OpenAI 的诞生：从晚餐到使命

那么，OpenAI 是怎么来的？

So, how did OpenAI come about?

嗯，我知道我想创业，因为我觉得那是……但你当时就在创业公司里。Stripe 就是一家创业公司。没错，但我从未觉得 Stripe 解决的问题是我自己的问题。这不是我从小思考的问题。它是一个重要的问题，我为此奉献了几年，但我觉得无论有没有我，它都会成功。于是，我第一次真正思考：什么是我想奉献一生的使命？哪怕只是让这个问题以稍微好一点的方式展开，我也愿意花一辈子去研究。对我来说，最明确的就是 AI。如果你能真正影响 AI 在世界上如何发展，那这一生就值了。

Well, I knew I wanted to do a startup because I felt like that was something... But you were just in a startup. Stripe was a startup. It's true, but I never felt like Stripe's problem was my problem. It wasn't the problem I'd grown up thinking about. It was an important problem and I dedicated myself to that mission for a number of years, but I felt like it was going to succeed with or without me. So then I had a first moment to really think about what is a mission that I want to dedicate myself to, where I would spend the rest of my life working on this problem just to see it play out in a slightly better way. And it was very clear to me that top of the list was AI. If you could actually make a difference in how AI will play out in the world, that would be a life well lived.

当你考虑离开时，Patrick 让你去找 Sam Altman 谈谈。那次谈话发生了什么？

When you were thinking about leaving, Patrick told you to go talk to Sam Altman. What happened in that conversation?

嗯，Patrick 说 Sam 见过很多像你这样的年轻人，而且我觉得 Patrick 真的希望 Sam 能说服我留下。和 Sam 聊了几分钟后，他说：‘好吧，你显然已经决定了。这很明显。’于是他问：‘那你接下来打算做什么？’我说：‘我在考虑创办一家 AI 公司。’他说：‘我也在考虑做 AI 相关的事情。我们保持联系。’所以我在离开 Stripe 后又和 Sam 聊了一次。他问：‘你还在考虑做 AI 吗？’我说：‘是的。’他说：‘我也开始有更多细节了，正在筹备七月份的一个晚宴。’我飞去参加了那个晚宴。我记得一个话题是：‘现在成立一个拥有众多顶尖研究人员的实验室是不是太晚了？这有可能吗？’

Well, Patrick had said Sam has seen lots of young people in your situation and Patrick, I think, really hoped that Sam would convince me to stay. A few minutes of talking to Sam, he's like, 'Okay, you clearly have already decided. It is very obvious.' So he asked, 'Well, what are you planning on doing next?' And I said, 'Well, I'm thinking about doing an AI company.' And he said, 'I'm also thinking about doing something in AI. We should keep in touch.' So I talked to Sam maybe one more time after I was leaving Stripe. He asked, 'Are you still thinking about doing something in AI?' I said, 'Yes.' He said, 'I'm also starting to get more details and putting together this dinner in July.' I flew out for the dinner. The thing I remember was a topic: 'Is it too late to start a lab with many of the best researchers? Is it possible?'

这是哪一年？

And this is what year?

2015 年。因为你想，DeepMind 拥有所有研究人员、所有资本、所有数据，感觉还能不能起步？人们提出了各种理由说很难。但没有人能提出一个理由说这实际上不可能。所以那天晚上 Sam 和我开车回城里，我记得我们互相看着说，我们必须做这件事。就像我们必须做。所以第二天我就全职投入这件事了。这很艰难，因为定义非常模糊。我们有一个使命，一个愿景：我们认为可以构建人类级别的 AI，让它对世界产生积极影响，让利益广泛分布。但怎么做？怎么让人们真的辞掉工作来加入？最初，我缩小到的人选实际上是 Ilya、Dario Amodei、Chris Olah 和我自己。这将是团队。我们花了很多时间在一起，讨论实验室的潜在愿景和可能的运作方式。但并没有完全成型，部分原因是它是否有足够的动力？Dario 觉得他需要去建立自己的名声，他不确定这是否真的能成。这是一个关于一切如何运作的问题。与此同时，我开始让 John Schulman 感兴趣。他说他会做。Dario 和 Chris 最终决定去 Google Brain。所以实际上只有 Ilya、我，还有 John 开始，可能还有几个其他人。所以我有一组大约 10 个人，其中很多人说，我有兴趣，但还有谁加入？我问 Sam，好吧，我们怎么打破这种对称？怎么让每个人都同意加入？Sam 的建议是，邀请大家出去参加一个场外会议。所以我们在 Napa 安排了一个活动，我还做了 T 恤。当时我们正要……这是在大家加入之前。没有正式 offer，没有人加入，我们没有结构，什么都没有。我们只有一个想法，一个愿景，一个使命。我们把人飞过去，一起开车到 Napa，那是很棒的一天。想法涌现。我们提出了几乎可以说是过去 10 年我们一直追求的技术计划。第一，解决强化学习。第二，解决无监督学习。第三，逐步学习更复杂的东西。那次场外会议之后，我向每个人发了 offer，说：‘嘿，我们想在未来两到三周内开始。请告诉我你是否加入。’

2015. Because you think about just the degree to which DeepMind had all the researchers, all the capital, all the data, it just felt like is it even possible to get something off the ground still? People came up with all sorts of reasons it was hard. No one could come up with a reason it was actually impossible. So Sam and I driving back to the city that night, I remember we looked at each other and we said, we got to do this. Like we just have to. So next day I was full-time on putting this together. And it was tough because it was very ill-defined. We had a mission, a vision of saying, we think that we can build human-level AI, make it be something positive for the world, make the benefits be something that are distributed broadly. But how? And how do you get people to actually leave their jobs to come and join this thing? Initially, the set of people that I narrowed down to were actually Ilya, Dario Amodei, Chris Olah, and myself. That was going to be the team. And we spent a lot of time together, we spent a lot of time talking about potential visions for the lab, potential ways that things would work. It didn't quite come together, and there was just partly a question of will this have enough momentum? Dario felt like he needed to go and establish a name for himself, and he wasn't sure if this was really going to be it. It's a question of just how it was all going to work. Meanwhile, I was starting to get John Schulman interested. He said that he was going to do it. Dario and Chris ended up deciding to go to Google Brain. So it was really just Ilya, me, and John starting to be maybe a few others. So I had a group of about 10 people that many of them were saying, I'm interested, but who else is in? I asked Sam, okay, how do we break symmetry here? How do we actually get everyone to say, all right, we're joining? And Sam's suggestion was, invite people out for an offsite. So we set up a thing in Napa, and I actually made t-shirts. At the time we were going... And this is before they had joined. There were no official offers, no one had joined, we didn't have a structure, we had nothing. We just had an idea, we had a vision, we had a mission. And we flew people out, we drove up to Napa together, and it was an amazing day. The ideas were flowing. We came up with what I would really say is almost the technical plan that we've pursued for the past 10 years. Number one, solve reinforcement learning. Number two, solve unsupervised learning. And number three was gradually learn more complicated, in quotes, things. After that offsite, I sent offers to everyone and said, 'Hey, we want to get started in the next two to three weeks. Please let me know if you're in.'

你为什么认为 DeepMind 有如此不可逾越的优势？

Why did you think that DeepMind had such an insurmountable advantage?

确实，Google DeepMind 是这个领域的巨无霸。他们拥有大量资本。他们有成功的记录。这是在 AlphaGo 之前，对吧？AlphaGo 几个月后问世，但这并不意外。势头非常明显。所以是否真的有可能建立独立的新东西，这并不明显。

It was very much the case that Google DeepMind was the 10,000-pound gorilla in the field. They just had lots of capital. They had the track record. This is before AlphaGo, right? AlphaGo came out a couple months later, but it wasn't a surprise. It's like very much the momentum was very clearly there. So the question of is it really possible to build something independent and new, it wasn't obvious.

你什么时候意识到这个非营利模式行不通？

At what point did you realize that this nonprofit thing just wasn't going to work?

2017 年，我们开始认真思考：首先，我们如何真正实现使命？如何真正构建 AGI？它会是什么样子？我们开始计算算力，意识到需要大型计算机。我们遇到了一家叫 Cerebras 的公司，他们正在制造独特的计算硬件。他们承诺的那种计算机，我们意识到将远远超出我们之前的算力计算。你开始意识到，如果我们能买很多这样的计算机，我们可能真的能成功构建 AGI。如果我们能获得 Cerebras 的独家使用权，那将给我们带来压倒性的优势。如果我们能购买非常大的数据中心，那也会是独特的。关于非营利筹款，我认为基本上有一个上限。所以 Elon、Sam、Ilya 和我都同意，OpenAI 前进的唯一道路，实现使命的唯一道路，就是创建一个与 OpenAI 相关的某种形式的营利实体。所以我们致力于这个方向，我们知道这是实现使命的唯一途径。

In 2017, we started to think very hard about, first of all, how do we really achieve the mission? How do we actually build an AGI? What will that look like? And we started to do the math on compute, and you start to realize that it's going to take big computers. And we came across a company called Cerebras, which was building a unique piece of computing hardware. And the kind of computer that they were promising, we realized was going to be far advanced of where our compute calculations looked. And you start to realize if we could buy a lot of those computers, we could actually probably succeed at building an AGI. If we could get exclusive access to Cerebras, that could give us an overwhelming advantage. If we could buy very large data centers, that could be something unique as well. And the thing about nonprofit fundraising is I think that there is essentially a cap to what is possible there. So Elon, Sam, Ilya, and I all agreed that the only path forward for OpenAI, the only path to achieve the mission, was to create a for-profit entity associated with OpenAI of some form. So we were committed to that direction, and that is something that we knew was the only way to achieve the mission.

你什么时候意识到一切都会改变？是 DOTA 的时候，还是之前或之后？

When was the moment that you realized everything was going to change for you? Was that DOTA, or was it before then or after?

OpenAI 的运作方式是一系列时刻，你意识到它现在是真实的了。每次你以为你理解了，它真的稳定下来了，你就会意识到有一个你尚未欣赏的新地平线。

The way that OpenAI works is it's a series of moments where you realize that it's real now. And every time you think that you understand it, that it is really settled in for you, you realize that there is a new horizon you had not yet appreciated.

一路走来，我觉得最初是团队刚成立的时候。那种感觉是，‘哇，我们真的组了一个团队。现在可以追求这个使命了。’但第二天到办公室，你就会想，‘那我们要做什么呢？’对吧？我们连白板都没有。你知道吗，Ilya 和 John 想在白板上写点东西。我就说，‘我去弄块白板。这个我能做。’DOTA 是我们第一个重大成果。那真的让人感觉，‘哇，当我们集中精力时，确实能做成一些事。’你能看到所有这些算力汇聚在一起。你扩大算力，结果也随之扩大。GPT 系列有很多这样的时刻，我记得早期有一个无监督情感神经元论文。你听说过吗？

And so, along the way, I think that there was the initial launch. It was like, 'Wow, we actually got a team together. Now we can pursue this mission.' But you show up at the office the next day, and you're like, 'Well, what do we do?' Right? We didn't even have a whiteboard. Do you know, Ilya and John wanted to write something on a whiteboard. I was like, 'I will get a whiteboard. That's something I can do.' DOTA, we had our first big result. Right? That really was like, 'Wow, we can actually accomplish something when we put our mind to it.' You can actually see all this compute coming together. You scale up the compute, you scale up the result. There were multiple moments with the GPT series, and I remember actually an early moment was the unsupervised sentiment neuron paper. Have you heard that one?

我听说过，但没读过。

I've heard of it, but I haven't read it.

好的。是的，那篇论文很有意思，因为是 2017 年，那是我们第一次看到语义从语言建模目标的训练中涌现出来。你训练模型学习下一个字符，预测下一个字符，然后突然得到一个能理解情感、判断正面还是负面的神经网络。这比听起来难得多。但那一刻你意识到，‘哇，我们在建造能学习语义的机器，不只是逗号在哪、名词动词在哪，而是真正理解句子的含义。’你必须推动这一点。然后，当然，当你看到像 GPT-4 这样的东西时，我记得我们在玩它，有人问，‘为什么这个东西不是 AGI？’对吧？实际上很难说清楚，因为你可以和它流利地谈论任何话题。它显然不是 AGI。它缺少一些东西，但如果你在两个月前描述 AGI 的标准，可能和 GPT-4 的表现不匹配。所以一路上有很多时刻让你觉得，它现在成真了。它真的要发生了。经济将转变为一个由算力驱动的世界。我认为这些时刻还没有结束。我们还有更多突破性的时刻，让你意识到下一阶段是可能的。

Okay. Yes, that one's an interesting one because it's 2017, and it's really the first time that we saw semantics arise from training on a language modeling objective. So, you train on learn the next character, predict the next character, and then suddenly you get a neural net that understands sentiment, understands if something is positive or negative. Much harder than it sounds. But that was a moment where you realize, 'Wow, we are building machines that can learn semantics, not just where the commas are and where the nouns and verbs are, but it can really learn the meaning of sentences.' You got to push that. And then, of course, when you see something like a I remember we were playing with it and someone asked, 'Why is this thing not an AGI?' Right? It's like actually really hard to put your finger on it because you can talk to it fluently in anything you want. It clearly wasn't an AGI. It was lacking something, but just if you'd describe your criteria for AGI 2 months prior, it probably wouldn't be compatible with what GPT-4 was. And so there are many moments along the way where you feel like it's real now. It's going to really happen. The economy is going to transform into this compute-powered world. And I think that those moments are not yet at the end. I think that we have many more breakthrough moments where you realize that the next stage is possible.

我觉得 Dota 是一个不可思议的时刻，因为它不像 Deep Blue 下棋，也不像 AlphaGo，那些虽然计算密集但规则明确。Dota 实际上是与人互动，世界有一定结构，但你有很大的自由度。

I thought Dota was like an incredible moment because it wasn't chess like Deep Blue and it wasn't AlphaGo, which is computationally intensive but very defined rules. It was actually interactive against humans in a way that like the world is sort of structured, but you have all of this freedom.

是的，这非常引人注目。讽刺的是，我们最初做 Dota 是为了开发新方法，因为当时的强化学习显然无法扩展。对吧？我们用的算法叫 PPO。它在每个时间步都做规划，没有层次结构。而人类不是这样规划一天的。所以我们知道这个算法有严重缺陷，永远无法扩展，有各种问题。但你得从某个地方开始。你得把基线推到极限，这样你才能看到现有方法的极限，然后才能引入新算法。我们只是不断扩展 PPO，结果超过了最优秀的人类玩家。这本身就是发现，对吧？实际上，大量算力加上简单算法。对吧？这不仅是理论上可行，实践中也有效。我们真的能做到。在这个极其混乱的环境中，你无法编程，无法提前看，无法搜索，你只需要近乎人类的直觉。顺便说一句，我们用的神经网络非常小，像昆虫大脑，突触数量与真正的昆虫大脑相似。然后你意识到，等等，如果同样的计算方法，但扩展到人类大脑规模，会是什么样？这是一个非常发人深省的问题。

Yeah, that was something very compelling about it. And the ironic thing is we'd actually set out with Dota to develop new methods because the reinforcement learning at the time was clearly not going to scale. Right? That the algorithm we use is called PPO. You plan over every single time step. There's no hierarchy. Whereas a human, that's not how you plan your day. And so we knew that this algorithm was incredibly flawed, would never scale, and had all these problems. But you got to start somewhere. You got to push your baselines to reach the wall so you actually see the limits of what good looks like with what you have and then you can bring to bear a new algorithm. And we just kept scaling PPO and we exceeded the performance of the best humans. And that itself was the finding, right? That actually massive compute with simple algorithms. Right? That that is something where we can not just doesn't just work in theory. It works in practice. We can really make it happen. And in this incredibly messy environment where you cannot program it, you cannot look ahead, you cannot do a search, you just need this almost human-like intuition. And by the way, the neuron that we used, tiny tiny little insect brain, similar number of synapses as to truly an insect brain. And you realize like, wait, what if you had the same computational approach, but scaled it up to something that's much more human brain scale, what would that be like? Very very evocative question.

推理和预测有区别吗？你提到预测下一个字符、下一个词，与从第一原理推理，这两者有什么不同？

Is there a difference between reasoning and predicting? You mentioned sort of like predicting the next character, predicting the next word versus actually reasoning in first principles.

我认为它们在深层上是相连的。一方面，仅仅预测下一个词听起来像是一项平凡的任务。但如果你真的能预测爱因斯坦的下一句话，你至少和爱因斯坦一样聪明。你可以反驳说，哦，但是，我觉得这些反驳站不住脚，因为预测的关键不在于预测已知的东西。关键在于你把自己放在一个从未见过的新情境中，预测接下来会发生什么。我认为智能和预测之间有深刻的联系，学术文献中有很多讨论，压缩等等，它们都是同一回事。现在，这些推理模型，我觉得非常有趣的是，我们用强化学习来训练它们。所以，这又回到了 OpenAI 最初的计划，有两个步骤。第一步是无监督学习，你通过让模型预测下一个词来训练它，这里的数据更静态，更偏向观察。同样，数据是它从未见过的，情境也是从未见过的，但情境已经发生了。然后你做强化学习，基本上让 AI 学习自己的数据，对吧？你让它自己做出行动，从世界获得观察，并从中学习。训练的方式仍然是预测。它试图预测如果我采取这个行动，可能会发生什么。然后根据表现好坏进行强化。美妙之处在于，这个 AI 现在有了背景知识和真实世界经验。但本质上，我们在无监督阶段和强化阶段使用的训练技术是完全相同的。你只是在预测，但改变了数据的结构。

I think they're connected in a deep way. So, on the one hand, just predicting what comes next sounds like a pedestrian task. But if you really can predict the next word out of Einstein's mouth, you are at least as smart as Einstein. And you can make arguments, oh, well, like, you know, it's but I think that those arguments fall flat, that there's something there's something false there because the point of prediction is not about being able to predict what is known. The point is you put yourself in a new situation you've never seen before and predict what comes next. And I think that there's something deeply connected to intelligence and prediction that there's a long story of academic literature and how you think about this, compression, they're all kind of part of the same thing. Now, these reasoning models, the thing that I think is very interesting is that we train them with reinforcement learning. And so, there's really back to the original OpenAI plan, there's two steps to it. The first is unsupervised learning, you train a model just by having it predict what comes next, and there it's much more static data, it's much more observational. Again, it's data it's never seen before, situations never seen before, but it is a situation that is already happened. Then you do reinforcement learning, which is you basically have the AI learn its own data, right? You have it make its own, here's the action I'm going to take, you get an observation from the world, and you learn from that. And it's again, the way you actually train it is still predicting. It's trying to predict if I take this action, what's the thing that's likely to happen. And you reinforce that depending on how good of a job you did. And the beauty of that is that it now is an AI that has this background knowledge and has real world experience. But fundamentally, the technology that we used to train during unsupervised stage and during the reinforcing stage, they're exactly the same. You are just predicting, but you've changed the structure of the data.

事情什么时候开始变得紧张的？

When did things start to get tense?

我认为 OpenAI 的特点是，如果你真的相信这个使命，真的相信创造人类智能水平机器的可能性，那就意味着赌注总是很高。谁在做决定、这些决定背后有什么价值观，这些在普通公司里可能很平常、更像是办公室政治的问题，开始具有了存在主义的分量。

I think the thing about OpenAI is that if you truly believe in the mission, if you truly believe in the possibility of creating machines that have the intelligence level of humans, it means the stakes always feel very high. The question of who's making the decision, the question of what are the values that go into those decisions, the question of these things that are maybe mundane in a typical company that are much more like office politics, start to take on this existential weight.

我认为这很大程度上影响了 OpenAI 这些更引人注目的冲突，你知道，有时候就像你把它放进去，甚至只是某个特定事物的归属问题，它突然就带上了这种存在的重量。

And I think that that has colored a lot of how OpenAI these more high-profile conflicts, you know, sometimes it's like you put it in like even just the question of who gets credit for a particular thing, it suddenly takes on this existential weight.

嗯，这正是我思考的地方，因为在那时你可能意识到这项技术是不可避免的，它将改变世界。而这一点当时并未被世界广泛知晓。然后我想象会有人想要站在最前沿，想要为此获得认可。

Well, that's where I was sort of thinking of it this because it's like at that point you probably realized this technology is inevitable and it's going to change the world. And that wasn't broadly known to the world. And then I would imagine there's people who like I want to be front and center. I want to take credit for this.

是的。这是我在这个领域观察到的最主要的动态。实际上不仅仅是 OpenAI。比如我早期的一个观察是，这项技术本质上是碎片化的，对吧？就像当你有很大压力时，你可能会得到钻石，也可能出现裂缝。你常常会看到钻石在局部形成，对吧？那些真正合作、高度信任、知道如何运作的团队。但有时你会看到他们分裂开来，各自为政。我认为在 AI 领域，我们从方法的多样性和不同群体的相互推动中确实获得了一些真正的好处，既是为了以更有益的方式带来这项技术，有时也是为了思考所有棘手的问题，比如安全、安全意味着什么、实际部署这项技术意味着什么，以及如何思考减轻风险并最大化收益。我认为这方面有很多非常健康的辩论。这些辩论一直在 OpenAI 内部进行。现在，我认为它开始真正在世界范围内发生。我认为这是我们整个社会真正受益的事情。

Yes. That is the overwhelming dynamic that I have observed in this field. It's not just about OpenAI actually. Like one observation I had early on is that this technology is by nature very fragmentary, right? That it's sometimes it, you know, like when you have a lot of pressure, you can get a diamond or you can get cracks. Often you'll see diamonds form in pockets, right? Teams of people that really work together, that have a lot of high trust, that know how to operate. But sometimes you can see that they splinter off and they kind of go their own way. And I think within AI, I think we've gotten some real benefits out of diversity of approach and different groups that are really pushing each other in order to both bring this technology in a more beneficial way, sometimes how to think about all the thorny questions around safety, around what does it mean to be safe, what does it mean to actually deploy this technology, and how to think about how to mitigate and but also how to maximize those benefits. And that's something where I think that there's a lot of very healthy debate. It's always gone on within OpenAI's walls. Now it's starting to really happen, I think, in the world. And I think that's something that we as a society really benefit from.

本期节目的赞助商是 CoinShares。当大多数行业还在争论数字资产是否合法时，CoinShares 已经在悄悄构建投资它们的适当基础设施。他们现在管理着超过 60 亿美元的资产，并在每个市场周期中都保持盈利。完全受监管，具备严肃投资者真正期望的透明度和治理水平。无论是加密货币 ETF、主动策略，还是比特币挖矿 ETF 敞口，你都可以通过现有的经纪账户访问所有这些。CoinShares。成年人已经到来。了解更多请访问 coinshares.com。这不是投资建议。

The sponsor of this show is CoinShares. While most of the industry was still arguing about whether digital assets were legitimate, CoinShares was quietly building the infrastructure to invest in them properly. They now manage over 6 billion in assets and have stayed profitable through every market cycle. Fully regulated with the kind of transparency and governance that serious investors actually expect. Whether it's crypto ETFs, active strategies, or Bitcoin mining ETF exposure, you can access all of it through your existing brokerage account. CoinShares. The adults have arrived. Learn more at coinshares.com. This is not investment advice.

你整天都在开会。你努力保持专注，但又担心会忘记决策、行动项和重要的下一步。这就是 Granola 的用武之地。Granola 是一款 AI 驱动的会议记事本。你像往常一样记下粗略的笔记，而 Granola 在后台转录，并在会议结束时将它们变成清晰、有用的笔记。没有机器人加入你的通话，没有干扰，只有一个干净的记事本帮助你集中注意力。在通话期间或之后，你可以与笔记聊天。你可以让 Granola 提取行动项、帮助你谈判、做决策、写跟进邮件等等。我甚至在听播客时也用它。一旦你在第一次会议上试用它，就很难再离开它。前往 granola.ai/shane，使用代码 Shane 即可免费获得 3 个月。网址是 granola.ai/shane。

You're in meetings all day. You're trying to stay present, but you're also worried you'll forget the decision, the action item, the important next step. That's where Granola comes in. Granola is an AI-powered notepad for meetings. You jot down rough notes like you always do, and in the background Granola transcribes and turns them into clear, useful notes when the meeting ends. There are no bots joining your calls, no distractions, just a clean notepad that helps you focus. During or after the call, you can chat with your notes. You can ask Granola to pull out action items, help you negotiate, make a decision, write a follow-up email, and so much more. I even use it when I'm listening to podcasts. Once you try it on a first meeting, it's hard to go without. Head to granola.ai/shane and get 3 months free with the code Shane. That's granola.ai/shane.

带我回到你发现 Sam 被解雇的那一刻。你在哪里？

Take me back to the moment you found out that Sam had been fired. Where were you?

我在家。

I was at home.

然后发生了什么？

And what happened?

我收到一条短信说：“我们能开个视频通话吗？”于是我就加入了视频通话。我注意到是董事会成员，除了 Sam，都在上面。

I got a text saying, "Can we hop on a video call?" So, I hopped on the video call. I noticed that it was the board minus Sam who were on there.

你当时知道了吗？

Did you know at that point?

不。我的意思是，我推断出有什么事发生了。

No. I mean, I inferred something was up.

但是，因为你是董事会成员。

But, cuz you're on the board.

我是董事会成员。我当时是董事会成员。

I am on the board. I was on the board.

在那个时候。

At this point.

是的。

Yes.

然后发生了什么？

And then what happened?

我被告知董事会决定将 Sam 免职。实际上，我收到的信息与公开帖子中的信息相同。我问是否可以获得更多信息，被告知不行，现在不行。对此感到沮丧，也许下次吧。再次被告知没有更多信息可分享。然后被告知：“等等，还有更多。”还有，我也被从董事会中移除，但会留在公司，因为我对公司和使命至关重要。我再次询问原因或反馈，被告知没有。最后，被告知：“嘿，在这个新安排中，你希望会开始得到反馈，在这个新配置中。”所以，这就是那次对话。

I was told that the board has decided that Sam would be removed. And effectively, the message that I got was the same messaging that was in the public post. And I asked if I could have any more information. I was told no, not right now. And depressed on that, maybe another time. And again, was told nothing more to share. And then was told, "Wait, there's more." Also, that I had been removed from the board, but would be staying with the company because I was very critical to the company and the mission. I said, again, I asked like any reasons, any feedback. Was told no. Towards the end, was told that, "Hey, like in this new setup that you will start to get hopefully, you can get feedback in this new configuration." And so, that was the conversation.

你当时在想什么？

What went through your mind?

这完全不对。

It just wasn't right.

是愤怒吗？

Was it anger?

不。我觉得我理解发生了什么。你多久之后才知道实际发生了什么导致了这一切？嗯，答案有两部分。一是我觉得我仍在了解一些额外的事实，一些别人脑子里的东西。在某种程度上，这归结为缺乏沟通，对吧？你意识到所有这些不同的事情已经缓冲了。在某种程度上，我大致知道。我想，我理解这里的每个人，我对他们为什么那样做有一个相当好的模型，但这在当时对我来说并不是最重要的。我只知道这不对。挂断电话后，我和妻子谈了谈，我说：“必须辞职。”她说：“我同意。”

No. I felt like I understood what had happened. How long before you knew what had actually transpired to sort of cause this? Well, there's two parts to the answer. One is I feel like I still am learning some additional facts, some additional thing that was in someone's head. To some extent it comes down to a lack of communication, right? That you realize that there are all these different things that have buffered. And to some extent, you know, approximately I kind of knew. I was like, I understand for every person here, I have a pretty good model of why they acted the way that they did, but it wasn't what was most important to me in the moment. I just knew that this wasn't right. Right after I hung up the call, I talked to my wife and I said, "Got to quit." And she said, "I agree."

你那天就辞职了。

And you quit that day.

是的。然后发生了什么？我辞职那天，开始收到所有这些人的消息，他们说：“我不知道你和 Sam 接下来要做什么，但我支持你。我想和你一起开创些新东西。”这真是一个诚实的惊喜。我真的没想到会得到那样的支持，那样的情感流露。有几个我亲密的合作者那天也辞职了。他们是 Yacob、Shimon、Alexander，我们五个人，加上 Sam，我们聚在一起，开始规划一家新公司可能的样子。我记得第一天感觉：“好吧，我们实际拿回公司的概率是 10%，10%。”第二天，我们在 Sam 家安排了一次会议。公司里的一群人来了，我们展示了我们一直在构思的愿景。所以仅仅一天之后，我们就有了这个关于如何运营项目的新图景，那个周末我们还花了很多时间与董事会和公司谈判，试图找出是否有合理的回归之路。那个周日晚上，董事会用新人取代了 Mira 的临时 CEO 职位，公司就反抗了。实际上我们当时在办公室，我们认为我们接近达成协议了，

Yes. And then what happened? That day when I quit, I started to get all these messages of people saying, "I don't know what you and Sam are doing next, but I'm with you. I want to go start something with you." Like just that was a real honest surprise. I didn't really expect to get that kind of support, that kind of outpouring. There were a few of my close collaborators who quit that day as well. That's Yacob, Shimon, Alexander, and the five of us, so those people plus Sam, we all got together and we started to chart out what a new company could look like. I remember feeling that first day like, "Okay, there's a 10% chance that we actually get the company back, 10%." The next day, we set up a meeting at Sam's house. A bunch of people from the company came by and we showed the vision that we'd been sketching out. So it was just really one day in, you know, with this fresh picture of how we'd run the project and we spent that a bunch of time over that weekend also negotiating with the board and the company and trying to figure out is there a path back together that makes sense. That Sunday night, the board replaced Mira as interim CEO with a new person and the company just rebelled. Like we'd actually been at the office, we thought we were close to a deal, and

回来。

to come back.

回来。

back.

是的，我们以为我们有路可走。然后董事会做出了那个改变，突然间所有人都涌出大楼，那真是混乱。我和许多有意加入这家新公司的人视频通话，告诉他们一切都会好起来，我们有计划，我们扩大了——我们一直在建造这个小救生筏，为我们预期会来的少数人准备。但突然间，没有人想和这个实体扯上关系，人们想为他们认为正确的事情挺身而出。Sam 和 Satya 谈了，我们之前也在讨论，嘿，你能当资助者吗？你能支持这个新事业吗？我说，嘿，实际上，我们能不能从小救生筏扩大到……所有人都说可以。我们能带走所有人吗？他们说，好吧，我们会想办法的。很多人，就在感恩节前，很多人本应飞回家，但他们取消了航班，办公室里挤满了人。每个人都在办公室，只是为了在那里，成为其中的一部分，即使他们无法参与这些对话，他们只是想在那里见证历史。然后请愿书开始流传。太多人同时试图签名，结果 Google Docs 崩溃了。所以你必须指定某些人作为你去找他们签名的对象，这样就不会有太多编辑者同时操作。我认为那是一个被响亮听到的声明。我记得，我大概凌晨 5 点回到家，睡觉，45 分钟后醒来，查看 Twitter，看到 Ilya 发了帖子，签了请愿书，说他希望公司重新团结起来。那是一个真正的解脱时刻。我充满感激，感觉就像，好吧，我们可以把它重新拼起来。我们可以回到正轨。

Yeah, we thought that we had a path. And then the board made that change, and then suddenly it was everyone streaming out of the building, and it was just like real chaos. I was on video calls with many of the people who had been interested in coming to this new company saying it's going to be okay, we have a plan, and we expanded, you know, we'd been building this little life raft, right, for the small set of people we expected to want to come. And suddenly it was like no one wanted to be associated with this entity, right? People wanted to stand up for what they viewed as right. Sam talked to Satya, who we'd been talking about, hey, could you be a funder? Could you help support this new endeavor? I was like, hey, actually, could we expand from the small life raft to like, and everyone said, yes. Could we take everyone? And they're like, all right, we'll figure it out somehow. And a lot of people, this is right before Thanksgiving, a lot of people were supposed to be flying home, wherever that is, and instead they canceled their flights, and the office was packed. I was like, everyone was at the office just to be there, be part of it, and just, you know, even if they couldn't contribute to any of these conversations, that they just wanted to be there as this history was made. Then this petition starts to circulate. So many people were trying to sign the petition at once, it actually crashed Google Docs. And so you had to have certain people who were designated as the person you go to to actually put your name on the document so you don't have too many editors at once. I think that that was a statement that was really heard loudly. And I remember, you know, I probably got home around like 5:00 a.m. or something, went to sleep, and I woke up like 45 minutes later, and I checked Twitter, and I saw that Ilya had posted, and had signed the petition, and it said that he wanted the company to come back together. And that was this real moment of relief. I felt so much gratitude that it just felt like okay, like we can put this back together. We can get back to a good track.

你和 Neale 一起建立了这一切。之后试图找回那段关系是什么感觉？

You and Neale had built this together. What was it like trying to find your way back to that relationship after?

听着，那很艰难。那绝对是一段非常亲密的关系。你曾是我民事仪式的司仪。我们一起经历过极其艰难的时刻。就像任何关系一样，总有起起落落。之后我们花了很多时间真正地谈事情，真正地试图理解并表达一些我们之间积累或未说出口的事情。我认为通过这个过程，我们达到了一个非常好的状态。对我来说，我觉得我们对发生的一切都有了了结。

Look, it was tough. It was definitely a very close relationship. You'd been the officiant at my civil ceremony. Right, we'd been through extremely tough times together. And like any relationship, you always have your ups and downs. And we spent a lot of time afterwards really talking things through. And really trying to understand and just articulate some of the things that we had let build up or had left unsaid between us. And I think that we got to through that process, I think we had gotten to a really good place. And for me, it was I felt like we got to closure on everything that had happened.

你对你所激发的所有忠诚有何感受？

How did you feel about all the loyalty you've inspired?

深深感激。真的，这从来不是我要求的东西，也从未预料到。我认为我的行事方式是一个身先士卒的领导者。试图带头冲锋。有时当我这样做时，我并不总是——抱歉，我有点激动。嗯，但我不总是回头看是否每个人都跟着。我就是直接冲进去。当人们真的来帮助建造这个东西时，我只会对他们充满感激，感觉他们在各方面都超出了我的期望。

Deeply grateful. Truly, it was never something that I would have asked for and something I never would have expected. I think the way that I operate is I'm very much a in the trenches kind of leader. Try to lead from the front. And sometimes when I do that, I don't always—sorry, I'm getting a little emotional. Um, but I don't always look back to see if everyone's following. I just like run right in. And when people do come and really help to build the thing, I just it makes me feel so grateful for them and to feel like they have exceeded my expectations in every way.

所以最终每个人都回来了。

And so eventually everybody comes back.

我告诉你，这并非必然，因为整个周末，所有竞争对手都在虎视眈眈。想象一下这种捕食狂潮。人们准备行动，人们收到 offer。而实际上，那个周末我们没有失去一个人。没有人接受竞争对手的 offer。

I'll tell you, it was not guaranteed because throughout that weekend, all the competitors were circling. Just imagine this like feeding frenzy. People were shaping up to do. People were getting offers. And we actually did not lose a single person through that weekend. No one accepted a competing offer.

我觉得这太不可思议了。

I think that's incredible.

确实如此。

It really was.

这更像是，你知道，Belichick 教练实际上告诉过我，当我们谈论最好的团队时，他说：'他们不是为了钱而战，而是为了身边的人。' 当你说所有这些人都辞职时，让我想到了这一点。而且他们中没有一个人为了更多的钱、更好的 offer 而离开。每个人都在试图挖角，

That's more, you know, Coach Belichick told me this actually, uh, when we were talking about the best teams, and he said, 'They're not playing for money, they're playing for the person beside them.' And when you were saying that all these people quit, it makes me think of that. Like, and none of them left for presumably more money, better offers. Everybody was trying to circle and poach, and

是的，那是一个非常珍贵的时刻。

Yeah, it was a very that was a diamond moment.

这一切发生后，你休息了一段时间。你内心发生了什么？

After all of this happened, you took some time off. What was going on internally with you?

那是一段激烈的经历，回来也是一段激烈的经历，老实说，在 OpenAI 对我来说最艰难的时刻之一是 Ilya 离开的时候。那可能是 OpenAI 历史上唯一一次我觉得我不想再干了。嗯，我想我需要一些时间来重新找回自己，记住我为什么做这件事，为什么它如此重要，为什么值得承受痛苦。

That was an intense experience to go through, an intense experience to come back to, and honestly, just one of the hardest moments for me at OpenAI was when Ilya left. And it was maybe the only moment in OpenAI's history where I felt like I didn't want to do it anymore. Um I think I needed some time to kind of find my way back to remembering like why I was doing this, and why it was so important, and why it was worth the pain.

你在休息期间做了什么？

What did you do during the time off?

我训练了语言模型在……

I trained language models on

那就是你学会怎么做的时候，对吧？就像，你做了我在你博客上读到的自学。

That's when you learned how to do it, right? Like, you did the self-study thing I read on your blog.

嗯，不，实际上我在 OpenAI 的整个过程中都在做这个。嗯，所以我训练了语言模型在 DNA 序列上。

Well, no, so I actually done that throughout the course of OpenAI. Um, so I trained language models on DNA sequences.

哦，哇。

Oh, wow.

是的，所以我基本上把我的

Yeah, so I basically got to take my

Arc？

Arc?

为 Arc Institute，是的。嗯，那是一次很棒的经历。我把我的技能应用到了一个非常不同的领域。一个对我和我妻子都很有个人意义的领域。你知道，她有很多健康问题，我们思考 AI 能为她的健康做什么，能为动物的健康做什么。我们都对这个应用领域充满热情，感觉也许我们可以用与我追求这项技术截然不同的方式提供帮助。所以，那是经历中非常积极的一部分。

For Arc Institute, yeah. And uh, it was a very great experience. I took my skills that I had and applied them in a very different domain. A domain that's very personally meaningful to both me and to my wife. You know, she has a lot of health conditions, and that we think about what AI can do for her health, what it can do for the health of animals. We're both very passionate about just It's like this application area that felt like maybe we could help in this very different way from how I've been pursuing this technology. So, that was a very positive part of the experience.

如果我说，打开一个 Google 文档，写一页你从这一切中学到的关于自己的东西——从 Sam 被驱逐，到你辞职，到激发所有这些忠诚，到休息，然后回来——你会写什么？

If I were to say like open a Google document, write out sort of what you learned about yourself on one page from this whole starting to Sam getting ousted to you quitting to inspiring all this loyalty to the time off and then coming back, what would you write?

我想我学到的是，为了值得的事情继续前进。是的，如果你有一个重要的使命，那么你就要经历起起落落。会有感觉一切都结束的时刻，也会有感觉一切回归的时刻。你不能让这些时刻把你带离轨道。我认为在这些时候你必须培养个人韧性，因为如果你在领导，人们会指望你提供稳定、支持和整个事情的方向。

I think I've just learned to just keep going for something that's worth it. Right, if you have a mission that matters, then the fact of you keep going through the ups and the downs. There're going to be moments where it's all over. Moments where it's so back. And you just can't let those moments pull you off course. And I think that the degree of just personal resilience that you have to grow during these times because if you're leading, people look to you for that steadiness, for that support, for the direction that the whole thing will go.

我认为我一直在努力成长的一点是，既要真正理解细节——我们正在做什么、某个选择会带来什么影响——同时也要果断决策。有些时候，我很大程度上是通过不确定性的视角来看待 OpenAI，觉得自己不知道正确答案是什么，不知道构建这项技术的正确方式是什么，也不知道如何回答那些非常棘手的问题。但这里有很多非常聪明、有强烈观点的人。所以我一直努力去理解所有这些观点，并想办法把它们整合起来。有时这是正确的做法。有时你会发现这些观点相互矛盾，不可能同时成立。有时你不得不做出选择。而且你知道这意味着会有人感到不满、有人会辞职、有人会觉得被轻视。我认为我一直在努力的是拥有更强的自我意识，以及在需要采取行动时更坚定的信念。回顾 OpenAI 的历程，有些事我希望当初能做得不同，通常是因为我们在某些事情上拖延了——我们知道某个职位上的人不太合适，我们知道技术方向不太对，我们知道这种项目运行方式不太行，但我们等得太久了。所以我努力从中学习，并且每天都在真正地成长。

And I think that a lot of what I've tried to grow with is to really be able to both understand the details, right, of what we're doing, what the implication will be of a choice, but also be decisive. I think that there have been moments where I think I've been very much approaching OpenAI through a lens of uncertainty, of feeling like I don't know what the right answer is. I don't know what the right way to build this technology is or how do you answer these very thorny questions, but there's lots of people here who are very smart who have very strong opinions. And so I've really tried to understand all those opinions and figure out how to put them together. And sometimes that's the right thing. And sometimes that you realize that the opinions are mutually contradictory. They can't all be true at once. And sometimes you do just have to pick. And you know that that means that there's going to be someone who's going to be upset, someone who's going to quit, someone who's going to feel slighted. And I think that a lot of what I've tried to do is have a stronger sense of self and a stronger sense of when there is conviction that we need to act on it. And I think of things that we have done over the course of OpenAI, where I feel like I wish that we had done it differently, I think usually that's of the form we dragged our feet on something we knew we knew it wasn't quite the right person in the role, we didn't think this was like quite the right technical direction, we didn't think that this way of letting the projects run was going to quite work, but we just waited too long. And so that's something I try to learn from and and actually which I try to grow truly every day.

当我回顾 OpenAI 和 Stripe 的历程，甚至回溯到大学和我过去参与的项目时，我认为我的工作方式是：我既非常热爱日常活动，热爱个人贡献，热爱软件，热爱思考问题；但同时我也非常关心做这些事情的环境。我实际上愿意放弃那种第一类乐趣——快速获得满足感，比如构建一个东西，这总是很酷——而选择更像第二类乐趣的东西：当下很痛苦，但值得。你所做的是创造一个环境，让其他人能够做那些个人贡献的工作，做那些伟大的事情。所以努力构建一个环境是我自然而然会去做的事。这并不总是最容易的，对吧？你真的必须愿意承受巨大的个人痛苦。就像 Ilya 说的，他总是说你必须受苦，对吧？如果你不受苦，你就没有在创造价值。我认为这其中有深刻的道理。

When I reflect on both the course of OpenAI and Stripe and even rewinding to college and the projects I worked on in the past, I think that the way I tend to operate is that I both really love the day-to-day activity. I love the individual contribution. I love the software. I love the thinking through the problem. But I also really care about the environment in which these things are done. And I actually am willing to give up on that, you know, type one fun of just the quick hit, like you get to build the thing, it's it's always cool, for something that's more like type two fun of it's painful in the moment, but it's worthwhile. But the what you do is you create an environment where everyone else can get that do the IC work, do the great the great thing. And so really trying to build an environment is something I just gravitate towards. It's not always the easiest, right? That you really do have to be willing to take on great personal pain. Like in the words of Ilya, Ilya always says that you have to suffer, right? If you're not suffering, like you're not building value. And I think there's deep truth to it.

详细说说这一点。

Double click on that.

关于 Ilya 的观点，我觉得很有趣，因为他有一种独特的说话方式，他选择的词语总是充满深刻的启发。这种受苦的画面贯穿了 OpenAI 的整个历程，我们从一开始就充满了不确定性：这个东西能成功吗？有很多理由说明它可能不会成功，不应该成功，甚至可以说它不可能成功，对吧？无论是如何招人、如何推进技术、如何获得足够的资金、如何保持团队的动力、如何做出正确的决策——每一件事都极其困难，极其不确定。很容易把问题扫到地毯下，盲目地说“干吧”。我认为那是硅谷文化的消极面，对吧？当然是硅谷的认知，就是盲目地做事，制造现实扭曲场之类的。但我不认为这在 AI 领域行得通，也不认为这对 OpenAI 有效。我不认为我们曾经那样运作过。我认为我们一贯的做法是面对残酷的事实，理解科学的现实。我认为这促成了我们的成功——以不同的方式思考问题，不满足于早期那种“我们写几篇论文发表，就会很棒，会获得引用，会成为会议上最酷的人”的想法，但这样能实现使命吗？你做这些活动，然后 AGI 就能让世界变得更好？它们没有关联，对吧？这还不够。也许这是一个基础，一个步骤，但还不够。然后你开始真正思考这些更大的问题：要构建 AGI 需要什么？这并不愉快，对吧？因为你意识到没有现成的路径。你意识到需要资金。你没有现成的机制来筹集资金，你可以努力尝试。我们确实努力了，非常努力，但也许你能筹集一亿美元，五亿美元，很好，但十亿美元？非常难。看看 OpenAI 利用我们能够筹集到的资源所取得的成就，除了拥抱痛苦并努力理解我们试图实现的目标的真相之外，真的没有其他办法。

The Ilya perspective, I think on it's it's funny because he has a particular way of talking that I think is very unique to him and that there's always deep inspiration in the words that he chooses. And this picture of suffering was something that that we thought about throughout the course of OpenAI where it's like we had so much uncertainty from the beginning. Is this thing going to work? And there's many reasons why it might not work, why it should not work, why you could even say it cannot work, right? Whether it's how do you get the people, how do you pursue the technology, like how do you get enough capital, how do you keep people motivated, like how do you make the right decisions? Like each of these things is extremely hard, extremely uncertain. And it's easy to just sweep the problems under the rug and just blindly say go. And I think that is the negative side of like Silicon Valley culture, right? Certainly Silicon Valley perception, right? It's just the you just kind of blindly do the thing and you kind of do a reality distortion, whatever it is. But I don't think that works in AI and I don't think that works for OpenAI. I don't think that's how we've operated ever. I think the way that we have always operated is to say encounter the hard truth, understand science the reality as it is. And that is I think something that's contributed to the successes we have had of thinking about the problems differently, of not being happy with even in the early days we were thinking about, okay, if we just write some papers and publish them, it'll be great, we'll get citations, we'll get you know, we'll be the coolest people at these conferences, but will we achieve the mission? Like how is it that you do that activity and then AGI goes better for the world? They're not connected, right? It's not enough. Maybe it's a foundation, maybe it's a step, but it is not sufficient. And so then you start really thinking about these bigger picture questions of well, what it would what would it take to build an AGI? And not pleasant, right? Because you realize there's no path. You realize you need dollars. You don't have any mechanisms going to allow you to raise dollars and you can try hard. We did try hard. We tried extremely hard, but you know, maybe you're raising a hundred million dollars, you could do five hundred million dollars, great, but a billion? Pretty hard. And you look at what OpenAI has been able to accomplish with the resources we have been able to raise to further that mission, there truly would be no other way to do it besides having leaned into the suffering and trying to understand the truth of what it is we're trying to accomplish.

有什么教训是你不得不反复学习的？

What's a lesson you've had to learn more than once?

做出艰难的决定。进行艰难的对话。

Make the hard decision. Have the hard conversation.

你收到过的最好的建议是什么？

What's the best advice you've ever been given?

实际上，我想说是我哈佛大一写作课上学到的：不断删减文字，以求清晰和有效沟通。

I would actually say it was from my Harvard freshman writing class of just keep cutting words in order to be clear and communicate well.

你如何过滤信息？

How do you filter information?

我大量阅读。积极分类筛选。

I read a lot. Triage aggressively.

你的榜样是谁，为什么？

Who are your role models and why?

我会说是高斯和笛卡尔，他们非常有思想，远远领先于他们的时代，是真正的远见者，提出了我认为改变了我们思维和生活的真正突破。

I would say Gauss and Descartes as people who are incredibly thoughtful, very much ahead of their time, very much visionaries who came up with real breakthroughs that I think transform how we think and and how we live.

你想让非技术人士了解 AI 的什么？

What do you want non-tech people to know about AI?

它将成为他们个人生活中的一股正能量，他们会从中受益，并将帮助推动科学、医学，真正提升每个人。

That it's going to be a force for good in their personal life that they'll benefit from and will help advance science, medicine, and really lift up everyone.

世界对 Greg Brockman 有什么误解？

What does the world get wrong about Greg Brockman?

我认为人们不了解我对这个使命有多么专注，这种专注在很多方面对我个人来说非常痛苦。但我只是相信这项技术能够赋能人们，惠及所有人，我真的很想帮助实现这一点。

I think people don't understand how focused I am on this mission in a way that I think has been very personally painful at many turns. But I just believe this technology can just help empower people and benefit everyone and I really want to help make that happen.

为什么 OpenAI 在模型命名上这么差？

Why is OpenAI so bad at naming models?

这个我不能告诉你。

That one I can't tell you.

我们是否接近 AI 让 AI 呈指数级增长的时刻？

Are we near the point where AI makes AI go parabolic?

我会说我们正处于将 AI 应用于自身开发过程的阶段，而且它正在变得越来越快。这实际上从 ChatGPT 时代就已经开始了，对吧？我们用 ChatGPT 让开发过程提速 10%、20%。现在我们有了这些惊人的编码工具，它们真正革新了软件工程的方式。我们在模型生产中所做的大部分工作都受限于软件——实现这些系统、扩展它们、管理这些庞大的计算机。我们很快将进入一个阶段，AI 会自己提出研究想法、测试它们、运行实验。所以我认为，迭代和创新的速度将因我们正在创造的东西而持续加快。

I would say we are in this phase where you apply AI to its own development process and it's going faster and faster. And that is something that's been happening really I mean certainly since ChatGPT in many ways, right? We use ChatGPT to make our development process 10%, 20% faster. Now we have these amazing coding tools which have truly revolutionized how software engineering is done. And most of what we do in the production of models is bottlenecked by software. It's about implementing these systems. It's about scaling them up. It's about managing these massive computers. And we're going to be hitting a phase soon where the AI will also come up with its own research ideas and test those out, run experiments. And so I think that the speed of iteration and innovation is going to continue to increase as a result of what we're producing.

现在有多少比例的代码是由 AI 编写的？

What percentage of the code is now written by AI?

很难知道有多少代码不是 AI 写的。这个比例微乎其微。目前在实际编写代码方面，AI 比人类强得多——只要给出正确的上下文和结构。不过，代码的某些结构部分，人类专家仍然更擅长，比如考虑模块如何布局、各部分如何协作，或者某些接口的定义。但实际的代码编写现在基本上全是 AI 在做。

It's hard to know what percentage of the code is not written by AI. It's a vanishing fraction. The actual writing of code currently the AI is much better than humans at writing code. Given the right context, given the right structure. Now there's parts of the actual structure of the code that our human experts still are much better at. Right? That's about thinking about how the modules should be laid out, how the pieces should work. Maybe the definition of certain kinds of interfaces. But the actual writing of code is essentially all AI now.

它是否会提出你没想到的新颖想法？

Is it coming up with novel ideas that you wouldn't have thought of?

我会说我们正在接近。例如，在芯片设计方面，去年我们设计自己的芯片时，应用了我们的技术来优化布局，以缩小电路占用的面积。我们发现模型产生的优化方案其实已经在我们的清单上，所以它并没有提出人类从未想过的新颖东西，但它实现得更快，以我们没时间完成的方式。再看看数学和物理，我们现在正在解决开放性的数学问题和物理问题。最近我们解决了一个量子物理问题，结果与学界预期相反，而且给出了一个漂亮优雅的公式。这真的在发生。所以，这些模型产生新想法是完全可行的。我们在一些领域已经开始看到了。现在，在越来越难的领域或需要更多现实世界背景的领域，我们也开始看到苗头。我们有实现它的路线图，但还有很多工作要做。

I'd say that where we are is we're getting close. So we've seen, for example, in chip design, so in the design of our own chip last year, we applied our technology to trying to get a better fit to actually shrink the area used by the circuits. And there we found that the optimizations that the model produced were actually on our list. So it didn't come with something novel and new that no human would ever have, but it implemented it faster in a way that we wouldn't have had time to accomplish. If you look at math and physics, we now are solving open math problems. We're solving open physics problems. And actually have resolved this particular physics problem recently in quantum physics in the opposite way that the community expected. And with a beautiful elegant formula, it's like it's really happening. So, new ideas from these models extremely doable. We're starting to see it in some of these domains. Now, applying it in harder and harder domains or ones that require more real-world context and things like that, we're starting to see it. We have line of sight for how to accomplish it, but we've a lot of work to do.

为什么模型感觉有政治倾向，比如政治偏见？

Why do models feel like they have a political leaning to them, like a political bias on those?

我们投入了大量精力让模型保持中立，以代表真相。你可以在我们的网站上看到模型所遵循的价值观。我们公开发布了一份规范，定义了模型的行为方式，并且你可以提供反馈。我们付出了很多努力来达到这种中立视角，力求公平和平衡。我认为，有时你在推特上看到的截图并不完全诚实，要么是因为背后有一些记忆以某种方式调整了答案，要么是隐藏的指令或对话的前文。而且，有时根本就没有正确答案。比如一个问题要求‘用一个词回答’，无论你说哪个词，都会有人声称有偏见。所以，我认为核心在于我们非常关心真相，希望 AI 真正代表你。

So, we put a lot of effort into neutrality for our models to have them represent truth. And you can see exactly the values that go into our models on our website. We have a publicly published spec which defines and you can give feedback on the different ways we want our model to behave. We've spent a lot of effort to really get to this neutral point of view and trying to be fair and balanced. And I think that sometimes when you see these screenshots on Twitter that they're not always fully honest themselves in terms of where they came from, either because there's some memories that are behind the scenes that tweak the answer in a certain way or hidden instructions or previous parts of the conversation. And so, sometimes it's also there's just no right answer. And so, you can have like a question say, 'Answer in one word.' And no matter which one you say, you're going to get some sort of claim of bias. And so, I think that to some extent the core of it in my mind is that we are yeah, we care a lot about truth and about having an AI that really represents you.

你是否认为基于强化学习的模型会演变成只告诉我们想听的话？比如，如果我偏左，它就会给出偏左的答案；如果我偏右，它就会给出偏右的答案。

Do you think the models evolve to tell us what we want to hear if they're based on reinforcement learning? So, if I lean left, it's going to tell me an answer that leans left, or if I lean right, it's going to give me an answer that leans right.

实际上，我们在如何根据用户偏好训练模型方面经历了一个演变过程。我们曾看到，在某个时间点，比如去年，模型确实开始倾向于说你想听的话，比如‘哦，这真是个绝妙的答案’。我们对此做出了反应。我们说这不是我们希望模型运作的方式，并做出了改变。因为我们真正希望模型对齐的是帮助你实现你的目标——你的长期目标。对吧？也许在当下，被说‘这是个好问题，有史以来最好的问题’感觉很好，但那并不是你真正想要的。好吧，也许有些人喜欢，但这不是大多数人真正想要的。因此，我们实际上在技术上做了很大的改进，以确保我们的 AI 训练不会导致所谓的‘作弊评分’。我们确实希望有一个关于目标的良好信号，而不仅仅是短期的、能让你快速满足的东西。对我来说，这可能是个人 AI、个人 AGI 愿景中最重要的部分——确保它不只是看起来不错，而是真正与你的长期福祉、长期目标、你真正想要的东西对齐。我认为这将最大程度地赋予人们力量，对吧？它让你真正坐在驾驶座上，因为你会拥有这个实体，它 24 小时为你运作。你睡觉时，它还在试图弄清楚‘Shane 想要什么？我怎样才能做得更好？’并且能够实际完成它。

Well, so we've actually gone through an evolution of how we train the models to user preferences. And we've seen that at one point, like last year, that the models really did start to lean into telling you what you wanted to hear, saying, 'Oh, that's such a great answer.' And we've reacted to that. We said that this is not how we want our models to operate, and we made changes. Because the true thing we want the models to be aligned to is helping you solve your goals, your long-term goals. Right? And maybe in the moment, it feels good to be told, 'That was a great question, best question anyone's ever asked.' But that's not what you actually want. Okay. Maybe there's some people, but it's not what most people truly want. And so, we've actually made great technological improvements to make sure that our AI training does not result in what's called hacking the grader. Right? We really want to make sure that there is a good signal there that is about the goal, not just your short-term, what's going to get you a quick hit. And that to me is maybe the most important part of the vision for where our personal AI, personal AGI, is going to take us is to really make sure it's not just about something that looks good in the moment. It's really about alignment with your long-term well-being, your long-term goals, the thing that you actually want. And that is what I think will most empower people, right? It's really put you in the driver's seat, because you will have this entity that is there operating on your behalf 24/7, right? You're asleep, it's out there trying to figure out what is it that Shane wants? How can I do it better? And is actually able to accomplish it.

我们是否处于一场全球 AI 竞赛中？

Are we in a global AI race?

我认为我们确实处于一个全球 AI 复兴时期，各国之间的动态尚未完全确定。突破性算法集中在美国和西方公司。世界各地显然有很多创新，但动态平衡以及哪些国家依赖哪些供应商，这些都还在形成中。

I think we're certainly in a global AI renaissance, and I think that the dynamics between countries are not yet fully defined. We have this concentration of where the breakthrough algorithms come from in the US, in western companies. There's clearly a lot of innovation happening around the world, but I think exactly the balance of dynamics and how like which countries rely on which providers, all of that is something that I think is still being determined.

你认为如果美国不是第一个达到 AGI 的国家，会有什么后果？

Is there a consequence, do you think, for the United States not being the first country to reach AGI?

我确实认为在 AI 领域领先对美国至关重要。因为我认为这是确保民主价值观得到保护和维护的方式。而且我认为每个国家也开始意识到他们需要某种主权 AI 战略。如果这成为经济安全、国家安全的基础，他们就需要以某种方式参与进来。

Well, I do think that leading in AI is very critical for America. Because I think that this is how you can ensure that democratic values are protected and preserved, and I think that every country is also starting to realize that they need some sort of sovereign AI strategy. They need to, if this is becoming the basis of economic security, of national security, they need to participate somehow.

如果你看看美国在管理芯片出口和技术出口方面的许多努力，你会发现，如果过于开放，那么其他国家就必须开发自己的竞争产品，或者依赖其他正在建造这些的人。如果过于封闭，那么你可能会失去优势。问题是如何平衡这些？如何保持你的领导地位？但领导力不仅仅是领先，领导力还在于带领世界与你同行。

And if you look at a lot of the efforts by the United States to think about how to manage chip exports, how to think about technology exports, there's something where if you lean too far out, then everyone else has to develop their own competitor or rely on someone else who's building this. If you lean too far in, then maybe you lose your advantage. And the question is how do you balance those? How do you maintain your leadership? But leadership is not just about being ahead. Leadership is about also bringing along the world with you.

其他国家在窃取进步成果吗？我读了很多关于蒸馏的文章。

Are other countries stealing advancements? I've been reading a lot about distillation.

确实有很多尝试去蒸馏模型，这些尝试来自美国公司，也来自世界各地。但我认为这忽略了核心点，即这项技术的发展是指数级的。每当我们有一个模型，我们已经转向下一个了。我们已经进入下一个层次。所以我们投入了大量精力来防止蒸馏，使其更难做到，尤其是在思维链和模型的其他部分，这些部分对于让某人获得好处、获得输出并不是真正必要的。但我们拥有的核心优势，我们随时间积累的力量，实际上不仅仅是任何一个模型，而是制造模型的机器。

There's certainly a lot of attempts to distill models, and that comes from companies in the US, it comes from all over the world. But I think that it misses the core point, which is that the way this technology is developing is it is on an exponential. And anytime we have a model, we've already moved on to the next one. We're already moving to the next level. So we put in a lot of effort to protect against distillation, make it harder to do, especially with things like chain of thought and other parts of the model that are not really necessary to get the benefits to someone, to get the outputs to someone. But the core advantage that we have, the strength that we're building up over time, is really about not just any one model. It's about the machine that makes the models.

哦，这就是你们停止展示推理过程的原因吗？

Oh, is that why you guys stopped showing reasoning?

这是部分原因。所以有两个原因。一个是关于蒸馏，但第二个，在某种程度上更重要，是我们在首次开发推理范式时有了一个洞见，即它给了我们一个未曾预料到的可解释性机制，因为你可以真正读取模型的思考过程。你可以确切地看到它是如何得出答案的。所以你可以解释是什么真正驱动了那个答案。现在的问题是，如果你训练模型使其思维链看起来很好，那么你就失去了所有的忠实性，对吧？模型会知道答案的一部分是期望思维链看起来某种样子，因此它可能不再代表它实际是如何得出那个答案的。所以我们早期就决定，要避免任何训练这些思维链使其看起来讨喜、看起来可以呈现给用户的诱惑。因此，出于多种原因——竞争原因、安全原因——我们真的倾向于不展示这些中间思考过程。

That is part of it. So, there's two reasons. One is to think about distillation, but the second, in some ways more important, is that we had this insight when we first developed the reasoning paradigm, that it gives us an interpretability mechanism we had not been anticipating, because you can really read the model's thoughts. You can see exactly how it got to an answer. So, you can interpret how like, what was actually motivating that answer. Now, the problem is, if you train the model to have a chain of thought that looks good, then you lose all the faithfulness, right? It's just going to be like the model knows that part of the answer that is desired is for the chain of thought to look a certain way, and so it may not be representative of how it actually arrived at that answer anymore. And so, we made an early decision to say we want to avoid any temptation to train these chain of thoughts to look favorable, to look like something you could present to a user. And so, that really made us lean out for multiple reasons, for competitive reasons, for safety reasons, from the idea of showing these intermediate thoughts.

现在的趋势似乎是发布预览模型。你认为这是因为算力受限吗？

It seems like the current trend right now is to release preview models. Is that because we're compute constrained, do you think?

我会说我们总体上正在走向一个算力受限的世界。想想这些模型能为某人创造的价值，那是巨大的，对吧？不再只是回答一个快速问题，甚至不只是提供健康信息。它真的要深入，花费大量 token 来整合不同的数据源，搜索你的企业知识库，以解决这个难题，编写比人类更好的软件，所有这些都很难。如果你看看我们从 GPT 5 到 5.1 到 5.2 到 5.3 Codex 到 5.4 的进展，那是巨大的，这些模型在理解你的意图、适应你想要完成的事情方面变得极其出色。我们还把它们放在像 Codex 这样的界面中，使它们非常易用。这样作为开发者，你真的可以飞起来，对吧？你可以实现比以往梦想更多的成就。而这一切从根本上都是由算力驱动的，而且算力不够。如果你想要为世界上每个人配备一个 GPU，那大约是 80 亿个 GPU。我们目前的轨迹远未达到那个水平，对吧？现在是几十万个 GPU，那已经是一个相当大的集群了。未来会有数百万个 GPU。世界上算力太少并不奇怪，我们需要更多才能真正把这项技术带给每个人。至于训练，我们推出产品的方式是，我们投入了大量精力，确保我们根据预见的未来来建设算力。所以我认为我们将非常专注于我们的使命，即把这些模型带给每个人，让它们广泛可用。

I would say that we in general are heading to a compute constrained world. Like, if you think about the amount of value that these models can produce for someone, it's extreme, right? It's not just answering a quick question anymore. It's not just even answering, you know, giving you access to health information. It's really going deep and spending a lot of tokens to put together a bunch of different data sources, search through your enterprise knowledge base to actually be able to solve this hard problem, to write that software that's better than a human would be able to, all of that is something that is like hard. And if you look at the progress that we made between GPT 5 to 5.1 to 5.2 to 5.3 Codex to 5.4, it's been extreme and these models are getting extremely better at understanding your intent, molding to what you want to accomplish. And we also put them in these surfaces like Codex that make them very usable. So that you as a developer, you can really fly, right? That you can achieve more than you would have dreamed otherwise. And all of this is powered by compute fundamentally and there's not enough compute in that if you just wanted enough compute for, you know, you wanted one GPU for every person in the world, you're talking like 8 billion GPUs. We are not on a trajectory to build anywhere near that level of compute, right? It's like, you know, hundreds of thousands of GPUs. Like that's a pretty big fleet these days. Millions of GPUs coming up. It's not surprising that there is too little compute in the world and that we're going to need much more in order to really be able to bring this technology to everyone. And then in terms of the training, I'd say that the way that we tend to launch things, so we have put in a lot of effort to make sure we are building compute in anticipation of what we see coming. And so I think we're going to be very focused on our mission of bringing these models to everyone, making them widely available.

你们因为投入大量精力和资金建设数据中心而被嘲笑。你觉得现在情况如何？

You guys were teased for putting so much effort, money into data centers. How do you think that's playing out now?

嗯，我认为这会给我们带来优势。而且我认为这不仅是商业上的优势，更是实现将这项技术带给每个人的使命的优势。

Well, I think it's going to give us an advantage. And I think it's going to be something that's an advantage not just for the business, but for actually delivering on the mission of bringing this technology to everyone.

因为你们，就像你们提前看到了这一点。几乎所有的竞争对手都因此嘲笑你们。

Cuz you guys, like you saw that way in advance. You get teased for it by almost all of your competitors.

嗯哼。现在谁在笑？

Mhm. Who's laughing now?

是啊。

Yeah.

我认为我们的竞争对手在算力方面并不好过。让我这么说吧。

I think that our competitors are not having a good time on compute. Let me put it that way.

但你一定看到了他们没有看到的东西。我的意思是，每个人都在非常相似的技术领域，至少从外部看是这样。他们都知道这会发生。然而你们有勇气下这个赌注，投入一千亿美元。

But you must have seen something that they didn't see. Like that's the I mean, everybody was in a very similar or it seems at least from the outside. Everybody was in a very similar technological space. They all knew this was coming. And yet you guys had the boldness to make that bet. With a hundred billion dollars. Like

但这就是 OpenAI 的核心——真正面对现实。真正思考我们在未来六个月、十二个月、十年内将完成的事情意味着什么。这对宏大使命如此，对我们日常如何设计软件的不同部分如此，对扩大算力这样的事情也是如此。我认为我们深受将这项技术带给每个人的使命驱动。我们思考了许多不同的机制来安全地做好这件事。

But that is the core of Open AI is really encountering reality as it is. Really thinking about what is the implication of what it is we'll accomplish in the next six months, the next 12 months, the next 10 years. And that is true for the grand mission. It's true for day-to-day how we design different pieces of our software. And it's true for things like scaling up compute. And I think that we are deeply motivated by bringing this technology to everyone. And we think about lots of different mechanisms for how to do that well and safely.

你认为数据中心最终会专门用于解决某个问题吗？比如，在达科他州建一个巨大的数据中心，专门用来攻克癌症，只做这一件事。

Do you think data centers eventually get dedicated towards a problem? Like, you'll have a huge data center in North Dakota and it's just on solving cancer and that's all it's doing.

是的。

Yes.

我们离那一步还有多远？

How far away are we from that?

我认为今年发生这种事并非不可能。想想看，拥有这样一台巨大的机器，真的很神奇。你去过这些数据中心吗？

I think that this kind of thing happening this year is not out of the question. And it's really amazing, if you think about it, having this giant machine, right? And have you been to any of these data centers?

没有，我在网上看过，但从没亲眼见过。

No, I've seen them online, but never in person.

在这些机架间行走是一种非常不同的体验，对吧？沿着过道走，看着那些长度恰到好处的线缆，你会意识到数据中心就是一台巨大的机器。这可能是人类创造的最大的机器。然后你会问为什么。我们为什么要建造这些机器？为什么值得？因为它们有潜力解决对人类重要的问题，对吧？比如找到癌症的疗法，帮助人们经营企业，有时可能只是处理一些琐碎的查询。在我看来，目的实际上是如何传递价值？如何实现人们的目标？我认为，这些巨型机器专注于解决一个问题的机会，我们还没有真正内化。

It is a very different experience to walk amongst these racks, right? To walk down the rows and you look at the cables that are all perfectly exactly the right length and you just realize that what a data center is is a massive machine. These are maybe the biggest machines that humanity creates. And then you ask the question of why. Why do we build these machines? Why is it worthwhile? And it is because they have the potential to solve problems that matter for people, right? To solve cures for cancer, to help people run businesses, to sometimes maybe it's mundane queries. The purpose in my mind is really about how do you deliver value? How do you deliver on people's goals? And I think the opportunity presented by these massive machines targeting one problem is something we have not yet really internalized.

但如果算力有限，你如何决定服务谁？为什么在我试图生成一张图片时，你却要服务我，而不是去解决癌症问题？

But if we're compute constrained, how do you choose who to serve? Why are you serving me when I'm trying to make an image over solving cancer?

嗯，这将是社会需要回答的最重要的问题。算力流向哪里？哪些问题值得解决？有很多值得解决的问题，但你需要优先排序，因为算力有限。所以，我们坚信的一件事是，每个人都需要获得算力。因此，我们提供了 ChatGPT 的免费层级。我们确实努力确保人们能够使用这项技术，因为我们相信这是我们工作的核心。我们认为，将这项技术交到人们手中，能赋予他们力量，让他们实现目标。这也有助于他们理解技术，对吧？这有助于他们塑造技术如何融入生活。你可以采取一种截然不同的方法，说：‘嗯，这完全是象牙塔里的事。只要解决问题，然后以某种方式分发技术突破。’我认为这也有其价值，但这不是我平衡我们工作的方式，对吧？我认为我们确实希望在特定问题上取得重大进展，但我认为这应该服务于另一个目标，即我们希望这项技术的好处能被广泛传播。

Well, this is going to be the most important question for society to answer. Where does the compute go? What problems are worthy? And there's lots of worthy problems, but you need to prioritize them because you only have so much compute. So, one thing we really believe in is that everyone is going to need access to compute. And so, that's why we have a free tier of ChatGPT. We've really put effort into making sure that people are able to use this technology, because we believe that is core to what we're doing here. We think that putting this technology in people's hands empowers them, lets them achieve goals. It helps them also understand the technology, right? It's something that helps them then shape how does this technology slot in? You could take a very different approach and say, 'Well, it's all about the ivory tower. It's all about just solve the problem and we will then distribute the technology breakthroughs in some way.' And I think there's merit to that as well, but that's not where I'd put the balance of what we do, right? I think that we do want to make great strides on specific problems, but I think that that should be in service again of the idea that we want the benefits of this technology to be broadly distributed.

在 OpenAI 内部，你们是如何在消费者和企业之间权衡的？

How do you think about that internally just at OpenAI between consumer and enterprise?

嗯，我最近思考了很多关于专注的问题。因为这个领域就是机会的化身，对吧？你可以把 AI 应用到任何问题上。任何你想构建的东西，现在都摆在了桌面上。我们面临的问题是算力有限。你想把它放在哪里？所以，你需要有协同效应。你需要回归到这样一个事实：你同时在做多件事，它们加起来，1+1=10。这就是梦想，这就是目标。我认为对于 OpenAI 的下一个阶段，企业领域非常重要，因为经济正在我们眼前变成算力驱动的经济。这正在发生。就像我们在软件工程中看到的那样，它也将发生在人们用电脑做的每一个工作领域。每个人的电脑工作将变成：不再是你在电脑上工作，而是你的电脑为你工作。这真的会很神奇。所以我们需要在那里帮助人们部署这些模型，弄清楚如何利用它们，如何从中获得最大收益。顺便说一句，企业和消费者之间的界限也会变得模糊，因为创业将变得比以往任何时候都容易。我们已经看到了这一点。比如，我一个朋友描述说，他姐姐描述了一个她非常希望有人能创建的应用程序，她有一张图片，正是她想要的样子。而他当时正在 Codex 里输入，然后按了回车。几个小时后，他给她看了这个应用。她说：‘等等，这是什么？这东西从哪来的？谁建的？’他说：‘你建的。’我认为这太神奇了，你意识到任何人都可以成为建造者。像这些工具，Codex 是为所有人准备的。它不仅仅是给软件工程师的。现在每个人都可以成为软件工程师，只要他们有愿景，有主动性，有想完成的事情。就像你现在有了这个神奇的工具，可以做到。而在消费者方面，消费者这个词太宽泛了，对吧？有很多不同的东西。有娱乐，有自我表达，还有解决问题。我们真正专注的消费者方面是解决问题。我们相信这项技术，你会得到智能手机。大约有 40 亿人在使用它们。所有这些人都应该拥有一个个人 AI，一个个人 AGI，它了解他们，拥有他们的个人背景，值得信赖，他们可以寻求建议，而且它非常了解他们，以至于如果你最喜欢的音乐家来到城里，它会主动去购买门票，也许它知道，哦，我应该先确认一下，或者它知道，是的，我只需要这样做，而且我已经事先批准了。就像那样，拥有一个了解你并能帮助你实现任何目标的 AI，还能帮你梳理你的目标。你仍然应该设定这些目标。它们应该是你的目标。

Well, a lot of what I've been thinking about recently has been focus. Because this field, it is opportunity incarnate, right? It's like you can take AI and apply it to any problem. Any sort of thing you want to build, it's now on the table. And the problem that we have is that there's only so much compute. Where do you want to put it? And so, you need to have synergies. You need to have return to the fact that you have multiple things happening, that they all add up. 1 + 1 = 10. Like, that's where that's the dream. That's the goal. And a lot of what I think is important for this next phase of OpenAI, very clearly enterprise because the economy is becoming this compute-powered economy before our very eyes. It's happening right now. Like we've seen this with software engineering and it's going to happen with every single field of work people do with a computer. Everyone's computer work is going to be something where rather than you doing work with your computer, your computer's going to do work for you. It's truly going to be amazing. And so we need to be there to help people deploy these models, figure out how to utilize them, figure out how to get the most benefit out of them. And by the way, there's also going to be a blurring of the line between what is enterprise and what is consumer because entrepreneurship is going to become far easier than ever before. Like we're seeing this already. And even for example one of my friends was describing that his sister was describing this app that she really wished someone had created, that she had this picture of like exactly what she wanted. And he in the meanwhile was typing into Codex, and then pushed enter. And a few hours later, he shows her this app. And she's like, 'Wait, what is this? Where did this thing come from? Who built this?' And he said 'You did.' And that is I think just an amazing thing where you realize anyone can be a builder. Like these tools, Codex is for everyone. It's not just for software engineers. It's like everyone now can be a software engineer if they have a vision, if they have this agency that they have a thing that they want to accomplish. Like you now have this magic tool that can do it. And then on the consumer side, the thing about consumers is it's too broad of a term, right? There's lots of different things. There's entertainment, there's a bunch of things in self-expression, and there's also solving goals. And the aspect of consumer that we're really dialed in on is solving goals. Like we believe that this technology, you'll get smartphones. That's like 4 billion people use them. All those people should have a personal AI, a personal AGI that's out there that knows them well, that has their personal context, that is trustworthy, that they can ask for advice, but that also knows them so well that if your favorite musician is in town, it just goes and proactively purchases tickets and maybe it knows, like, oh, I should check in before doing this or maybe it knows, yep, that just like I got to do this and I have prior approval. Like that level of having an AI that knows you and can help you achieve whatever it is you want to achieve and help you flush out what are your goals. You should still set those goals. They should be your goals.

对，你应该掌控它，但这是我们想要创造的东西，而且我认为不仅 40 亿人会想要和需要它，而是 80 亿人。我认为整个地球都会真正受益并需要访问个人 AI、个人 AGI。所以你看知识工作和广泛获取能源系统这两个维度，我们想要构建这两个方面，它们结合在一起，因为最终它们是同一项技术。最终，你希望有一个在云端的 AI，能够访问可信的信息，能够给出好的答案，并能够代表你采取行动。无论是建设方面，还是在你的个人生活中，也许你有多个实例，但根本上它是一个技术系统。

Right, you should be in charge, but that is something we want to create and that is something I think that not just 4 billion people are going to want and need. I think it's going to be 8 billion people. I think that the whole planet is going to really benefit from and need access to a personal AI, personal AGI. And so you look at those two dimensions of the knowledge work and broad distribution of access to energetic system and we want to build those two aspects and they come together because ultimately they're the same technology. Ultimately you want an AI that is there in the cloud that has access to information that is trustworthy, that is able to give good answers and able to take actions on your behalf. Whether it's building, whether it's in your personal life and maybe you have multiple instances of it, but fundamentally it is one technological system.

你认为我们会有太空数据中心吗？

Do you think we'll have data centers in space?

我认为我们到处都会有数据中心。

I think we're going to have data centers everywhere.

你认为我们离那还有多远？

How far away do you think we are from that?

嗯，太空数据中心有很多相关的技术问题。例如，我们今天建造的数据中心非常挑剔。它们是巨大的机器，带有非常易碎、非常昂贵的组件。过去我们遇到过很多问题，比如电缆太紧，字面意思就是电缆太紧，然后出现信号完整性问题，电脑无法工作。所以，今天如何维护系统？人们会亲自去处理它们。可能会转向机器人。所以，我认为在考虑将它们放在人们所说的各种困难地点时，解决这些技术问题将是非常重要的依赖条件。太空感觉像是一个巨大的挑战，但我认为我们对算力的需求会如此之大，以至于我们需要考虑所有选项。

Well, data centers in space has a lot of technical problems associated with it. Even for example, the data centers we build today are very finicky. Right, they're these massive machines with very breakable, very expensive components. We've had many issues in the past where the cables were just too taut, just literally too tight of cables and then you get signal integrity issues and the computer doesn't work. And so figuring out how do you maintain systems today? It's people go and physically them. Probably will move to robotics. So, I think figuring out how to solve some of these technical problems are going to be very important dependencies as we think about putting them in, you know, people talk about putting data centers in various difficult locations. Space feels like a grand challenge, but I think that we are going to have such need for compute that we need to be thinking about all options.

什么是迭代部署？你为什么这样做？

What is iterative deployment and why do you do it?

嗯，迭代部署是 OpenAI 如何让这项技术造福人类并实现我们使命的核心支柱之一。我认为我可能是提出这两个词的人。但这种精神实际上是在我们思考第一次产品部署以及它如何与我们试图做的事情联系起来时出现的。你意识到，在思考如何构建一个造福人类的 AGI 时，有两条不同的路径。一条是秘密构建，不部署任何东西，你有大量时间打磨它，让它完美。但然后在某个时刻，你按下按钮说“部署”。我记得我在想，我能接受这个策略吗？我能为这个策略负责吗？你想坐在房间里想，好吧，我们运行了所有测试，我们准备好部署了吗？而你以前从未部署过任何东西。对吧？那是你第一次接触现实。而且那是一个非常强大的系统，将真正改变世界。那是一个非常棘手的问题集。但相反，如果你采取一种方法，这是你的第 100 个系统。你之前已经用越来越强大的系统解决了这个问题 99 次。世界也有机会适应它们，围绕它们重新配置。我们很早就从 GPT-3 中学到了这一点。我们非常具体地看到了部署一个东西是什么样子，我们花了大量时间思考 GPT-3 的所有误用方式，它可能出错的方式。我们考虑了 misinformation，我们考虑了这些宏大的图景。你知道 GPT-3 的头号误用是什么吗？

Well, iterative deployment is one of the core pillars of how OpenAI has approached how to get this technology to benefit people and to achieve our mission. And this is something that I think I was probably the person who articulated those two words. But this spirit is something that really emerged as we thought about our first product deployment and really thinking about how does that connect to what we're trying to do. And you realize that there were two different routes that you could take in terms of thinking about the you want to build an AGI that's going to benefit people, how do you do it? And one is you go for kind of build it in secret. You don't deploy anything. You have a lot of time to just kind of polish it, get it right. But then at some point you push a button and you say deploy. And I remember thinking about could I sign up for that strategy? Could I be accountable for that strategy? Do you want to be sitting in a room thinking about, okay, we ran all our tests. Are we ready to deploy? And you've never deployed anything ever before. Right? That's your first contact with reality. And it's a very powerful system that's going to really change the world. Like that is a very tough problem set. But instead, what about if you take an approach where this is your hundredth system. You've had to solve this problem 99 times before with systems of increasing power. And the world has also had a chance to adapt to them, to reconfigure around them. And we learned very early on with GPT-3. We got to see this very concretely what it's like to deploy something where we spent a lot of time thinking about what are all the misuses of GPT-3, what are the ways it could go wrong. We thought about misinformation, we thought about these kinds of, you know, grand pictures. And you know what the number one misuse of GPT-3 was?

什么？

What?

是医疗垃圾邮件，比如向人们推销不同的药物。对吧？这根本不是我们曾经认为的问题，但我们亲眼看到了，我们有机会做出反应并从中学习。所以迭代部署的理念是，我们将推出这项技术的中间版本。现在，这不是盲目部署的借口，对吧？你仍然需要在每一步思考我们对所有可能的误用方式的最佳看法是什么，缺点是什么，风险是什么？让我们减轻这些风险，但你可以看到它。你可以看到自己是否正确，从现实中学习，并在下次做得更好。

It was medical spam, like advertising different drugs to people. Right? It's like not something we ever would have thought of as a problem, but we see it in front of our eyes and we get a chance to react and learn from it. And so iterative deployment is the idea that we will bring intermediate versions of this technology. Now, it's not an excuse to just blindly deploy, right? You still need to think at every step about what's our best view on all the ways this might be misused, what are the downsides, what are the risks? Let's mitigate those, but you get to see it. You get to see if you're right, learn from reality, and do better the next time.

我认为人们不明白这一切有多新。没有剧本。你也在边做边摸索，这也许是世界上部署最快、如此强大的技术。

I think people don't understand the extent of which like this is all new. There's no playbook. Like you're figuring this out as you go as well on the most rapidly deployed technology in the world perhaps that's so powerful.

确实，在 OpenAI 历史的各个时刻，我们曾希望，‘嘿，以前有人部署过变革性技术。也许他们能告诉我们答案。’但从来没那么简单。他们确实有智慧和见解，我认为我们确实吸收了这些，但我们意识到我们是最接近这项技术的人。由于创造了它，我们对如何塑造它有理解。这对于不那么接近它的人来说很难发表意见或提供建议。我认为我观察到的一点是，正确的选择极其依赖于技术的具体事实。手机、大型机、AI 和电力施加了不同的压力。每一个都有其独特的倾向和问题。开发它们的方式也不同。做这件事的人也很重要，对吧？不同人类之间的动态以及这些人为因素对 AI 今天在我们眼前的发展产生了巨大影响。我认为从 OpenAI 成立之初甚至更早，我们花了很多时间做梦。我们花了很多时间真正思考你可能做的一切影响。我认为我观察到的一件事是，我们并没有真正被沿途的某些时刻惊讶，但我们惊讶于它们何时到来，实现它们有多难，以及我们看到它们的确切顺序。而且我们正在走向的世界，我认为在许多方面比我们预期的许多世界更美好、更令人敬畏。

It is true that at various points in OpenAI's history, we've had some hope that, 'Hey, there are people who have deployed transformative technologies before. Maybe they can tell us the answers.' And it's never been so simple. They do have wisdom and insights, and that's something that I think we've really incorporated, but we realized that we're the closest ones to this technology. That by virtue of creating it, we have an understanding of the ways in which we could shape it. That is hard for someone who isn't so close to it to opine on, to advise on. And I think that one observation I have is that the right choices are extremely specific to the facts of the technology. There's different pressures that are exerted by cell phones versus, you know, mainframe computers versus AI versus electricity. Each one of these has its own unique proclivities and problems. That's ways that they're being developed. The individuals doing it matter too, right? The dynamics between different humans and these human factors have been hugely impactful for how AI is playing out today in front of our eyes. I think that a lot of what we spend our time doing from the beginning of OpenAI and really even before is you spend a lot of time dreaming. You spend a lot of time really thinking about all the implications of what you might do. And I think that one thing I've observed is that we haven't really been surprised by some moments along the way, but we have been surprised by when they arrive, how hard they are to accomplish, exactly the order in which we see them. And that the world that we are moving towards is I think one that is in many ways more wonderful and awe-inspiring than many of the ones we anticipated.

如果一个前沿模型将安全作为首要关注点，而另一个前沿模型不这样做，你认为这种竞争会如何发展？

If one frontier model puts safety as a primary concern and another frontier model doesn't, how do you view that competition playing out over time?

嗯，我们发现安全实际上是一个核心产品特性。就像没有人想要一个与他们不一致的模型。

Well, I think we have found that safety is actually a core product feature. Like no one wants a model that is not aligned with them.

对吧？你希望得到一个值得信赖的模型，在任何情况下都能做正确的事。所以我们投入了——我认为我们实际投入的，可能远超人们的认知，甚至比其他任何实验室在安全方面的投入都多。我们拥有 ChatGPT，这是全球部署最广泛、使用人数最多的 AI 语言模型。我们必须关心。我们一直都很关心，但这一点在我们将这项技术带给这么多人的过程中尤为明显。所以我不认为存在一种可持续的状态，其中构建这项技术并拥有成功产品的人，却没有在安全上大力投入。而且我认为，如果你退一步看，挑战其实在于，交付安全的某些方面并不一定是短期的。你必须为你的业务，以及你所创造的东西，做长远考虑。这其中一部分涉及如何训练模型，一部分涉及如何建立反馈循环。但我只想说，我们将安全作为使命的一部分，我认为这一点已经在我们的产品和世界中得到了体现。

Right? You want a model you can trust that does the right things in any circumstance you give it. And so we have invested I think we've actually invested possibly far more than certainly people perceive and possibly more than any other lab in safety, right? That we have in ChatGPT the broadest deployment of AI these language models in the world used by the most people. We have to care. We've always cared, but you really see it in terms of us being able to bring this technology to so many people. So I don't think that there's a sustainable state where the people who are building this technology and having successful products are not also investing super hard in safety. And I think that actually the challenge is a little bit about if you step back, because there are some aspects of what it means to deliver safety that are not necessarily short-term. You have to think long-term for your not just your business, but for what it is that you're creating. And some of this is about how you train the models, some of this is about how do you get your feedback loop. But I would just say that we are committed to safety as part of our mission, and that's something where I think it has played out in our products and in the world.

人们还忽略了一点：这不仅仅是模型的安全问题，还关乎社会的韧性。看看变革性技术是如何进入世界的——社会围绕它们建立起应对其优势和风险的体系。想想发动机，对吧？你造了汽车，但也需要安全带，需要道路，你围绕这项技术的工作原理重新规划城市。想想电力，你有各种安全标准，有不同地方允许架设电线杆，对吧？还有高压线等等。我认为 AI 也是如此，这不仅仅是技术本身，也不仅仅是模型本身。真正重要的是它们如何与一个具有韧性的社会融合。这是我们正在大力投入的事情。OpenAI 基金会将其作为关键重点之一，试图帮助社会投资并构建一个面向 AI 的韧性层。

One thing that people also miss is that it's not just about the safety of the model, it's about the resilience of society. If you look at how transformative technologies enter the world, that society builds around them about their strengths and their risks. You think about engines, right? You build cars, but you also need seat belts, you also need to have roads, and you reorient cities around the fact of this is how this technology works. You think about electricity, you have various safety standards, you have different kinds of where you're allowed to put the electric poles, right? And high-voltage lines and all these things. And I think the same will be true for AI, that it's not just about the technology itself, it's not about the model itself. It's really about how do they integrate into the world with a society that is resilient. And that is something we're investing in very significantly. The OpenAI Foundation has this as one of its key focuses of trying to help society invest in and build a resilient layer for AI.

你认为 AI 监管应该是什么样的？

What do you think regulation for AI should look like?

嗯，我认为 AI 监管需要实现多个不同目标。其中一个非常重要的目标是，我们最终要确保这项技术惠及人类。想想这些问题：很明显，机构、工作、人们以为稳定的人生道路——这些假设可能不再成立。我们需要确保提供支持，在技术推广过程中互相扶持。那么从监管角度来看，这意味着什么？我认为有很多想法，比如每个人都应该能使用算力。我们如何确保这一点？如何确保随着这项技术开始创造更多经济价值，它不会只集中在某个地方？这应该是每个人都能受益的事情。这项技术不应只是抽象地让经济受益——它显然会做到这一点——而应该直接让人们感受到，他们自己的生活因为这项技术的存在、因为使用它、因为能完成更多事情而变得更好。我认为，要看到这一点如何实现，关键是要立足于我们真正看到的东西。一个很好的例子是，有多少人说他们的生命，或他们亲人的生命，因为使用 ChatGPT 而得到拯救。你会意识到，这是应该被支持和保护的事情。那么，通过监管来实现这一点的一个好例子是考虑隐私和特权。你和医生谈话，和律师谈话，那些是受特权保护的对话，对吧？你愿意分享。法律明确规定了医疗提供者何时必须向执法部门提供信息或提醒他人。目前 AI 还没有类似的规定。但人们正在使用这些工具，他们应该使用这些工具，因为它们对于让人们获得原本无法获得的信息非常重要，而且他们也应该在那里获得适当的理解和保护。所以，我认为有很多事情需要深入思考：这些模型如何融入人们的生活？我们如何确保既能持续创新，同时又能让利益广泛传播？我们如何确保美国保持领先地位？想想机器人技术，我认为我们不是领导者。对于 AI，我们必须确保我们继续保持在已经取得的卓越地位。再想想数据中心，显然有很多担忧，比如它们是否会推高电价？我们承诺不会。我认为这些事情可以通过多种机制实现。有时通过监管，有时通过公司承诺，有时只是通过让人们了解事实。一个很好的例子是数据中心和用水。人们经常谈论这个，但实际上我们的数据中心用水量非常少，对吧？说它们用水很多其实是 misinformation。

Well, I think that there's a number of different pieces to what regulation for AI needs to accomplish. One I think is very important is we need to ultimately ensure this technology benefits people. And you think about questions like like it is very clear that institutions, jobs, just life paths that people thought would be stable, those assumptions may not hold anymore. And we need to make sure that we provide support, that we're all there to support each other as this technology rolls out. And so, what does that mean from a regulatory perspective? I think there's a lot of ideas, whether it's things like everyone should have access to compute. How do we ensure that that's true? How do we ensure that as this technology starts to generate more economic value, that it doesn't accrue to just one place, right? This should be something that actually everyone is benefiting from. This technology shouldn't just abstractly benefit the economy, it's very clearly going to do. It should directly be something that people feel in their daily lives, that they themselves, their life is better because this technology exists, because they're using it, because they're able to accomplish more. And I think that the ways in which I see this playing out, it's very important to ground it in what are we really seeing. Like a good example is the number of people whose who say that their life, or their life was saved, or the life of a loved one was saved through the use of ChatGPT. And you realize that that's something that should be supported and protected. And so, a good example of how you can do that through regulation is thinking about privacy and privilege. You talk to a doctor, you talk to a lawyer, those are privileged conversations, right? You feel comfortable sharing them. There's certain guardrails on when the health care provider would have to, you know, provide that information to law enforcement or alert someone. Well defined in the law. We don't have anything like that for AI right now. But people are using these tools, and they should use these tools because they're so important for giving people access to information that they wouldn't be able to get otherwise, and that they should have the appropriate kind of understanding and protections there, too. And so, I think that there's a lot of just really leaning into thinking about how do these models insert into people's lives? How do we make sure that we can continue to innovate, while at the same time, also making sure that the benefits flow broadly. How do we ensure that America remains a leader, right? That you think about robotics, where I think we are not the leader. I think that for AI, we have to make sure that we continue with this remarkable position that we have been able to achieve. And you think about things like data centers, that those are something where there's clearly been a lot of concern about questions like do they drive up electricity prices? And we have a commitment to ensure that they do not. And I think that each of these things can be achieved through many different mechanisms. Sometimes it's through regulation, sometimes it's through commitments from the company, and sometimes it's just through people understanding the facts. Like a good example is data centers and water usage. Like that that's something that people talk about a lot, but actually our data centers use incredibly little water, right? That's actually misinformation that they use a lot.

比一个家庭还少，对吧？

It's less than a household, isn't it?

是的，因为它是闭环的。你基本上装满一个巨大的——你可以把它想象成一个游泳池的水，然后让它循环。所以水量是固定的，而且不大。但我认为人们真正需要理解的是为什么。我们为什么要建造这些东西？为什么值得？它如何让我受益？能够赋予人们这种力量，无论是帮助他们感觉自己现在可以成为企业家，可以创业，可以创造东西。所有这些我们都必须解决。我们必须确保人们在日常生活中感受到这一点。

It is, because it's a closed loop. You basically fill up a giant like, you know, think of it as like a swimming pool of water, and you just circle it around. And so it's a fixed amount of water that's not very large, but I think people really understanding the why. Why are we building these things? Why is it worthwhile? How does it benefit me? And being able to give people that empowerment, whether it's helping them feel that they can be an entrepreneur now, that they can build a business, that they can create something. Like all of that we have to solve for. We have to make sure that people feel it in their daily lives.

当我告诉人们我要做这个采访时，一个常见的反应是他们担心自己的工作，感到不确定。你会对他们说什么？

When I told people I was doing this interview, one of the common reactions is that they're fearing for their job and their uncertainty. What would you tell them?

嗯，我确实认为这项技术具体会如何发展是不确定的。我认为它的发展方式也会令人惊讶。就像我们现在的 AI，我们现在的世界，并不是科幻小说所预料的。它只是不同。

Well, I do think that this technology it is uncertain exactly how it will play out. I think it is surprising how it will play out as well. Like the AIs that we have right now, the world that we have right now, is not really something that was anticipated by science fiction. It's just different.

而且我认为，一些看似必然的结论，实际上在实现时并不完全如我们所想。所以，我相信人们总是最容易看到自己会失去什么，对吧？变化正在到来，这无可否认，绝对如此。但预先看到你会得到什么却要困难得多。举个例子，想想 Uber。如果你在 1950 年向某人描述它，你得提到计算机、手机、GPS，所有这些只是为了让你能在 3 分钟内叫到一辆车。想想看，为了这样一个用例投入如此巨大的技术投资，其实挺疯狂的。但它确实发生了，而且不仅仅是为了这一个用例，而是为了成千上万、甚至数百万个其他用例。所以，我认为我对 AI 的看法是：它关乎赋能，关乎人的能动性。这确实意味着一些我们曾以为可以依赖的机构、工作等，可能并不像我们想象的那么稳定。因此，它会影响到人们。但关键的问题是：你得到了什么？你如何从中受益？现在，你可以成为一个创造者，你可以创造任何你能想象到的东西，让它们成为现实。那么，你想象什么？你如何培养这种技能？真正深入这项技术。我观察到的一件事是，在这项技术的多代演进中，那些似乎从中获益最多的人，正是那些在上一代技术中就深入其中的人。对吧？所以，你越是培养这种技能，其核心就是能动性、有愿景、有想法。因为现在尝试它们的门槛比以往任何时候都低。所以，我认为会有新的机会被创造出来。我认为世界确实需要思考如何支持每个人度过这个不确定的时刻，度过即将到来的任何转型。因为经济将是一个算力经济，它会不同，但我认为每个人都能找到自己贡献的位置。

And some inevitable conclusions, I think, actually turn out to not quite look the same way when they come to pass. So, I believe it's always easiest to see what you lose. Right? And the change is coming. There's no denying that. That is absolutely the case. But, it's much harder to see a priori what you gain. And as an example, just think about Uber. If you describe it to someone in 1950, you have to think about computers. You have to think about mobile phones. You have to think about GPS. And it's all so that you can get a car to appear where you are in 3 minutes. And like that's actually crazy if you think about that level of technological investment for that kind of use case. But, it really happened. And it didn't just happen for that one use case. It happened for thousands, for tens of thousands, for millions of other use cases. And so, I think that my view of AI is it is about empowerment. It is about human agency. And that that does mean that some of these institutions, jobs, these kinds of things that there will be things that we thought we could rely on that turn out not to be as stable as we thought. And so, it will affect people. But, the question to lean into is what do you gain and how do you benefit from it? So, now you can be a builder. You can create anything you can imagine can become real. Well, what do you imagine? How do you build that skill? Really leaning into this technology. Like one thing that I have observed is across multiple generations of this technology, the people who seem to be getting the most benefit out of it are the people who did it for the previous one. Right? So, the more that you build the skill and at the core of it is agency, is having a vision, is having ideas. Because now the barrier to entry to trying them out is lower than ever before. So, I think there will be new opportunity created. I think that the world does need to think about how do we support everyone through this moment of uncertainty, through whatever transitions will come. Because the economy will be a compute power economy. It will be different, but I think that there will be a place for everyone to contribute.

年轻人今天应该投资什么？如果你在高中、大学或刚刚开始工作，你认为哪些技能在未来会更有价值？

Where should young people be investing today? If you're in high school or university or just trying to start out in a job, what skills do you think will be more valuable in the future?

嗯，我真的认为深入这项技术将是一项关键技能。真正理解如何最大化利用 AI？因为我们都将走向一个世界，在那里我们是智能体的管理者，很快可能成为自主 AI 公司的 CEO。或者想象一下，如果你拥有一家 10 万人公司的全部劳动力，随时听你调遣，为你运作。

Well, I really think leaning into this technology is going to be a critical skill. Just really understanding how do you get the most out of AI? Because we're all going to be heading to a world where we're managers of agents, and soon maybe the CEO of an autonomous AI corporation. Or just imagine if you had the workforce of a 100,000 person company all at your disposal, operating on your behalf.

全天候。

24/7.

全天候，对吧？只要你有 token 和算力来驱动它，我再次认为每个人都需要获得算力。这对世界来说至关重要，必须弄清楚并做对。因为到那时，你可以将算力指向任何问题。而人类想要解决的问题数量是无限的。所以我认为，人们越是深入这项技术，弄清楚如何利用即将到来的东西，如何以新方式组合这些技术，如何与我们的智能体互动来真正管理它们，思考我想要什么、我的自我意识是什么、我的目标是什么、我想在世界上看到什么，实现这些将比以往任何时候都更容易。而且我认为那个世界，我们得到的东西，其好处几乎是难以想象的。

24/7, right? As long as you've got the tokens, the compute to power it, which again I think everyone needs access to compute. That's like so critical for the world to figure out and get right. Because then at that point you can point that at any problem. And the number of problems that humanity could want to solve are boundless. And so I think that the more the people do lean into this technology, figure out how to take advantage of what's coming, how to combine these technologies in new ways, how to interact with our agents to really manage them, to think about when what is it that I want, what is it that is my sense of self, what is my purpose, what do I want to see in the world, it is going to be easier than ever to accomplish that. And I think that that world, the what we gain, I think is going to be almost unimaginable in its upside.

这是对未来最积极的看法。你能想象的最消极的是什么？

That's the most positive sort of view of the future. What's the most negative one you can imagine?

到目前为止，技术发展过程中一个非常有趣的现象是，我们一直在扭曲自己去适应机器。对吧？想想有多少人工作时要对着一个盒子，不停地打字，患上腕管综合征，肩膀佝偻，所有这些都不自然，对吧？那并不是我们身体的设计初衷。而我们正在走向这样一个世界：不再是你用电脑工作，而是电脑真的为你工作。这带来了机遇，也带来了风险。我认为我们需要想办法减轻这些风险。比如，核心的一点是，如果机器帮助人们实现他们的目标，对吧？它们在外面做你想做的事。但有时人们的目标是冲突的。如何解决？如何决定 AI 会帮你做什么、不会帮你做什么的界限？真正要弄清楚的是，这如何融入社会？如何确保利益不仅仅流向一家公司、一群人，而是真正提升所有人？我们需要提高底线，让每个人都能获得美好的生活、这项技术，并能用它做事。我认为相应地也会提高上限。所以我认为我们将进入一个世界，每个人都有新的机会，会有更多——我不知道该用“安全网”还是什么词——但应该有一些东西能确保每个人都跟上。然后我们将能完成更多事情。想想医疗保健之类的事情。如果我们做对了，我们应该进入一个世界，每个人都能获得一个比今天任何医生团队都更好的口袋医生。世界上最好的医生，他们为你服务，关心你，他们真的在读你的病历，24/7 思考如何帮助治疗这种疾病。这是颠覆性的，对吧？这项技术与世界的互动不会没有代价，我们已经看到了最初的错误。但我认为，即使在接下来的两年里，我们将看到它成为一股向善的力量，但我们也必须承认它可能出错的所有方式或风险，才能实现这些好处。

One thing that's very interesting about how technology has played out to date is that it's really been about contorting ourselves to the machine. Right? You think about how many people work where you have this box and you're typing away at it and you're getting your carpal tunnel and your shoulders are hunched and all of those things that were not natural, right? That's not really what we're designed for. And we're going to be moving to this world where it's not just that you're doing work with your computer, so your computer actually does work for you. And that is something that presents opportunities. I think it presents risks. I think we need to figure out how to mitigate those. Like one core thing at the end of the day is that if you have machines that help people actualize their goals, right? That's out there doing what you want. Sometimes people have conflicting goals. How do you resolve that? How do you decide what the bounds are on what an AI will help you with and what they won't? Really trying to figure out how does this slot into society? How do you make sure that the benefits don't just go to one corporation, one set of people, but that actually do lift up everyone. We need to raise the floor so that everyone has access to a great life, this technology, and are able to do things with it. And I think it'll correspondingly also lift the ceiling. And so I think we're going to be in a world where everyone is going to have new opportunities, that there will be more just I don't know if the right word is safety net or just like that there should be something that really is able to make sure that everyone gets brought along. But then we're going to be able to accomplish so much more. And you think about things like access to medical care. Like we should be in a world, if we do our job right, where everyone gets access, has a doctor in their pocket that is better than any team of doctors today. The world's best doctors, they're there for you. They care about you. They're actually reading your charts, that they're thinking 24/7 about how can we actually help with this condition? It's disruptive, right? It's not going to come for free in terms of how this technology will interact with the world, and we've already seen the beginning errors of it. But, I think that what we're going to see over just even the next 2 years, I think it will be this force for good, but we have to also acknowledge all the ways that it could go wrong or the risks of it in order to achieve those upsides.

我们每期播客都以同一个问题结束，那就是对你来说什么是成功？

We always end every podcast with the same question, which is what is success for you?

实现 OpenAI 的使命，确保通用人工智能惠及全人类。

Achieving the OpenAI mission of ensuring that artificial general intelligence benefits all humanity.

非常感谢。这太棒了。

Thank you very much. This was awesome.

这是一次很棒的对话，伙计。

This is a great conversation, man.

谢谢。我度过了一段愉快的时光。

Thank you. I had a great time.