格雷格·布罗克曼（第一部分）

当然，你知道的，它会以各种不同的方式和形式表现出来。

And of course, you know, it plays out in all sorts of different ways and forms.

但从根本上说，我认为对我而言，建立一个步调一致、领导团队团结、朝着同一个方向前进、拥有明确愿景并付诸执行的公司，这才是最重要的。

But fundamentally, I think that for me, getting to a aligned, board leadership team company that's moving in one direction, that has a vision and is executing on it, that's been the most important thing.

在那一刻，我感觉我们的生命像走马灯一样在眼前闪现。

And I felt like in that moment that we had our life flash before our eyes.

所以从那以后，你绝不能再把它视为理所当然。

And so I think since then, you can never take it for granted.

你现在觉得自己对当时发生的事情有了很好的理解吗？

And do you feel like you have a good understanding of what happened now?

是的。

I do.

是的。

You do.

我回来的一部分条件就是说，让我们对所有事件进行一次独立审查。

Part of the deal for coming back was to say, let's do an independent review of all events.

对吧？

Right?

所以会有一家律师事务所进来，全面审查所有事情，并判断这里是否有什么需要改变的地方？

So there's someone some law firm's gonna come in, take a look at everything, and say, is there something that needs to be different here?

一家律师事务所确实这么做了，并表示我们理解董事会为何这样做。

And a law firm did that and said that we understand why the board did this.

这是他们的权利，但他们绝对不是非得这么做。

It was within their right, but they definitely didn't have to.

于是，所有旁观者都把自己的解释投射到所发生的事情上。

And so there are all these onlookers who then project their own explanation onto what happened.

我认为到目前为止，他们对所看到的一切都相当失望。

I think they've all been pretty disappointed so far in terms of what they've seen.

这让人感到不舒服。

And it's uncomfortable.

这确实让人不舒服，因为我觉得这是一个我们的内部动态外溢到外界的时刻。

It is uncomfortable because I think that it's a moment where our internal dynamics spilled out into the outside world.

在我整个OpenAI的经历中，我的一个做法是秉持着‘保持团队团结’这一信条。

One of my approaches throughout OpenAI has been, I had this mantra of keep the band together.

是的。

Yeah.

在我们经历的各种分裂中，我采取的策略就是：我们有这么多人，有不同的观点，存在冲突，想朝不同方向推进，不同的人认为自己应该掌权，不管怎样，但这些人都是聪明人。

And through various splits that we had, that was the approach I took is to say, hey, we've got all these people, we have different views, we're having conflict, we wanna push in different directions, different people think they should be in charge, whatever it is, but these are all smart people.

我希望将这些不同的观点提炼成一个关于该做什么的一致性图景。

And I'd love to distill all these views into consistent picture of what needs to happen.

对我来说，这正是我在OpenAI的核心工作：试图弄清楚，我们有这样一个宏大的使命，但该如何实现它？

And that to me is the core thing that I feel like I do at OpenAI is try to figure out, we have this grand mission, but how do we do it?

正确的决定是什么？

And what is the right decision?

在每一个这样的时刻——这不仅仅是关于解雇事件，还包括回顾我们与埃隆共事的时期，以及与Anthropic创始人们共事的时光。

And in each of those instances, and it's really not just for the firing, it's like looking back to our time with Elon, to our time with the Anthropic founders.

对于每一个这样的时刻，我都认真思考过：我们如何才能共同努力，达成一个卓越的结果？

For each of these moments, I have really gone through each of them trying to say, like, how can we all work together to get to a great outcome?

是的。

Yeah.

最终，我们总是以人员离职告终，要么是因为对方决定不再继续，而这通常会带来后续的痛苦。

In the end, we've ended up with departures either because the other person decides they don't want to do it anymore and that there's usually downstream pain.

如果你观察发生了什么，通常我们会让别人来讲述这个故事。

That if you look at what happens, normally we let other people tell the narrative.

比如，我们继续前进，部分原因是我认为我们只是不希望——如果我们是推动公司前进的人，我们希望别人能掌握自己的叙事。

Like, we keep going and partly, I think we just don't want you know, if if we're the ones who have who are pushing on the company, we want other people to be able to own their story.

是的。

Yeah.

但有时人们会利用这一点来攻击我们。

But then sometimes people use that to kind of whack us.

对。

Yeah.

告诉我你对我们关于大卫和歌利亚的对话有什么记忆。

Tell me what you remember about our David and Goliath conversation.

因为我感觉这有关联。

Because I feel like this relates.

是的，我们确实谈过。

Yeah, we did talk about it.

我正努力回想具体是在什么背景下讨论的，但我觉得核心是，OpenAI，如果你看看我们所处的行业，有谷歌、Meta，这些都是拥有巨大市值、众多员工、大量用户和海量算力的成熟公司。

And I'm trying to remember exactly what context it was in, but I think that the core is that OpenAI, like if you look at the industry that we're in, right, there's Google, there's Meta, and these are established companies with very, very huge market caps and many people and all sorts of users and lots of compute.

而我们是挑战者。

And we are the challenger.

我们是刚刚起步的那一个。

Like we are the ones who are just getting started.

从零开始。

Start to start.

没错。

That's right.

不知为何，我认为由于ChatGPT如此成功，人们就认为我们——既然我们是品类领导者，那我们就是歌利亚，是既得利益者。

And somehow I think that because ChatGPT has been so successful that people then think of us, well, because we're the category leader, therefore we're the Goliath, we're the establishment.

而在我看来，AGI才是真正的目标。

And where I sit, I say the AGI, that's the real prize.

这项技术可能会发展成真正有利于这些现有巨头的形式。

And this technology is something that I think could play out such that it really benefits these incumbents.

我们的理念始终如此，我们创办公司的全部原因，就是思考如何为所有人引导这条道路走向更好的方向。

And our ethos has always been, the whole reason we started was to think about how can we steer this in a better direction for everyone.

我认为，在这个由巨无霸企业主导的行业中，我们恰恰是大卫。

And that is something where I think we are very much, David, in an industry of Goliaths.

我认为埃隆的情况也是如此，对吧？

And I think the same is true with Elon, right?

埃隆·马斯克，世界首富，也是世界上最强大的人物之一，却在起诉我们，对吧？

That Elon Musk, the richest man in the world, one of the most powerful men in the world, and that he's suing us, right?

说我们欺骗了他。

Saying that we tricked him.

当我回望我们所做的事情时，我和伊利亚投入了大量精力，试图与他合作成功。

And when I look back at what we did, I spent so much effort, Ilya and I spent so much effort to try to make it work with him.

我们当时非常透明，什么

And we were very transparent What

愿景上的差异在哪里？

was the difference in vision?

所以是控制权。

So control.

我们无法与他达成协议的原因是，我们都同意，我们需要比通过慈善途径所能获得的多得多的资金。

The reason that we could not get to a deal with him, because we all agreed that we're gonna need far more capital than we can get through philanthropic means.

所以我们需要一个营利性实体。

So we're going to need a for profit entity.

我们就条款进行了谈判，而他

And we negotiate over the terms and he And

他也同意这一点吗？

he agreed with that as well?

哦，他同意这一点。

Oh, he agreed with that.

是的。

Yeah.

他告诉我们，现在是行动的时候了。

He told us the time is now to go.

他说，这就是关键的转折点。

He said, this is the triggering event.

一旦我们取得了第一个重大成果，我们都意识到这是唯一的前进方向。

Once we did our first big result, we all got that this was the only way forward.

然后，关于细节的谈判开始了，他要求拥有多数股权，要求绝对的初期控制权，并且要求担任首席执行官。

And then the negotiation over the details began and he needed majority equity, he needed absolute initial control, and he needed to be CEO.

但我们无法让步于控制权。

And we could not give on the control.

我记得我和伊利在思考这个问题。

I remember Ily and I were thinking about it.

我们是否该退让？

Do we fall?

我们把这一切交给他吗？

Do we give this to him?

我们把公司交给他吗？

Do we give him the company?

然后你开始想，假设它真的成功了。

And then you start thinking about, well, imagine it actually works.

想象一下，你真的实现了目标。

Imagine that you actually achieved the mission.

想象一下，你真的造出了通用人工智能，而有一个人对它拥有绝对控制权。

Imagine you actually build an AGI and there's one person who has absolute control over it.

不管那个人是谁。

Doesn't matter who that person is.

你感觉好吗？

Do you feel good?

太多了。

Too much.

太多了。

Too much.

所以我们大力推动某种否决机制，以便在其他人全都反对、我们根本无法推进时能够阻止。

And so we pushed hard for an override of some form, so if everyone else was voting against, and we just couldn't get it done.

所以我们拒绝了。

And so we said no.

当时交易没谈成，最终是不欢而散的吗？

That Did it end on bad terms right in that moment when the deal wasn't made?

我以为会这样。

I thought it would.

那真的非常、非常艰难。

It was very, very tough.

那是一个情感上极其耗竭的时刻，因为我们花了六周——五周半到六周——完全没有做任何工作，只专注于谈判这些条款。

It was a very emotionally draining moment because we'd spent six weeks, five and a half, six weeks, not doing any work, just really negotiating these terms.

我们进行了许多深刻的情感对话，我以为一切都泡汤了。

We had these deep emotional conversations and I thought that it was all blown up.

但并没有。

But it wasn't.

我们仍然努力寻找共同前进的道路。

And that we still tried to find a path forward together.

但他回来提出的条件是：要么承诺加入非营利组织，这意味着给埃隆再增加两个董事会席位，那样的话，就是我lya、Sam、Greg、埃隆、埃隆、埃隆。

But the thing that he came back with was to say, hey, either commit to the nonprofit, which means giving Elon two more board seats, but it would be, you know, Ilya Sam, Greg, Elon, Elon, Elon.

你们必须承诺不挖角，且一年或两年内不离职，类似这样的条件。

You'd have to commit to a non solicit and to not quitting for a year or two or something like that.

因此，这些承诺有着非常具体的要求。

So there were very specific terms for the commitment.

我们说，我们会考虑一下。

And we said, we're, we were going to think about that.

他回来提出的另一件事是：你知道吗？

The other thing that he came back with was he said, you know what?

你们为什么不直接并入特斯拉呢？

Why don't you just merge into Tesla?

我对特斯拉没有完全的控制权。

I don't have full control over Tesla.

这解决了你的顾虑。

It solves your concern.

在这里打造通用人工智能。

Build the AGI here.

秘密进行。

Do it in secret.

你必须这么做，因为股东们不会喜欢。

You gotta do it because the shareholders won't like it.

这是他极力推动的关键点。

And that was the big thing he pushed on.

所以你得到了你的营利性业务，得到了特斯拉这个现金奶牛——这就是当时使用的说法。

So you get your for profit, you get the cash Tesla at the cash cow was the term that was used.

对于这一点，我们完全没有兴趣。

And that one, we had no desire to do.

我们真的认真考虑过。

Like, we really thought about it.

我们只是觉得，这不太对劲。

We were just like, just doesn't seem right.

只是不一样。

Just a different

这是不一样的事情。

It's a different thing.

这是不一样的事情。

It's a different thing.

感觉它不太可能成功。

It doesn't feel like it's likely to succeed.

而且这并不是以一种透明的方式推动AGI，从而造福人们。

And it's not like pushing on AGI in like this transparent way that's gonna benefit people.

所以我们进行了多次迭代。

And so we went through multiple iterations.

我们讨论过，是否可以再次开放利润谈判？

We talked about, could we open up this for profit negotiations again?

当时有点觉得，也许未来我们可以这么做。

And it was kind of like, yeah, maybe we could in the future.

然后我们真的想出了一个主意。

And then we actually came up with an idea.

所以我们开始筹办一个非营利性募捐活动。

So we pursued a nonprofit fundraiser.

那我们就先这么做，但我们需要看看实际能筹到多少资金。

So let's do this for now, but we need to see what we can actually raise.

在这个过程中，我们想到或许可以进行一次首次代币发行，通过某种加密货币来筹集资金。

And along the way that we had an idea for maybe we could do an ICO, and we could actually raise capital through doing some sort of crypto coin.

我们还有一些建议，关于如何将这与我们未来生产的算力或AGI联系起来，他对此非常兴奋。

And we had some ideas for how that would tie to the future compute or AGI that we produce, And he got so excited about that.

我记得我们有过一次对话，他说：你已经解决了这个问题。

I remember we had a conversation where he said, you've solved it.

这都很好。

It's all great.

我们将筹集100亿美元。

We're going to raise $10,000,000,000.

他说他同意了。

He said that he was on board.

我们花了很多时间思考，好吧，让我们真正实际地规划一下。

And we spent a lot of time thinking about, okay, let's like really practically chart this out.

这是个好主意吗？

Is this a good idea?

我们想做这件事吗？

Do we wanna do it?

我们开始对这个想法有些犹豫了。

We kinda started to get iffy on it.

然后他发了一封邮件，说：你知道吗？

And then he sent an email saying, you know what?

我决定不支持这个ICO。

I've decided I will not support the ICO.

这一切都在2018年1月发生了。

This all kind of played out in January 2018.

那时他说，我们真的需要一条筹集更多资金的路径。

And then at that point he said, we really need a path to raising more money.

我们告诉他，嘿，我们有个不涉及ICO的其他想法，但依然会是一个盈利性实体。

And we told him that, hey, we have another idea that doesn't involve an ICO it would still be, you know, we're gonna talk about this like for profit entity.

所有这些想法都在不断涌现，我们尝试了各种不同的方案。

Again, all these ideas were just swirling around and different different things we could try.

他说他根本不相信这个能成功。

And he said that he just didn't believe it was gonna work.

所以他宣布退出。

So he said he's out.

他的计划是在特斯拉打造通用人工智能，并做那件事。

His plan was to build AGI at Tesla and do that thing.

他实际上试图招募我们的一大批人。

He actually tried to recruit a bunch of our people.

没人跟他走。

No one went with him.

但在他离开之前，他召集了整个团队开了一次全员会议。

But on his way out, he did this all hands with the team.

在这次全员会议上，他说，格雷格、萨姆和伊利亚看到了筹集数十亿美元资金的路径。

And in that all hands, he said that, hey, Greg Sam and Ilia see a path to raising billions of dollars.

我不认为这能成功，但如果你看到了一条路，你就得去走。

I don't think it's going to work, but if you see a path, you got to follow it.

所以我支持他们。

And so I support them.

他说他还会继续提供建议并保持参与。

He said he'd even keep advising and stay involved.

他支持所有这些，但他只是觉得，你知道，这需要每年数十亿美元的资金。

He supports all of this, but he's just like, you know, this requires billions of dollars per year.

他还谈到他将开展特斯拉AGI项目，不会专注于安全问题，因为按照他的说法，如果羊群在立法要求安全，而狼群却不考虑安全，那也就没必要提谷歌了。

He also then went to talk about that he's gonna do Tesla AGI, that he wasn't gonna work on safety because, in his words, if the sheep are legislating safety and the wolves aren't having safety, then there's no point referring to Google.

所以，你知道，就是这种语气。

So, you know, it's like that kind of tone.

于是，在接下来的一年里，我们开始思考，究竟该如何把这些庞大的资金转化为实际资本。

And so then we began over the next year to figure out, you know, how do we actually raise these large pawns into capital?

那时候伊隆对此完全接受吗？

Was it all cool with Elon at that point?

当然，没错。

For sure, yes.

我的意思是，这非常明确。

I mean, I think it was very explicit

每个人都清楚，这就是未来的发展方向。

that everyone understood this was what was happening moving forward.

是的。

Yes.

好的。

Okay.

接下来一步是什么？

And what was the next step?

我们需要设计一个能够帮助我们筹集更多资金的结构。

Well, we needed to figure out some structure that would actually let us raise more capital.

我们考虑了各种不同的想法。

And we thought about different ideas.

最终我们设立了名为OpenAI LP的实体，并在2019年中期正式启动。

We ended up implementing this entity called OpenAI LP, and in mid-twenty nineteen that we actually got started.

在此期间，我们一直让埃隆了解进展情况。

And along the way, we kept Elon apprised of what was happening.

我们把条款清单发给了他。

We sent him the term sheet.

期间有过几次通话，萨姆向他汇报了我们的进展。

There were a bunch of calls where Sam would talk to him about what we were doing.

我记得2018年12月收到他的一封邮件，那封邮件令人沮丧。

And I remember getting an email from him in December 2018 that was devastating.

他发了一封邮件，主题大概是‘我需要提醒你吗？’

He sent an email saying the subject was something like, need I remind you?

邮件正文大概是：如果没有资源和执行上的巨大改变，OpenAI相对于谷歌成功的概率为零。

And the body was something like, OpenAI has a zero percent chance of success relative to Google without a dramatic change in resources and execution.

连百分之一都没有。

Not a one percent.

我希望情况能不同，但事实就是这样。

I wish it were different, but that's true.

这需要每年数十亿美元，甚至数亿美元——而这正是我们告诉他我们计划筹集的金额远远不够的地方。

This is going to require billions of dollars per year, even hundreds of millions of dollars, which is what we told him we were planning on raising isn't enough.

我记得我当时正在度假，那天整个人都被击垮了。

And I remember I was on vacation and I just was like, kind of knocked out for that whole day.

是的。

Yeah.

很明显，他认为我们注定会失败。

It was very clear that he thought that we were destined to fail.

对。

Yeah.

如果他不觉得你们注定会失败，他是不会发这封邮件的。

I don't think he would have sent it if he didn't think you were destined to fail.

是的。

Yeah.

对。

Yeah.

我同意。

I agree.

那一年他一直持这种观点。

That'd been his theme for that whole year.

我认为，当我们开始取得成功时，情况才真正发生了变化。

And I think that things really changed as we started to be successful.

而且我认为，现在他试图讲述一个完全不同的故事。

And that I think that now there's a very different tale that he's trying to tell.

你上一次和他说话是什么时候？

When's the last time you spoke to him?

几个月前。

A couple months ago.

愉快吗？

Pleasant?

非常愉快。

Very pleasant.

每次都挺愉快的。

It's always pleasant.

是的。

Yeah.

随着二十世纪中期营养科学的发展，研究人员开始意识到，现代饮食模式、食物种类单一、加工食品以及时间紧张，可能导致每日微量营养素摄入出现微小但重要的缺口。

As nutrition science advanced through the mid twentieth century, researchers began to understand that modern eating patterns, limited variety, processed foods, and time constraints could leave small but meaningful gaps in daily micronutrient intake.

如今，大规模人群研究证实，大约百分之九十的成年人在一种或多种必需维生素或矿物质的摄入上不足。

Today, large population studies confirm that approximately ninety percent of adults fall short on one or more essential vitamins or minerals.

AG1正是针对这些发现而研发的。

AG1 was formulated in response to those findings.

每份每日摄入量提供超过75种维生素、矿物质和全食物来源的营养素，包括维生素A、C、E、B6、B12，以及镁、锌、硒和碘。

Each daily serving provides more than 75 vitamins, minerals, and whole food sourced nutrients, including vitamins A, C, E, B6, B12, magnesium, zinc, selenium, and iodine.

这是一种由营养学研究支持、以人体易于吸收的形式配制的科学配方。

A scientifically developed formula backed by nutrition research and delivered in forms the body can readily access.

由于吸收决定利用率，AG1还含有益生菌，以支持肠道健康和消化功能，帮助身体有效利用所摄入的营养。

Because utilization depends on absorption, AG1 also contains probiotics to support gut health and digestive function, helping the body make effective use of what it receives.

AG1不含麸质或乳制品，不使用任何转基因成分或人工甜味剂，适用于多种饮食方式，包括植物性、生酮、低碳水化合物和古法饮食。

AG1 contains no gluten or dairy, uses no genetically modified ingredients or artificial sweeteners, and is suitable for a wide range of dietary approaches, including plant based, paleo, keto, and low carbohydrate diets.

提供单份包装，方便随身携带使用。

Available in single serving packets for when you're on the go.

每一批AG1都在严格的质量控制下生产，并经过检测，以确认其成分、营养含量及无污染物，确保每一份产品的可靠性。

Manufactured under strict quality controls, every batch of AG1 is tested to confirm its composition, nutrient levels, and the absence of contaminants, ensuring reliability from one serving to the next.

难怪AG1受到全球顶尖运动员和专家的信任。

It's no wonder AG1 is trusted by the world's top athletes and experts.

曾经是未来概念的每日营养支持，如今已成为应用科学的成果。

What was once a futuristic concept, daily nutritional support in a glass, is now the result of applied science.

了解更多，请访问 drinkag1.com/petra。

Learn more at drinkag1.com/petra.

今天就行动吧。

Do it today.

你是怎么开始与微软合作的？

How did you end up getting into business with Microsoft?

微软是我们的早期合作伙伴。

So Microsoft was an early partner.

他们为我们Dota项目捐赠了计算资源。

We got compute donated from them for our Dota project.

这就是一切的开端。

So that was the very beginning of it.

随着时间的推移，我认为萨姆与萨提亚、凯文·斯科特一直保持着良好的关系。

And over time, I think Sam had kept a good relationship with Satya, with Kevin Scott.

他们真的相信我们正在做的事情。

And they really believed in what we were doing.

他们真正看到了通用人工智能和构建这些强大系统的愿景。

And I think really saw the vision for AGI and building these powerful systems.

他们什么时候真正成为合作伙伴的？

And when did they actually become partners?

他们最终为这家有限责任公司提供了资金吗？

Did they end up funding the LLC?

是的。

Yeah.

所以他们向有限合伙企业注资了十亿美元，随后又进行了20亿美元的投资，接着又进行了100亿美元的投资。

So they funded the billion dollars into the LP and then they did a $2,000,000,000 investment and they did a $10,000,000,000 investment.

因此，这笔资金实际上非常可观，对吧？

So that funding was in fact very significant, right?

当时没有人投入如此规模的资金。

That no one else was putting in that, you know, that quantum of capital at the time.

现在情况不同了，但那时他们是这么做的。

Different now, but at that time, they were doing it.

因此，我非常感激他们帮助我们推进使命和所从事的事业。

So I was grateful for them really helping us further the mission and what we were doing.

在此过程中，我们当然也进行了技术授权并建立合作伙伴关系，我们的目标是：一起来打造这个东西。

And along the way that of course, that we had licensing of our technology and we would partner and, you know, the goal was, Hey, let's build this.

如果我们还能将这项技术部署到微软内部，为微软创造价值，那么我认为这最终将推动我们实现使命和模型的追求。

And then if we can also have this technology be deployed within Microsoft and create value there for Microsoft, then, you know, again, I think it is the thing that ultimately unlocks our pursuit of the models of the mission.

当你们获得大量资金注入时，业务发生了怎样的变化？

How did your business change when you had the influx of cash?

我认为我的一贯理念是，永远不要认为我们已经赢得了这一切。

I would say that my ethos is always to remember we have not earned any of this.

我们还不是一家盈利的公司。

We are not a profitable company.

所以我们花在计算上的钱非常多，这正是我们的目标。

And so we spend a ton on compute and that that's our goal.

所以我认为发生巨大变化的是，我们开始敢于畅想，某种程度上，这并不只是关于资金，而是关于超级计算机。

And so I think that the big thing that changed is that we could dream, it wasn't really about the cash in some ways, it was really about the supercomputers.

因此我们花了大量时间设计超级计算机，梦想着那些庞大的集群，并提出了各种设计方案。

And so we spent a lot of time designing supercomputers and we're dreaming of these like massive clusters and we had all sorts of design.

这非常有趣，因为对方交易的一个核心观点是：我们不确定这些模型是否真的有用，但至少OpenAI能帮我们更好地构建用于深度学习的超级计算机。

And it was very interesting because one of their theses of the deal was, hey, we don't know if these models will be useful, but at the very least OpenAI can help us figure out how to build supercomputers better for deep learning.

我们非常渴望为深度学习打造出色的超级计算机。

And we were very eager to build great supercomputers for deep learning.

于是，这不再仅仅是一个问题：我们会不会拿到一个数据中心？

And so that was something that it became not just, you could kind of like be like, are we going to get a data center?

因为在2017年，我们真的在纠结：我们怎么才能获得一个数据中心？

Because in 2017, it was very much like how are we gonna get a data center?

数字根本对不上。

The numbers just don't add up.

我们无法达到所需的规模。

There's no way to get to the scales that we needed to.

然后，由于这个合作，这种情况开始变得可能。

And then it started to become possible because of this partnership.

听起来你们的目标是一致的。

And it sounds like you were aligned.

对。

Yes.

这很合理。

It made sense.

是的。

Yes.

ChatGPT相信上帝吗？

Does ChatGPT believe in God?

我会说ChatGPT没有一致的人格。

I would say ChatGPT does not have a consistent personality.

我们对它的看法是，你应该能够根据自己的偏好来塑造它。

And the way we think about it is that you should be able to shape it to your own preferences.

因此，你可以拥有一个持有这种信念的版本，也可以拥有一个不持有的版本。

And so you can have one that has such a belief and you can have one that does not.

对我来说，重要的一点是认识到ChatGPT并不是一个独立的实体。

To me, one thing that's important is to recognize that ChatGPT is not an entity.

对吧？

Right?

但它更像是一个由多种代理组成的集合体，这些代理可以被塑造成不同的形式。

But it's almost this plurality of, you know, sort of agents that that can be shaped in different form.

它有情感吗？

Does it have emotions?

我认为这是一个开放性问题，还存在一个被称为模型福祉的概念。

I think that this is an open question, and there's a whole idea of what is called model welfare.

对吧？

Right?

这些模型有情感吗？

Are these models do they have emotions?

它们过得开心吗？

Are they having a good time?

我认为，随着模型变得越来越聪明，我们必须面对这些问题。

And I think that we're going to have to encounter these questions as the models get smarter and smarter.

我认为，目前我们还没有一个关于什么是情感的完善理论。

And I think that right now we don't have a good theory of what it means to have emotions.

你能确定另一个人是否有情感吗？

Can you tell even another person does?

你相信它们有情感，是因为你自己有情感，但我们还没有科学的方法来测量它。

You believe that they do because you yourself have them, but we don't have a scientific measurement of it.

因此，我认为随着时间推移，我们必须认真思考我们所创造的东西究竟是什么，不仅从功能上，还要理解这个模型实际上是如何运作的。

And so I think that we're going to have to pay some serious attention as time goes on to figuring out what is it that we're creating, not just functionally, but how is this model actually operating?

它会睡觉吗？

Does it sleep?

ChatGPT从不睡觉，这正是它出色的原因，对吧？

ChatGPT never sleeps, which is part of what makes it great, Right?

它就像一部24小时在线、随身携带的设备。

It's a 20 fourseven in your pocket.

它可以是医生，也可以是老师。

It can be a doctor, it can be a teacher.

你认为最近的突破是否正在变得越来越快？

Would you say that the breakthroughs coming are coming faster and faster?

还是说变化的速度自始至终都保持一致？

Or is the rate of change the same as it's been from the beginning?

指数增长仍在继续。

The exponential continues.

因此，翻倍周期没有变化，但感觉突破的速率实际上一直相当稳定。

And so the doubling period is the same, but it feels like the rate of breakthroughs has actually been fairly constant.

有没有什么你原本以为AI会擅长，但实际上并不行的事情？

Is there something that you thought that AI would be good at that it's not?

有没有什么你原本以为AI不擅长，但实际上却很擅长的事情？

And is there something that you didn't think it would

当然有，是的。

be good at that it is?

当然有，是的。

For sure, yes.

我认为这些模型在软件方面变得非常出色，不仅仅是编码部分，还包括所有的调试、测试和端到端的流程。

I'd say that the models have gotten so good at software, not just the coding part, but all of the debugging, the testing, the end to end of that.

对我们来说，从2022年5月开始，这种转变发生得非常迅速，简直是天壤之别。

And it really felt like it transitioned very quickly for us between 05/2002, night and day difference.

就在一两个月前，这些功能还根本行不通。

And just that's in a month, two months worth of it wasn't working.

它从勉强能用，一下子变得非常出色。

It was like kind of okay to it's great.

看到这一点真的让人非常兴奋。

And that was a real exciting thing to see.

我认为模型未来让我期待的地方，是目前还达不到的，那就是更深入的人类判断。

And I think where I'm excited to see the models get great, I don't feel they are yet, is more in, call it human judgment.

比如，我非常期待的是，模型能帮助我们每个人成为更好的自己。

Like the kind of thing I'm very excited about is the idea that you can have models that help us all be better versions of ourselves.

你能否拥有能够帮助调解有分歧的人们的模型？

And can you have models that help negotiate people who have differences?

你知道，如果在早期的OpenAI时期就有这样的技术，结果可能会不同。

You know, maybe if we had this in earlier times at OpenAI, would have different outcomes.

但想想如何实现像世界和平这样的目标，对吧？

But think about achieving outcomes like how do you have world peace, right?

如何真正实现不同国家之间更好的贸易协议？

And how do you actually have better trade deals between different countries?

像这种能在如此高层面上进行协商的能力，目前我还看不到模型能带来实际价值。

Like, being able to negotiate things at that level, I don't really see the models yet adding value.

我认为没有什么能阻止它们达到这个水平，但我觉得我们还有很长的路要走。

I don't think there's anything that stops them from getting there, but I think that we have some ways to go.

你觉得它们只是在这些领域训练或测试得还不够吗？

Do you feel like they just haven't been trained up or tested enough in those realms?

是因为这个原因吗？

Is that why?

还是因为有什么缺失的东西？

Or because there's something missing?

我觉得我们只是需要弄清楚如何训练它们来完成这些任务。

Think we just really need to figure out how to train them for those tasks.

AI 最擅长的是当前信息、历史信息，还是理论信息？

Is AI best with current information, historical information, or theoretical information?

我认为 AI 在理论信息方面最擅长，因为真正希望它具备的是推理能力。

I think that AI is best with theoretical information in the sense that what we really want out of it is to be a reasoner.

我们希望它能够深入思考，解决新问题。

We want it to be able to think hard, solve new problems.

因此，其中包含的所有历史信息和当前信息都是次要的。

And so any of the historical information, current information that's in there is incidental.

所以我们训练它们的方式，使得它们学到了很多历史事实。

So the way we train them, that it does learn a lot about historical facts.

一个刚训练好的模型无法访问当前信息。

A model that you've just trained has no access to current information.

你需要将它连接到工具，才能获取实时信息。

You need to hook it up to tools in order to be able to get that real time.

但这样做的核心目标是你想要一个聪明的系统，对吧？

But the whole point of this is you want something that's smart, right?

你希望它能够应用于新领域，并在该领域中完成你关心的任何任务。

You want it to be able to apply to a new discipline and be able to accomplish whatever it is that you care about in that discipline.

你还在写代码吗？

Are you still coding?

我在写。

I am.

像你这样职位的人还继续写代码，这不太寻常吧？

It's unusual for someone who is in your position to continue coding, no?

是的。

Yes.

你只是对它有一种共鸣吗？

Do you just feel a connection to it?

我觉得，要确定正确的方向，最好的方式就是身先士卒、冲锋在前。

I feel that the way that I know the right direction to lead is by leading from the trenches, leading from the front.

而我发现，在这个领域，因为变化实在太快，要弄清楚该做什么——比如，什么才是正确的决定？

And the thing that I have found is that in this field, because things change so quickly, figuring out what to do, like what is the right decision?

有时候是公司层面的战略决策，但有时候只是非常微观的，比如我们该如何设计这个API或者这个特定的软件模块？

And sometimes it's a strategic decision for the company, but sometimes it's just very micro, like what is the way that we should architect this API or this particular piece of software?

你必须真正地感受到它。

You need to really feel it.

你必须亲自尝试过各种方法，才能明白哪些有效、哪些无效。

You need to have tried things and have a sense of what works and what doesn't.

因此，我一直以来的做事方式是，比如当我们想要创建API——这是我们第一个产品时，我们知道必须做出一个产品才能融资，但我们也在思考：我们到底可以构建哪些不同的东西？

And so a lot of how I have operated has been, like, for example, when we wanted to create the API, which was our first product, we knew we needed to create a product in order to raise capital, but we thought about, well, what are the different things we could build?

我们究竟如何从无到有？

How do we actually go from nothing to something?

这实际上是我所参与过的最困难的项目，因为整个过程感觉完全颠倒了。

And it was actually the hardest projects I've ever worked on because it felt totally backwards.

通常构建产品的做法是：你有一个想要解决的问题，人们并不关心背后的技术。

The way that you're supposed to build a product is that you have a problem you want to solve, no one cares about the technology behind it.

但在我们的情况下，我们拥有的是GPT-3，一项在寻找问题的技术。

But in our case, we had GPT-three, which was a technology in search of a problem.

所以我们决定，干脆通过API让企业能够使用它，由他们自己决定如何应用。

And so we said, let's just make it accessible to businesses through an API and they can decide what to do with it.

这是一段长达六个月的艰苦历程。

And it was a six month grind.

我记得在一月份，我开车穿梭于旧金山，寻找愿意尝试这项技术的公司。

I remember January, I was driving around San Francisco trying to find companies that would be willing to try out this technology.

一些朋友只是出于帮忙才参加了会议，二月份我们做了一些这样的尝试，这非常幸运，因为到了三月，整个世界因新冠疫情而停摆。

Some of my friends took the meeting just as a favor and we did a little bit of that in February, which was very good because in March, the whole world shut down for COVID.

所以不再需要开车四处奔波了，但那是一段极其艰苦的时期，我们要从零开始构建这个产品，并获取第一批测试用户。

So there's no driving around, but it was just this like intense grind of building this thing from nothing and getting the first beta users.

我记得当时告诉我们的团队，我们有两个目标。

And I remember telling our team that we had two goals.

第一个是获得第一个付费客户。

One was to get the first paying customer.

也就是有人第一次给我们付钱。

So the first dollar that someone would give to us.

第二个目标是找到一个我们团队每个人每天都会使用的使用场景。

The second was get a use case that we would all use internally every single day.

是的。

Yeah.

第一个目标我们很快就达成了。

And that first one we got very quickly.

但第二个目标又花了大约两个月的时间。

It actually took another two months or so.

但第二个目标，我们直到ChatGPT才真正实现，所以又花了大约两年时间。

But that second one, we didn't really get till ChatGPT, so that took like another two years.

所以这真的是一次大胆的冒险。

So it was really a leap of faith.

但我之所以能做到，是因为我亲自推动技术发展，以一种我认为别人无法整合的方式。

But the way that I did it was by building and really moving the technology in a way that I don't think that anyone else could have marshaled.

你觉得，如果你从建设工作中休息一年，还能再回来吗？

Do you feel like if you took a year off from the building part, could you come back?

还是说，如果你不一直紧盯着，它发展得太快，就会把你甩在后面？

Or is it something that if you're not always on top of it, is it moving so fast that it would run away?

其实，我确实这么做过。

Well, I actually did this.

所以我有答案。

So I have an answer.

2025年是我第一年全职投入管理工作。

So 2025 was my first year where I was really full time management.

那时，我手下大约有25名直接下属，正在处理一系列非常多样且复杂的任务，包括我们的数据中心、GPU基础设施以及大规模训练任务。

And at that point, I think I had like 25 direct reports and was really running a very diverse and complicated set of problems, including our data centers, including GPU infrastructure, including large training runs.

我需要支持一批非常优秀的工程师。

And there were really, really excellent engineers that I needed to support.

今年，我聘请了一些非常出色的经理来帮我。

And this year I've hired some really excellent managers to help me.

因此，我实际上又重新开始写代码了。

And so I've actually been able to get back to writing code.

感觉怎么样？

How did it feel?

感觉既有生疏又有熟悉，因为这个领域的核心问题在某种程度上其实没有改变——我们依然在训练神经网络，神经网络依然还是前向传播、反向传播和优化器步骤。

It felt there was a mix of rustiness with familiarity because the problems in this field, in some ways they actually do not change because we're still training neural nets and the neural nets still are forward, backward and an optimizer step.

所以我们所做事情的基本框架是相同的。

And so that the fundamental shape of what we're doing is the same.

但在某些方面，一切又都不同了。

But in some ways, everything is different.

尤其是今年，工具完全不一样了。

And this year in particular, the tools are entirely different.

不过有一件好事是，其他人也面临同样的变化。

Now, the one nice thing is they're different for everyone else too.

所以大家都在一起学习，我不觉得别人已经熟悉这些新东西，而我却在摸索老生常谈的内容。

And so everyone is learning together and I don't feel like that everyone moves and that I'm discovering something that is old to everyone.

告诉我ChatGPT刚推出时的情况。

Tell me about when ChatGPT first came out.

看到人们与它对话是什么感觉？

What was it like seeing people experiencing talking to it?

这完全是件全新的事情。

Was a totally new thing.

当时，我们训练了GPT-4。

At the time, we trained GPT-four.

我非常清楚GPT-4会大获成功，我们应该把它变成一个聊天系统。

It was very clear to me that GPT-four was going to be a huge hit and we should turn it into a chat system.

我们之前有一个更早的模型，GPT-3.5，我们让不同的测试者试用过。

And we had an earlier model, GPT-3.5, that we'd actually been having different beta testers try out.

我们付了数百名承包商的费用来使用它。

We were paying like hundreds of contractors to use it.

但它就是感觉不够好。

It just didn't feel like it was great.

我非常支持发布ChatGPT，意思是：先搭建好聊天系统的基础设施，然后再接入真正的模型，那样就会很棒。

I was very supportive of releasing ChatGPT where it's like, let's get the infrastructure out for doing a chat thing and then we'll put it in the real model and it'll be great.

所以真正的意外是，我们对这个新模型太习以为常了，以至于忘了之前的模型对人们来说其实是前所未见的。

And so the real surprise was that we'd gotten so used to this next model that we had forgotten that the previous one was actually something people had never seen before.

只要让它变得可访问、可用，就能让人恍然大悟。

And just making it accessible and available, that that's something that would then cause it to click.

因为对你来说，它已经有点老生常谈了。

Because for you it was already kind of Old hat.

哇。

Wow.

是的。

Yes.

在OpenAI工作最令人惊叹的一点就是，你生活在未来。

And that's one of the amazing things about working at OpenAI is you live in the future.

在一个充满人工快感和强烈刺激的世界里，有一种截然不同的东西——纯净、精准、运动型尼古丁。

In a world of artificial highs and harsh stimulants, there is something different, something clean, something precise, athletic nicotine.

不是便利店柜台后那些原始产品，也不是那种让人焦躁不安的强烈刺激。

Not the primitive products found behind convenience store counters, Not the aggressive buzz that leaves you jittery.

而是一种精心调配的纯净能量与专注清晰的结合。

But a careful calibration of clean energy and focused clarity.

运动型尼古丁，市面上最低剂量的无烟草尼古丁产品，完全在美国制造。

Athletic Nicotine, the lowest dose tobacco free nicotine available, made entirely in The USA.

不含人工甜味剂，只有纯粹而明确的提升。

No artificial sweetener, just pure purposeful elevation.

运动型尼古丁是一种表现型促智剂。

Athletic nicotine is a performance nootropic.

运动尼古丁是一种转变心态的工具。

Athletic nicotine is a tool for shifting mindsets.

运动尼古丁是追求卓越的伙伴。

Athletic nicotine is a partner in pursuit of excellence.

缓释、低剂量、渐进提升、持久能量、平稳过渡、激发成果。

Slow release, low dose, gradual lift, sustained energy, soft landing, inspired results.

运动尼古丁，更多专注。

Athletic nicotine, More focus.

更少杂念。

Less static.

运动尼古丁。

Athletic Nicotine.

更多清晰。

More clarity.

更少干扰。

Less noise.

Athletic Nicotine.

Athletic Nicotine.

更精准。

More accuracy.

更少焦虑。

Less anxiety.

Athletic Nicotine.

Athletic Nicotine.

从不断突破极限的顶级运动员，到追求愿景的艺术家，Athletic Nicotine 提供了你一直在寻找的提升。

From top athletes pushing their limits to artists pursuing their vision, athletic nicotine offers the lift you've been looking for.

了解更多，请访问 athleticnicotine.com/tetra，体验 Athletic Nicotine 带来的进阶表现。

Learn more at athleticnicotine.com/tetra and experience next level performance with Athletic Nicotine.

警告。

Warning.

本产品含有尼古丁。

This product contains nicotine.

尼古丁是一种成瘾性化学物质。

Nicotine is an addictive chemical.

告诉我关于Sora的事情。

Tell me about Sora.

Sora是技术树上的另一个分支。

Sora is a different branch of the tech tree.

有趣的是，人类并不生成视频，对吧？

And the thing that's interesting is that humans do not generate videos, right?

我们并不直接输出像素。

We do not directly output pixels.

我们可以创建工具来实现这一点，但我们没有像Sora那样有效的东西。

We can create tools to do it, but we don't have something that works quite at Sora.

但也许在我们的想象中，我们确实能做到，对吧？

But maybe we do in our imagination, right?

所以，也许这其中有一些与人类智能紧密相关的东西。

So maybe there is something that is deeply tied to human intelligence there.

而且这当中肯定有一件非常重要的事，那就是世界模型。

And there's definitely something very important, which is a world model.

我认为关于GPT文本模型是否具备世界模型，存在很多争论。

And I think that there are lots of debates about do the GPT text models have a world model?

如果你问它们一些问题，比如，有一个球在桌子上，它滚下了桌子，现在它在哪里？

And if you ask them things about, well, there's a ball on the table and it rolls off the table, where is it now?

它们能够回答出来。

They can answer it.

所以我认为，其中包含的世界知识比你预期的要多得多。

So I think there's much more world knowledge in there than you'd expect.

但对于Sora来说，这一点却显得格外突出，对吧？

But for Sora, it's so front and center, right?

它具备对物理规律的理解，当你看到它犯错时，比如掉下一部iPhone，如果它没有在正确的时刻碎裂，早期版本的Sora会完全搞错，这会彻底破坏体验。

That it has an understanding of physics and you, you really feel it when it makes a mistake, like dropping an iPhone and that if it doesn't shatter at the right moment, early versions of Sora would get that totally wrong and it just breaks the experience entirely.

但在我看来，Sora所代表的是创造力的潜力。

But I think that what Sora represents in my mind is creative potential.

去年我们发布ImageGen时，人们开始把他们自己和家人的照片转化为吉卜力风格的画像，这一点我们看得非常清楚。

And we saw very much with last year when we released our ImageGen and people started to turn photos of themselves and their family into like Studio Ghibli portraits.

当时最让我感兴趣的是，人们并不真的只想生成那些与现实毫无关联、缺乏人性的图像。

And it was this moment where the thing that I found most interesting was that people didn't really want to just generate images that were not connected to reality in some way, where there's no humanity in it.

但只要加入了一点人性的元素——比如是你和家人的照片，而不是一些随意的人物——人们就会立刻产生共鸣，这变得真正重要。

But as soon as there was an element of humanity, it's a picture of you and your family, it's not just some arbitrary people, suddenly people connect with it and it really matters.

因此，我认为我们仍处于这些模型未来潜力的非常早期阶段。

And so I think we're still at the very early days of what these models will be capable of.

但对于需要世界模型的应用，比如机器人技术，或物理世界中的其他任何领域，拥有这样的模型将非常有帮助。

But I think for applications that require a world model, you think about robotics, you think about anything else in the physical world, having such models can be extremely helpful for that.

它是如何学会物理世界的？

How did it learn the physical world?

和

Same way that

我们的文本模型学习方式一样。

our text models learned.

它的学习方式是通过观察视频，对吧？

The way that it learns is through observing video, right?

通过观察世界并尝试预测接下来会发生什么。

Through observing the world and trying to predict what will happen next.

你认为人工智能有神秘的一面吗？

Do you see a mystical side to AI?

是的。

I do.

跟我讲讲吧。

Tell me about it.

这里有一个关于扩展定律的概念，这是一种经验性观察：当我们向模型投入更多计算资源时，模型规模会以某种方式增长，这些参数、数据集大小之间存在某种关系，模型的表现会遵循一条确定的曲线，并且可以非常精确地进行预测。

Well, there's this idea of scaling laws, which are an empirical observation of as we pour more compute into models that their size has increased certain ways, you get the relationship between all these parameters, the size of the data sets, that there's some deterministic curve of how good those models are, and you can project it out very precisely.

关于扩展定律，有一些现象，你可以在公开论坛上看到人们说‘扩展定律已经失效了’，但他们是错的。

And there's something about the scaling laws, and you can go on public forums and see people saying, Oh, scaling laws are dead, but they're wrong.

他们完全错了。

They're absolutely wrong.

规模定律仍在持续发展。

Scaling laws continue unabated.

更困难的是实际实现这一点的工程挑战。

The thing that's harder is the engineering to actually realize that.

这非常困难。

That is difficult.

但在我看来，我们正在做出一些根本性的科学发现。

But to me, there is some fundamental scientific discovery that we are making.

这是对未知领域的探索。

This is an exploration into the unknown.

这是一片前沿领域。

This is a frontier.

你可以看到这一点，因为我们正在引入这些深刻的科学原理和数学方法，并看到了实证结果。

And you can see it because we're bringing these like deep scientific principles and mathematical approaches, and you're seeing the empirical results.

因此，这几乎就像我们的首席科学家雅各布喜欢描述的那样：构建更大的模型就像建造更大的火箭，其中的容差会变得越来越严格。

And so it's almost like the way that Jakob, our chief scientist, likes to describe it, is that building a bigger model is like building a bigger rocket, where the tolerance will all become more tight.

但你仍然受到同样的火箭方程、同样的引力以及其他这些因素的限制。

And you're still constrained by the same rocket equation and the same gravitational pull and all those things.

但我觉得，随着我们不断推进，我们不仅对模型有了更多了解，也对我们自己有了更多认识。

But I think that what we're doing as we keep going is that we learn more, not just about the models, but we learn more about ourselves.

我们更深入地理解了什么是智能。

We learn more about what intelligence is.

我认为这一点非常深刻。

And I think that that is something that is just very profound.

现在OpenAI有多少员工？

How many employees are there at OpenAI now?

我们大约有5000人。

We're about 5,000 people.

他们主要从事什么工作？

And what do they do?

做很多不同的事情。

A lot of different things.

所以我们大约有一千到一千五百人从事研究，再加上我所在的部门，叫做扩展团队。

So we have about maybe a thousand, maybe 1,500 people that are in research and then my organization, which is called scaling.

因此，这一部分团队正在全力推动深度学习的发展，探索其极限，将技术应用于芯片，建设数据中心，构建整个系统。

And so together, that part of the organization is really pushing forward what deep learning can do, trying to explore its limits, engineering into silicon, building the data centers, that whole apparatus.

我们还有大约五百人，或者再增加一千人，专注于部署工作。

We have a number of people, maybe it's 500 people, maybe another thousand who are focused on deployment.

这部分工作涉及应用，真正将像ChatGPT和Codex这样的产品落地实现。

So this is on applications and actually bringing products like ChatGPT, Codex to life.

所有这些领域之间都有着紧密的合作。

And there's deep partnership across all of these.

我们无法独自完成这一切。

Like we cannot do this alone.

这真正需要从研发、部署到从已部署产品中学习的全流程协作。

It really requires the end to end of building, deploying and learning from what we've deployed.

这正是我们取得进展的方式。

That is really how we make progress.

还有人负责销售、财务、公关、法律等众多职能，所有这些职能共同协作，使整个体系得以运转。

And then there's people who work on sales, finance, communications, legal, really a whole host of functions that all come together to make this apparatus work.

你是如何学会领导这样一家大公司的？

How did you learn to lead such a big company?

这对我来说是全新的。

It is new to me.

这确实是全新的。

It is new.

而且这也不是我原本打算去做的事情。

And it's not what I set out to do either.

是的。

Yeah.

在邓巴数字之内，与每个人保持紧密联系，了解每一个人，这曾吸引了很多

Being under Dunbar's number and having a tight connection across everyone, knowing everyone, like that was something that appealed to many of

我们。

us.

是的。

Yeah.

我认为发生的情况是，这个使命非常宏大，对吧？

And I think that what's happened is that the mission is big, right?

要真正实现一种能惠及每个人的通用人工智能，仅靠一小部分职能是不够的。

To really be able to deliver an AGI that's going to benefit everyone, you don't just do it with a small set of functions.

你真的需要思考如何打造一家大型企业？

You really need to figure out how do you build a large business?

你如何实现大规模的收入？

How do you get large revenue?

因为你需要这些资金来支持算力、筹集资本，并支付员工薪酬，尤其是在我们所处的这种极其竞争激烈的市场中。

Because you need that to fund compute and to be able to raise the capital and to be able to pay the employees, especially in like the insanely competitive market that we are in.

因此，我认为这种必要性迫使我去适应。

And so I think that the necessity of it has forced me to then adapt.

在OpenAI的每一个阶段，我的运作方式都让我不断学习、改变和成长。

And at every single stage of OpenAI, the way that I operate, I've learned and I've changed and I've grown.

我觉得，我今天的工作方式中，有很多地方让我回看去年的自己、五年前的自己，都会觉得：我那时候什么都不知道，真的什么都不知道。

And I think that there's so much of how I work today that I look back at last year me, I look back at five year ago me and I'm like, I knew nothing, like really nothing.

所以我认为，这完全是通过持续迭代实现的，我始终回到最初的状态，保持谦逊，不认为自己什么都知道。

And so I think that it's just through constant iteration and constant, I go back to that very beginning of having the humility to not believe that I know everything.

我觉得，我很多做事方式就是不断提问，努力厘清方向，真正理解我们为什么要做这件事？

And I think that I, a lot of how I operate is I ask a lot of questions and try to drive clarity and try to really understand why are we doing this?

这合理吗？

Does this make sense?

这些部分是怎么关联起来的？

How does this fit together?

谁在负责这件事？

Who are the people who are working on this?

然后深入这些细节。

And get into those details.

在人工智能的所有领域中，你是如何决定把时间花在哪些事情上的？

Of all the things to focus on in AI, how do you decide where your time goes?

这始终是一个非常棘手的问题。

This is always a very tough question.

我认为这需要直觉与判断的结合。

And I think there's a combination of intuition.

多年来，我成功做到的一点是，能够感知到何时我们取得了突破，或看到突破的迹象，从而意识到这是一个重要的方向，我们必须全力投入。

And I think that what I have successfully done over the years is kind of have a sense of when we have a breakthrough or the signs of life on a breakthrough, that this is the significant direction and that we really need to focus on this.

有时这意味着我要亲自投入，有时则意味着我们需要让整个公司聚焦于此，而我自己则转向另一个问题。

And sometimes it means I focus on it personally, and sometimes it means it's just, we really need to get the company on it, but I am dedicated to a different problem.

但这就是我的工作方式。

But that is how I operate.

因此，目前我投入了大量精力在数据中心和机器学习工程上。

And so right now I'm spending a lot of my effort on the data centers, the ML engineering.

当然，我有优秀的领导者和工程师帮助我，正是他们让这一切成为可能。

Again, I have great leaders helping me with this and great engineers who make it all possible.

同时，我也在协助Codex项目，花大量时间思考未来正确的方向是什么。

And I'm also helping out with Codex and spending a lot of time really trying to think about what is the right future shape.

比如，哪些产品是人们尚未意识到，但在未来六到十二个月内将成为每个人都在使用的主流产品？我们要确保以最快、最好的方式朝这个方向推进。

Like what's the product that people have not yet realized is going to be the product everyone's going to be using in six months, twelve months, and make sure we're running at that as fast as we can and as well as we can.

而这通常需要调动公司各个部门的人力。

And that's usually marshaling people from across the company.

但很大程度上，这取决于我如何看待模型的现状与未来走向，并将其与我们想要交付的价值相匹配。

But a lot of it is driven by where I see the models are, where they're going and mapping it to where is the value that we're trying to deliver.

你会说，每隔几个月或每年，你专注的方向都会与之前完全不同吗？

Would you say every few months or every year you're focused on something completely different than what you were before?

是的。

Yes.

你喜欢这样吗？

Do you like that?

这有趣吗？

Is that fun?

我喜欢。

I do.

是的。

Yes.

这从来不会无聊。

It's never boring.

你如何看待公司五年后和十年后的发展？

Where do you see the company in five years and then in ten years?

在这个领域，这些时间线很难预测，因为我认为世界在那段时间内将发生巨大变化。

Well, in this field, those timelines are so hard to project because I think the world is going to be massively different in that time.

我一年前就开始说，2025年、2026年和2027年将是变革性的时刻。

I started saying this a year ago, that 2025, 2026, 2027 are going to be these transformative moments.

我认为这些将是关键的时刻。

And I think they're going to be these critical moments.

而且我真的能看出来。

And I think you can really see it.

今年，我们将为每一个知识工作职能配备智能代理。

This year, we're going to have agents for knowledge work for every single function.

而且

And

这是新的。

that's new.

这仅仅发生在过去一年。

That's just been in the past year.

这也很新。

That's just new.

我对这个预测有点提前了。

I was a little bit early on this prediction.

我去年以为会是代理元年，但显然今年才是。

I thought last year would be the year of agents, but it is very clearly this year.

公平地说，我认为12月基本上成为了代理的转折点。

And to be fair, I think December basically turned into the agent moment.

我们就在这一刻。

Here we are.

第二件事是科学发现。

The second thing is scientific discovery.

我认为我们将开始看到科学发现以真正的方式发生革命。

I think we're going to start seeing scientific discovery revolutionize in a real way.

我认为会有一些早期采用者组织，能够预见这股浪潮并围绕它进行全面调整。

And I think that are going to be organizations that are early adopters that see this wave coming and really retool around it.

我们正在OpenAI所做的就是围绕这一点进行调整。

And what we're doing with OpenAI is we're retooling around this.

我们希望成为最前沿的AI组织，不仅因为我们认为这对我们的运营有帮助，更因为我们希望弄清楚如何将这项技术带给每个人？

We want to be the most AI forward organization, both because we think it's helpful for our operation, but because we want to figure out how do we bring this technology to everyone?

比如，你如何用这些如此强大的工具来构建一家公司？

Like how do you build a company with all of these tools that are so accelerative?

我认为到2027年，这些工具将开始被广泛普及。

And I think that what we're going to see in 2027 is these tools start to be diffused very, very widely.

很有趣的是，考虑到公司形态将会发生变化，因为现在创建公司变得如此容易。

And it's very interesting to think about that the shape of what a company is is going to change because it's so easy to create companies now.