world that matches your interests and whether or not you follow the people.

有了生成式人工智能，我認為我們將很快進入這樣一個領域：不僅你今天在 Instagram 上看到的大部分內容都是由世界上符合你興趣的東西推薦給你的，而且無論你是否關注這些人。

I think in the future, a lot of this stuff is going to be created with these tools, too.

我認為，在未來，很多東西也將通過這些工具來創建。

Some of that is going to be creators using the tools to create new content.

其中一部分將由創作者使用這些工具來創建新內容。

Some of it, I think, eventually is going to be content that's either created on the fly for you or pulled together and synthesized through different things that are out there.

我認為，其中一些內容最終將是為你即時創建的，或者是通過現有的不同內容整合在一起的。

So I think that's just one example of how the core part of what we're doing is just going to evolve.

是以，我認為這只是一個例子，說明我們正在做的事情的核心部分將如何發展。

And it's been evolving for 20 years already.

它已經發展了 20 年。

Well, very few people realize that one of the largest computing systems the world has ever conceived of is a recommender system.

很少有人知道，世界上有史以來最大的計算系統之一就是推薦系統。

Yeah.

是啊

I mean, it's this whole different path, right?

我的意思是，這是一條完全不同的道路，對嗎？

It's not quite the kind of gen AI hotness that people talk about, but I think it's all the transformer architectures and it's a similar thing of just building up more and more general models.

這並不像人們談論的那種人工智能的熱潮，但我認為它是所有的變壓器架構，也是建立越來越多通用模型的類似事情。

Embedding unstructured data into features.

將非結構化數據嵌入特徵。

Yeah.

是啊

One of the big things that just drives quality improvements is it used to be that you'd have a different model for each type of content, right?

過去，每種類型的內容都有不同的模式，對嗎？

So a recent example is we had one model for ranking and recommending reels and another model for ranking and recommending more long form videos.

最近的一個例子是，我們有一種模式用於排名和推薦卷軸，另一種模式用於排名和推薦更多的長視頻。

And then it takes some product work to basically make it so that the system can display anything in line.

然後，還需要進行一些產品工作，才能使系統能夠顯示排成一行的任何內容。

But the more you just create more general recommendation models that can span everything, it just gets better and better.

但是，你越是創建能夠涵蓋一切的通用推薦模型，它就會變得越來越好。

So, I mean, part of it I think is just like economics and liquidity of content and the broader of a pool that you can pull from.

所以，我的意思是，我認為其中一部分原因在於經濟學、內容的流動性以及你可以從中獲取的更廣泛的內容。

You're just not having these weird inefficiencies of pulling from different pools.

這樣就不會出現從不同資源庫中提取資源的怪異低效現象。

But yeah, I mean, as the models get bigger and more general, that gets better and better.

不過，我的意思是，隨著模型越來越大，越來越普遍，效果也會越來越好。

So I kind of dream of one day, like you can almost imagine all of Facebook or Instagram being like a single AI model that has unified all these different content types and systems together that actually have different objectives over different timeframes, right?

是以，我夢想著有一天，你幾乎可以想象 Facebook 或 Instagram 的所有內容都像一個單一的人工智能模型，將所有這些不同的內容類型和系統統一在一起，在不同的時間範圍內實現不同的目標，對嗎？

Because some of it is just showing you, you know, what's the interesting content that you're going to be, that you want to see today.

因為有些內容只是向你展示，你知道，你今天想看的有趣內容是什麼。

But some of it is helping you build out your network over the long term, right?

但從長遠來看，這其中有一些是在幫助你建立自己的網絡，對嗎？

People you may know or accounts you might want to follow.

您可能認識的人或想關注的賬戶。

These multimodal models tend to be much better at recognizing patterns, weak signals and such.

這些多模態模型在識別模式、弱信號等方面往往更勝一籌。

And so one of the things that people always, you know, it's so interesting that AI has been so deep in your company.

是以，人們總是認為，人工智能在貴公司的應用如此深入，實在是太有趣了。

You've been building GPU infrastructure, running these large recommender systems for a long time.

你們構建 GPU 基礎設施、運行這些大型推薦系統已經有很長一段時間了。

Now you're, now you're...

現在你，現在你...

I'm a little slow on it, actually.

事實上，我對它有點遲鈍。

Getting to GPUs.

使用 GPU。

Yeah.

是啊

I was trying to be nice.

我只是想表示友善。

I know.

我知道

Well, you know, too nice.

嗯，你知道，太善良了。

I was trying to be nice, you know, you're my guest.

我只是想表示友善，你知道，你是我的客人。

When I was backstage before I came on here, you were talking about like owning your mistakes or something.

我上臺之前在後臺的時候，你說過要勇於承擔錯誤之類的話。

Right?

對不對？

So...

所以...

You don't have to volunteer it out of the blue.

你不必突然主動提出來。

I think this one has been well tried.

我認為這一點已經得到了很好的驗證。

Yeah.

是啊

It's like I got raked over the shoulder for a while, you know?

就好像我的肩膀被耙了一下，你知道嗎？

But as soon as you got into it, as soon as you got into it, you got into it strong.

但是，只要你投入其中，只要你投入其中，你就會很投入。

Let's just put it...

讓我們把它...

There you go.

這就對了。

There you go.

這就對了。

Now, the thing that's really cool about generative AI is these days when I use WhatsApp, I feel like I'm collaborating with WhatsApp.

現在，生成式人工智能最酷的地方在於，如今當我使用 WhatsApp 時，我感覺自己是在與 WhatsApp 合作。

I love imagine.

我喜歡想象。

I'm sitting here typing and it's generating the images as I'm going.

我坐在這裡打字，它就在我打字的同時生成影像。

I go back and I change my words, it's generating other images, you know?

我回去修改我的文字，它就會產生其他影像，你知道嗎？

And so, the one that old Chinese guy enjoying a glass of whiskey at sundown with three dogs, golden retriever, golden doodle, and a Bernese mountain dog.

於是，就有了那個在日落時分帶著三隻狗（金毛尋回犬、金毛嘟嘟犬和伯恩山犬）喝著威士忌的中國老頭。

And it generates, you know, a pretty good looking picture.

你知道，這幅畫看起來很不錯。

Yeah.

是啊

Yeah.

是啊

We're getting there.

我們快成功了

That's me.

這就是我。

Every month.

每個月

Yeah, yeah.

是啊，是啊。

And then now you could actually load my picture in there and it'll actually be me.

現在你可以把我的照片放進去，那就真的是我了。

Yeah.

是啊

That's as of last week.

這是上週的數據。

Yeah.

是啊

That's me.

這就是我。

I know.

我知道

I'm spending a lot of time with my daughters imagining them as mermaids and things over the last week.

上週，我花了很多時間和女兒們在一起，把她們想象成美人魚什麼的。

It's been a lot of fun.

這很有趣。

Yeah.

是啊

But yeah, I mean, that's the other half of it.

但是，是的，我的意思是，這是它的另一半。

I mean, a lot of the gen AI stuff is going to...

我的意思是，很多人工智能技術都將...

On the one hand, it's, I think, going to just be this big upgrade for all of the workflows and products that we've had for a long time.

一方面，我認為這將是對我們長期使用的所有工作流程和產品的一次重大升級。

But on the other hand, there's going to be all these completely new things that can now get created.

但另一方面，現在可以創造出所有這些全新的東西。

So, meta AI, you know, the idea of having, you know, just an AI assistant that can help you with different tasks in our world is going to be, you know, very creatively oriented, like you're saying.

是以，元人工智能，你知道，在我們的世界裡，擁有一個能幫你完成不同任務的人工智能助手的想法，你知道，就像你說的那樣，將是非常有創造力的。

But, I mean, they're very general, so you don't need to just constrain it to that.

但是，我的意思是，它們非常籠統，所以你不需要僅僅侷限於此。

It'll be able to answer any question.

它能回答任何問題。

Over time, I think, you know, when we move from, like, the LLAMA3 class of models to

隨著時間的推移，我想，當我們從 LLAMA3 類模型轉向

LLAMA4 and beyond, it's going to, I think, feel less like a chat bot where it's like, you give it a prompt and it just responds, and then you give it a prompt and it responds, and it's just, like, back and forth.

我認為，LLAMA4 及更高版本將不再像哈拉機器人那樣，你給它一個提示，它就回復，然後你再給它一個提示，它再回復，就這樣來來回回。

I think it's going to pretty quickly evolve to, you give it an intent, and it actually can go away on multiple timeframes, and, I mean, it probably should acknowledge that you gave it an intent up front, but, I mean, you know, some of the stuff, I think, will end up, you know, it'll spin up, you know, compute jobs that take, you know, weeks or months or something, and then just come back to you when, like, something happens in the world, and I think that that's going to be really powerful.

我認為它很快就會發展成，你給它一個意圖，它實際上可以在多個時間框架內消失，而且，我的意思是，它可能應該承認你在前面給了它一個意圖，但是，我的意思是，你知道，有些東西，我認為，會最終，你知道，它會旋轉，你知道，計算工作需要，你知道，幾個星期或幾個月或什麼的，然後只是回到你身邊，就像，在世界上發生的事情，我認為這將是非常強大的。

So, I mean, I'm quite...

所以，我的意思是，我很...

Today's AI, as you know, is kind of turn-based.

眾所周知，如今的人工智能是一種回合制。

You say something, it says something back to you, but, obviously, when we think, when we're given a mission or we're given a problem, you know, we'll contemplate multiple options or maybe we come up with a, you know, a tree of options, a decision tree, and we walk down to the decision tree simulating in our mind, you know, what are the different outcomes of each decision that we could potentially make, and so we're doing planning, and so in the future, AIs will kind of do the same.

你說了什麼，它就會回你什麼，但很明顯，當我們思考時，當我們被賦予一項任務或被賦予一個問題時，你知道，我們會考慮多個選項，或者也許我們會想出一棵，你知道，一棵選項樹，一棵決策樹，然後我們走到決策樹下，在腦海中模擬，你知道，我們可能做出的每個決定的不同結果是什麼，所以我們在做規劃，所以在未來，人工智能也會做同樣的事情。

One of the things that I was super excited about, when you talked about your vision of creator AI, I just think that's a home-run idea, frankly.

當你談到你對人工智能創造者的願景時，有一件事讓我非常興奮，坦白說，我認為這是一個全壘打的想法。

Tell everybody about the creator AI and AI studio that's going to enable you to do that.

向大家介紹一下人工智能創造者和人工智能工作室，它們將使你們能夠做到這一點。

Yeah, so we actually, I mean, this is something that we're, you know, we've talked about it a bit, but we're rolling it out a lot wider today.

是的，所以我們實際上，我的意思是，這是我們，你知道，我們已經談論過它一點點，但我們今天推出它更廣泛。

You know, a lot of our vision is that...

你知道，我們的許多願景是...

I don't think that there's just going to be, like, one AI model, right?

我認為不會只有一種人工智能模型，對嗎？

I mean, this is something that some of the other companies in the industry, they're, like, you know, it's like they're building, like, one central agent, and yeah, we'll have the meta AI assistant that you can use, but a lot of our vision is that we want to empower all the people who use our products to basically create agents for themselves.

我的意思是，這是行業內其他一些公司正在做的事情，就像，你知道的，就像他們正在建立一箇中央代理一樣，是的，我們會有你可以使用的元人工智能助手，但我們的很多願景是，我們希望讓所有使用我們產品的人基本上都能為自己創建代理。

So whether that's, you know, all the many, many millions of creators that are on the platform or, you know, hundreds of millions of small businesses, we eventually want to just be able to pull in all your content and very quickly stand up a business agent and be able to interact with your customers and, you know, do sales and customer support and all that.

是以，不管是平臺上數以百萬計的創作者，還是數以億計的小企業，我們最終都希望能夠拉入你的所有內容，並快速建立一個業務代理，能夠與你的客戶互動，進行銷售和客戶支持等所有工作。

So the one that we're just starting to roll out more now is, we call it AI studio, and it basically is a set of tools that eventually is going to make it so that every creator can build sort of an AI version of themselves as sort of an agent or an assistant that their community can interact with.

是以，我們現在剛剛開始推出的是，我們稱之為人工智能工作室，它基本上是一套工具，最終將使每個創作者都能建立一個人工智能版本的自己，作為他們的社區可以與之互動的代理或助手。

There's kind of a fundamental issue here where there's, there's just not enough hours in the day, right?

這裡有一個根本性的問題，就是每天的時間不夠用，對嗎？

It's like, if you're a creator, you want to engage more with your community, but you're constrained on time.

這就好比，如果你是一名創作者，你想更多地參與社區活動，但時間有限。

And similarly, your community wants to engage with you.

同樣，您的社區也希望與您互動。

But it's tough.

但這很難。

I mean, there's just, there's limited time to do that.

我的意思是，這樣做的時間有限。

So the next best thing is, is allowing people to basically create these artifacts, right?

所以，下一個最好的辦法就是，允許人們創造這些人工製品，對嗎？

It's sort of, it's an agent, but it's, you train it to kind of, on your material to represent you in the way that you want.

它有點像經紀人，但你要訓練它，讓它根據你的材料，以你想要的方式代表你。

I think it, it's a very kind of creative endeavor, almost like a, like a piece of art or content that you're putting out there.

我認為，這是一種非常有創意的努力，幾乎就像一件藝術品或你要發佈的內容。

And it's going to be very clear that it's not engaging with the creator themselves, but I think it'll be another interesting way, just like how creators put out content on, on these social systems to be able to have agents that do that.

它將會非常清楚地表明，它並不是與創作者本身互動，但我認為這將是另一種有趣的方式，就像創作者如何在這些社交系統上發佈內容一樣，能夠讓代理來做這件事。

Similarly, I think that there's going to be a thing where people basically create their own agents for all different kinds of uses.

同樣，我認為未來還會出現這樣一種情況：人們基本上都會為各種不同的用途創建自己的代理。

Some will be sort of customized utility, things that they're trying to get done that they want to kind of fine tune and, and train an agent for.