Tetragrammaton with Rick Rubin - 格雷格·布罗克曼(第二部分) 封面

格雷格·布罗克曼(第二部分)

Greg Brockman (Part 2)

本集简介

继续参与第二部分对话的格雷格·布罗克曼是OpenAI的联合创始人兼总裁,OpenAI是ChatGPT背后的公司。 ------ 感谢赞助商对我们播客和团队的支持: AG1 https://DrinkAG1.com/tetra ------ Athletic Nicotine https://www.AthleticNicotine.com/tetra 使用代码 'TETRA' ------ LMNT电解质 https://DrinkLMNT.com/tetra 使用代码 'TETRA' ------ 注册接收 Tetragrammaton 通讯 https://www.tetragrammaton.com/join-newsletter

双语字幕

仅展示文本字幕,不包含中文音频;想边听边看,请使用 Bayt 播客 App。

Speaker 0

四字神名。

Tetragrammaton.

Speaker 1

我第一个获得用户访问的网站是一个令人惊叹的体验。

The first website I ever built that got users was this amazing experience.

Speaker 1

我有一个想法,要构建我称之为逆图灵测试的东西。

I had this idea to build what I call the reverse Turing test.

Speaker 1

在图灵测试中,为了判断机器是否具有智能,会让一个人与另一个人对话,同时与一个AI对话。

So in the Turing test to determine is a machine intelligent, you have a human who talks to another human and talks to an AI.

Speaker 1

目标是分辨出哪一个是人类,哪一个是AI。

And the goal is to figure out which of these is the human, which is the AI.

Speaker 1

于是我建立了一个网站,将这个过程变成了一场竞赛:两个真人彼此对话,同时各自与一个AI对话。

So I built a website that turned this into a competitive game where you have both humans are talking to each other and they're each talking to an AI.

Speaker 1

他们不知道哪个终端对应哪一方。

They don't know which terminal is which.

Speaker 1

目标是比对方更快地判断出哪个终端连接的是另一个真人。

And the objective is to figure out which of your terminals is the other human before the other human does.

Speaker 1

因此,最佳策略是提出一些能分辨出你是在和人类还是机器人对话的问题,但同时又要表现得有点像机器人。

And so the optimal strategy is to ask questions that kind of discern, am I talking to a human or a bot, but while still acting kind of bot like.

Speaker 1

如果你表现得太像人类,就会输,因为对方会识破你的身份。

If you act too human like, then you'll lose because the other person will figure out who you are.

Speaker 1

哇。

Wow.

Speaker 1

那是2008年。

This was 2008.

Speaker 1

我当时自学了编程。

And I just taught myself how to code.

Speaker 1

我上网看了W3schools的教程,学习了HTML、JavaScript、PHP和CSS。

I'd gone online to W3schools tutorials, did HTML, JavaScript, PHP, CSS.

Speaker 1

我记得我开发了这个游戏,这是一个双人游戏。

And I remember that I built this game, and it's a two player game.

Speaker 1

我当时正坐在大厅里。

I was sitting in the lobby.

Speaker 1

所以如果有人来了,他们能有良好的体验,有人可以对战。

So in case anyone showed up, they'd have a good experience and have someone to play against.

Speaker 0

在哪个大厅?

Lobby where?

Speaker 1

我让我的网站上有一个类似游戏大厅的地方。

I made it so that there was like a game lobby just on my website.

Speaker 0

我明白了。

I see.

Speaker 1

于是我坐在那里,就开着屏幕等着,可怜地等着有人出现。

And so I just sit there just waiting, waiting with this open screen, just sadly waiting for someone to show up.

Speaker 1

整整两周,都没人来。

And for like two weeks, no one showed up.

Speaker 1

但有一天,那是最辉煌的一天,我从StumbleUpon获得了1500次访问。

But then one day, it was the most glorious day, I got 1,500 hits from StumbleUpon.

Speaker 1

哇。

Wow.

Speaker 1

是的,如果你还记得StumbleUpon,它就像早期的一种工具,会把人们送到随机的网站上。

Yeah, if you remember StumbleUpon, it was like an early like, you know, it would send people to random websites.

Speaker 1

那一天真是令人惊叹,总是有三四局游戏在同时进行。

And it was an amazing moment where that day, there were like always three or four games going constantly.

Speaker 1

我会坐在大厅里,几分钟内就有人加入。

I'd sit in the lobby and someone joined within a couple of minutes.

Speaker 1

我记得那种感觉,这个想法原本只在我脑子里,现在却变成了现实。

And I remember this feeling that this thing was in my head and now it's in reality.

Speaker 1

现在这些人都在享受我创造的东西,我想继续追求下去。

And now these people are all enjoying what I built and I want to keep chasing it.

Speaker 0

对。

Yeah.

Speaker 0

你玩得越多,会不会在这款游戏里变得更强?

The more you played, would you get better at that game?

Speaker 1

会。

Yes.

Speaker 1

是的。

Yes.

Speaker 1

我玩得相当不错。

I got quite good at it.

Speaker 1

实际上,这很有趣。

And it was interesting, actually.

Speaker 1

我花了很多精力改进这个机器人。

My I focused a lot on improving the bot.

Speaker 1

我的机器人非常原始,它的运行方式是:我保存了所有过往对话的数据库,然后在任何特定对话中,尝试将其与最相似的对话匹配,并用当时人类的回复来回应。

My bot was, like, very rudimentary, and that the way that it would work is I kept a database of all the previous games, and then I tried to, in any particular conversation, match that conversation to the most similar one and then reply with what the human had said then.

Speaker 1

它实际上还真管用,对吧?

And it actually kind of worked, right?

Speaker 1

对于任何类似闲聊的内容,只要你有足够丰富的回复数据库,就能应对得很好。

For any sort of chitchatty thing that kind of it's done already, then you have a pretty good database of replies.

Speaker 1

但一旦涉及更复杂的内容,它当然就会崩溃——

But anything more sophisticated, and of course, it just would fall on its- Can

Speaker 0

你还记得那个游戏的想法是从哪里来的吗?

you remember at all where the idea came from for that game?

Speaker 1

我之所以对人工智能产生兴趣,是因为读了艾伦·图灵关于图灵测试的论文。

Well, the way that I got excited about AI was by reading Alan Turing's paper on the Turing test.

Speaker 1

这是他1950年发表的论文《计算机器与智能》。

So this is his 1950 paper called Computing Machinery and Intelligence.

Speaker 1

我在开发这个游戏前不久读了这篇论文,里面充满了最具启发性的想法。

I was reading it shortly before building this game and it had the most inspirational ideas in it.

Speaker 1

因为首先他提出了一个问题:机器是否可能具备智能?

Because first he asked the question, can a machine ever be intelligent?

Speaker 1

他说:我不知道‘智能’到底意味着什么。

And he says, look, I don't know what intelligence means.

Speaker 1

所以让我们来定义一个测试。

So let's define a test.

Speaker 1

他提出了图灵测试,但接着又问:你该如何编程来应对这个测试呢?

He defined the Turing test, but then he says, how will you ever program an answer to this test?

Speaker 1

你永远无法直接编程实现它。

You will never program it.

Speaker 1

要写出所有‘这个人这么说、那个人那样说’的规则实在太难了。

It's just too hard to write down all the rules of this person says this and that.

Speaker 1

相反,你需要构建一个能够自行学习答案的机器。

Instead, you will need to build a machine that can learn its own answer to this.

Speaker 1

你必须打造一台学习型机器。

You have to build a learning machine.

Speaker 1

所以,如果你能建造一台像人类儿童那样学习的‘儿童机器’,再由人类在它做对事时给予奖励、做错事时给予惩罚,那么你就能解决这个测试。

So if you can build a machine that is a child machine that learns like a human child, and you can then have a human who gives it rewards and punishments as it does good things and bad things, then that's how you will solve this test.

Speaker 1

这里有个疯狂的地方。

And here's the wild thing.

Speaker 1

这正是我们一直在做的事情。

That is exactly what we've been doing.

Speaker 1

这就像艾伦·图灵在1950年预言了我们如何首先构建这些无监督学习模型——它们能观察世界,并积累如此丰富的知识。

This is like Alan Turing in 1950 projecting how we will first build these unsupervised models that learn, that sort of observe the world, have all this knowledge in them.

Speaker 1

然后我们进行这种强化学习过程。

And then we do this reinforcement learning process.

Speaker 1

我们通过给予机器奖励和惩罚来实现你为它设定的目标。

We give the machine rewards and punishments in order to achieve the objective that you have in front of it.

Speaker 1

我记得曾和我的联合创始人伊利亚讨论过这个,我当时说,图灵是怎么知道的?

And I remember talking to my co founder Ilya about this, and I was like, how did Turing know?

Speaker 1

他说,图灵之所以被称为图灵,是因为他实在太聪明了。

And he said, there's a reason that Turing is Turing, right, that he is just so smart.

Speaker 1

所以,正是因此,他他

And, that's why, yeah, he he

Speaker 0

你为什么觉得从1950年到现在花了这么长时间?

was Why do you testing think it took so long from 1950 to now?

Speaker 1

答案很简单。

Very simple answer.

Speaker 1

是计算力。

It's compute.

Speaker 1

当时计算能力根本不够。

There just was not enough compute.

Speaker 1

不管图灵有多聪明,他都没有一台可以实现他测试的计算机。

It's like no matter how smart Turing was, he did not have a computer that he could implement his test on.

Speaker 0

在早期,你有没有觉得人工智能似乎是一种失败的东西?

Was there a feeling in the early days for you that AI was sort of a failed thing?

Speaker 0

这些想法从五十年代就存在了,但一直没真正成功。

The ideas have been around since the fifties, and it hasn't really worked.

Speaker 1

对我来说,这一直是一个巨大的谜团。

So this for me was always a great mystery.

Speaker 1

这很明显。

It was so clear.

Speaker 1

图灵测试的愿景非常清晰。

The Turing test vision was so clear.

Speaker 1

这正是你必须去做的事情,因为作为一名程序员,我必须理解问题的解决方案。

It was like, that's the thing you need to do because as a programmer, I have to understand the solution to the problem.

Speaker 1

但图灵的观点是,你可以拥有一台能够自行找出问题解决方案的机器。

But the point that Turing makes is that you can have a machine that comes up with its own solution to the problem.

Speaker 1

我当时就想,我有那么多根本不知道怎么解决的问题,也许机器能做到。

And I was like, think of how many problems I have no idea how to solve and maybe the machine could do it.

Speaker 1

这就是我想做的事。

That's what I wanna do.

Speaker 1

我想帮助这种东西得以实现。

I wanna help that thing come into existence.

Speaker 1

我记得在开发完这个游戏之后,我去了哈佛大学,那时还是2008年,我非常兴奋地想跟一位自然语言处理教授做研究。

And I remember after building this game, I showed up at college at Harvard and it was still 2008, and I was very excited to do research with a natural language processing professor.

Speaker 0

你在哈佛学数学吗?

Were you studying math at Harvard?

Speaker 1

我本来打算学数学,结果被计算机科学迷住了。

I was gonna study math and I got computer science nerd snipe.

Speaker 1

我最想学的本来是数学。

My number one thing was was gonna be math.

Speaker 1

我原本打算主修数学、化学和哲学三重专业,但后来我意识到,我真正热爱的是用计算机进行实际构建。

I actually was originally intending to do math, chemistry, and philosophy triple major, but I I ended up realizing that I just loved the practicality of of building with computers.

Speaker 1

于是我去找这位教授,问他我能否跟他做研究。

And so I went to this professor and I asked him if I could do research with him.

Speaker 1

他说:当然可以,没问题。

He said, yes, no problem.

Speaker 1

他给了我一个语法树的问题。

And he gave me this parse trees problem.

Speaker 1

我记得当时看着那些语法树。

And I remember looking at the parse trees.

Speaker 1

心想:这根本不可能扩展。

Was like, this is never gonna scale.

Speaker 1

这太

This is

Speaker 0

不是,那是什么?

not What is that?

Speaker 0

我不知道那是什么。

I don't know what that is.

Speaker 1

解析树是一种老派的自然语言处理方法。

Parse trees, they were like an old school NLP, like natural language processing approach.

Speaker 1

也就是说,想象一下,你取一些句子,找出其中的宾语、名词等,就像小学生学的语言分析那样。

So the idea of like, imagine you take sentences and you figure out like where the object is and where the noun is, kind of like the things that people do in elementary school.

Speaker 0

就像语言学。

Like linguistics.

Speaker 1

是的,语言学是一种用于自然语言处理的方法。

Yeah, like linguistics as an approach for how you're going to do natural language processing.

Speaker 1

你可以从中得到一些看起来简单的结果,因为它能生成看起来合理的句子。

And you can get some kind of simple looking stuff out of that because it'll make reasonable looking sentences.

Speaker 1

但这种方法永远无法扩展到进行这种对话的程度。

But that's never gonna scale to having a conversation like this.

Speaker 1

对我来说,这一点非常清楚。

It was just so clear to me.

Speaker 1

这根本不是托尔所谈论的内容。

Was like, this is not what Tore was talking about.

Speaker 1

我会去做一些更有用的事情。

I'll go do things that are useful.

Speaker 1

而实际上,我进入了编程语言领域。

And instead I actually got into programming languages.

Speaker 1

对此我非常兴奋。

So I was very excited about that.

Speaker 1

我选了一门课。

I took a class.

Speaker 1

我想在那里做更多研究。

I wanted to do more research there.

Speaker 1

关于编程语言的一点是,编译器是一个计算机程序,它接收另一个程序并以某种方式使其更好。

And the thing about programming languages is you have all of this power of like a compiler is a computer program that takes a different program and makes it better in some way.

Speaker 1

因此,它通常将程序从高级形式转换为机器能够理解的形式,通常还会对其进行优化,使其更快,真正将人类的意图转化为机器能够处理细节的形式。

So it usually takes it from a high level form and puts it into a form the machine can understand, usually optimize it so it's faster and really take the intent of a human and translate into a form that the machine can then take care of some of the details.

Speaker 1

因为对我来说,这种精神就是我希望有一台机器,能够解决我无法解决的问题,以我无法达到的方式赋予我力量,带我达到新的高度。

Because for me, that was the spirit is I want a machine that can solve problems that I can't, that will empower me in ways that I am unable to reach, that will bring me to new heights.

Speaker 1

不只是为我,而是为每个人。

Not just for me, but for everyone.

Speaker 1

所以我就会去这么做。

And so I would do that.

Speaker 1

我深入参与了哈佛计算机协会,我们为哈佛社区构建各种服务。

I got very into the Harvard Computer Society where we would build services for the Harvard community.

Speaker 1

所以我们提供电子邮件和网页托管以及其他技术服务。

So we would host email and web hosting and other technical services.

Speaker 1

为社区开发了不同的网页应用。

Built different web applications for the community.

Speaker 1

这正是我当时的信念。

And so this was very much the ethos that I had.

Speaker 1

而直到我加入Stripe、在做创业项目并构建产品时,我才真正开始关注社区。

And it really wasn't until I was already at Stripe where I was doing a startup and building things, but paying attention to the community.

Speaker 1

我不断看到人们谈论深度学习。

And I just kept seeing people talking about deep learning.

Speaker 1

感觉每天只要上Hacker News——那个很多工程师发布内容的网站——都会看到关于深度学习的讨论。

It felt like every day, if you went on Hacker News, which was this website that many engineers would post content on, you'd see something about deep learning for axe.

Speaker 1

我记得当时在想,深度学习到底是什么?

And I remember wondering what is deep learning?

Speaker 1

那时候,想弄明白几乎不可能,因为你去deeplearning.tom或.org之类的网站,上面只说深度学习是一种新的AI方法。

And at the time it was basically impossible to figure out because you go to deeplearning.tom or .org or whatever it was, and it just said deep learning is a new approach to AI.

Speaker 0

但并没有说明它到底是什么。

But didn't say what it was.

Speaker 1

也没说它到底是什么。

Didn't say what it was.

Speaker 0

是的,

Yeah,

Speaker 1

对。

yeah.

Speaker 1

这完全说不通。

It made no sense.

Speaker 1

但我记得我认识这个领域的一个朋友,于是我去找他,他开始引荐我认识领域里的其他人。

But I remember that I had a friend in the field, so I went and talked to that person, and he started introducing me to other people in the field.

Speaker 1

我不断被介绍给大学里所有最聪明的朋友,因为他们现在都进入了这个领域。

And I just kept getting introduced to all my smartest friends from college because they were all on the field now.

Speaker 0

当时世界上这个领域有多少人?

How many people were in the field at that time in the world?

Speaker 1

这是一个很小的群体,非常小的群体。

It was a small community, very small community.

Speaker 1

我不知道确切的总数,最多一千人左右。

I don't know the overall number, a thousand at most.

Speaker 1

而且它正在迅速增长,因为我了解到,2012年有一个关键时刻,真正引爆了当前的深度学习革命。

And it was rapidly growing because the thing that I learned is that there was this moment in 2012 that really unleashed the current deep learning revolution.

Speaker 1

在很多方面,一切都在为那个时刻积累,但正是那个时刻让许多人确信:这里真的有东西。

And in many ways, everything had been building up to that moment, but this was the moment that really cemented the there's something real here for many people.

Speaker 1

这就是AlexNet的诞生,这是一篇在该基准测试中参赛的图像识别论文。

And this was the creation of AlexNet, which was a image recognition paper that competed on this benchmark.

Speaker 0

解释一下这是什么。

Explain what that is.

Speaker 1

所以,我认为在2006年或2008年,斯坦福的一个实验室发起了一项竞赛,他们从互联网上收集了数百万张高分辨率图像。

So the idea is that in, I think, 2006, 2008, a lab at Stanford created a competition where they gathered millions of high resolution images from across the web.

Speaker 1

在当时,我们会认为这是一个小数据集。

At this point, we would consider a small data set.

Speaker 1

但在当时,这已经是规模巨大且前所未有的了。

At the time, it was massive and unprecedented.

Speaker 1

他们将这些图像分类为一千个不同的类别,由人类标注,比如这是某种特定的猫,这是某种特定的鸟,这是某种特定的飞机。

And it categorized this image into a thousand different categories that humans would label them and say, this is a specific type of cat, and this is a specific type of bird, and this is a specific type of airplane.

Speaker 1

所以是一千种不同类别的图像。

So a thousand different categories of images.

Speaker 1

目标是创建一个机器、一个程序,能够将一张新图像归类到这千个类别中的某一个。

And the goal was create a machine, create a program that can categorize a new image into one of these thousand buckets.

Speaker 1

所以你能识别出图像中是猫还是狗吗?人们会在这个比赛中激烈竞争,这需要整合四十年来的计算机视觉研究思路,就像那些语法树的感觉一样,对吧?

So can you recognize whether they're a cat or dog in an image and that people would compete in this and it would compete hard and it would take all of these forty years worth of computer vision research ideas, very similar to like the feeling of those parse trees, right?

Speaker 1

你会使用各种不同的技术,比如边缘检测之类非常具体的方法。

You would have these different techniques that would do like edge detection and things like that that are like very specific.

Speaker 1

如果你思考一下识别猫的规则,可能是找两只眼睛,但你怎么知道那真的是眼睛呢?然后你再找鼻子,但即便如此,物体的朝向也会让编程实现变得非常复杂。

And if you think about what are the rules for recognizing a cat, it's like, well, maybe you look for a eye and another eye, but how do you recognize that there's an eye and you look if there's a nose and then, but okay, the orientation could make it very complicated to actually program this.

Speaker 0

关系。

Relationships.

Speaker 1

没错。

Exactly.

Speaker 1

所以你必须以非常分层的方式来描述这些关系。

So you have to talk about these relationships in very hierarchical.

Speaker 1

如果你想想这个过程,你需要理解所有这些不同部分是如何组合在一起的,以及这些部分之间又是如何相互关联的,非常复杂。

If you think about the process of you have to see how all these different pieces fit together and then how those pieces relate to other pieces, very complicated.

Speaker 1

因此,人们在这里并没有取得很好的成果。

And so people were not getting very good results here.

Speaker 1

这感觉完全不可能,而且真的,

It felt like a total impossible And really,

Speaker 0

这是对现实世界的观察。

it's a study of the real world.

Speaker 0

它其实并不技术性。

It wasn't really technical.

Speaker 0

它是观察性的。

It was observational.

Speaker 1

是的。

Yes.

Speaker 1

因为人们会想,人类是如何做到这一点的?

Because the thing that people would do is that they would say, how do I think humans do this?

Speaker 0

或者

Or

Speaker 1

我头脑中的过程是什么?

What's the process I have in my head?

Speaker 1

或者我们写下一些符号化的方法,来告诉机器如何执行一个过程。

Or Let's write down some symbolic way to tell the machine how to pursue a process.

Speaker 1

这种方法在某些领域非常成功,对吧?

And this was very successful in some domains, right?

Speaker 1

例如,90年代我们成功开发出强大的机器来下棋,但在计算机视觉等其他领域却非常不成功。

For example, chess is one that in the '90s that we built great machines to solve it, but very unsuccessful in other domains like computer vision.

Speaker 0

规则只有一页。

There's only one page of rules

Speaker 1

没错。

Exactly.

Speaker 1

因为象棋规则简单,搜索空间也小。

For the game of Simple rules and a small search space.

Speaker 1

所以计算机可以简单地列出所有可能性,然后从中选择胜出的路径——但这种方法对围棋来说是不够的,对吧?

So you could have a computer that would basically just say, you what, I'm gonna look through all the possibilities and that is how I will win, which by the way, was not enough for Go, right?

Speaker 1

围棋规则简单,但搜索空间极其庞大。

Go, simple rules, but massive search space.

Speaker 1

因此,你需要在搜索之上加入类似人类直觉的东西。

And so you needed something more like human intuition on top of the search.

Speaker 1

而对于计算机视觉,你需要的完全是直觉般的东西。

And for computer vision, you needed something that was entirely like intuition.

Speaker 1

于是,神经网络就此登场。

And so that's where the neural nets came into play.

Speaker 1

于是,由杰弗里·辛顿、伊利亚·苏茨克弗和亚历克斯·赫塞夫斯基组成的研究团队创建了一个神经网络,并赢得了这场竞赛。

And so a team of researchers who were Jeffrey Hinton, Ilya Sutskever, Alex Hersevsky, created a neural net that won this competition.

Speaker 1

它不仅仅是略微胜出。

And it didn't just slightly win it.

Speaker 1

它彻底碾压了其他所有方法,实现了巨大的飞跃。

It just blew everything else out of the water, like a massive jump.

Speaker 0

他们对它的看法跟其他人不同吗?

Did they have a different vision of it than everyone else?

Speaker 1

他们的方法就是神经网络。

Their approach was neural nets.

Speaker 1

没人相信它。

No one else believed in it.

Speaker 0

我明白了。

I see.

Speaker 1

这个结果背后的内幕其实非常有趣,因为亚历克斯·赫塞夫斯基是杰夫·辛顿实验室的一名研究生,他当时正在为GPU开发非常快速的卷积内核。

And it's really it's actually very funny, the inside story on how that result came to be, because Alex Hersevsky was a grad student in Jeff Hendon's lab, and he was working on very fast convolutional kernels for GPUs.

Speaker 1

所以他实际上是在编程图形处理器(GPU),而如今这些GPU正是人们用于深度学习的设备。

So he basically was programming graphics processing units, GPUs, which now are what people use for for for deep learning.

Speaker 1

每个人都对他感到同情。

And everyone felt bad for him.

Speaker 1

他只是说,这不过是个工程类项目。

He was like, that's just an engineering project.

Speaker 1

他只是在编写这些超快的内核。

He's just writing these very fast kernels.

Speaker 1

谁会在意呢?

Who cares?

Speaker 1

我们都在做这些酷炫的研究。

Like, we're off doing all this cool research.

Speaker 1

他只是一个工程师。

He's just an engineer.

Speaker 1

这没什么价值。

That's not valuable.

Speaker 1

他在这个小图像识别数据集上取得了一些不错的成果,但人们并不太在意。

And he had some cool results on like this small image recognition dataset and people didn't really care.

Speaker 1

但伊利亚看到了这些成果,立刻就知道该怎么利用这些内核,对吧?

But Ilya saw that and he instantly knew what to do with these kernels, right?

Speaker 1

他意识到这正是一项即将突破的成果,因为当ImageNet这个大型数据集出现时,他觉得这是一个看似不可能实现的伟大挑战。

That he realized this is a breakthrough in the making because when ImageNet had come out, this big dataset, he had felt like this was this grand challenge that was just so impossible.

Speaker 1

如果你能解决它,那将意义非凡,但你只是需要投入足够的计算资源。

If you could solve it, it'd be so great, but you just need to be able to put enough compute into it.

Speaker 1

他看到了这些内核,它们能让我们非常高效地使用配备GPU的计算机。

And he sees these kernels that we're gonna use a computer with GPU very efficiently.

Speaker 1

他说,我们需要把这两件事结合起来。

And he said, we need to put these two things together.

Speaker 1

别把它用在你正在用的其他数据集上。

Don't apply it to this other dataset that you're using.

Speaker 1

图像数据才是关键。

Image data is the thing.

Speaker 1

杰夫·辛顿的贡献是一种管理技巧,因为亚历克斯·克里扎夫斯基非常讨厌写论文。

And then Jeff Hinton's contribution was a management trick because Alex Krzyavsky really hated writing papers.

Speaker 1

他有一篇综述论文即将提交。

He had a review paper coming up.

Speaker 1

杰夫告诉他:每周你在数据集上取得1%的改进,我就把你的综述论文截止日期推迟一周。

And Jeff told him, each week that you get a 1% improvement on the dataset, I will push back the deadline on your review paper by one week.

Speaker 1

他这样做了十几二十次,连续不断。

And he did this like a dozen times, two dozen times in a row.

Speaker 1

所以这就像亚历克斯一直坚持不懈地钻研这个问题,结果数字越来越好。

And so it was just one of these things where it just like Alex just kept grinding at the problem and the numbers got better and better.

Speaker 1

所以你会看到,这个领域的发展方式是,你需要正确的理论。

And so you see the way the progress in this field happens is you need the right theory.

Speaker 1

你需要正确的目标,正确的底层方法。

You need the right objective, the right sort of underlying approach.

Speaker 1

你需要正确的工程,对吧?

You need the right engineering, right?

Speaker 1

你需要真正地实现它。

You need to really implement it.

Speaker 1

你需要对这个问题全力以赴。

You need to push hard on the problem.

Speaker 1

即使感觉不可能,你也绝不能放弃。

You need to not give up even when it feels impossible.

Speaker 1

你还需要正确的精神,对吧?

And you need the right spirit, right?

Speaker 1

你需要知道这是值得的,并且即使面对一切困难,也要有继续前进的渴望。

You need to know that it's worthwhile and you need to have that desire to keep going even in the face of everything else.

Speaker 1

所以我认为,正是这三者共同作用,才在这一时刻解锁了这一特定成果。

And so I think those three things together were what unlocked this particular result in this particular moment.

Speaker 1

他们将成果提交到了竞赛中。

And they submitted to the competition.

Speaker 1

在计算机视觉领域,人们都在问:刚才是怎么回事?

In the computer vision field was like, what just happened?

Speaker 1

对吧?

Right?

Speaker 1

这个原本被认为不可能解决的问题,如今基本上已经被解决了。

That this impossible problem has basically now been solved.

Speaker 1

对此的反应如何?

What was the reaction to that?

Speaker 1

这在计算机视觉圈内堪称一场地震,对吧?

It was one of these things where within the computer vision community, it was seismic, right?

Speaker 1

我认为人们很快从认为神经网络是死路一条,做神经网络的人简直是在骗人,转变为只相信神经网络。

I think people very quickly went from saying neural nets are a total dead end, like you're kind of a fraud if you're doing neural nets, to only neural nets.

Speaker 0

但曾经有一段时间,如果你做神经网络,你就是个骗子。

But there was a time when you were a fraud if you were doing neural nets.

Speaker 0

没错。

Absolutely.

Speaker 0

是的。

Yes.

Speaker 0

直到突破出现。

Until the breakthrough.

Speaker 1

没错。

That's right.

Speaker 1

而且这里的背景历史也非常有趣。

And actually, the history here is also very fascinating.

Speaker 1

所以有一篇1995年的论文,你可以在维基百科上找到,它谈到了深度学习繁荣与萧条的历史。

So there's this paper you can find on Wikipedia somewhere from 1995 that talks about the history of the deep learning booms and busts.

Speaker 1

这实际上是在所有当前浪潮之前的事情。

So it's really before all the current waves.

Speaker 1

如果你读过它,就会发现里面所说的内容,和我们在OpenAI整个历史中常听到的话完全一致。

And if you read it, that the things they're saying in there are the exact same things people would say to us throughout the whole history of OpenAI.

Speaker 1

这些神经网络研究者,在1965年时就被说成毫无新意。

These neural net people, it's in 1965, they would say these neural net people have no new ideas.

Speaker 1

他们只是想造更大的计算机。

They just want to build bigger computers.

Speaker 1

这让你意识到,历史是由胜利者书写的,当时曾有一场有组织的运动极力反对神经网络方向,推崇符号系统,否定神经网络。

And so it makes you realize that history is written by the victors, that there was a very concerted campaign waged against this whole direction saying symbolic systems, yes, neural nets, no.

Speaker 1

而神经网络研究者们清楚自己想做什么。

And that the neural net people knew what they wanted to do.

Speaker 1

他们想要更大的计算机、更深的神经网络,而符号系统的人则与资助机构关系密切,直接把整个领域污名化,说这完全是胡扯。

They wanted bigger computers, deeper neural nets, and that the symbolic systems people got in very cozy with funding agencies and really just poisoned the well and said, this whole thing is kind of BS.

Speaker 1

因此,七十年代神经网络的发展被扼杀,正是因为人们指责它过度炒作。

And so that's what killed it for the seventies was that they claimed that it overhyped.

Speaker 1

当时有众多团队在研究,却声称没有任何成果。

There are all these groups working on it, that there's no results.

Speaker 1

他们钉上了最后一颗钉子,因为有一项研究证明单层神经网络无法解决某个特定问题。

That they put the nail in the coffin that there was this result that showed that a single layer on neural net couldn't solve a particular problem.

Speaker 1

因此,整个领域就此终结了。

So therefore the whole thing's dead.

Speaker 1

当然,神经网络研究者们说:只要让我们用多层的就行,我们知道该怎么做,但这一切都像是体制扼杀了它。

And of course the neural net people were like, but just let us go to not single layer, like we know what to do, but it was all like one of these things where the establishment killed it.

Speaker 1

他们逐渐集中权力,直接说不行。

They sort of centralized, they said no.

Speaker 1

最疯狂的是,八十年代,这篇论文说神经网络卷土重来的原因是计算资源的民主化,对吧?

And then the crazy thing is in the eighties, what this paper says is the reason neural nets came back was because of the democratization of compute, right?

Speaker 1

从以前只有教授掌控所有计算资源,突然间,所有博士生都拥有了自己的电脑。

It went from being that you have these professors who guard all the compute to suddenly all these PhD students have their own computer.

Speaker 1

所以教授再也无法禁止他们做神经网络研究了。

And so professor can't tell them they're not allowed to do neural nets.

Speaker 1

于是,人们又开始重新投入这项研究。

And so suddenly people are doing it again.

Speaker 1

所以对我来说,这个主题具有普遍性,这非常有趣。

And so to me, that is, it's so interesting that this theme is universal.

Speaker 1

它一直如此。

It's like always been true.

Speaker 1

这并不是过去十年才出现的新现象。

It's not a new thing over the past ten years.

Speaker 1

它已经存在了六七十年。

It's something that's been there for sixty, seventy years.

Speaker 0

那么,在采用之后发生了什么?

So what happened after the adoption?

Speaker 1

在AlexNet取得成果之后,计算机视觉领域之外的人仍然轻视它。

Well, so after the AlexNet result, that people outside the field of computer vision would still poo poo it.

Speaker 1

他们会说,哦,它在计算机视觉上有效,但神经网络和机器翻译毫无关系,对吧?

They would say, well, oh, it works for computer vision, but neural nets have nothing to do with machine translation, right?

Speaker 1

因为你们有固定大小的图像之类的东西。

Because you have these fixed size images and stuff like that.

Speaker 1

对于机器翻译来说,有这些可变窗口之类的东西。

And for machine translation, have these variable windows and things like that.

Speaker 1

2014年,出现了序列到序列模型。

'20 14, you have sequence to sequence.

Speaker 1

虽然没有像AlexNet那样带来巨大的飞跃,但你能明显看到,是的,你会继续推进它,而这将成为你唯一需要的东西。

Didn't get the same massive step function you did with AlexNet, but you could just see, yeah, you're going to push that and this is going be the only thing you need.

Speaker 1

结果是,各部门之间的壁垒被彻底打破了。

And what happened is you effectively had the walls between departments being torn down.

Speaker 1

这是一种美妙的统一:你原本以为计算机视觉、机器翻译、语音识别是完全不同的领域,但其实不然,你只是拥有AI。

And it's this beautiful unification of you thought that you had all these different domains of computer vision, machine translation, speech recognition, it's, nope, you just have AI.

Speaker 1

你只是拥有深度学习。

You just have deep learning.

Speaker 0

听起来,每当边缘群体能够汇聚在一起,就会产生更有趣的结果。

It sounds like any time that the fringe groups can come together, something much more interesting can happen.

Speaker 1

是的。

Yes.

Speaker 1

你听说过亚瑟·C·克拉克吗?

You familiar with I think it's Arthur C.

Speaker 1

克拉克的第一定律?

Clarke's first law?

Speaker 1

没有。

No.

Speaker 1

如果一位年长而德高望重的科学家告诉你某件事是可能的,那他几乎肯定是对的。

It's if an elderly but distinguished scientist tells you that something is possible, they're almost certainly right.

Speaker 1

但如果一位年长而德高望重的科学家告诉你某件事是不可能的,那他几乎肯定是错的。

But if an elderly but distinguished scientist tells you that something is impossible, they're almost certainly wrong.

Speaker 1

这太棒了。

That's great.

Speaker 1

我非常喜欢这句话。

I love that.

Speaker 0

OpenAI 是在你旧金山的客厅里创立的。

OpenAI started in your living room in San Francisco.

Speaker 0

描述一下那个客厅给我听。

Describe the living room to me.

Speaker 1

那是一个宽敞的开放式空间,我们有一张黑色木制桌子,形状是大椭圆形的。

It was a big open space, and we had a black wood table that was this big oval shape.

Speaker 1

还有一些沙发。

Had some couches.

Speaker 1

我用了一台大屏幕电视。

I did big screen TV.

Speaker 1

第一天,根本没有白板。

Day one, there was no whiteboard.

Speaker 1

我记得有两个研究员在争论某件事。

And I remember two researchers were debating something.

Speaker 1

他们转身想在白板上写点什么。

They turned to write something on the whiteboard.

Speaker 1

但那里根本没有白板。

There wasn't one.

Speaker 1

我当时就想,我可以弄块白板。

And I was like, I can get a whiteboard.

Speaker 1

所以我从第一天起就觉得我在创造价值。

And so I felt like I was adding value from day one.

Speaker 0

是的。

Yeah.

Speaker 0

第一天在房间里的人有哪些?

So who was in the room that first day?

Speaker 1

第一天在房间里的是萨姆·阿尔特曼、伊利亚·萨茨克弗、沃伊泰克·扎伦巴。

In the room that first day would have been Sam Altman, Ilya Sitzkever, Wojtek Zaremba was there.

Speaker 1

我想维姬·钟、帕姆·维加塔、约翰·舒尔曼、安德烈·卡帕西也当时在场。

I think Vicky Chung, Pam Vegata, John Schulman, Andre Karpathy probably would have been there at the time.

Speaker 1

他们中的一些人当时正在完成 elsewhere 的学业,比如博士论文。

Some of them were finishing up their work elsewhere, their PhDs.

Speaker 1

如果我漏掉了其他人,我向你们道歉。

I apologize if I'm forgetting anyone else.

Speaker 1

但那是创始团队,创始氛围是我们有一个伟大的目标:我们真的希望帮助构建通用人工智能,并让它成为对人类有益的力量。

But that was, you know, the founding team, the founding vibe was we had this great objective of we really wanted to help build AGI and have it be something that is a positive force for humanity.

Speaker 1

但我们并没有一个明确的理论来说明如何实现它。

And we did not have a thesis on how we would do it.

Speaker 1

所以我们就从这里开始。

And so that was where we started.

Speaker 0

当你在那个房间里时,我们处于人工智能革命的哪个阶段?

And at what stage of the AI revolution were we when you were in that room?

Speaker 0

当时哪些是已知的,哪些是未知的?

What was known, what was not known?

Speaker 1

那是2016年初。

So this was the very beginning of 2016.

Speaker 1

当时,

At this time,

Speaker 0

这并不算太久以前。

it's not- Not so long ago.

Speaker 1

没那么久以前。

Not so long ago.

Speaker 0

十年前。

Ten years ago.

Speaker 1

是的,十年,十年。

Yeah, ten years, ten years.

Speaker 1

时间过得真快。

Time flies.

Speaker 1

没错。

Yes.

Speaker 1

太不可思议了。

That's wild.

Speaker 1

在这个领域里,时间会压缩。

And time compresses in this field.

Speaker 1

太疯狂了。

It's crazy.

Speaker 1

到那时,我们已经经历了四年的深度学习革命。

We had gone through, at that point, four years of this deep learning revolution.

Speaker 1

因此,很明显,当时正处于一个早期阶段,成果唾手可得——你只要拿一块GPU、一个神经网络,对准一个新问题,它就能奏效,并带来惊人的结果。

And so one thing that was clear was that it was like there was this early phase where the fruit was just hanging on the ground because you could just take a GPU, take a neural net, you point it at a new problem and it's going to work and it's going to get you awesome results.

Speaker 1

因此,在很多方面,新架构的出现是基础研究的黄金时期。

And so new architectures was kind of the heyday of basic research in a lot of ways.

Speaker 1

单个研究人员能在几个月内提出一个新颖的想法,验证它,并发表一篇出色的论文。

So individual researchers being able to come up with a novel idea in a few months, prove it out, get an awesome paper.

Speaker 1

这在以前是前所未有的事。

It would be an unprecedented thing.

Speaker 1

这种成果甚至能定义一个领域。

Would kind of like define a field.

Speaker 1

所以我们当时正处在这个时刻。

So that was the moment that we were in.

Speaker 1

但那还不是大规模工程的时期。

It wasn't yet the moment of grand engineering.

Speaker 1

那时还不是大规模计算的阶段,因为你希望拥有尽可能多的计算资源,但你并不能通过增加更多GPU来获得显著提升,对吧?

It wasn't yet the moment of large scale compute because you wanted as much compute as you could get, but you couldn't really get more out of more GPUs, right?

Speaker 1

实际上,你只有一块GPU,而协调多块GPU协同工作是当时的挑战。

It really was that you had one GPU and orchestrating many of them together.

Speaker 1

我们还没有好的方法来从这种配置中获得理想的回报。

We didn't have good techniques for how to actually get good returns from that.

Speaker 1

我记得在最早期,我参与了首批支持研究人员的工程项目。

And I remember in the very early days working on the first engineering projects to support the researchers.

Speaker 1

我观察到两位研究人员与两位工程师合作的场景。

And I observed two researchers building with two engineers.

Speaker 1

通常的流程是,研究人员会说:这是我想要的系统,这是我的需求。

Way that it would go is that researchers would say, here's the system I want, here are my requirements.

Speaker 1

然后工程师们会去开发,几天后回来,把成果投射到我的电视上。

And then the engineers would go off and build something and come back a few days later and they would project it up on my TV.

Speaker 1

接着他们会逐行检查,花整整一个下午争论每一个代码行。

And then they would go line by line, spend a whole afternoon just debating every single line.

Speaker 1

我记得看着那一幕,心想这永远都不会结束。

I remember looking at that and thinking this is never going to end.

Speaker 1

太慢了。

It's so slow.

Speaker 0

花的时间太长了。

Takes too long.

Speaker 1

太久了。

Too long.

Speaker 1

行不通。

Not going to work.

Speaker 1

所以我最终转而参与了这个项目,并与研究员保持非常紧密的协作循环。

So instead I ended up working on the project and I would work in a very tight loop with the researcher.

Speaker 1

我会说,这里有五个想法。

And I would say, here are five ideas.

Speaker 1

她会说,这四个都不行。

She would say these four are bad.

Speaker 1

我会说,太好了。

And I'd say, great.

Speaker 1

这正是我想要的。

That's exactly what I wanted.

Speaker 1

所以我并不是试图强加自己的想法,而是真正努力去理解对方的视角、对方对世界的看法,并尝试思考:如何以各种不同的方式来转化它。

And so just really not trying to push my own ideas, but really trying to learn the other person's perspective, the other person's view on the world, and try to then say, okay, could translate it in all these different ways.

Speaker 1

试图厘清什么是真相,什么是现实。

And to try to just tease out what truth is, what reality is.

Speaker 0

通常来说,这些研究人员也是工程师吗?

Typically, are the researchers engineers as well or no?

Speaker 1

在这个领域,他们更接近工程师。

In this field, they are much closer to engineers.

Speaker 1

有一些人确实处于这个交叉点上。

And there are some people who are really at that intersection.

Speaker 1

例如,我们的首席科学家雅各布·帕乔基,他的一大特点就是双脚踏在两个世界中,他拥有深厚的理论理解。

For example, Jakob Pachoki, who's our chief scientist, one of the things that has really distinguished him is that he really has his foot in both worlds, that he has deep theoretical understanding.

展开剩余字幕(还有 480 条)
Speaker 1

他拥有优化领域的博士学位,但同时也非常擅长构建系统,并且已经多次实践过。

He has a PhD in optimization, but he also really knows how to build systems and has done it many times.

Speaker 1

因此,这是一种独特的技能组合。

And so that is a unique skill set.

Speaker 1

这非常有价值。

It's very valuable.

Speaker 1

所以我发现,工程师要想在这个领域创造价值,门槛相当高,因为这些研究人员都懂编程。

And so the thing that I found was that for engineers to add value in this field, you have a pretty high bar because these researchers, they all know how to code.

Speaker 1

他们能自己搭建东西。

They can build their own things.

Speaker 1

所以你必须比他们自己做得更好。

So you have to do better than they would on their own.

Speaker 1

这与你为医生构建系统的情况形成对比,对吧?

And that's a contrast if you're just building for doctors, say, right?

Speaker 1

大多数医生可能并不懂编程。

Most doctors probably don't know how to code.

Speaker 1

所以,为他们构建比他们自己能做的更好的东西,这个门槛就降低了。

So the bar to do better at building something for them than they could build on their own is relaxed.

Speaker 1

因此,我们关注的很多内容是如何确保真正赋能并推动研究人员前进。

And so that a lot of what we focused on is how do you make sure you're empowering and moving forward the researchers?

Speaker 0

那天在房间里的人,彼此熟悉程度如何?

How well did the handful of people in the room that day know each other?

Speaker 1

有一部分人曾经一起完成过博士课程。

So there was a subset of people who had, they'd all gone through PhD programs together.

Speaker 1

他们中有些人曾一起实习过。

Some of them had interned together.

Speaker 1

因此,有一群人彼此认识,还有一群人则是刚认识。

So there was a set of people who knew each other and there was a set of people who were newer to each other.

Speaker 1

但到那时,我们其实已经经历了一些奠基性的活动。

But at this point, we'd actually already gone through some formative events.

Speaker 1

所以从2015年下半年开始,我一直负责招聘,寻找这个领域中最优秀的人才。

So we had really throughout the back half of 2015, I'd been doing all the recruiting to find just who are the best people in the field.

Speaker 0

第一次会议是OpenAI的会议,还是一次演变成OpenAI的聚会?

Was the first meeting the meeting of OpenAI or was it a get together that turned into OpenAI?

Speaker 1

我认为真正促成这件事的第一次聚会是2015年7月的一次晚餐。

I'd say that the very first moment that was really the get together that set things in motion was a dinner in July 2015.

Speaker 1

那次晚餐有萨姆、伊利亚、埃隆,还有其他几位人士参加。

And that that was with Sam, that was with Ilya, that was with Elon, that was with a handful of others.

Speaker 1

当时的问题是,现在开始一个真正能实现通用人工智能的实验室,是不是太晚了?

And the question there is, is it too late to start a lab that can actually really get to AGI, right?

Speaker 0

为什么说太晚了?

Why would it be too late?

Speaker 1

当时感觉DeepMind已经占尽优势了,对吧?

Well, it felt like DeepMind kind of had it, right?

Speaker 1

DeepMind拥有谷歌提供的所有人才和计算资源,感觉通用人工智能似乎已经近在咫尺。

That DeepMind had all the talent that they had as part of Google, all the compute, that it felt like maybe AGI was very close.

Speaker 1

你真的能聚集一群顶尖人才,全力去实现这个目标吗?

And can you actually get together a group of great people and really go for this?

Speaker 0

你会说你是出于与谷歌的竞争才启动这个项目的吗?

Did you start it in competition with Google, would you say?

Speaker 1

我不觉得这是竞争,但我认为这是一种互补,对吧?

I don't think of it as competition, but I do think of it as complementary, right?

Speaker 1

我认为关于人工智能应该如何发展的观点,这是一个非常根本的立场:我认为每个人都有权参与人工智能的发展。

That I think that my view on how AI should go, and this is very foundational, is that I think that AI is something that everyone deserves to participate in.

Speaker 1

对我来说,我们即将构建出极其强大的系统,而这些系统如何帮助全人类共同进步,是可能发生的事情中最重要的事情。

And to me, it felt like we're going to build these incredibly powerful systems and that how they play out for humanity to uplift everyone is something that is the single most important thing that can happen.

Speaker 1

参与其中并帮助引导这一方向,确保它真正对每个人都有益,这才是我想做的事情。

And contributing to that and helping steer that in a direction that makes sure it actually is beneficial to everyone, like that's the thing that I want to do.

Speaker 1

所以对我来说,这并不是关于谁的基准测试表现更好来回较劲。

And so to me, it felt like it's not about the back and forth on who has the best benchmark.

Speaker 1

真正重要的是,我们如何构建系统,以及如何让整个社会与这些系统融合,对吧?

It's really on how do we build systems and overall society that integrates with those systems, right?

Speaker 1

这些事物将共同演化。

These things are going to co evolve together.

Speaker 1

这是一个比我们今天所处的世界更好的世界。

That is a much better world than the one that we have today.

Speaker 1

但这不是任何一个群体能够单独完成的。

And that isn't something that any one group can do on their own.

Speaker 0

你当时知道这会是一场多大的革命吗?

How big of a revolution did you know it was then?

Speaker 0

你当时能预见我们现在所处的境地吗?

Could you see where we are now then or no?

Speaker 1

那时我觉得,如果这件事要成功,就必须是这样的感觉。

Then it felt like if it was gonna work at all, it'd kinda have to feel like this.

Speaker 1

我认为我们还没有完成。

And I think we're not done.

Speaker 0

描述一下房间里每个人的性格、优势和劣势。

Describe the personalities and strength and weaknesses of every person in the room.

Speaker 1

我想说,萨姆是一位富有远见的人。

Well, I'd say that Sam, I think, is a visionary.

Speaker 1

我认为萨姆是一个鼓励人们畅想宏大愿景的人。

And I think that Sam is someone who gives permission to dream big ideas.

Speaker 1

而且我认为他非常关心他人。

And I think he also cares a lot.

Speaker 1

我认为他非常关心人们。

I think he cares a lot about people.

Speaker 1

我认为他总是对如何找到解决任何问题的方法持非常乐观的态度。

And I think that he is someone who is, like, always very optimistic about how we can find a way to configure any solution to a problem.

Speaker 1

因此,我认为他是一个让你觉得‘这根本不可能成功’的人。

And so I think that he is someone where you feel like, hey, this is never gonna work.

Speaker 1

但他总会找到解决方案。

He will find a solution.

Speaker 1

是的。

Yeah.

Speaker 1

但他始终是一个——

But he is always someone-

Speaker 0

他让你相信这个问题是可以解决的。

He opens your mind that it can be solved.

Speaker 1

没错。

That's right.

Speaker 0

而且也许这还不是最好的解决方案。

And that maybe it's not the best solution yet.

Speaker 1

那是

That's

Speaker 0

对。

right.

Speaker 0

但一旦你知道它能被解决,你就可以去寻找更好的方案。

But once you know it can be solved, then you can work on a better one.

Speaker 1

没错。

That's right.

Speaker 1

而且这并不是脱离现实的,对吧?

And it's not detached from reality, right?

Speaker 1

他就像一个与之紧密相关的角色,我认为他是一位极其出色的促进者,帮助研究人员并将这项技术整体引导到世界中。

It's like kind of connected to, like, I think he's been, he's like an excellent sort of facilitator for researchers and for this overall sort of shepherding of this technology into the world.

Speaker 1

伊利亚,我认为他是一位富有远见的人。

Ilya, again, I think is a visionary.

Speaker 1

我记得在最初那几天,他说他有一个想法,关于如何解决所谓的无监督学习问题,如何通过观察世界来实现。

I think he is someone who I remember in that very first couple of days, he said, I have this idea I've been thinking about for how we can solve what was called unsupervised learning, how we can observe the world.

Speaker 0

那是什么?

What is that?

Speaker 0

所以如果

So if

Speaker 1

你可以想想人类婴儿是如何仅仅通过观察世界来学习的,对吧?没有人告诉他们什么是对的。

you think about how a human baby learns just by observing the world, right, there's no one saying this is the right thing.

Speaker 1

没有人提供输入。

No one's doing input.

Speaker 1

没错。

Exactly.

Speaker 1

它就这样发生了。

It just happens.

Speaker 1

它就这样发生了。

It just happens.

Speaker 1

对我来说,这一直是个疯狂的概念:机器如何能在没有人告诉它做得好坏的情况下学会呢?

This was alwaysto me, this always felt like a crazy concept of how can the machine ever learn without someone telling it whether it's doing a good job or not.

Speaker 1

是的。

Yeah.

Speaker 1

但我们找到了方法。

But we figured it out.

Speaker 1

我记得他有很多想法,关于如何真正将其融入机器中。

I remember that he had a lot of ideas on how to really push it into machine.

Speaker 0

即使在那个房间里,也会有人觉得这太超前了吗?

Even in that room, would there be people who think that's too far?

Speaker 1

在那个房间里,我们立刻开始写下各种想法,那种能量是实实在在的。

So in that room, we immediately started trying to write down ideas and that the energy was just palpable.

Speaker 1

在大家齐聚这个房间之前,其中一个关键步骤是2015年11月的那次外部会议。

One of the steps along the way to everyone coming to that room was this off-site in November 2015.

Speaker 1

于是我梳理了整个领域以及所有顶尖人才,并不断询问别人:你认识哪些特别出色的人?

So I mapped out the whole field or all the best people and kind of been asking people for who do you know who's great?

Speaker 1

然后他们就会把我介绍给那些人。

And then they would introduce me to people.

Speaker 1

所以我一直留意着,不断有人向我推荐一个叫沃伊切赫的人。

And so I just kind of kept track of I kept getting introduced to this guy called Wojciech.

Speaker 1

我当时就想,好吧,沃伊切赫应该是我必须争取的人。

I was like, all right, Wojciech's probably someone I should go after.

Speaker 1

我们已经缩小了目标人选范围,但他们都在问:还有谁会加入?

And we had narrowed down to a set of people, but they were all kind of like, okay, well, who else is joining?

Speaker 1

他们说:我很感兴趣,但还有谁在?

Like, I'm interested, but who else is in?

Speaker 1

那你该怎么促成这件事呢?

And you're like, how do I collapse this?

Speaker 1

我问了萨姆该怎么办。

And I asked Sam what to do.

Speaker 1

他建议把大家召集到一次场外会议。

And he suggested to bring everyone to an off-site.

Speaker 1

而这时,我真的很感谢约翰·舒尔曼,因为他表示他会参加。

And at this point, actually, remember it was very grateful to John Schulman who had said that he would be in.

Speaker 1

所以至少我不是唯一一个已经承诺的人。

So I was at least not the only one who had committed.

Speaker 1

我想可能还有一两个人也在,但当时确实还是一群尚未凝聚成团队的人。

And I think maybe there was one or two others who were kind of there, but it really was a group of people who had not yet coalesced into a team.

Speaker 1

是的。

Yeah.

Speaker 1

我们把大家都召集了出来。

And we brought everyone out.

Speaker 1

我们当时在我家公寓里。

We were in my apartment.

Speaker 1

我们上了巴士。

We got into the bus.

Speaker 1

我们开车去了纳帕。

We drive up to Napa.

Speaker 1

就是那一天,所有人都默契十足,对吧?

And it was just this day where everyone clicked, right?

Speaker 1

那种能量简直太顺畅了,对吧?

That the energy was just so smooth, right?

Speaker 1

那就是人类形态下的心流状态。

It's that flow state in human form.

Speaker 1

我记得我们在白板上写下了这个计划,我有一张这张白板的照片,上面是我们的计划。

And I remember that we wrote up on this flip chart, and I have a picture of this flip chart, the plan.

Speaker 1

这是一个三步计划。

There's a three step plan.

Speaker 1

第一步是解决强化学习,第二步是解决无监督学习,第三步是逐步学习更复杂的东西。

Step one was solve RL, Reinforcement Learning two: Solve UL three: Gradually learn more complicated things.

Speaker 1

这实际上就是我们过去十年一直在做的事情。

And this actually is what we've been doing for a decade.

Speaker 1

这简直太疯狂了,对吧?

It's actually crazy, right?

Speaker 1

我们真的确立了这个愿景。

Like we really set out the vision.

Speaker 0

所以这与最初的愿景非常贴近。

So It's very tight to the original vision.

Speaker 0

这只是最初愿景的延伸和发展。

It's just the growth of the original vision.

Speaker 1

确实如此。

It is.

Speaker 1

真的就是这样。

It really is.

Speaker 1

当我回顾我们所做的一切时,发现所有工作都是为了实现同一个目标,采用了几乎相同的技术路径和底层理念。

When I look at everything we have done, it has been in service of the same goal with really the same almost technical approach and the same ethos underlying it.

Speaker 0

房间里还有其他人具有特别的专业知识、观点,或者与其他人不同,值得描述一下吗?

Any of the other people in the room who had particular either expertise or views or that were different than the others that are worth describing?

Speaker 1

是的。

Yeah.

Speaker 1

我会说 Wojciech,我至今仍与他密切合作,他也是一个独特的人物。

I'd say Wojciech, and I still work very closely with him, is also a unique character.

Speaker 1

他在创意生成方面非常出色,能为任何问题提出极具创造力的想法。

He is extremely good at idea generation, and he will come up with very creative ideas to any problem.

Speaker 1

而且他完全不执着于自己的想法。

And then he is also someone who is not at all attached to his own ideas.

Speaker 1

对。

Yeah.

Speaker 1

对吧?

Right?

Speaker 1

因为如果你容易生气或敏感,就很难保持如此开放的创造力。

Because if you're someone who's going to take offense, then it's really hard to be that generative.

Speaker 0

当然,你希望它尽可能好,但不可能所有想法都是对的。

Well, you want it to be as good as it can be, That's they can't all be your right.

Speaker 1

我通常在和他们合作时,会认真思考:我们该考虑的边界在哪里?

What I usually do when working with them is really think about, okay, well, what are the bounds on what we should think about, right?

Speaker 1

这些是我们不希望涉足的领域,或者这是我们最终需要达成的目标。

Here are the places that we don't want to go or here's kind of the place that we need to end up.

Speaker 1

然后,创意生成的过程就会形成一个强大的良性循环。

And then the idea generation process just ends up with this great flywheel.

Speaker 0

你是怎么加入Stripe的?

How did you end up at Stripe?

Speaker 1

当我还在麻省理工学院时,我一直在创业。

So while I was at MIT, I was building more.

Speaker 1

我做了更多的初创公司。

And I did more startups.

Speaker 1

从每一次经历中,我都学到了另一件不该做的事。

And from each one, I felt like I learned another thing not to do.

Speaker 1

最终,我感觉自己已经足够了解如何成功创办一家初创公司,但我缺少一个关键要素——一个好点子。

And eventually, I kind of felt like I knew enough that I could be successful at doing a startup, but I was missing a key component, which is having an idea.

Speaker 1

我一直在观察我的朋友们,他们来自计算机俱乐部,自己创办了初创公司,他们读了硕士,在那里想出了一个很棒的点子。

And I was pattern matching off of my friends who had been in the Computer Club, started their own startup, they'd gone to a master's program, come up with a cool idea there.

Speaker 1

我当时想,显然这就是你该做的事。

I was like, well, clearly that's what you need to do.

Speaker 1

你需要去读研究生。

You need to go to grad school.

Speaker 1

你看到了这条路径。

You saw the path.

Speaker 1

是的。

Yes.

Speaker 1

但我看到的这条路径实际上太过常规了,对吧?

But the path that I saw was actually too much of the beaten path, right?

Speaker 1

常规路径就是:在你的博士项目中发明一些东西,然后把这项技术转化为一家初创公司。

The beaten path is in your PhD program, you invent something there and you turn that technology into a startup.

Speaker 1

我记得当时就觉得,好吧,我才21岁。

And I remember just feeling like, okay, I'm 21.

Speaker 1

我太年轻了,做不了世界上真正重要的事。

I'm just too young to do real things in the world.

Speaker 1

这条路是唯一可能的选择。

This path is the only thing that's possible.

Speaker 1

但至少,我可以开始认识那些在做创业的人,因为我上过一门创业课,但根本没用。

But at the very least, I can start meeting people doing startups because I had taken a startup class and it was just not useful.

Speaker 1

我想,好吧,这根本无法让我达到我想去的地方。

I was like, all right, like, this is not going to get me to where I want to go.

Speaker 1

于是我决定去认识那些在做创业的人。

And so I decided I'd meet these people doing startups.

Speaker 1

就在第二天,我收到了一封来自帕洛阿尔托一家支付创业公司人员的邮件。

And literally the next day, I got an email from some people working on a payment startup in Palo Alto.

Speaker 1

我当时就想,好吧,我的新目标就是去认识这些人,慢慢学会识别模式。

And I was like, well, my new thing is to meet these people and learn to pattern match over time.

Speaker 1

我记得当我见到帕特里克时,我们一见如故。

And I remember when I met Patrick, we just clicked.

Speaker 0

他的背景是什么?

What was his background?

Speaker 1

嗯,他之前在麻省理工学院。

Well, so he had been at MIT.

Speaker 1

所以他和他哥哥——是

And so he and his brother- Did

Speaker 0

你是在麻省理工认识他的吗?不是,

you know him from MIT at No,

Speaker 1

我没有。

I did not.

Speaker 1

是的,但我们有共同的朋友,因为他上过麻省理工,约翰上过哈佛。

Yeah, but we had mutual friends because he had gone to MIT, John had gone to Harvard.

Speaker 1

所以是那个圈子里的人在互相打听。

So it was that community that they were asking around.

Speaker 1

当然,我的名字在两个圈子里都被提及了。

And my name, of course, came up in both circles.

Speaker 1

我记得我飞过去的时候,正在下雨,天气特别糟糕。

And I remember I flew out, and it was raining and kind of miserable.

Speaker 1

我记得开门后,我们一刚开始聊天,就立刻产生了默契。

And I remember opening the door and just like, when we first started talking, it was just an instant connection.

Speaker 1

对吧?

Right?

Speaker 1

我觉得我们的背景在很多方面其实很不一样。

I think we just, like, had this, you know, different backgrounds in many ways.

Speaker 1

对吧?

Right?

Speaker 1

但我们也拥有非常相似的技术视角。

But also, we had a very similar technical perspective.

Speaker 1

就连一些特别极客的事情,比如我们用的同款Kinesis分体键盘,都一模一样。

And like even for very nerdy things, I mean, had the same like split keyboard that we use, this Kinesis keyboard.

Speaker 1

我们俩都用Dvorak布局。

We both were Dvorak.

Speaker 1

我们实际使用的布局是不同的。

The actual layout that we used was different.

Speaker 1

我们讨论了如何为他们系统的防火墙做设计,还聊到了内核里的各种东西。

And we were talking about how to like build the firewalls for the systems that they had and you're talking about, you know, different things in the kernel.

Speaker 1

所以我们就像是在技术上产生了共鸣。

And so it was just like we had this, like, technical connection.

Speaker 1

我觉得我特别欣赏帕特里克和约翰的一点是,他们和我同龄,却已经在外头干出名堂了,对吧?

And I think that what I also really appreciated about Patrick and John is that they were my age and they were already out there doing things, right?

Speaker 1

他们已经创办了初创公司。

That they were already doing a startup.

Speaker 1

我当时觉得,这不可能吧。

And I was like, I don't think that's possible.

Speaker 1

我们俩之中,肯定有一个人搞错了。

One of the two of us is wrong.

Speaker 1

我真的很想知道他是谁。

I really want to know who it is.

Speaker 0

是的。

Yeah.

Speaker 0

然后发生了什么?

And then what happened?

Speaker 0

所以

So

Speaker 1

周末过得很棒。

weekend was great.

Speaker 1

他们说你应该加入。

They said you should join.

Speaker 1

我说,让我想想。

I said, let me go think about it.

Speaker 1

我回到了学校。

I went back to school.

Speaker 1

约翰周四恰好在城里。

John happened to be in town on Thursday.

Speaker 1

所以我就想,你知道的,我大概该做决定了。

And so I was like, you know, I I probably should make my decision.

Speaker 1

然后我想,你知道吗?

And I was like, you know what?

Speaker 1

我想做这件事。

I want to do this.

Speaker 1

因为再说一遍,回到那个关于‘如果这能成功’的梦境算法。

Because again, back to that algorithm of dream of the what if this works.

Speaker 1

我觉得我对支付一无所知。

I'm like, I don't know anything about payments.

Speaker 1

这不是我从小热衷的问题。

Like, this is not the problem that I've grown up passionate about.

Speaker 1

但这些人但

But these people But

Speaker 0

你能解决一个真正的问题。

you get to solve a real problem.

Speaker 0

是的,解决一个

Yeah, solve a

Speaker 1

真正的问题。

real problem.

Speaker 1

没错。

Exactly.

Speaker 1

这些是我觉得可以向他们学习或与之合作的人。

And these are the people that I feel like I can learn from or that I can work with.

Speaker 1

我记得约翰问过我:‘你愿意做吗?’

And I remember John was like, will you do it?

Speaker 1

我当时说:‘好吧。’

I was like, alright.

Speaker 1

行吧。

Fine.

Speaker 1

我会做的。

I'll do it.

Speaker 1

我只记得当时心想,好吧。

And I I just remember feeling like, okay.

Speaker 1

是的。

Yeah.

Speaker 1

就像他们需要我一样。

Like, that They want me.

Speaker 1

有这么一个人。

There's someone exactly.

Speaker 1

他们需要我。

They want me.

Speaker 0

这种感觉真好。

That's a good feeling.

Speaker 1

那是一种非常好的感觉。

It was a really good feeling.

Speaker 1

是的。

Yeah.

Speaker 1

所以我决定在星期四这么做。

And so I decided that on Thursday.

Speaker 1

星期五,我花了一整天时间告诉老师们我不去了。

Friday, I spent telling my teachers that I was out.

Speaker 0

那进展得怎么样?

How did that go?

Speaker 1

有点困难。

It was a little tough.

Speaker 0

对你来说有情感上的冲击吗?

Was it emotional for you?

Speaker 1

这 definitely 感觉像某种事情的终结。

It was definitely it felt like the end of something.

Speaker 0

而且

And

Speaker 1

你知道吗,当我告诉哈佛我要离开时,他们说

you know, Harvard, when I told them I was leaving, they said

Speaker 0

你会回来的

You're coming

Speaker 1

回来的。

back.

Speaker 1

你会回来的。

You're coming back.

Speaker 1

没错。

Exactly.

Speaker 1

我想这大概是为了他们的数据统计之类的吧。

And I think it's probably for their numbers and, you know, that kind of thing.

Speaker 1

但 definitely 感觉到,好吧,我明白这背后的逻辑了。

But it definitely felt like, okay, I see how this goes.

Speaker 1

麻省理工则稍微不一样,他们跟相关的人谈了,说你每六个月得报到一次,因为如果太久没联系,可能就得重新申请,我们也不太确定。

MIT was a little bit more like, Okay, like talking to the relevant person that they said that, you know, you have to check-in every six months or so because if too long goes by, then, you know, maybe you'll have to reapply and we're not really sure.

Speaker 1

而且那种氛围完全不同。

And it just was a very different vibe.

Speaker 1

我记得和教授们交谈过,他们都很支持我。

And I remember talking to the professors, the professors were supportive.

Speaker 1

我想这是他们经常看到的情况,但他们也确实有点难过,因为我学期中就离开了。

And I think this is a thing that they see and but they were also, you know, I think a little sad to see me go like mid semester.

Speaker 1

是的,我非常感激他们提供的指导。

Yeah, I really appreciated the mentorship they provided.

Speaker 1

当时我正在上操作系统课程,这是MIT一门著名的课程,由一群非常优秀的教授授课。

Like, was in the middle of the operating systems course, which is this famous course at MIT taught by these, like, extremely good professors.

Speaker 1

这门课实在太棒了,我非常遗憾没能继续完成更多的项目。

And it was just such a cool class, and I was very sad to not get to implement further projects.

Speaker 1

所以这确实意味着放弃了一些宝贵的经历。

And so it was definitely kind of giving up on some sort of exposure to an experience.

Speaker 1

而且和这些人告别也很艰难,对吧?

And it was also hard to say goodbye to the people, right?

Speaker 1

那里有这么多我原本去学习和合作的人。

There were all these people that I'd gone there to learn from and work with.

Speaker 1

有趣的是,他们中的许多人后来也来到了硅谷,我在Stripe有机会和他们共事,也以其他方式相处了很长时间。

And the funny thing is many of them ended up coming out to the valley, got to work with them at Stripe, got to spend a bunch of time with them in other ways.

Speaker 1

所以,这实际上并没有像我当时以为的那样,是一次真正的告别。

So it was much less of a goodbye than I thought it was at the time.

Speaker 1

是的。

Yeah.

Speaker 0

在你长大的地方成长,并且如此迅速地搬迁,你觉得自己不一样吗?

Growing up where you grew up and moving as quickly as you did, did you feel different?

Speaker 1

我确实觉得自己不一样。

I definitely felt different.

Speaker 0

对。

Yes.

Speaker 0

你感到孤独吗?

Was it lonely?

Speaker 0

你觉得自己是个局外人吗?

Did you feel like an outsider?

Speaker 1

确实感觉很不一样。

Like, definitely felt different.

Speaker 1

我就是完全融入不进去。

I definitely didn't fit in.

Speaker 1

对吧?

Right?

Speaker 1

还有很多其他孩子喜欢的东西,我完全不懂。

And there was like a lot of things that other kids were into that I just didn't understand.

Speaker 1

我记得上幼儿园时坐校车,其他孩子都在跟着收音机唱歌。

Like, I remember being on the school bus in kindergarten and other kids were like singing along to the radio.

Speaker 1

但我一句歌词都不会。

And I didn't know any of the words.

Speaker 1

我就觉得,根本不知道该怎么跨越这个鸿沟。

And I just felt like I don't even know how to bridge this gap.

Speaker 1

所以我有过很多这样的时刻,总觉得我身上有什么地方就是不太一样。

And so I had a number of moments like that where I just felt like there was just something about me that, like, doesn't quite match.

Speaker 1

有些是因为活动,比如很多孩子都去打猎,而我家根本从不这么做。

Some of it was about activities like a lot of the kids would hunt, and that was not something that my family did at all.

Speaker 1

所以就是有一种非常不同的感觉。

And so there was just something very different.

Speaker 1

我曾经短暂地打过冰球。

I did play hockey for a little bit.

Speaker 1

我打得不太好,但我当守门员。

I was not very good, but I was goalie.

Speaker 1

于是我开始尝试做一些能让我更融入的活动。

So I started to try to, you know, do some activities that would match.

Speaker 1

但我记得,我特别在意的一件事是为自己塑造一个身份,因为如果你与众不同,又没有明确的身份,就会很孤独。

But I remember I was one thing I really cared about was carving an identity for myself Because if you're different and you don't really feel like you have an identity, then it's lonely.

Speaker 1

但如果你与众不同,同时拥有一个定义你的身份,那你就是在走自己的路。

But if you're different and you have an identity, something that defined you, you're charting your own path.

Speaker 1

对我来说,就是当一个聪明的孩子。

And for me, it was being the smart kid.

Speaker 1

我记得在小学时,每周都有一次拼写测验。

Like I remember in elementary school that we had a weekly spelling quiz.

Speaker 1

流程是这样的:周一,老师会发给你本周的10个或20个单词。

And the way it would work is on Monday, the teacher would give you the 10 or 20 words for the week.

Speaker 1

到了周末,老师就会考这些单词。

And then at the end of the week, you'd be tested on them.

Speaker 1

周一的时候,老师念单词,你得把它们写下来,这算是一次预考。

And you'd kind of have this pretest on Monday where you'd have to write them down as the teacher says them.

Speaker 1

如果你写错了,就得去问其他写对的同学该怎么拼。

And if you got it wrong, you'd have to go ask one of the other kids who got it right how to spell it.

Speaker 1

通常我都能全对。

And normally I'd get them all right.

Speaker 1

但我记得有一周,我有一个单词拼错了。

But I remember one week I got one of the words wrong.

Speaker 1

我记得班上另一个聪明的孩子,他答对了。

I remember another one of the smart kids in the class, he got it right.

Speaker 1

我感到非常羞愧,不得不去问他正确答案,因为这会侵蚀我一直以来的核心身份。

And I was so ashamed that I would have to go and ask him for the right answer because that would be just eroding this core identity that I had.

Speaker 0

哇。

Wow.

Speaker 0

不过听起来这其实挺健康的,你经历了这些,从而不再需要事事都懂。

That sounds like it was really healthy, though, that you got to do that and it set you up to not have to have all the answers.

Speaker 1

是的。

Yes.

Speaker 1

对。

Yes.

Speaker 1

我觉得这件事未必是坏事。

I think it was not necessarily a bad thing that happened.

Speaker 0

这是个好故事。

It was a good story.

Speaker 1

没错。

That's right.

Speaker 0

所以当你遇到像帕特里克这样的人时,是不是有种感觉:‘他和我一样’?

So then when you got to meet someone like Patrick, was it a feeling of, Oh, he's like me?

Speaker 0

那一定是一种很棒的感觉。

Must have been a great feeling.

Speaker 1

确实如此。

It really was.

Speaker 1

确实如此。

It really was.

Speaker 1

而且他之前创办过一家初创公司,所以他和约翰以前就

And he had done a startup before, so he and John had previously

Speaker 0

他当时才21岁,就已经创办过一家公司了。

And he was 21, and he had already done a startup.

Speaker 0

没错。

Exactly.

Speaker 0

是的。

Yes.

Speaker 0

那这件事是怎么发生的?

How did that happen?

Speaker 1

所以他们创办了一家初创公司,哦,天哪,我好长时间没想过了。

So they did a startup called oh, man, I haven't thought about this for a while.

Speaker 1

我不记得所有细节了,但我想是这样的。

I'm not gonna remember all the details, but I think okay.

Speaker 1

帕特里克和约翰在爱尔兰长大,我认为帕特里克通过LIFPS社群认识了保罗·格雷厄姆,他是创业加速器Y Combinator的创始人,而LIFPS是他们非常热衷的一种编程语言。

So Patrick and John grew up in Ireland, and I think that Patrick had met Paul Graham, who runs Y Combinator, this startup incubator, through the LIFPS community, so through the programming language that they were very into.

Speaker 1

我认为这就是他后来加入Y Combinator、创办初创公司的契机。

And I think that that was his connection to then doing YC, doing a startup.

Speaker 1

这家初创公司被一家名为Live Current Media的公司收购了。

That startup got sold to a company called Live Current Media.

Speaker 1

他们为收购方工作了一段时间,但显然这并不是他们想用一生去做的事情。

They worked for the acquirer for some time, but clearly, was not the thing that they wanted to do with their lives.

Speaker 0

嗯嗯

Mhmm.

Speaker 0

你加入Stripe的时候有多早?

How early were you in Stripe?

Speaker 1

在Stripe的早期阶段,我们一共四个人。

So early days of Stripe, there were four of us.

Speaker 1

有约翰,有帕特里克,有达拉·巴克利,还有

There was John, there was Patrick, there was Dara Buckley, and there was

Speaker 0

我。

me.

Speaker 0

嗯嗯

Mhmm.

Speaker 1

我刚去的时候,已经有了一些基础设施和一个支付处理器。

And when I first was there, there was some infrastructure and there was a payment processor.

Speaker 1

当时我们也不太清楚接下来该怎么做,其中一个想法是:如果我们开发一些应用,然后用我们正在构建的支付系统来支持这些应用呢?

It also wasn't clear exactly how we were going to proceed because there was one idea was, well, what if we build some apps and then use this payment processing we're building to power those apps?

Speaker 1

这就是你实际构建东西的方式。

And so that's how you actually build something.

Speaker 1

最终,支付处理器可能会成为核心,但这些应用也可能成为方向。

And eventually the payment processor will probably be the thing, but these apps could be too.

Speaker 1

当时帕特里克正在开发一个时间追踪工具,这也是其中一个潜在的想法。

And so there was a time tracking thing that Patrick had been working on that was one of the potential ideas.

Speaker 1

所以它仍处于初期阶段,但我们已经能看到它的发展方向。

So it's still in this nascent form, but still something that we could see, like the direction of travel.

Speaker 1

有趣的是,我之前提到我在麻省理工参加过一个创业课程,课程要求你模拟创办一家初创公司。

And the funny thing, by the way, is that I mentioned that I did this startup class in MIT and that as part of that, you're supposed to build like a mock startup.

Speaker 1

而我最终创建的正是一个支付处理器。

And the one that I ended up building was a payment processor.

Speaker 1

所以我花了大量时间研究如何在线实现支付处理,结果发现这简直糟透了。

And so I spent all this time trying to figure out how do you do payment processing online and realizing it's horrible.

Speaker 1

这太痛苦了。

It's painful.

Speaker 1

这就像试图去 PayPal 并阅读他们的文档,想办法注册某个服务。

It's like trying to go to PayPal and trying to like read their documentation, trying to figure out how to sign up for something.

Speaker 1

一切都如此晦涩难懂,你不禁会想:怎么可能这么糟糕?

It was just so opaque and you just realize how can it possibly be so bad?

Speaker 1

是的。

Yeah.

Speaker 1

而这就是 Stripe 的根本突破:我们意识到支付处理完全可以做得更好。

And that was the fundamental unlock for Stripe was this realization that you can do payments processing better.

Speaker 1

我记得最初的网站上写着:支付处理没必要这么糟糕。

I remember that the initial website said payment processing doesn't need to suck.

Speaker 1

对吧?

Right?

Speaker 1

这正是我们的理念:根本没必要这样糟。

And it was just like, like, that is the ethos of just like, it just doesn't need to.

Speaker 1

它可以变得更好。

It can be better.

Speaker 1

我真的很喜欢这种精神。

And I really like that spirit.

Speaker 1

我记得曾经和一位风险投资人谈过。

I remember talking to a VC at some point.

Speaker 1

那是上线前的事,但我们已经获得了不少关注。

This was prelaunch, but we had gotten a lot of buzz.

Speaker 1

他说:好吧,你们的核心优势是什么?

And he was saying, okay, what is your secret sauce?

Speaker 1

我当时说:我们只是做得更好。

And I was like, well, we just do it better.

Speaker 1

他回应:不不,别开玩笑了。

And he's like, no, no, come on.

Speaker 1

我知道你对每个人都这么说,但她问的是,真正的核心优势到底是什么?

I know that's what you say to everyone, but she's like, what's the actual secret sauce?

Speaker 1

我说:我真的不知道该怎么跟你说。

And I'm like, I don't know what to tell you.

Speaker 1

我们就是专注于每一个细节,把它们做到位。

Like, that is what we do, is just focus on every single detail and get it right.

Speaker 0

所以这其实并不是在创造什么新东西。

So it really wasn't creating something new.

Speaker 0

它本来就是PayPal。

It was already PayPal.

Speaker 0

只是更好的PayPal。

It was just a much better PayPal.

Speaker 1

没错。

Exactly.

Speaker 0

是这样吗?

Is that right?

Speaker 1

确实是。

It was.

Speaker 1

而且我们真的专注于整个体验的每一个细节,从头到尾。

And it was really focusing on the details of the whole experience end to end.

Speaker 1

对吧?

Right?

Speaker 1

而且真的要思考,因为如果你自己经历过,就会意识到,比如,这个部分,你该怎么注册账户?

And really thinking about just because if you've gone through it yourself and you realize, like, this part, like, how do you sign up for an account?

Speaker 1

你该怎么实际连接到API,连接到计算机之间通信的方式?

How do you actually even connect to the APIs to the actual way the computers talk to each other?

Speaker 1

你需要针对哪些不同的参数进行编程?

What are the different parameters you program against?

Speaker 1

你该怎么决定使用哪种编程语言或哪种封装库?

How do you figure out which programming language to use or which wrapper to use?

Speaker 1

所有这些事情。

All these things.

Speaker 1

每一件事都可能带来巨大的阻力。

Each one of those can add a ton of friction.

Speaker 1

所以如果你只是专注于做好这一项,做好那一项,做好另一项,为自己做好,为朋友做好。

And so if you just focus on, okay, I'm gonna make this one good, this one good, this one good, make it good for myself, make it good for my friends.

Speaker 1

实际上,这对每个人都会是一件很棒的好事。

Actually, this will be something nice and good for everyone.

Speaker 1

那你为什么离开Stripe呢?

And why did you leave Stripe?

Speaker 1

嗯,我在Stripe待了将近五年,四年半左右。

Well, I had been at Stripe for almost five years, four and a half years.

Speaker 1

大约在四年左右的时候,我开始思考这是否是我想要长期从事的事业。

And about four years in, I think I started to consider whether or not this was what I wanted to do for the long term.

Speaker 1

我看待事情的方式是,假设你的职业生涯以五年为一个阶段。

And the way that I kind of viewed things is like, okay, if you think of your career in five year chunks.

Speaker 1

五年是一个合适的长度,因为少于五年的话,很难做出显著的成就。

And five years is about the right length because less than that is kind of hard to do something significant.

Speaker 1

我记得当时觉得,我已经把这家公司带到了一个即使没有我也会成功的地步。

And I remember feeling like, okay, I'd gotten this company to a place where it was going to succeed with or without me.

Speaker 1

那么问题就来了:我是继续留下,还是去尝试一些新的东西?

And then the question was, do I continue or do I go and do something new?

Speaker 1

我对创业这个想法感到非常兴奋。

I was very excited about doing the idea of a startup.

Speaker 1

但我记得和帕特里克谈过,他似乎有一些非常有说服力的理由让我留下。

But I remember talking to Patrick, I think he had some very convincing reasons to to stay.

Speaker 1

他提到的一点是,要组建一个能够对世界产生重大影响的团队非常困难。

That one point he made is that it's very hard to assemble a group of people that can do significant things in the world.

Speaker 1

最理想的情况是,你去创办一家初创公司并取得一些成功。

And best case scenario, you go start some startup and have some success with it.

Speaker 1

五年后,你才可能建立起一个能够做事的团队,但你在这里已经拥有它了,对吧?

And then five years later, you'll have formed that group of people that can do stuff, but you already have it here, right?

Speaker 1

我们已经拥有一个能够高效完成任务的紧密团队。

We already had this tight group that was able to accomplish things.

Speaker 1

那为什么还要离开呢?

And so why walk away from that?

Speaker 1

所以这非常艰难。

So it was very tough.

Speaker 1

这并不容易。

It was not easy.

Speaker 1

我记得当我告诉帕特里克我要离开时,我哭了。

I remember I cried when I told Patrick that I was out.

Speaker 1

是的,这真的非常艰难。

Like, yeah, it was very, very tough.

Speaker 1

他和约翰为我举办了一场果汁派对。

He and John threw a juice party for me.

Speaker 1

这是一次非常温馨的送别。

It was a very, very nice send off.

Speaker 1

我回忆起当时思考这个问题时的感觉:如果真的很难组建起一群能成就大事的人,那我必须现在就开始行动。

And the reason that I decided, I remember feeling like, as I thought that through, I was like, well, if it really is so hard to build that group of people who can accomplish significant things, I got to get started now.

Speaker 0

对。

Yeah.

Speaker 0

听起来你离开是因为Stripe的使命并不是你希望专注的使命。

It sounds like you left because the mission of Stripe wasn't the mission that you wanted to focus on.

Speaker 1

这是真的。

That is true.

Speaker 1

是的,这是一个美好的使命,我非常支持。

Yeah, it's a beautiful mission and it's one I very much support.

Speaker 1

但要说清楚的是,这是否是我必须以任何形式都去追求的使命呢?

But it's different to say, is this one where I will just need to pursue it in any form?

Speaker 0

你离开的时候,知道自己接下来要做什么吗?还是只是打算边走边看?

And did you leave knowing what was going to be next or did you leave thinking I'm going to figure it out?

Speaker 1

更偏向于后者。

More the latter.

Speaker 1

我列出了三个可能专注的领域。

I had a list of three different areas I might focus on.

Speaker 1

第一个是人工智能。

Number one was AI.

Speaker 1

第二个是虚拟现实和增强现实。

Number two was VRAR.

Speaker 1

第三个是编程教育。

Number three was programming education.

Speaker 1

对我来说,如果我能以某种方式为人工智能做出贡献,那我就去做。

And for me, it was very clear if I can contribute to AI in some way, okay, I'm doing that.

Speaker 1

但我不确定自己是否具备这些技能。

But it wasn't clear to me, do I have the skills?

Speaker 1

现在是合适的时机吗?

Is it the time?

Speaker 1

所有这些问题。

All those things.

Speaker 1

所以另外两个只是备选方案。

And so the other two were kind of backup options.

Speaker 0

我知道你曾经写过一本化学教科书。

I understand there was a point in time that you wrote a chemistry textbook.

Speaker 0

那是真的。

That is true.

Speaker 0

那是什么时候?

When was that?

Speaker 1

那也是2008年。

That was also 2008.

Speaker 1

所以高中毕业后,我休学了一年。

So after high school, I took a year off.

Speaker 1

我在高中期间主要专注于学业,既修了大学课程,也参加了各种竞赛。

And I had spent much of my high school doing academic, both college courses, but also competitions.

Speaker 1

我非常投入数学竞赛,也深深迷上了化学竞赛。

I got very into math competitions, got very into chemistry competitions.

Speaker 1

我记得十年级时,因为九年级已经学过化学,我妈妈在网上找到了一个化学竞赛,我就随便参加了。

And I remember tenth grade, because I'd taken chemistry in ninth grade, I took some chemistry competition that my mom had found online and just kind of did it on the mark.

Speaker 1

我获得了我所在地区的第一名。

And I got best in my region.

Speaker 1

我当时想,嗯,这挺酷的。

I was like, okay, that's cool.

Speaker 1

我所在的州并没有这个比赛,所以我去明尼苏达州参加。

I got my state didn't really have it, so I was in Minnesota to do it.

Speaker 1

我参加了全州的比赛,并获得了州第一名。

I took the statewide one and I got best in the state.

Speaker 1

他们邀请我参加国际化学奥林匹克竞赛的培训营。

And they invited me to the training camp for the International Chemistry Olympiad.

Speaker 1

也就是说,这是全国前20名的学生。

So it's top 20 kids in the nation.

Speaker 1

为此,他们会提前寄一些教科书给你。

And for this, they send you some textbooks ahead of time.

Speaker 1

他们说:请阅读这本有机化学教材的第一到第八章。

They say, please read chapters one through eight of this organic chemistry textbook.

Speaker 1

这些就是培训营将使用的教材。

These are the textbook we're going use for the camp.

Speaker 1

于是我读完了第一到第八章。

And so I read through chapters one through eight.

Speaker 1

我没太当回事。

Didn't take it that seriously.

Speaker 1

我当时想,看吧,我注定会成功。

I was like, look, I am just destined to succeed.

Speaker 1

就是会顺利进行的。

Like, it's just going to work.

Speaker 1

到目前为止都顺利发生了。

Like, it happened so far.

Speaker 1

我记得我去了化学竞赛的训练营。

And I remember I showed up at the chemistry competition at the chemistry camp.

Speaker 1

其他孩子读的不只是那前八章。

And the other kids had read not just those first eight chapters.

Speaker 1

他们读完了所有书。

They'd read all the books.

Speaker 1

他们读的不只是这一整本书,还有那些厚厚的物理化学和其他书籍。

They'd read not just the whole one book, but all these big fat, like physical chemistry, all these other ones.

Speaker 1

我当时就想,等等,什么?

And I was just like, wait, what?

Speaker 1

你能做到那样?

You can do that?

Speaker 1

那些人真的在做这些?

These other people are doing that?

Speaker 1

我记得在为期两周的培训期间,我感到无比沮丧和挫败,因为那些人懂的所有东西我都不懂。

And I remember just feeling so demoralized and crushed for the two weeks of the training because it was just like these people knew all these things that I didn't know.

Speaker 1

我们本该为期末考试复习的。

And we were supposed to be studying for the final exam.

Speaker 1

我在房间里玩手机游戏。

I was playing cell phone games in my room.

Speaker 1

我就放弃了。

I just gave up.

Speaker 1

我当时想,根本没希望了。

I was like, there's just no hope.

Speaker 1

然后他们宣布了前四名去参加国际赛。

And so they announced the top four to go to the international.

Speaker 1

他们说,这四个人都不是我。

They said, these four people, they weren't me.

Speaker 1

他们宣布了两名亚军。

They announced two runners up.

Speaker 1

他们公布了那两个人。

They announced those two.

Speaker 1

不是我。

It wasn't me.

Speaker 1

我当时想,我真是二十岁了。

I was like, I am honestly 20.

Speaker 1

幸运的是,他们不会告诉你其他人的具体排名。

Fortunately, they don't tell you the ranking of the rest.

Speaker 1

我当时想,太差了。

I was like, so bad.

Speaker 1

我记得新学年开始时,回望自己那个夏天是如何度过的。

And I remember at the beginning of the next school year, looking back at how I'd spent that summer.

Speaker 1

我觉得自己面前有一个绝佳的机会,却白白浪费了。

And I felt like I had this amazing opportunity in front of me, and I'd squandered it.

Speaker 1

那种感觉非常糟糕。

And it was a horrible feeling.

Speaker 1

我有一种感觉,自己只是随波逐流,以为单靠天赋就能自然达到我想去的地方。

It was this feeling that I had just coasted, that I just sort of believed that talent alone would just get me to where I wanted to go.

Speaker 1

我不需要努力。

I didn't have to work hard.

Speaker 1

我当时就想,我再也不想有这样的感觉了。

And I was like, I never want to feel that again.

Speaker 1

因此,那一年我认真对待了这件事。

And so I took it seriously that year.

Speaker 1

我开始学习物理化学。

I started taking physical chemistry.

Speaker 1

我选修了有机化学。

I took organic chemistry.

Speaker 1

我选了所有这些课程。

I took all these things.

Speaker 1

我花了大量时间研究过去的化学竞赛题目,认真学习,并查阅了大量资料,以成为一名真正的竞争者。

I spent a lot of time on looking at old chemistry competitions and learning them and looking through a bunch of different material in order to become a real competitor.

Speaker 1

那一年,我入选了前四名的团队,参加了国际化学奥林匹克竞赛,并获得了一枚银牌。

And that year, I made the team of the top four, went to the International Chemistry Olympiad, got a silver medal.

Speaker 1

这是一次非常非常棒的经历。

It was very, very awesome experience.

Speaker 1

我真的很享受这个过程。

I really enjoyed it.

Speaker 1

但对我来说,最大的收获是:你必须始终努力付出。

But for me, the big takeaway was this feeling of you always have to work hard.

Speaker 1

我最喜欢的一句名言是关于自行车的:事情永远不会变得更容易。

And one of my favorite quotes is the cycling quote, It never gets easier.

Speaker 1

你只是变得更快。

You just go faster.

Speaker 1

我认为这一点是我一直奉行的准则。

And I think that that is something I very much live by.

Speaker 0

那你是怎么决定写这本书的呢?

So how did you end up writing the book?

Speaker 1

我觉得我找到了一种看待化学的独特方式,一种基于数学第一性原理的方法。

Well, I felt like I'd come up with a very unique way of looking at chemistry, a very mathematical first principles approach to it.

Speaker 1

因为如果你读大多数化学教科书,它们只是说:好吧,记住这些反应就行了。

Because if you read most chemistry textbooks, it just says, well, basically memorize these reactions.

Speaker 1

这些是化学性质。

Here are these chemical properties.

Speaker 1

这些是化合物。

Here's these compounds.

Speaker 1

但你总会问为什么。

But you always ask why.

Speaker 1

比如,为什么会是这样?

Like, why is it that way?

Speaker 1

这些原子必须以这种方式相互作用吗?

Does it have to be that these atoms interact in this way?

Speaker 1

这个化合物必须是这个颜色吗?

Or does this compound have to be this color?

Speaker 1

你是如何从第一性原理推导出来的?

Like, how do you derive it from first principles?

Speaker 1

所以,为了在比赛中表现优异,我采取的方法是努力找出背后的规律,而不是仅仅

And so the approach that I'd taken in order to be good at the competitions was to really try to figure out the underlying rules, not the

Speaker 0

这在你读的教科书里并没有提到。

Which wasn't in the textbook that you read.

Speaker 1

教科书里没有这些内容。

It's not in the textbook.

Speaker 1

你必须自己提炼出来。

You have to distill it down yourself.

Speaker 1

对吧?

Right?

Speaker 1

也许他们试图以某种方式沟通,但那并不是他们花时间去做的事情。

And maybe they try to communicate in some way, but it's like, it's just not the thing that they spend their time on.

Speaker 1

所以我整理了这本书的结构。

And so I structured the the book.

Speaker 1

我觉得我对化学的教学方式与众不同,我按照一些朋友在数学领域——特别是名为‘问题解决的艺术’的数学论坛——所做的方式来组织内容,那里采用的是非常苏格拉底式的教学法。

I felt like I had a different way of of teaching chemistry, and I structured it in a way that was inspired by some of my friends who had done something similar in math, in this math forum called the art of problem solving, where it's a very Socratic method.

Speaker 1

所以这本书里全是问题,但这些问题经过精心设计,每一个都层层递进。

So the book has just questions, but they're intentionally scoped so that each one builds up.

Speaker 1

第一个问题从你本应掌握的知识出发,只要你稍加思考,就会明白其中的道理。

And so the first one starts from knowledge you should have, if you just think a little bit, you're like, oh, I can see how this works.

Speaker 1

下一个问题是建立在前一个问题的基础上,再下一个又建立在再前一个问题的基础上。

And the next one builds on the previous thing, the next one builds on the previous thing.

Speaker 1

在化学领域,你需要一些实验结果。

In the case of chemistry, you need some experimental results.

Speaker 1

你会说,比如这个双缝实验。

You say like, here's this double slit experiment.

Speaker 1

然后你会想,好吧,这对粒子和波来说意味着什么?

And then you're like, okay, well, what does it mean for particle versus wave?

Speaker 1

接着,如果它既是粒子又是波,那么,你知道的,以此类推。

And then, okay, if it's both a particle and a wave, then, you know, and so forth.

Speaker 1

这正是我非常想传达给他人的东西,我真的很在意这一点,不只是让这种教学方式只停留在我的脑海里,而是希望其他人也能从中受益。

And this was something that I really wanted to communicate to others, that I really cared about, not just this approach living in my head, but other people being able to benefit from it.

Speaker 1

不过,我最终没有完成。

Now, I never finished.

Speaker 1

我写到了大约一百页左右。

I made it through about a 100 pages.

Speaker 1

如果你感兴趣,可以在我的网站上找到它。

You can find it on my website if you're interested.

Speaker 1

但我认为这种精神是我真正想延续下去的。

But I felt like that ethos was something I really wanted to carry forward.

Speaker 0

你实际上写出了你希望自己当初能读到的那本书。

You basically wrote the book that you wish you could have read.

Speaker 1

没错,就是这样。

That's exactly right.

Speaker 0

太棒了。

That's great.

Speaker 1

是的。

Yes.

Speaker 0

编程与其他活动有哪些相似或不同之处?

How is coding similar or different to other activities?

Speaker 1

我认为编程的方式是,你深刻理解某个过程。

So the way that I think about coding is that you deeply understand some process.

Speaker 1

你用一种非常晦涩的方式——我们称之为程序——把它写下来,然后任何人都能从中受益,对吧?

You write it down in a very obscure way we call a program, and then anyone can get the benefit of that thinking, right?

Speaker 1

人们不需要自己写代码。

People don't need to write the code.

Speaker 1

他们不需要了解其中涉及的机械原理。

They don't need to understand the sort of mechanics of what went into it.

Speaker 1

我认为还有其他一些非常脑力密集的领域也是如此,比如数学,对吧?

And I think that there are other very cerebral domains that are like this, like mathematics, right?

Speaker 1

你认真思考一个问题,然后以一种晦涩的方式写下来,我们称之为证明,但没人会去读这些证明,只有少数五位关注特定领域的数学家才会深入阅读。

Where you think hard about a problem, you write it down in an obscure way, we call it proof, but no one reads those, Only like the five mathematicians who care about a particular domain will really deeply read it.

Speaker 1

所以在我看来,编程之所以与众不同,是因为它几乎像魔法一样。

And so I think that what makes coding stand apart to me is it's almost like magic.

Speaker 1

你脑海中有一个构想,只需通过描述,它就真的实现了。

You sort of have this vision in your head and you just, by describing it, somehow it comes to be.

Speaker 1

因此,在某种程度上,这就像管理一样,你有一台计算机,它会按照你脑海中的愿景,以非常字面的方式执行任务,当你编写程序时。

And so in some ways it's like management, right, that you have a computer that is there to perform the function that you have in mind, the vision that you have, and that it carries it out in a very literal fashion when you write a program.

Speaker 1

我认为,这在我所知的其他任何传统领域中都从未见过。

And I think that that to me is something I've never seen in really any other traditional domain.

Speaker 1

就像你做的许多其他事情,都无法获得这种同样的杠杆效应,对吧?

Like it feels like many other things that you might do that you just don't get that same leverage, right?

Speaker 1

如果你心中有一个愿景,而它 somehow 变成了现实,你却不需要在现实中物理地移动任何东西。

If you have this vision and somehow it comes into reality and that you don't have to physically move things in the world.

Speaker 1

它就这样自然而然地实现了。

It just comes to be.

Speaker 0

你会更倾向于把它描述为一种语言,还是一种数学?

Would you describe it more like a language or more like math?

Speaker 1

我会更倾向于把它描述为一种数学。

I would describe it more like math.

Speaker 1

但关于数学,我认为人们有一个误解,那就是数学并不是一加一,对吧?

But the thing about math that I think there's a misconception for is that math is not one plus one, right?

Speaker 1

它不是关于这些机械式的计算。

It's not about these like mechanical calculations.

Speaker 1

数学是关于宇宙的底层结构,对吧?

Math is about the underlying structures of the universe, right?

Speaker 1

它是关于理解不同对象、不同概念之间如何以一种深层的概念方式相互关联,以及这些对象之间的对称性和关系看起来有多么不同。

It's about understanding how different objects, different ideas relate to each other in this deep conceptual way and the sort of symmetries and the relationships between objects look very, very different.

Speaker 1

我认为编程也是如此。

And I think that programming is like that.

Speaker 1

它真正关乎理解一个网站应该如何工作,用户想要什么,以及在各种边缘情况下事物应该如何表现。

It's really about understanding how should a website work or what does someone want and what are all the different ways that something in a corner case should behave.

Speaker 1

如果出现错误,我们该如何处理?

If there's an error, how should we handle it?

Speaker 1

所有这些事情看起来有些平凡,但如果你深入观察其底层架构,你会意识到:你有多个系统彼此通信,数据以不同形式存储,像加密这样的概念也被运用其中。

All of these things, they feel somewhat mundane, but if you really look at the underlying architecture, you're thinking about, you have all these systems that are talking to each other, you have data stored in these different forms, you have ideas like encryption that are brought to play.

Speaker 1

你如何协调这一切,以交付有用的东西?

And how do you orchestrate all of that in a way that delivers something useful?

Speaker 1

因此,在我看来,这正是数学之美被具象化为实用形式的过程。

And so to me, I think it's about the beauty of mathematics that's reified into useful form.

Speaker 0

数学是自然的上层结构,还是自然是数学的上层结构?

Is math an overlay on nature, or is nature an overlay on math?

Speaker 1

这是个很好的问题。

That's a great question.

关于 Bayt 播客

Bayt 提供中文+原文双语音频和字幕,帮助你打破语言障碍,轻松听懂全球优质播客。

继续浏览更多播客