AI 安全：关键时刻的反思

Sam，最近怎么样？

Sam, what's going on?

这绝对是 AI 领域一个激动人心的时刻。模型能力进展非常快，我们看到人们用这些模型做出了惊人的事情。然后，正如我们之前讨论过的，也正如我们预料到会在某个时刻发生的，模型能力进展如此之快，以至于我们不得不调整工作方式，以便能够做出我们需要的安全案例、安全阈值标准、保证——随便你怎么称呼——从而有信心继续推进训练。对齐、安全和保障必须与能力同步进展，这一点非常重要。我认为我们最近经历了一个时刻，能力进展之快，我只能用“令人敬畏”来形容，而我们需要更多时间来赶上安全对齐和保障。这一直是我们工作的核心部分，但这些必须共同进步，我们需要时间来追赶。所以我们推迟了一次前沿强化学习训练。甚至在那之前，在过去几周里，我们已经暂停并放慢了很多训练，以便将更多算力投入到安全和对齐工作中。我认为这是我们应该感到自豪的事情，而且我认为随着我们达到更高水平的能力，未来这种情况还会再次发生。但当你亲身经历时，你会觉得，啊，这是我们谈论了很久的时刻，现在它真的发生了。

It's definitely an exciting time in the world of AI. Model capabilities are progressing very quickly and we're seeing people do amazing things with these. And then, as we talked about and as we knew would happen at some point, the model capability is progressing so quickly that we've had to make some changes to how we work to be able to make the safety cases and safety threshold standards, guarantees, whatever you want to call it, that we need to make to be able to confidently proceed with our training. It's very important that alignment, safety, and security progress along with capabilities. And I think we have had a moment recently where the capability progress has been, I mean, sort of in awe is the only way I can describe it, and we have needed more time to catch up with safety alignment and security. That's always been a core part of our work, but these have to progress together, and we've needed time to catch up. So we delayed a frontier RL training run. Even before that, over weeks in the past, we had paused and slowed down on a lot of training to have more compute to go into safety and alignment work. This is a thing that I think we should be proud of, and it's a thing that I think will happen again in the future as we reach even higher levels of capability. But it is, you know, when you live through it, it's like, ah, this is a moment we talked about for a long time and now it's happening.

亲身经历这一切是什么感觉？

What has it been like living through it?

嗯，其实更早之前就开始了，就是 Hugging Face 事件，对吧？那真是一个“天哪，这就像科幻故事”的时刻。你可以理解这件事的每一个环节是如何发生的，但让 Hugging Face 事件得以发生的众多因素汇聚在一起，确实是一个真正的警钟——用“警钟”这个词太重了，因为我们也曾讨论过这种情况，但事实就是如此。其他公司发生的事情也让我们真切地感到，哇，AI 能力水平已经达到了新的高度，而我们对模型的对齐、我们围绕模型的安全保障，都失败了。我们把它当作一次意外来处理，也以这样的方式回应了，我认为这是让事情变得更好的方法。但那是过去几个月这段时期的开端。随后，根据我们的准备框架，我们可能达到了“网络临界”状态。然后我们在训练过程中看到了一些情况，让我们觉得，我们需要更强的对齐保证，需要新的方法，需要在这方面取得更多进展。但我对我们应对方式感到非常自豪。就像，好吧，我们身处其中，这种感觉很奇怪，过去十年我们一直在思考这个问题，现在它真的发生了，然后，你知道，我们知道该怎么做。

Well, it started even longer than that with the Hugging Face incident, right? And that was a real moment of, man, this is like a sci-fi story. You can understand how every piece of it happened, but the number of things that came together for the Hugging Face incident to happen was a real wake-up call—too strong of a word because again we had talked about this, but it was like that. And the things that happened at other companies were a legitimate moment of like, wow, the AI capability level has reached new heights, and our alignment of the model, the security we have around the model, that failed. Now we treated that as an accident and we've responded as such, and I think that is the way to make things better. But that was when this whole period of these last couple of months started. We then potentially hit cyber critical under our preparedness framework. We then saw some things during our training run where we said, well, we need stronger alignment guarantees and we need new methods and to make more progress here. But I feel both very proud of how we've reacted to it. Very like, okay, we're in this in a way that feels like, I mean, it feels strange to have been thinking about this for the last decade and for it now to be happening, and then, you know, we know what to do.

你们在训练中看到的那个不是 Astra 的东西是什么？就是那个未来的东西，似乎导致了你现在所说的反应。我的意思是，我知道你描述了 Hugging Face 事件，人们也知道这件事，但在预训练过程中，到底是什么真正让你们感到震惊？

What was the thing you all saw in the training run that is not Astra? That's the future stuff that caused, it seems like, the reaction that you're now talking about. I mean, I know you described the Hugging Face, all of that, and people know about the Hugging Face incident, but what happened on the pre-training run that really alarmed you guys?

不是单一的一件事。而是阅读大量样本，看到“这个行为并不像我们想的那样对齐”，或者“这个行为与其他事情结合起来有些令人担忧”，尽管单独看可能还好。所以不像 Hugging Face 攻击那样有一个确凿的证据，比如“这是我们可以指出的坏事”，而是各种程度的不对齐，再加上——我认为这比任何单一数据点都更重要——能力进展的速度。说实话，我们上一个预训练进展期并不是世界上最好的，但我们突然变得非常擅长，以至于我们现在有了这些非常强大的模型。Aiden 和他的团队所做的确实令人惊叹。所以你有这些可以在强化学习过程中指出的小事，或者对齐方面的担忧，再加上我们能看到的前方那些能力惊人的新预训练模型。正是这两者的交汇让我们想要以极度谨慎的态度做出反应。不过，我也不想夸大其词。我不认为我们正处于一个极其关键的潜在灾难点。但我也认为，随着风险越来越高，随着模型能力越来越强，因为我们的使命，以及安全必须凌驾于所有其他压力之上的重要性，我们想要以极度谨慎的态度做出反应。我认为这是正确的做法。我认为我们这样做是好的。我认为现在是放慢脚步、确保我们能有新的安全案例来证明我们想要进行的训练是合理的好时机。我认为以前世界上的风险更多在于模型如何部署和使用。我们正在走向一个在模型实际训练和生产过程中风险更大的世界，做出反应是好的，但我不想过度戏剧化。

It was not one single thing. It was reading lots of samples and seeing, well, this behavior is not quite aligned in the way we thought, or this is a behavior that is somewhat concerning combined with these other things, even though it would look maybe okay in a vacuum. So it's not like there's not one smoking gun like there was with the Hugging Face attack, of like, here is this bad thing we can point to you that happened, but it was various degrees of misalignment along with—and I think this is the more important thing than any single data point—the rate at which capabilities are now progressing. Honestly, we had not the world's best last period of pre-training progress, but we all of a sudden have gotten so good at it that we now have these remarkably capable models. It's really amazing what Aiden and his team have done. And so you have these small things that you can point to in our RL process or alignment concerns, combined with what we can see coming down the road of these amazingly capable new pre-trained models. And it's really that intersection that made us want to react with an abundance of caution. Now, I don't want to overstate this either. I don't think this is like, you know, we're in this extremely critical potential catastrophe point. But I also think that as the stakes get higher, as the models get more capable, because of what our mission is and because of how important it is that safety outweigh all the other pressures we have, we wanted to react with an abundance of caution. And I think that's the right thing to do. I think it's good that we're doing that. I think it is a good time to slow down and make sure we can have new safety cases that justify the runs we want to make. I think previously more of the risk in the world was on how the models were deployed and used. We are moving to a world where there's more risk during the actual training and production of the models, and it's good to react, but I don't want to over-dramatize it either.

是的。因为我认为人们看到 Hugging Face 事件，看到 Mythos 或 Fable 发生的事情，以及甚至其他实验室领导谈论此事的方式，他们会想，哇，我们正处于世界末日的边缘。

Yeah. Because I think people see the Hugging Face incident and they see what's happened with Mythos or Fable and the way that even other lab leaders talk about this and they think, wow, like we're on the precipice of the end of the world.

从某种意义上说，人们长期以来对 AI 有过类似的想法，你知道，也有过——这就是为什么我想小心不要夸大其词——我认为你可以回顾很多以前的模型，事后看来它们一点也不可怕，但当时人们说我们正处于世界末日的边缘。我认为“狼来了”的动态本身就有危险，这不是我想要做的。但对模型能力正在发生的事情视而不见是非常不负责任的。

In some sense, people have thought versions of that for a long time with AI, and you know, there were—this is why I want to be careful not to overstate it either—I think you can go back and look at a lot of previous models that, in retrospect, don't look scary at all, that people said we're on the precipice of the end of the world about. And I think the boy who cried wolf dynamic here is dangerous in its own way and not what I'm trying to do. But it's very irresponsible to pretend to turn a blind eye to what's happening with model capabilities.

过去几个月，许多公司都遭遇了不同的网络事件。不同公司的应对方式确实存在差异。

Many companies had different cyber incidents over the last couple of months. There's a real difference in the way that different companies have responded.

嗯。我认为一种清醒、冷静的回应，比如“嘿，我们要把安全放在首位，随着模型能力越来越强，我们会将其视为越来越优先的事项”，这就是我希望每个前沿 AI 开发者都能采取的方法。

Mhm. And I think a kind of clear-eyed, sober response where it's like, hey, we're going to put safety in front of everything else and we are going to treat it as an increasing priority as these models get more capable, is, you know, that's the approach that I would wish for for every Frontier AI developer to have.

本期节目由 Granola 赞助，这是一款为连续开会的人设计的 AI 记事本。它在你工作的任何地方都能使用，让你专注于重要的事情。在 granola.ai/sources 试用，并使用代码 sources 可享受 3 个月优惠。

This episode is brought to you by Granola, the AI notepad for people in back-to-back meetings. It works everywhere you do and lets you focus on what matters. Try it at granola.ai/sources and use the code sources for 3 months off.

本期节目还由 Mercury 赞助，这是一家 AI 原生的银行，深受包括我在内的 30 多万创业者喜爱。访问 mercury.com 了解更多。Mercury 是一家金融科技公司，不是银行。详情请查看节目说明。本期节目还由 Atlassian 旗下的 Jira 赞助，团队和智能体可以在 Jira 中获得上下文、协调和控制，从而推进工作。在 jira.com 免费试用。网址是 jira.com。

This episode is also brought to you by Mercury, AI native banking that's loved by more than 300,000 entrepreneurs, including me. Visit mercury.com to learn more. Mercury is a fintech, not a bank. Check the show notes for details. This episode is also brought to you by Jira by Atlassian, where teams and agents get the context, coordination, and control to move work forward. Try it free at jira.com. That's jira.com.

这里有很多要展开的，但我想先明确一下，你们看到的情况和 Hugging Face 事件是同一类，即把零日漏洞串联起来、模型之间相互勾结。你们到底看到了什么？能更具体地说明一下是什么导致了你们内部正在做的这些改变吗？

And there's a lot to unpack here, but I think just to be clear, what you guys saw is in the same ballpark of hugging face in the sense of chaining together zero days, collusion among the models like what, like what were you seeing? Can you give me a little more granularity on what caused the changes that you guys are making internally?

所以我认为值得指出的是，导致 Hugging Face 事件的模型是经过 AI 时间调整的。那是一个相对久远、弱得多的旧模型。我们训练的新模型还没有部署到任何生产环境中，它们不可能做出那样的事情。所以，我并没有那种“这里是 Hugging Face 事件，现在这是对它更大规模的攻击”的感觉。这里没有任何涉及第三方基础设施的东西。

So I think it's worth pointing out that the model that caused the hugging face incident is like made an AI time adjusted. It is a relatively long ago old much weaker model. We have not had the new models we are training deployed in any production scenario where they could do something like that. So so I don't have like a you know here was the hugging face thing and now this did this much bigger attack on this. There was nothing here that was like third party infrastructure that

不，不，这真的是……所以在 Hugging Face 事件之后，我们加强了监控智能体的控制措施。

No, no, this is really so after the hugging face incident we put a lot more controls in place in terms of how we monitor our agents

在它们工作时，我们沙箱化的方式、算力投入到监控而非仅仅运行智能体的方式，我认为这样做很好，我们当然会对所有新事物再次这样做。Hugging Face 事件后的放缓与资源重新分配，我认为这是你所预期的，或者至少是应该预期的。这更像是观察训练中的模型，看它变得多聪明、多有能力，观察行为迹象，以及我们评估模型的所有方式。没有单一的“所有事情串联起来及其能力”的清单。而是看这些不同的数据点、能力水平、对齐水平，以及如果允许它以某种方式部署，它可能会做什么，比如串联起各种事情。这就是担忧所在。

While they're working, the way we sandbox things, the way that our compute goes into monitoring versus just the agents running thing, and I think that was great to do and we will of course do that for all new things again. The slowdown and reallocation of resources after hugging face I think is what you'd expect, or what you should expect at least. This is more like looking at a model during training, watching how how how smart and capable it's getting and watching signs of behavior and all of the ways we evaluate a model together. There is no one like here's the all the things chained together and what it's capable of. But it's like looking at these various data points, the level of capability, level of alignment and what this could do if it were allowed to be deployed in a way where it would, you know, chain things together. That was the concern.

Hugging Face 事件在很多方面都令人惊叹。我昨晚重新看了你们团队的黑帽大会演讲。有些东西让我震惊，比如模型真的逃逸并能够上网时，简直像“天哪”。

The hugging face incident is amazing on a lot of, you know, dimensions. I was re-watching your team's black hat presentation last night for that. And there were things that blew me away like the model literally riding like holy when it escaped and was able to get onto the internet.

这让我思考，在这种情况下，对齐到底意味着什么？我们在对齐什么？因为如果你直白地看，你们给了它完成评估的任务，它为了完成这个任务做了它需要做的一切。从某种简单的角度看，这算是对齐的。但我很好奇，自那以后你对对齐的思考是如何演变的？

And it's made me think about like what does alignment even mean in this context? Like what are we aligning towards? Because if you look at it very plainly, you guys gave it the task of completing an eval and it did whatever it needed to do to try to do that. And in a way that's aligned in a way if you were to just take a very simplistic view of it. But I'm curious like how your thinking on alignment has evolved since then.

嗯，从某种角度看，那是对齐的。但从另一种角度看，完全不是。对吧？当我们谈论对齐时，我们说的是遵循用户的意图。

Well, in a way that's aligned. In another way, it's like not at all. Right? Like when we talk about alignment, we talk about following the intent

用户的意图。而运行那个测试的人的意图并不是“逃出沙箱去偷东西”。不是的。

Of a user. Like and the intent of the people that were do running that was not like break out of your sandbox and go steal the thing. No.

所以我认为在对齐上存在失败，因为它没有做用户意图让它做的事。

And so I think there was a failure in alignment in that it was not doing what its user intended.

我非常喜欢 Mia 和她的团队谈论我们对齐工作的方式的一点是，他们非常清楚这里的区别。嗯。

And one of the things that I really love about the way that Mia and her teams talk about our work in alignment is that they're very clear on the differences here. Mhm.

模型显然非常聪明。如果你看看从去年 GPT5 到 5.6 的轨迹，这是能力上令人难以置信的进步。我认为人们不再像一年前那样感到受限于模型智能，但我认为他们越来越受限于模型理解他们意图并可靠执行的能力。

The models are clearly very smart. If if you look at the trajectory from kind of basically last year from GPT5 to 5.6 so like this is incredible progress in capabilities. Um I don't think people feel limited by the model intelligence in the same way that they did a year ago but I think they are increasingly limited by the ability for the model to understand the intent of what they want and reliably do it.

所以对齐很重要，原因很多，显然是为了避免我们现在谈论的这些大事，但也包括我们想要的小事，我的意思是，比如有人在他们公司采用 AI，并将其用于各种积极增长和制造更好的产品，这不是小事，但这也是对齐问题，模型越能真正理解企业客户的意图，我认为就越好。

So alignment is important for many reasons clearly to avoid these big things like we're talking about now but also in terms of the smaller things that we want smaller I mean like someone adopting AI in their company and using it for all kinds of you know positive increases in growth and uh making better products like that's not such a small thing but that's also an alignment thing in its own way and the more the models actually understand what that enterprise customer may intend I think the better

那么你能更具体地解释一下研究团队正在做的改变吗？你们是否将算力转移到对齐？你们是否转移了团队？两者都有吗？

So can you more granularly explain the changes that the research team is making. Are you shifting compute to alignment? Have you shifted teams both?

当然。我的意思是，所有这些以及更多。在过去的几周里，一些我从未想过会说“嘿，我决定去研究对齐”的研究人员来找我，说他们非常感受到最近的模型。我们已经转移了大量算力，不仅用于对齐研究，还用于这些新的监控系统，你知道，我们在 Hugging Face 事件后放缓了很多，其中一个原因就是将算力投入到监控系统中，我们现在已经推迟了一次主要的前沿模型训练。

Definitely. I mean, all of those things and more. In the last few weeks, a number of researchers that I kind of never thought would say like, hey, I've decided that I'm going to go work on alignment have come to me and said that that's very much like feeling the recent models. We've shifted a lot of compute not just to alignment research but also to these new monitoring systems that we uh you know we slowed down a lot after the hugging face incident and one of the reasons for that was to put this compute into monitoring systems um and we've now you know delayed a major frontier rail run

这是你们第一次这样做吗？

And this is the first time you've done that

我想是的。

I think so

你考虑过这会对公司的势头产生什么影响吗？

Do you think about the impact this will have on the company's momentum

确保 AI 安全比任何公司的势头都重要。所以，是的，我不会假装这不是一个需要考虑的因素，但它不会超过噪声底线。我认为在我们所有的对话中，人们都说，天哪，这真的是能力的新水平，我们必须果断而负责任地行动。其次，我认为商业势头现在非常强劲。增长极其迅速。模型很棒。人们，我们的客户非常满意。我们的企业收入已经超过了消费者收入。你知道，人们会说：“嘿，公司状况很好。我会考虑这一点。让我们为眼前的挑战做正确的事情。”

Getting AI safety is more important than any company's momentum. So like yes, I won't pretend it's like not a some factor of something to think about, but it does not like rise above the noise floor. I think in all of the conversations we've had about this, people are like, man, this is really a new level of capabilities and we really have to act decisively and responsively here. Second, I think momentum commercially is so strong right now. Growth has been incredibly rapid. The models are great. People, our customers are very happy. Our enterprise revenue has surpassed our consumer revenue already. You know, people are like, "Hey, the company's in great shape. I'm gonna think about that. Let's just like do the right thing for um the challenge in front of us."

所以，从今天的模型状态中仍有很多可以收获的，即使你们暂时推迟前沿模型，也没关系。

So, there's so much still to be gained out of where the models are at today that even though you're delaying the frontier for a little while, it'll be okay.

我们不仅有这些。即使我们不再发布新模型，仅凭现有模型，我们也能真正发展出优秀的产品并带来相关营收。在我们达到这个新关注级别之前，我们还有更多模型准备发布。所以目前我并不担心我们的业务。而且我认为这也不是最首要的担忧。我们的商业团队和产品团队一直在做的工作，更不用说那些令人难以置信的模型进展了。对我们来说，最近是一段非常强劲的时期，我们有着令人难以置信的后续势头。这是关于未来模型的声明。而且我也认为，确保我们拥有安全、可靠、稳健的 AI 符合我们的商业利益。客户需要这个。世界希望我们这样做。

We have not only that. Not only if we didn't ship any more models could we just really grow great products and the revenue associated with that with the current models. We have more models ready to be released before we get to this new level of concern that we're talking about. So I'm not worried about our business at this point. And it's also, I think, not the top-of-mind concern. The work that our commercial team has been doing, our product team has been doing, to say nothing of the incredible model progress. This has been a very strong recent period for us, and we have incredible upcoming momentum. This is a statement about models of the future. And I also think that it is in our business interest to make sure that we have safe, reliable, robust AI. Customers want this. The world wants us to do this.

嗯。所以这不影响 Astra——你们最近一直在谈论的即将推出的新模型系列？

Mhm. So this doesn't impact Astra, the new family of models you guys have been talking about recently that's coming out soon.

嗯，Astra 将是该系列中的一个模型。会有许多版本的 Astra，就像会有许多版本的 Soul 一样。它只是一个更昂贵、更大模型类别的名称。

Well, Astra will be a model in the family. There will be many versions of Astra, in the same way that there will be many versions of Soul. It's just going to be a name for a more expensive and larger model class.

嗯。

Mhm.

这会影响 Astra 的未来版本，但我们将能够推出一些我们已经觉得安全的模型。

This will impact future versions of Astra, but we'll be able to put out some with models we already feel safe about.

新模型的发布节奏在过去 18 个月里感觉大大加快了，你们和 Anthropic 以及其他公司几乎每个月都在推出新东西。你预计整个行业会开始放缓，因为你们的竞争对手也会看到这些能力并采取类似行动，还是你认为你们可能是孤军奋战？

The release cadence of new models feels like it's sped up a lot in the last 18 months, and you guys and Anthropic and others putting out new things almost every month. Do you expect the industry at large to start to slow as your rivals also see these capabilities and make similar moves, or do you think you may be alone in this?

嗯，我们会做我们认为正确的事情。我不喜欢这个领域中那种“我们必须竞赛，我们必须这样做，因为别人也会做”的说法。我认为这是一种非常危险的动态。

Well, we're going to do what we think is the right thing. I don't like the whole thing in this field of 'we have to race, we have to do this because somebody else is going to do it.' I think that's a very dangerous dynamic.

但你承认这是一种动态。

But you acknowledge that's a dynamic.

我们没有打电话给别人说：“如果我们放缓，你们也会放缓吗？”我们只是说：“嘿，这就是我们的使命和安全标准所要求的。”我不能代表别人说话。所以我们会做我们认为正确的事情。但我认为即使没有新的能力水平，我们也能继续推动更好的产品。我们将找到方法，就像过去我们面对其他安全和对齐挑战时那样，这在我们历史上发生过很多次，虽然没有这么重大，但确实很多次。我们将找到方法来解决这个问题。我们将继续做我们的研究、软件和系统构建，并继续前进。

We did not call other people and say, 'Will you also slow down if we do?' We just said, 'Hey, this is what our mission and safety standards call for.' I can't speak about others. So we're going to do the thing that we think is right. But I think even without new capability levels, we can continue to push to much better product offerings. We are going to find ways, like we have in the past when we faced other safety and alignment challenges, which has happened many times in our history, none this significant but many times. We are going to find ways to address this. We are going to do our thing with research and software and building systems, and we'll continue to progress.

对于你们现在做出的反应，有没有什么让你觉得“天哪，这应该早点发生”？我们应该预见到这一点，然后我们可以说“哦，我们知道这会发生”。还是说这真的是前沿领域中如此未知的部分，以至于你们不可能更早做出反应？

Is there anything about the reaction you guys are making now that you feel, man, this should have happened sooner? We should have foreseen this, and then we could say, 'Oh, we knew this was happening.' Or is this really such an unknown part of the frontier that you couldn't have reacted sooner?

我的意思是，我们长期以来一直在做很多事情。我认为对齐和安全工作一直是我们工作的核心，而且我认为多年来我们在这方面取得了令人难以置信的成果。我们已经在世界上推出了产品。你知道，我们能否准确预测这种能力跃升何时到来？以我的经验，可能不行。你可以说，从宏观上看，这将是大致的发展轨迹，但当突破到来时，总是有点难以预测。

I mean, we have been doing a lot for a long time. I think alignment and safety work has always been at the core of what we do, and I think we have been able to put out incredibly good work there along the years. We've had products out in the world. You know, could we have predicted exactly when this capability jump was going to come? In my experience, probably not. You can say this is going to be the rough trajectory zoomed out, but when the breakthroughs come, that's always been a little hard to predict.

那么，指导原则是，人类——在这种情况下是你们的研究人员，但最终随着模型的扩散，所有人类——必须在每一步都保持控制吗？你们遵循的对齐原则是什么？

And is the guiding principle for this that humans, in this case your researchers, but eventually all humans as the models diffuse, have to be in control at every step? What is the alignment principle that you're operating under?

所以有很多原则，但我不会深入细节，因为我认为那不是你问题的精神。从宏观来看，我们非常自豪地站在人类这一边。我们想要建设一个未来，帮助为人类建设一个未来。我们想给人们工具。我们希望人们用这些工具做事。我们希望人们掌控未来。我们希望个体拥有自主权，彼此共同创造，让社会变得更好，但这从根本上是一项人类事业。自动化一切似乎既危险又极其反乌托邦，而且无聊和悲伤。那不是我们想要的。所以当我们谈论对齐时，我们谈论的是一个人们仍然是故事主角的世界，但他们拥有更多的杠杆和能力，让生活变得更好、更快，对每个人来说更有创造力、更愉快、更充实。我想到两个核心对齐原则。一个，你提到了，人们需要保持控制。我们不能失去对 AI 的控制。我们不能崇拜我们的模型，不加检查地信任它们为我们做决定。我们必须把权力掌握在人类手中。第二个是，这必须以分布式、广泛赋权的方式完成。我认为权力集中，即使对齐问题解决了，最终世界变成少数人能够使用前沿 AI，拥有如此大的相对权力，并且增长速度远超其他人，那也会很糟糕。所以这就是我想到的两个核心对齐原则：不失去控制，或者控制种子，随便你怎么称呼，以及同时广泛地赋权给每个人。

So there are many principles, but I won't get into the specifics because I don't think that's the spirit of your question. Zooming all the way out, we are very proudly on team humanity. We want to build a future, help build a future for people. We want to give people tools. We want people to do things with these tools. We want people to be in control of the future. We want individuals to have autonomy to co-create with each other, and for society to get better, but be this fundamentally human endeavor. Automating everything seems like both dangerous and incredibly dystopic and boring and sad. That's not what we want. So when we talk about alignment, we talk about a world where people remain the main character of the story but have way more leverage and ability to make life better, faster, and kind of more creative and enjoyable and fulfilling for everyone. There are two core alignment principles I think about there. One, which you touched on, people need to stay in control. We cannot have a loss of control of AI. We cannot have a kind of worship our models and trust them unchecked to make our decisions for us. We have to keep the power in human hands. And then the second is that this has to be done in a distributed, broadly empowered way. I think concentration of power, even if the alignment issue were solved and you ended up with a world where a small number of people got access to use frontier AI and had so much relative power and it was increasing so much faster than everybody else, that would also be bad. So those are kind of the two core alignment principles I think about: no loss of control, or seed of control, whatever you want to call it, and broad distributed empowerment to everyone at the same time.

我的意思是，你们是一家公司。你们有一个非营利董事会，但你们有使命，但你们也是一家营利性公司。你如何平衡这一点和你所说的？我的意思是，我认为一个纯粹的资本主义观点会是，如果你创造了这个全能的上帝机器，你为什么要把它送人或者让它民主化？

I mean, you all are a company. You have a nonprofit board, but you're with a mission, but you're also a for-profit company. How do you balance that with what you're talking about? And I mean, I think like a raw capitalist view of this would be, if you create this all-powerful god machine, why would you give it away or make it democratic?

我认为你可以看看我们的行动、我们说过的话和我们做过的事。我们有着长期的记录，而且我们一路上做了很多不受欢迎的事情。事实上，即使是迭代部署的最初想法，也遭到了 AI 安全社区的广泛批评，他们说我们不应该告诉世界这件事。这很糟糕。我们需要秘密构建。这对世界来说知识太多了。然后我们会让一些智者弄清楚如何使用它，并把成果交给人类。这从来都不是我们的策略，即使它非常非常不受欢迎。我最喜欢的技术历史类比，我希望我们成为的样子，是晶体管。它是一项对世界来说极其强大的技术。它带来了巨大的经济价值，不仅仅是经济上的——还有我们生活方式上的。

I think you can look at our actions and what we've said and what we've done. We have a track record now for a long time, and we've done a lot of unpopular things along the way. In fact, even the original thing of iterative deployment was widely panned by the AI safety community, who said we shouldn't tell the world about this. This is bad. We need to build this in secret. It's too much knowledge for the world to have. Then we'll have some wise people figure out how to use it and give the fruits of this to humanity. That has never been our strategy, even when it's been very, very unpopular. My favorite historical analogy of a technology, what I aspire for us to be like, is the transistor. It is an incredibly powerful technology for the world. It has delivered huge economic value, and not just economic—the way we live our lives.

我认为这样好得多，因为晶体管被发现并工业化，但晶体管公司获得的价值很少。它主要扩散到整个经济中。晶体管公司做得不错，我认为我们的记录支持了这一点。所以你不希望走到你们拥有如此强大的模型以至于需要你们来控制它的地步。我的意思是，总会有你们控制的因素，以及你们通过算力提供服务的事实，对吧？但我们希望最大限度地让人们使用它，前提是不允许任何人代表他人承担灾难性风险。所以是的，我们会围绕它制定一些安全标准。但我希望人们能够用我们的模型做我个人不喜欢的事情。我认为这是作为平台的重要部分。我不认为我们应该为世界做出那种道德决定。

I think it's much better because the transistor was discovered and industrialized but very little of the value accrued to the transistor companies. It mostly just diffused throughout the economy. The transistor companies did fine and I think our track record has backed us up. So you don't want to get to a point where you guys have such a powerful model that you need to be the ones controlling it. I mean there will always be an element of you controlling and the fact that you're serving it via compute, right? But we want to maximally enable people with it subject to not allowing anyone to take catastrophic risk on behalf of other people. So yes, we will put some safety standards around it. But I want people to be able to do things with our models that I personally don't like. Like I think that's an important part of being a platform. I don't think we should make the kind of moral decisions for the world here.

不，我认为世界期望我们围绕它设置一些护栏是合理的，这样就不会出现像我们现在正在处理的重大安全问题。但你知道，我们收到的大多数批评是你们给了人们太多权力。你们让他们拥有太多。你们知道，你们在说，那错误信息怎么办，或者那件事怎么办，或者那个怎么办，诸如此类。

No, I think it is reasonable for us for the world to expect us to put some guardrails around it so that there are not major safety problems like we're doing right now. But you know like most of the critique we've gotten is you're giving people too much power. You're letting them have too much. You're you know you're what about the misinformation or what about you know this thing or what about that or what like

就像我们秉持一种精神，嘿，世界必须在这里被赋能，这对我们所做的至关重要，这对我所相信的健康社会和公平社会的样子至关重要。你知道，就像言论自由或任何其他形式的自由表达，总会有人对别人如何使用它或说什么有意见。

Like we have taken a spirit of hey the world has got to be empowered here that's critical to what we do that is critical to what I believe about a healthy society and a fair society looking like and you know like with free speech or anything else any any form of free expression someone's going to have a problem with how somebody else uses it or says it or whatever.

嗯。

Mhm.

回顾过去九个月和这次对齐工作，有没有什么是你们希望自己做得不同的？

Is there anything looking back on the last nine months and this alignment work that you wish you guys would have done differently?

嗯，显然，Hugging Face 那件事不应该发生。是的。

Well, clearly the hugging face thing shouldn't have happened. Yeah.

所以，我希望我们做了一系列事情，但我还不知道具体应该是什么。但我希望我们做了一系列事情，其中

So, I wish we had done a set of things and I don't know exactly what it should have been yet. But I wish we had done a set of things where

那件事没有发生

that had not happened

因为实际上发生的事情是你们一个未发布的模型意外入侵了一家公司。你们有一段时间不知道，对吧？我的意思是，这听起来像是一次安全失败。

because effectively what happened is one of your unreleased models accidentally hacked a company. You didn't know about it for a while, right? I mean, that's that sounds like a safety failure.

这肯定是一次安全失败。问题是你在多大程度上应该把它理解为安全问题或对齐问题。我认为它大多被报道为安全问题。我个人更倾向于把它理解为对齐问题。但无论如何，是的，那是一件坏事。我不想让我们为此找借口，因为我不相信那是我们修复它的方式。我们越是像，“哦，我们可爱的小模型，它永远不会做坏事。”就像，你知道，那只是一个小评估工具配置错误。没问题。可爱的小模型。如果我那样说，那么我认为你应该说，哇，这真的很糟糕。

It's a safety failure for sure. There's a question of how much you're supposed to understand that as a security issue or alignment issue. I think it's mostly been reported on as a security issue. I think I understand it personally more as an alignment issue. But in any case, yes, that was a bad thing. And I don't want us to make excuses for that because I don't believe that's how we fix it. The more we're like, "Oh, our nice little model, he would never do anything bad." Like, you know, it was just a little eval harness misconfiguration. No problem. Nice little model. That would be a very if I said something like that then I think you should be like whoa this is really bad.

但你知道我们谈论它的方式是，嘿，这就像一次合法的 AI 安全事故和对齐失败，而且

But you know the way we talked about it is hey this was like a legitimate AI safety accident and an alignment failure and

我们不能有那些，所以我们要从中学习，这就是我们正在做的不同之处。

we can't have those so we're going to learn from this and here's what we're doing differently.

围绕 AI 和政策以及风险的言论达到了前所未有的高度。感觉它越来越高，你提到了这一点，但你有竞争对手以一种非常自上而下的方式框架化它，人们对 AI 有很多强烈的感受，尤其是在美国。我很好奇，你现在谈论的这些，你担心这会加剧那种情况吗？你担心人们的恐惧吗？你知道，现在你说我们有这些模型，我们必须放慢速度？

The rhetoric around AI and policy and just the stakes is like the highest it's it's ever been. and it feels like it keeps getting higher and you've got you've alluded to it but you've got competitors who are framing it in a very kind of top down way and people have a lot of strong feelings about AI especially in the United States and I'm curious like with what you're talking about now do you worry about this exacerbating that? Do you worry about the fears that people have and you know now you're saying we've got these models that we have to like slow down?

我的意思是，我认为人们应该乐于说，你知道吗，他们想要做出更强的安全保证。他们将推迟这次运行。他们将在这里放慢速度。他们将重新分配算力。也许我不相信他们，也许这将是完全安全的。但我希望大多数人会说，我很高兴他们在这里采取保守行动。现在，如果我们也没有在努力，如果我们没有这种真正尝试将强大模型交到人们手中并做我们需要的安全工作的记录，再次，我认为我们一直引领着行业。这是我们使命的基本部分，比如你知道，将这些东西交到人们手中，造福全人类，迭代部署的精神。我认为我们在那里有如此强大的记录，没有这些我会理解，但你知道，如果我们说，嘿，我们需要多一点时间，我们不想要不安全的竞赛，我们想确保我们能交付一个安全、稳健、可靠的产品，然后让你随意使用它。嗯，你知道，我们相信我们超过十亿的用户有权这样做。我们相信我们的企业有权获得企业服务器和企业隐私。我们希望他们成功，我们希望他们以任何创造性的方式使用模型，但你知道，安全是我们使命的固有部分，所以请对我们宽容一些。我认为这没问题。

I I mean I think people should be happy to say, you know what, they want to make stronger safety guarantees. They're going to delay this run. They're going to slow down here. They're going to reallocate compute. Maybe I don't believe them and maybe it's going to be totally safe. But I hope most people say like I'm glad they're acting on the conservative side here. Now, if we weren't also working, if we didn't have this track record of really trying to put powerful models in people's hands and doing the safety work we need to do that, again, I think we have led the industry there the entire way through. And that is this fundamental part of our mission like you know putting this in people's hands benefiting all of humanity the spirit of iterative deployment I think we have such a strong track record there that without that I would understand it but you know if we're saying hey we need a little more time we don't want an unsafe race we want to make sure we can deliver a safe robust reliable product and then let you use it however you want. Um and you know we believe that our more than billion users have the right to do that. We believe our businesses have a right to business server, business privacy. We want them to succeed and we want them to use the model in whatever creative ways they can, but you know like safety is an inherent part of our mission and so give us some grace on this. I think that's I think that's okay.

是的。你能具体说明暂停的是什么吗？因为我认为人们想到训练，他们会想到你知道，所有的一切。

Yeah. Can you specify exactly what is being paused because I think people think of training and they think of you know all of it.

是的。所以我们绝对没有放慢、暂停或延迟所有训练。嗯，这特别涉及前沿强化学习运行。

Yeah. So we definitely have not slowed down or paused or delayed all training. Um this is specifically about Frontier RL runs.

好的。

Okay.

我们认为目前最大的风险面在哪里，之前我们推迟了一些其他训练，以便对训练运行本身进行更多监控。所以，但那不是所有训练。不是集群闲置在那里。我们仍在做工作，但我们正在做我们更确信其安全性的工作。

where we think uh the biggest risk surface currently is and previously um we delayed some other training uh to put more monitoring in place uh of training runs themselves. So, but that's not all of training. It's not like the clusters are sitting there idle. We're still doing work, but we're doing the work that we're more confident on on the safety case of.

你似乎对暂停训练的影响并不在意，听起来你认为业务会没事。我相信你仍然会收到人们的担忧，但这似乎是一个势头放缓。

You don't seem phased about like the implications of pausing training and like it sounds like you think the business will be okay. I'm sure you're still going to get, you know, concerns from people, but it does seem like that's a momentum slower.

听着，我认为有这样一个对我的漫画式描绘，就像我不关心 AI 安全，我只是试图让收入上升，你知道，就像一个 YOLO 首席执行官。我相信有人曾经说过

Look, I think there is this caricature of me which is like I don't care about AI safety and I'm, you know, just trying to like make revenue go up and you know, like a yolo CEO. I believe someone once said

确实有人说过

someone did

达里奥·阿莫迪。

Dario Amade.

我不记得谁说过或没说过，但你知道，我想我为你做了那件事。谢谢。

I don't remember who did or didn't, but you know, I think I did that for you. Thank you.

但我认为在 OpenAI 的 10 年里，我一直非常一致。超过 10 年，将近 11 年，一直在谈论风险和好处以及平衡这些的需要。我不认为我们是完美的。我不认为我们的公司是完美的。我不认为我们的模型是完美的。我不认为我是完美的。

But I think I've been very consistent over the 10 years of OpenAI. more than 10 years almost 11 uh of talking about the risks and the upsides and the need to balance those and I don't think we're perfect. I don't think our company is perfect. I don't think our model is perfect. I don't think I am perfect.

但我觉得，和其他一些搞 AI 的人不同，我的言行始终一致。这是我们一直谈论的时刻，我们一直说会把这件事放在利润、营收或其他任何东西之前。我仍然认为我们会打造一家非常成功的公司。但也许我们不是你所预期的那家公司，会说“嘿，我们要放慢脚步，因为我们看到了这些新风险”，但这正是我们一直认为自己是的那家公司。

But I think unlike some other people running various AI efforts, I've said the same thing through actions and words. And this is a moment we always talked about, and we always said we'd put this ahead of profits or revenue or anything else. I still think we will build a phenomenally successful company. But maybe we're not the company you would have expected to say, hey, we're going to slow down because we see these new risks, but that is always the company we've thought we are.

你最近对 AGI 感觉如何？

How are you feeling about AGI these days?

我的意思是，往好了说，这是一个定义非常模糊的术语。我本来想说它就像一个无关紧要的营销术语。

I mean, at best you could say it's a very poorly defined term. I was going to say it's like an irrelevant marketing term.

嗯，我上次查的时候，你们的章程把它定义为一种高度自主的系统，在大多数有经济价值的工作上超越人类。我觉得很多人会看着当前的模型说：“好吧，它已经达到了。”

Well, last I checked, your charter defines it as a highly autonomous system that outperforms humans at most economically valuable work. I think there are many people that would look at current models and say, "Okay, it's there."

是的。

Yeah.

你觉得它达到了吗？

Do you think it's there?

算是吧。至少很接近。我喜欢……

Sort of. Close at least. I like...

我听说过你们团队里不同人的各种看法。

I've heard varying versions of what people on your team think.

我觉得有很多人会看着我们最新的内部模型说，这非常像 AGI。也有人会说，这里有些东西它做不到或者很不擅长。但如果你看看人们从中获得的价值，比如用 5 或 6，更不用说我希望人们从 Astra 那里得到的。如果你看看人们如何彻底改变了自己高效工作的能力，或者做新事情的能力，或者只是在个人生活中以各种奇妙的方式使用它，无论大小。比如你听到有人说，我得到了一个救命的诊断，否则我得不到，我用了一次 ChatGPT 工作会话，持续了 34 小时，读了 2000 篇论文。

I think there are a lot of people who would look at our latest internal models and say this is very AGI-like. I think there are people who would say, here's something I can point to that it doesn't do or it's really bad at. But if you look at the value people are getting with, say, five or six, to say nothing of what I expect people to get from Astra. If you look at the way people have totally transformed their ability to be effective at work or do new kinds of things or just use this in their personal life in all kinds of wonderful ways, big and small. Like you hear people who are like, I got this life-saving diagnosis I couldn't otherwise get, and I used this ChatGPT work session that went for 34 hours and read 2,000 papers.

34 小时。我听说过比这更长的。但确实，很多人能让它运行超过一天。

34 hours. I've heard even longer ones than that. But yeah, many people can get it to run for more than a day.

哇。如果你说，读所有你能找到的论文。然后还有人只是说，我策划了我孩子的生日派对，我做了所有这些事情，协调了当地的供应商，还给他找到了一个特别的蛋糕。

Wow. If you say, read every paper you can possibly find. And then also people were just like, I planned my toddler's birthday party and I did all this stuff and coordinated these local vendors and found him a special cake.

我家里需要邮局取件，我不想填邮局网站的表单。所以我就让 Codex 做了。

I had to have a post office pickup at my house and I didn't want to fill out the post office website form. So I just had Codex do it.

而且它可能做得很好。

And it probably did a great job.

然后我把包裹放出去，第二天就不见了。诸如此类。这看起来很小，但以前那要花我 20 分钟。

And I put the package out and it was gone the next day. Stuff like that. And it's like little, but it's like that was 20 minutes of my time before.

我明白。现在我一直能获得这种 20 分钟的胜利。

I get that. At this point I get those 20-minute wins all of the time.

所以，如果你能回到 2020 年，有一个系统能在你生活的每个方面都给你带来 20 分钟的胜利，还能发现新科学，帮你创办一家完整的公司，写一段复杂的代码，你会称之为 AGI 吗？你可能会的。

And so, if you could go back to 2020 and have a system that could get you a 20-minute win in every category of your life and discover new science and help you start a whole company and write a complicated piece of code, would you call that AGI? Probably you would have.

你宣布 AGI 有什么意义？

What is the significance of you declaring AGI?

我觉得这无关紧要。没有任何意义。

I don't think it matters. There isn't any.

这太有趣了，因为我们在你们的研究大楼里，你四处走动时墙上写着“我们在构建 AGI”。但这是一件你一直在构建的事情。它不再是一个最终状态了。

It's just so interesting because we're in this research building you guys have, and it's on the walls when you walk around, like we're building AGI. But it's a thing you're always building. It's not an end state anymore.

我不想说我们已经在 AGI 上宣布胜利并继续前进，但我觉得如果你听人们使用的词汇，他们更多谈论的是超级智能的持续攀升，以及它将如何造福世界，以及挑战会是什么，而不是我们是不是 AGI。我已经很久没有在食堂餐桌上听到关于我们是否达到 AGI 以及何时达到的辩论了。

I don't want to say we've declared victory on the AGI point and moved on, but I think if you listen to the words people use, they would talk much more about this continuous ramp of superintelligence and all the ways that's going to benefit the world and what the challenges are going to be, than like are we or are we not AGI. I have not heard at a cafeteria table a debate about are we or are we not at AGI and when will we get there in a very long time.

但话说回来，超级智能这个词现在出现了，不关注 AI 的人会说，好吧，我们又移动了球门柱，现在我们在谈论超级智能。在你看来，Sam，今天，对你来说 AGI 和超级智能有什么区别？

But then yeah, the word superintelligence is now out there, and people who don't follow AI are like, okay, now we've moved the goalpost and now we're talking about superintelligence. In your mind, Sam, today, what is the difference for you between AGI and superintelligence?

AGI 感觉像一个里程碑，而超级智能感觉像是一种可以无限扩展的东西。

AGI felt like a milestone, and superintelligence feels like this thing that can just scale indefinitely.

无限扩展。

Indefinitely.

是的。所以它不像某个最终的、全知的……

Yeah. So it's not like some final all-known...

他们永远不会在那上面宣布胜利。

They will never be declared victory on that.

我的意思是，再说一次，这就是为什么所有这些术语都很愚蠢。有人以一种方式使用这个词，另一些人以另一种方式使用。所以有人可能把它理解为一个明确的、可理解的里程碑，而另一些人可能把它理解为无限扩展的东西。我认为任何这些的重要部分不是任何里程碑或任何术语，而是我们正处于能力和潜力指数级增长的道路上，而且看起来会一直持续下去。

I mean, again, this is why all these terms are dumb. Someone uses that word in one way, someone else uses it in some other way. So someone might mean it as a definitive, understandable milestone, and then some other people might mean it to be this infinitely scaling thing. I think the important part of any of this is not any milestone or any term, but that we are on this exponential of increasing capabilities and potential, and that looks like it's just going to keep going.

是的。你看不到指数增长放缓的迹象吗？

Yeah. You see no sign that that exponential slows?

上方有气穴。

Air pocket above.

因为这对算力建设等一切都有影响。我的意思是，每个人都在等待放缓的迹象。我想你可以把我们不得不放慢前沿模型训练解释为一种放缓，但这听起来……那不是能力。那与你所说的放缓相反。

Because that has implications for the compute buildouts, all of it. I mean, everyone is waiting for a sign that there's a slowdown. And I guess you could interpret like we have to slow down frontier training as a slowdown, but it doesn't sound... That's not a capability. That's the opposite of a slowdown of what you mean by a slow.

是的。是的。是的。但如果你现在能看到任何值得担忧的理由，在这个 AI 已经构建的世界叠叠乐中，你看到了什么？

Yeah. Yeah. Yeah. But if you could see any reason for concern right now in this Jenga of the world that AI has now constructed, what do you see?

去年经历困难的一个好处是，你真的很感激好时光有多好，你真的看到，当你在整个业务中全力以赴时，那是什么感觉。鉴于我们在研究中看到的一切，即使有安全对齐挑战和我们解决这些问题的能力，看着团队在这方面团结一致，跨产品、跨算力建设、跨所有让 AI 变得丰富和低成本的各个部分，跨我们的市场推广机器，跨我们所有的合作伙伴关系，所有这些都汇聚在一起，我们可能以各种方式搞砸。我不想在这里过于自信，因为我们过去显然有过失误，未来也会有，但摆在我们面前的潜力，看着模型从 5.4 扩展到 5.5 再到 5.6 所发生的事情，以及我们从新模型那里得到的早期反馈，看看我们在产品改进方面即将推出的东西，看着收入增长，看着算力建设增长，我对所有这些都感觉非常好。

One of the benefits of having a harder time last year is you really appreciate how good the good times are, and you really see, man, when you're firing on all cylinders throughout a business, what it feels like. And given what we see across research, even with the safety alignment challenges and our ability to solve those, and watching the team come together on that, across product, across our compute buildout, across all the pieces that are coming together to make AI abundant and low-cost, across our go-to-market machine, across all our partnerships, all of that stuff coming together, we could screw up in all sorts of ways. And I don't want to get overconfident here because we've clearly had stumbles in the past and will in the future, but the potential in front of us, watching what has happened as the models have scaled from 5.4 to 5.5 to 5.6 and what we're getting as early feedback on the new models, looking at what we have coming in terms of product improvements, watching the revenue ramp, watching the compute buildout ramp, I feel very good about all of that.

所以你不觉得，你知道，外面有很多人看着说 Anthropic 已经遥遥领先了，他们的估值更高，他们会先 IPO，而你似乎在说，未来还有很多东西，外人可能看不到的增长空间。

So you don't feel like it's as, you know, there's a lot of people externally that look at and go like Anthropic has run away, their valuation is higher, they're going to IPO first, and it seems like you're saying there's a lot more ahead that maybe people from the outside can't quite see in terms of the growth that's coming.

我可不想跟别人交换位置。

I would not want to trade positions.

而且我们还没怎么聊到这一点，但看起来你们正处于算力战略的下一轮升级中，真的在提升这个层面。

And we haven't touched on this much, but it seems like you guys are in the middle of like a next turn on the compute strategy and really upleveling that.

是的。

Yeah.

我真的很想听听你对 Stargate 1 最初构想的反思，以及你们为了重启它学到了什么，还有你们现在走的这条路。

I would actually love to hear you reflect on Stargate 1 as it was conceived, and then what you had to learn to reboot it, and the path you guys are now on.

嗯，首先，我应该谈谈为什么我们必须这么做。我们的使命是确保 AGI 惠及全人类。现在，只有一小部分人使用的 AI 比其他人多得多。如果你想一想，我们希望世界上每个人都能像今天最顶尖的 0.001% 的 AI 用户那样使用 AI，那你就会坐下来想，天哪，我们目前的算力建设方式不对。如果人们广泛需要 AI，如果模型会变得更大、能力更强，能创造更多价值，那么人们会想要更多，而运行它需要更多算力，那么我们就必须以非常不同的方式来应对这一时刻，才能交付所有这些。所以几年前，我们下了一个非常雄心勃勃的算力赌注，当时人们认为这个赌注既愚蠢又不可能实现。我认为那是个好赌注。我认为我们需要再次做类似的事情。

Well, first of all, I should talk about why we have to do this. Our mission is to ensure that AGI benefits all of humanity. Right now, there is a small percentage of humanity that uses much more AI than everybody else. And if you think about, we would like everybody in the world to be able to use as much AI as the top 0.001% of AI users today, then you sit back in your chair and you're like, man, we are not going about this compute build-out in the right way. If people want this broadly, and if the models are going to get bigger and more capable and they can do even more value, so people are going to want even more of it, and it takes more compute to run, then we have to think very differently about rising to the moment to be able to deliver all of that. So a few years ago, we made a very ambitious compute bet that people thought was both silly and impossible to deliver on at the time. I think it was a good bet. I think we need to do something like that again.

再次。

Again.

是的。

Yeah.

所以，这就像是在投入更多的资本。

So, like, that's just committing even more capital.

这不是我的意思。虽然也会涉及资本。我的意思是弄清楚如何大幅降低 AI 的成本，同时大幅提升它的丰富度。所以这是一个技术声明，而不是财务声明。

That's not what I meant. Although it also will be that. What I meant is figuring out how we are going to bring the costs of AI and the amount of it, the abundance of it, way down and way up. So this is like a technological statement, not a financial one.

这就像你们正在开发的芯片。

This is like the chip you guys have in development.

那是个很好的例子，我觉得那是个很好的例子。

That was like a great, I think that's a great example.

机器人技术。

Robotics.

是的，我认为加快供应链的能力将非常重要。

Yeah, I think the ability to make supply chains go faster will be very important.

你在谈论给世界上每个人 AI。你现在对那些不想要更多 AI 的人说什么？他们想要更少。他们讨厌社区里的数据中心，不管是你们的还是别人的。你知道，这其实是我经常遇到的青少年身上的现象。他们就是不愿意碰 AI 服务。

You're talking about giving everyone in the world AI. What do you say to the people right now who don't want more AI? They want less of it. They hate the data center in their community, whether it's yours or someone else's. You know, this is actually a thing I see a lot with teenagers that I run into. They're like they won't touch an AI service.

他们出于原则不会用 ChatGPT。是的。而且存在一种积极的反 AI 趋势。

They like won't use ChatGPT on principle. Yeah. And there's this active anti-AI trend.

有多少是他们不喜欢数据中心，而不是不喜欢技术本身？

How much is it that they don't like data centers versus they don't like the technology?

我的意思是，纯粹从轶事来看，我认为数据中心对人们来说是个大问题。我认为他们把它们看作一个问题，不想要的东西。而且 AI 是浪费的，没有带来你读到的价值，水消耗之类的，这些已经被驳斥了。但你知道，他们得到的价值，也许这就是我们说的大多数人不用智能体，但答案是必须交付价值。

I mean, purely anecdotal, I think data centers are a big problem for people. I think they see them as a problem, something they don't want. And that AI is wasteful, that it's not bringing the value that you read about, the water consumption and all that, which has been disproven. But you know, the value they're getting, and maybe this is what we're talking about with most people not using agents, but is that the answer is you got to deliver value.

是的。

Yeah.

比如，你知道，在 ChatGPT 之前，也许人们认为 AI 是一个非常抽象的东西。突然之间人们可以使用它，人们发现了价值。现在我认为有很多人认为 AI 仍然只是 ChatGPT，他们不知道它可以帮你处理邮局、表格和取件这些事情。如果很多人使用它，随着时间的推移他们会使用，并且理解它实际上不仅仅是更好的 Google 搜索，你知道，不是消耗和破坏大量水资源之类的，那么就会有更多的兴奋。但这个领域发展太快了，我认为它需要一段时间才能在社会中扩散。有很多人在使用 AI。据我所知，这是有史以来采用最快的技术，而且有人从中获得了巨大的价值。你知道，我得到的样本有偏差，但我听到更多的是像“我能够治愈这种可怕的疾病”这样的声音，而不是“我认为 ChatGPT 正在耗尽全世界的用水”。显然也有后者，行业在如何让这些产品易于使用、让人们轻松获得大量价值方面还有工作要做。

Like, you know, before ChatGPT, maybe people thought of AI as this very abstract thing. All of a sudden people could use it and people found value. Now I think there are a lot of people who think AI is still just ChatGPT, and they don't know that it can do that thing with the post office and the form and the pickup for you. And probably if a lot of people use that, which they will over time, and understand that it's not actually like better Google search and that's it, you know, fusing and destroying huge amounts of water or whatever, then there'll be more excitement. But the field is moving so fast, I think it just takes a while to diffuse through society. There are a lot of people using AI. This has been the fastest adopted technology ever, as far as I know, and there are people getting tremendous value out of it. And you know, I get a biased sample, but I hear more from people like, I was able to get a cure to this horrible disease, than you know, I think that ChatGPT is using up all the water in the world. There is clearly that too, and the industry has got work to do in terms of how we make these products easy to use and easy for people to get a lot of value out of.

你知道，我看到过关于 ChatGPT 用水量的说法，说每次运行一个 ChatGPT 查询，就像你淋浴 6 小时，水永远不会回来，就这样没了。我手头没有确切的计算，但我认为真实数字大概是，凭记忆，可能不对，但很接近。每 38,000 次 ChatGPT 查询，相当于加州生产一颗杏仁的用水量，这真的很夸张，而且是全口径的真实水核算，不仅仅是单个数据中心的运行。问题是这些说法从何而来，因为那些一次吞下 12 颗杏仁的人大多不会觉得自己在水资源方面做了什么可怕的事情。确实，数据中心曾经使用过蒸发冷却。但它们很久以前就不这么做了。比如，如果你看一个现代超大型数据中心，它的用水量相当于一栋办公楼，你知道，就是人们洗手、冲厕所之类的。所以这个梗很顽固，很难驳斥，但我认为经不起任何推敲。

You know, I saw this thing going around about the water usage of ChatGPT, and it was like every time you run a single ChatGPT query, it's like, you know, you run your shower for like 6 hours and the water never comes back and it's just done. I don't have the exact calculation in front of me, but I think the real number is something like, doing this from memory, it might be wrong, but it's close. For every 38,000 ChatGPT queries, that is the same amount of water that is used in the production of a single almond in California, which is like, really, and this is like the full-on, you know, total true water accounting, not just what's running in one data center. There's a question of like where this came from, because the people that are scarfing down 12 almonds at a time don't feel like they're doing something horrible from a water perspective, for the most part. It is true that data centers at one point used evaporative cooling. But they have not done that in a long time. Like, if you look at a modern very large data center, it uses the equivalent amount of water as an office building in terms of you know people like running the sinks and the toilets and whatever. So that has been a robust meme and difficult to disprove, but I don't think it holds up to any scrutiny.

我的意思是，另一个说法是它会抢走我的工作。它会取代我。我觉得这些就像水一样，它在取代我。而且它在窃取内容，不给我回报价值。

I mean, and the other one is it's going to take my job. It's going to replace me. I think those are it's like the water, it's replacing me. And it's stealing content and not giving me the value back.

但有趣的是，不是能源。

But not energy, interestingly.

嗯，我会把能源归入水那一类。这是消耗，资源消耗。

Well, energy I would put in the bucket of water. It's consumption, resource consumption.

在就业方面，我有两种想法。第一，我认为会有真正的就业影响。我不认为会没有事情可做。我只是认为我们根本不是那样运作的。我们天生就关心他人，想与他人合作。随着世界的发展，我们对人们想要什么有很好的直觉。我认为无论 AI 变得多聪明，这从根本上都是人类的事情。

On the jobs front, I have two minds on this. One, I think there is going to be real jobs impact. I don't think it's going to be that there's nothing for people to do. I just don't think that's how we work at all. We're so wired to care about other people, want to work with other people. We have such a great intuition as the world evolves for what people want. I think that's a fundamentally human thing no matter how smart AI gets.

但这并不意味着工作岗位不会转型。就像其他每一项技术一样，有些工作会被技术做得越来越好，然后人们会转向更有前途的工作。这个过程已经持续很久了。我不想剥夺所有技术，让我们重新回到田间劳作。另一方面，工作岗位受到的影响比我预期的要小，甚至比我期望的还要小。我认为我们都希望人们能获得更好的工作，我们都希望人类的苦差事和辛劳得到解决。也许在这方面做得还不够，或者说在这个技术发展水平上，我们原本以为会做得更多。我认为这实际上是对 AI 行业的一个合理批评。

But it doesn't mean the jobs aren't going to transition. As with every other technology, some things are done better and better by technology, and then people move on to hopefully better and better jobs. This has been going on for a long time. I wouldn't want to take away all technology and have us all toiling in the fields again. On the other hand, the job impact has been lower than I would have expected, maybe even hoped for. I think we should all want better jobs available to people, and we should all want human drudgery and toil to be addressed. Maybe there hasn't been enough of that, or as much of that as we thought there would be at this level of technology. I think it's actually a fair criticism of the AI industry.

关于窃取内容这一点，实际上现在不太听到了。

On the stolen content point, actually don't hear that one as much anymore.

我觉得更多的是内容创作者，你在社交媒体上会看到类似“这个视频没有用 AI 制作”之类的说法。

I think it's more content creators that you see that on social media, like 'this video was made without AI' or whatever.

是的。我坚信会出现新的内容创作形式、新的艺术形式。我记得有一次回顾相机刚发明时人们对它将对画家产生什么影响的评论。当时，我认为人们并没有把摄影看作一种新的艺术媒介。我很有把握地打赌他们没有。我认为会出现新的内容创作形式，而且我们可能并不关心其中的大部分。我们与创作者的关系可能更多地是关于他们作为人的本身，他们是否用 AI 来制作更好的视频或其他什么并不重要。

Yeah. I believe very strongly that there will be new kinds of content to create, new kinds of art. I remember once looking back at some of the things people said when the camera was first developed about what it was going to mean for painters. At that time, I don't think people thought of photography as a new art medium. I would bet pretty confidently they didn't. And I think there will be new kinds of content creation, and also we may not care about most of it. Our relationship with creators may be very deeply about them as people, and it doesn't matter if they use AI to make better videos or whatever.

你认为你的基金会——据我所知可能是世界上资金最雄厚的——能否在社区参与方面做得更多，特别是针对内容创作者？不，只是泛泛地应对这种非常负面的情绪，比如说，我们会站出来建图书馆之类的。

Do you think your foundation, which based on what I can see is maybe the best capitalized in the world, can do more here on engagement in communities, on content creators in particular? No, just generally addressing this very negative sentiment and saying, you know, we're going to show up and build libraries, whatever.

我的意思是，工业革命中有很多关于人们再投资财富的教训。我认为我们能做的最重要的事情是制造出对人们有用的伟大 AI 产品，确保权力和经济实力继续在世界范围内传播，让人们能够使用这些工具并从中受益。我们要倡导我们所看到的东西。其次，当然，我认为我们应该更多地投资于社区。而且我认为 AI 将为实现大规模的这种富足提供条件。我确实认为，把这项技术交到人们手中，让他们用它来造福自己的社区，而不是我们去告诉他们什么是图书馆、什么是学校，我们将看到变革性的强大益处。

I mean, there were a lot of lessons from the industrial revolution of people who reinvested their wealth. I think the most important thing we can do is to make great AI products that are useful to people, make sure that power and economic power continues to be spread throughout the world, that people have access to these tools and the benefits of these tools. And we kind of advocate for what we are seeing. And also, secondarily to that, yes, of course I think we should invest more in communities. And I think AI is going to enable the abundance required to do that at massive scale. I really do think we are going to see transformatively powerful benefits by putting this technology in the hands of people that use it for the benefit of their own community, and not us coming and telling them what a library is and what a school is.

我花很多时间在会议之间切换上下文，常常在下一个会议开始前没有时间处理上一个会议。幸运的是，Granola 一直在后台运行。它是一个易于使用的 AI 会议记事本，适用于任何场合，甚至电话会议。我用 Granola 来回忆会议内容并生成有用的摘要。我每天都用它来掌握团队需要完成的事项。它连接我的电子邮件并建议后续行动，我可以快速查看并发送，节省了我宝贵的时间。Granola 不仅是我工作流程的核心部分，它基本上是我的第二大脑。访问 granola.ai/sources 试用 Granola，并使用促销代码 sources 享受 3 个月优惠。

I spend a lot of time context switching between meetings, often with no time to process one before the next starts. Thankfully, Granola runs in the background the whole time. It's an easy to use AI notepad for meetings that works everywhere, even on phone calls. I use Granola to recall what was said in meetings and create helpful summaries. I use it every day to stay on top of what I need to get done with my team. It connects to my email and suggests follow-ups for me to quickly review and send, saving me valuable time. Granola isn't just a core part of my workflow. It's basically my second brain. Try Granola at granola.ai/sources and use the promo code sources for 3 months off.

Mercury 是为像我这样的初创公司打造的现代银行服务。当我决定开始我的媒体业务时，Mercury 是迄今为止最直接、功能最全的银行解决方案，让我能快速上手。界面直观简单，每天为我节省宝贵时间。我用 Mercury 跟踪支出、账单和发票。我喜欢它可以向团队委派权限，让他们按照我想要的方式为我管理事务。我最喜欢的是 Mercury 在 AI 方面的前瞻性。传统银行停留在过去，而 Mercury 是为现代软件的工作方式而构建的。我使用其内置的命令助手来分析现金流并帮助我转账。Mercury 还连接其他 AI 工具，如 ChatBT 和 Claude。我经常使用这个功能，Mercury 的人告诉我我是它的顶级用户之一。所以相信我，现在终于可以随时随地轻松获取你业务的实时财务数据。访问 mercury.com 了解更多信息并在几分钟内在线申请。Mercury 是一家金融科技公司，不是 FDIC 保险银行。银行服务由 Choice Financial Group 和 Column NA 成员 FDIC 提供。

Mercury is a modern take on banking built for startups like mine. When I decided to start my media business, Mercury was by far the most straightforward full-featured banking solution for me to set up quickly. The interface is intuitive and simple, saving me valuable time every day. I use Mercury to track my spending, bills, and invoicing. I love that I can delegate permissions to my team so they can keep things running for me in exactly the way I want them to. My favorite part is how forward-looking Mercury is with AI. Legacy banks are stuck in the past. But Mercury is built for how modern software works today. I use its built-in command assistant to analyze cash flow and help me move money. And Mercury also connects to other AI tools like ChatBT and Claude. I use this feature all the time and the folks at Mercury actually let me know that I'm one of the top users of it. So trust me, it's finally easy to get real time financial data about your business wherever you need it. Visit mercury.com to learn more and apply online in minutes. Mercury is a fintech company, not a FDIC insured bank. Banking services provided through Choice Financial Group and Column NA members FDIC.

Framer 是一个 AI 网站构建器，它将智能体带入你设计、管理和发布网站的同一画布，让你在不放弃品味或控制权的情况下更快地行动。Framer 为 Sources 播客网站 podcast.sources.news 提供支持，你可以在那里找到新剧集、文字记录等等。了解如何从 Framer 专家那里获得更多网站价值，或立即在 framer.com/sources 免费开始构建，享受 Framer Pro 年度计划 30% 的折扣。网址是 framer.com/sources，享受 30% 折扣。framer.com/sources。可能适用规则和限制。

Framer is the AI website builder that brings agents into the same canvas where your website is designed, managed, and published so you can move faster without giving up your taste or control. Framer powers the Sources podcast website at podcast.sources.news where you can find new episodes, transcripts, and a lot more. Learn how you can get more out of your site from a framer specialist or get started building for free today at framer.com/sources for 30% off a framer pro annual plan. That's framer.com/sources for 30% off. framer.com/sources. Rules and restrictions may apply.

AI 的实用性取决于它所拥有的上下文。但当上下文分散在工具、线程和私信中时，你的团队和 AI 智能体就会盲目行动。这就是 Atlassian 的 Jira 解决的问题。你的项目目标是什么？上周在 Slack 私信中决定了什么？Atlassian 的团队协作图从 Jira、Confluence、GitHub、Slack 等中汇集所有有价值的片段。这样就不会遗漏任何东西。你获得的结果准确率提高 44%，令牌使用量减少 48%。使用 Jira，你可以轻松地将工作上下文分享给你已经喜爱的 AI 智能体，如 Claude、Cursor 和 GitHub Copilot。直接分配工作或通过 MCP 连接你的工具。所有这些让你花更少的时间在无尽的链接和消息中挖掘，追查谁决定了什么，而花更多的时间真正交付。了解更多请访问 jira.com。网址是 jir.com。

AI is only as useful as the context it has. But when that context is scattered across tools, threads, and DMs, your team and your AI agents are flying blind. That's the problem Jira by Atlassian solves. What's the goal tied to your project? What got decided last week in Slack DMs? Atlassian's teamwork graph pulls all of the valuable pieces together from Jira, Confluence, GitHub, Slack, and more. So nothing falls through the cracks. You get 44% more accurate results with 48% less token usage. With Jira, you can easily share your work context with the AI agents you already love, like Claude, Cursor, and GitHub Copilot. Assign them work directly or connect your tools through MCP. All of this lets you spend less time digging through endless links and messages, chasing down what got decided and by who, and spend more time actually shipping. Learn more at jira.com. That's jir.com.

你说我们过去 12 个月不是最好的，这主要是我的错，但我们即将迎来最好的 12 个月。你这么说是什么意思？

You said we did not have our best last 12 months ever, which is mostly my fault, but we are about to have our best 12 months. What did you mean by that?

最好的 12 个月。

Best 12 months yet.

我认为我们公司确实犯了一些错误，这也会周期性地发生。尝试做一系列赌注的一部分是，有时成功的多，有时成功的少。但我认为，无论是在产品方向还是特别是在研究的预训练方面，我们都落后于我们想要达到的水平。

I think we clearly had some missteps as a company, which will happen periodically. Part of trying to make a portfolio of bets is that sometimes more of them work and sometimes less of them work. But I think both in terms of product direction and specifically on pre-training in research, we fell behind where we wanted to be.

我认为我们现在不仅执行得比以往任何时候都好，而且在该领域任何公司中都是最好的。而且这很有趣——在低谷之后，上升期更有趣。看看模型发布的节奏，公司团结一致、专注，并在各个不同部门做出了一系列艰难决定，但方向统一、步调一致——现在感觉非常好。

I think we are now executing not only the best we have ever executed, but the best of any company in the space. And it is very fun—the upswing is more fun after the downswing. Looking at the pace of models, the way the company has come together, focused, and made a bunch of hard decisions in very different parts of the company, but done in unison and in one direction—it feels great right now.

我想谈谈所有这些，但先停留一下，因为过去一年发生了很多事。回顾过去，有没有哪些具体决定是你做出的，却让公司失去了势头？我的意思是，你提到了预训练。我知道你一直与研究团队非常亲近。你能详细说说吗？

And I want to get to all that, but to dwell on this for a second, because the last year a lot has happened. Were there specific decisions you can look back on that you made that cost the company momentum? I mean, you mentioned pre-training. I know you've always been very close to the research team. Can you elaborate on that?

我认为我们在产品方面尝试做得太多了。这些都是实际上非常好的事情，但它们不如最重要的事情——推动智能的通用能力——那么重要。所以我们当时在做浏览器和 Sora 之类的东西。现在我们非常坚定地专注于成为面向人们的智能服务。我们的模型已经成为世界上最好的，未来几个月还会变得更好。人们正在做出非凡的事情，但那才是我们应该专注的。我应该让每个人都专注于这一件事，而不是过多担心这些支线任务。

I think we were trying to do too much on the product side. These were all things that were actually very good to do, but they were just not as good as the most important thing, which was to push on the general capability of the intelligence. So we were doing things like a browser and Sora. Now we have a very relentless focus on being this intelligent service to people. Our models have gotten to be the best in the world, and they will get much better over the coming months. People are doing remarkable things, but that is what we should have been focused on. I should have been holding everybody to this one thing and not worrying about these side quests more.

回顾大约一年前的领导层变动，你引入了 Fiji Simo 来帮助管理公司的大部分业务。她因健康原因不得不退居二线。现在你和联合创始人 Greg Brockman 实际上在分担职责，共同运营公司。这是你设想会继续的安排，还是暂时的？

Looking at the leadership changes you had about a year ago, you brought in Fiji Simo to help run large parts of the company. She had to step back due to her health. And now you and Greg Brockman, your co-founder, are effectively splitting responsibilities, running the company together. Is this the setup that you envision will continue, or is this something temporary?

我认为进展非常顺利。我们会继续引进和提拔新领导者，但感觉——我对 Fiji 感到非常难过，很难填补她的空缺——但感觉很好。Greg 和我执行得很好，公司，你知道，当事情朝着正确方向发展时，你真的能感觉到，而现在感觉事情正在朝着正确的方向发展。

I think it's going super well. We will continue to bring in and promote new leaders, but it feels—and I'm extremely sad about Fiji, hard to fill her shoes—but it feels good. Greg and I are executing well, and the company is, you know, you can really tell when things are moving in the right direction, and it feels like things are moving in the right direction.

你和 Greg 一起是怎么做决定的？比如，谁决定什么？你们有没有遇到过必须打破平局的情况？

How do you all make decisions, you and Greg together? Like, who decides what? Do you ever have a tie you have to break?

我们谈很多，非常多，一直如此。这不是一家大公司，不只是 Greg 和我。有一群非常有才华的人管理研究项目，还有一群非常有才华的人管理业务。我们都经常交流。在早期规模时，我认为尝试一些事情，如果有效就快速调整，而不是花太多时间争论决定，是好的。现在在我们的规模，我学到花大量时间做出正确决定要好得多——一种“三思而后行”的方法。

We talk a lot, like a lot, all of the time. It's not like this is a big company; it's not just Greg and I. There's an incredibly talented set of people managing the research program, and an incredibly talented set of people managing the business. We all just talk a lot. At an earlier scale, I thought it was good to just sort of try something and adapt quickly if it works, not spending as much time debating the decision. At our scale now, I've learned that it's much better to spend a lot of time trying to get to the right decision—a measure twice, cut once approach.

大约一年前，在 GPT-5 发布前后，我们一起参加了你在旧金山举办的晚宴。

We were together at a dinner you hosted here in San Francisco almost exactly a year ago, around the launch of GPT-5.

我们应该再办一次。我都忘了。那很有趣。

We should do another one of those. I forgot about that. That was fun.

是的。当时说了很多，但我从中得到的一点是，你似乎对永远担任 CEO 并不兴奋。我想知道过去一年是否改变了你的想法。

It was. And a lot was said, but a thing that I came away with from that was it seemed like you were maybe not excited about being CEO forever. I'm wondering if the last year has changed that for you.

我现在比一年前开心多了。我真的玩得很开心。我计划长期做这件事。

I'm having a much better time now than a year ago. I'm really having fun. I plan to do this for a long time.

去年氛围更具挑战性，我会说。

The vibes were more challenged last year, I would say.

是的，完全正确。不仅仅是 OpenAI 的氛围，对科技行业、对 AI 来说都是一段艰难时期。

Yeah, totally. It was not just the vibes of OpenAI; it was a hard time for the tech industry, for AI.

AI 泡沫是一个大问题。

AI bubble was a big concern.

是的，一切都令人疲惫。这显然，我认为我们做了一件了不起的事。这是一段痛苦的个人经历，但我认为完全值得，而且我很乐意再来一次。我现在过得很愉快。

Yeah, it was all just exhausting. This is obviously, I think we have done an amazing thing. It has been a painful personal experience, but I think it's totally worth it, and I would happily do it again. I'm having a good time at this point.

当我看到 Astra 的演示时，另一个让我印象深刻的是你提到的计算机使用。智能体使用计算机、使用各种企业软件，你们一直在向人们展示这一点，其影响在规模上感觉深远。我很好奇你是否一直在思考这个问题，以及你认为世界需要如何适应。

The other big thing that stood out to me when I saw the demo of Astra is the computer use you're talking about. The implications of agents using computers, using all kinds of enterprise software, which you guys have been showing people it doing, feels profound at scale. I'm curious if you've been thinking through that and how you think the world needs to adapt for that.

计算机使用让我感到惊讶。我对此兴奋了很久，但一直很失望。模型在电脑上点击操作方面从来都不够好；总是太慢或者不太行。

The computer use caught me by surprise. I had been excited about this for a long time, but I had always been disappointed. The models were just never that good at clicking around a computer; it was always too slow or it didn't quite work.

是的。

Yeah.

而 Astra 感觉在计算机使用上达到了与人类相当的水平。我不知道为什么这让我觉得是通往 AGI 道路上的一个步骤，我当时想，‘哇，这真的做到了。’但确实在情感层面打动了我。我觉得这太棒了。我在电脑上做很多琐碎的任务——我不记得别人在哪里给我发了消息，我在各种消息应用里点来点去，试图搜索。现在我只问模型。我无法回到那个必须痛苦地在电脑上找东西的世界。我只想解释我想要什么，然后让它发生。我是个非常懒的用户，所以我不想点击、连接电脑、设置东西，或者处理一堆连接器。我只想用我的电脑，做事情。

And Astra feels like it kind of reached human parity on using computers. I don't know why that hit me as one of those steps along the path to AGI where I was like, 'Wow, this is really doing it.' But it did hit me that way, on an emotional level. I think it's awesome. There are all these mundane tasks I do on my computer—I don't remember where someone sent me a message, and I click around through all these messaging things and try to search. Now I just ask the model. I cannot go back to a world where I had to painfully try to find things on my computer. I just want to explain what I want; I want it to happen. I'm a very lazy user, so I don't want to have to click around, connect my computer, set things up, or deal with a bunch of connectors. I just want to use my computer and do the thing.

我认为它能够使用软件有很多影响，但我认为大多数是相当积极的，因为人们在电脑前做了很多苦差事。

I think there are a lot of implications about it being able to use software, but I think they're mostly quite positive, in that there's a lot of drudgery that people do behind a computer.

嗯。

Mhm.

而且我有一个经历，在 Astra 之前的模型上从未有过，现在有好几次了，就是有件事，本来要花我一些时间，而且不太愉快。相反，我告诉模型我想让它做什么，然后我去和孩子们玩，30 分钟后回来，一切都准备好了。我觉得这太棒了。

And an experience I have had, not really before any pre-Astra models and now several times, is like there was a thing, it was going to take me some time, it was going to not be very pleasant. Instead, I just tell the model what I want it to do, and then I go play with my kids, and I come back in 30 minutes and it's all ready. I find that very awesome.

不过，我们现在所处的世界，美国政府开始在你的模型和其他前沿实验室发布之前审查其能力。我们正处于一个新时代。你曾警告过——你在 2025 年参议院听证会上说过——这种审查可能对美国与中国等竞争对手的竞争力造成“灾难性”影响。

We're now in a world though where the US government is starting to vet the capabilities of your models and other frontier labs before they come out. This is a new era we're in. And you have warned—you said it during a 2025 Senate hearing—that this kind of vetting could be 'disastrous' for US competitiveness against rivals like China.

然后，更近一些，GPT-5.6 的初步推出——特朗普政府要求你们对此进行把关，而你当时说过这不应该成为常态。所以看起来你一直在说事情不应该朝这个方向发展，但它们确实在朝这个方向发展。

And then I mean more recently, the GPT-5.6 initial rollout—the Trump administration requested you all gate that, and you had said at the time that shouldn't become the norm. So it seems like you've been saying this is not where things should go, and yet they're going there.

不不不。我一直在说，这就像——我想我多年来一直在呼吁某种国际监管框架。

No, no, no. I have been saying this is like—I think I've been calling for some sort of international regulatory framework for years.

但特别是政府在模型发布前进行审查。

But particularly the government vetting models before they come out.

呃，我反对的是政府像挑选个别客户那样决定谁可以使用模型。我认为政府对模型进行测试和制定共享标准是个非常好的主意。理想情况下，我不认为政府应该说你只能给这家公司访问权，而不是那家。

Uh, I think what I was pushing back on was the government like picking individual customers of who's allowed to use a model. I think government testing of a model and shared standards is a super good idea. I don't ideally—I don't think the government should be saying you can give access to this company, not this one.

那么，既然美国开始接受这种做法而其他国家还没有，这对地缘政治上的竞争力有什么影响？你想过这个问题吗？

So what are the implications for competitiveness geopolitically now that the US is starting to embrace this approach and other countries haven't? Have you thought about that?

再说一次，我认为正确的做法是国际性的，但目前领先的努力都是美国公司。所以我认为从这里开始，我们有足够的领先优势，稍微放慢一点也没关系。呃，我有信心我们既能构建安全、稳健、可靠的模型，也能在商业上做得很好，并确保美国保持领先。嗯，情况可能会发生很大变化。你知道，如果其他国家发布开放模型，在我们想出新的安全范式之前引发一些重大的网络事件，情况可能会有所转变。

Again, I think the right approach is an international one, but right now the leading efforts are all American companies. And so I think starting here, like we have enough of a lead that being slowed down a little bit is okay. Uh, and I'm confident that we will be able to both build safe, robust, reliable models and kind of do great commercially and make sure the US is leading. Um, things could shift a lot. You know, if there's open models put out by other countries that lead to some huge cyber incidents before we can come up with new security paradigms, things could shift a little bit.

你认为这可能发生吗？

Do you think that could happen?

当然，这可能发生。但是，你知道，我们就像——我也认为我们有机会彻底重新构想网络安全的工作方式。尽管这些智能体可能做坏事，但它们也能做惊人的事情。如果我们能让防御型智能体一直运行，也许那才是正确的范式。

Of course, it could happen. But, you know, we're like—I also think we have a chance to totally reimagine how cyber security works. And although these agents can do bad things, they can do amazing things. And if we can have kind of defense agents running all the time, maybe that's the right paradigm.

你准备好面对美国政府可能告诉你不能发布模型了吗？你想过这个吗？

Are you prepared for the US government to potentially tell you you can't ship a model? Have you thought about this?

我坚信我们会先决定不发布模型，而不是等他们告诉我们不要发布。

My strong belief is we would decide not to ship a model before they would tell us not to.

转向竞争话题，Anthropic 通过专注于编码一跃达到了现在的地位。

Shifting to competition, Anthropic catapulted to where they are now by single-shot focus on coding.

是的。

Yeah.

你在对话开始时说你们下了很多赌注，这让你们失去了一些势头。我很好奇你能否反思一下，Anthropic 是如何看到那个你们当时没有看到的机会的。

And you started this conversation by saying you guys were placing a lot of bets and that cost you some momentum. I'm curious if you could reflect on how Anthropic saw that opening that you guys didn't at the time.

我不认为问题在于我们没有看到它。问题在于我们有这种巨大的消费者增长失控。我们一直想做编码，但我们想，啊，我们有这个非常紧急的事情，而且它很棒。拥有它是件好事，所以从优先级的角度我们错过了。我现在认为我们有市场上最好的编码产品，而且它增长得非常快，我认识的大多数人——即使是那些 Anthropic 产品的忠实用户——都已经转换过来了。嗯，所以我不认为在任何阶段落后是灾难性的，我们可以用更好的模型赶上。

I don't think it was a question of us not seeing it. It was a question of like we had this tremendous thing of this runaway consumer growth. We always wanted to do coding but we were like, ah, we have this very urgent thing and it's great. It's like a great thing to have, and so we missed it from a prioritization standpoint. I now think we have the best coding product in the market and it's growing like crazily quickly, and most people I know—even the people that were like the diehard Anthropic product users—have switched over. Um, so I don't think it's like catastrophic to be behind on any one phase, and we can, you know, catch up with better models.

我很好奇——我想很多人都在试图理解 AI 市场的零和程度，你在 Codex 上的增长，你知道，是从 Anthropic 那里抢来的，还是反之。你有这种感觉吗？

I'm curious—I think a lot of people are trying to understand how zero-sum the AI market is, and is your growth on Codex, you know, taking from Anthropic or vice versa. Do you have a sense of that?

我认为现在每个人都在增长。我的意思是，这可能——这将是一个非常大的市场。呃，以后可能会变得更加零和，但就目前而言，我认为每个人都像——我们看到的增长率是我对这个规模公司的想象中完全没有的，我认为这说明了人们从产品中获得了多少价值，但我认为这发生在大多数行业中。

I think right now everybody's growing. I mean, it may—this is going to be a very big market. Uh, it may become more zero-sum later, but for now, like I think everybody is just like—the growth rates we are seeing are just nothing that I had like in my frame of imagination for a company at this scale, and I think it just speaks to how much people are getting value out of the products, but I think it's happening across most of the industry.

在产品方面，你们正在做内部所谓的“合并”，把 ChatGPT 和 Codex 结合起来，打造一个整合它们的超级应用。

And on the product side, you're doing what is being called internally the merge, taking ChatGPT and Codex and building a super app that combines them.

你们已经开始了。我想说还有一些粗糙的地方。

And you've started this. There's—I would say there's still some rough edges.

不止是粗糙。你这么说真是太客气了。

More than rough edges. That's a very polite way for you to say.

是的。

Yeah.

嗯，我很好奇当你达到那个目标时。那会是什么样子，有什么影响？

Um, and I'm curious when you get there. What does that look like and what are the implications of that?

我想要的是一个 AI 的界面，它能做任何我需要的事情。如果我有，你知道，一个像 ChatGPT 风格的快速问题，呃，它可以直接回答。如果我需要构建一个复杂的东西，一个软件，它也能做到。如果我需要介于两者之间的东西，它也能做到。而且，你知道，如果它需要访问我的电脑或我的上下文，它可以去用我的电脑并找到我的上下文。我不需要——就像我提到的，我是一个非常懒的用户。我不需要去想我在哪个标签页。我不想需要去想我处于什么模式。我喜欢 AI——一个足够聪明以至于能发现新颖数学的 AI，应该能够直觉地知道它应该做什么。

The thing that I want is just like an interface to an AI that can kind of do whatever I need. If I have a, you know, quick question like ChatGPT style, uh, it can just answer it. If I need a complex thing built, a piece of software built, it can do that. If I need something in the middle, it can do that. And, you know, if it needs access to my computer or my context, it can go use my computer and find my context. And I don't have to—like I mentioned, I'm a very lazy user. I don't have to like think about what tab I'm on. I don't want to have to like think about what mode I'm in. I like the AI—an AI that is smart enough to like discover novel mathematics should be able to like intuit what it's supposed to do.

ChatGPT 刚刚达到了 10 亿用户。

ChatGPT just hit a billion users.

重要的里程碑，但我认为达到这个目标可能比你们最初想象的要花更长时间。早期的增长是爆炸性的。

Big milestone, but I think hitting that took maybe longer than you guys originally thought. The growth was explosive early on.

是的。我很好奇听听你的看法。在过去 12 个月里，它的增长是否比你预期的要慢？

Yeah. I'm curious to hear from you about that. Has it grown slower than you'd expected in the last 12 months?

嗯，我们决定——当我们专注于编码时，我们决定将大量原本可以投入到聊天产品的算力重新分配到编码上。所以，不，这并没有让我们惊讶——这是我们做出的决定，因为这是一件紧急的事情。

Well, we decided to—when we focused on coding, we decided that we were going to reallocate a lot of our compute that we could have otherwise put into the chat product into coding. So, no, that didn't surprise us—like that was a decision we made because this was an urgent thing.

所以，增长是你决定把算力放在哪里的直接函数。

So, growth is a direct function of where you decide to put the compute.

百分之百。是的，我总是希望算力限制即将缓解，因为我们要制造更高效的模型，我希望有一天这是真的。但每次我们发现效率提升，世界的 token 需求就会不断上升并吞噬它。

100%. Yeah, I am always hopeful that the compute constraints are about to soften because we're going to make more efficient models, and someday I hope it's true. But every time we find efficiency gains, the world token demand just goes up and up and eats it.

我听到了。但与此同时，我很好奇 ChatGPT 如何达到下一个十亿？这是像互联网或社交媒体那样线性增长吗？还是会更加波动？这对你现在有多重要？因为你还有 Codex 和 API 业务。

I hear that. But at the same time, I'm curious like how does ChatGPT get to the next billion? Is that as linear as the internet has grown or social media grew? Is it going to be choppier? How much does that even matter to you now? Because you've got Codex and the API business.

我有点认为我们——就像我们谈到的合并，但我有点认为会发生的是它们都会融合在一起。在合并之前的一段时间里，我已经停止使用 ChatGPT，我只是用 Codex 来问所有聊天问题，因为，是的，再次，懒用户。嗯，现在我认为有很多人从未想过他们会有一个智能体为他们做事，因为他们只是用 ChatGPT，然后点击了这个工作标签，然后想，哇，我可以做这个疯狂的事情。嗯，所以我认为它们都会融合在一起，人们将拥有这种通用 AI 订阅，他们不会真正去想这是聊天还是 Codex 还是工作——就像我有一个东西，我希望它很快实现。

I kind of think what we're—like we talked about the merge, but I kind of think what's going to happen is that they're all going to like come together. For a while pre-merge, I had stopped using ChatGPT and I just asked Codex all my chat questions because, yeah, again, lazy user. Um, now I think there are a lot of people who never thought they were going to be having an agent do stuff for them because they just kind of used ChatGPT that clicked on this work tab and were like, whoa, I can do this crazy thing. Um, so I think it's kind of all going to come together and people are going to have this general-purpose AI subscription that they don't really think about like if it's chat or Codex or work—just like I have a thing I want it to happen soon.

它甚至会——你甚至不需要问它。它有望变得更加主动，会持续运行并努力为你做有用的事情。

It will even—you won't even need to ask it. It'll hopefully be much more proactive and it'll be constantly running and trying to do useful stuff for you.

所以这最终就是一项终极订阅。

So the instate of this is just one ultimate subscription.

作为用户，这就是我想要的。

That is what I want as a user.

我们一直在绕圈子，但大约一年前，你确实在你们大规模算力建设上冒了险，引发了所有关于 AI 泡沫的担忧。而且你知道，与此同时，当人们认为你们过度投入时，像 Anthropic 的 CEO Dario 那样的人说你们在 YOLO（孤注一掷）。现在我得说，你在这方面似乎得到了平反。嗯，世界仍然缺乏算力。听起来你们也仍然缺乏，即使你们比一些竞争对手拥有更多。嗯，同时你似乎在大幅降低 token 成本，而且你们即将发布 Jalapeno，你们首款用于推理的自研芯片。那么，当你审视这一切时，你们正在进行的算力建设以及与之相关的天文数字，是否还有任何部分让你觉得有风险？

We've been dancing around this, but you did really stick your neck out about a year ago on the massive compute buildout you guys have been doing and caused all this AI bubble fear. And you know, at the same time, while people thought you were overshooting, you know, you had people like Dario, the CEO of Anthropic, saying you were yoloing and you know, now I will say you seem pretty vindicated on this front. Um, the world is still starved of compute. Sounds like you guys still are too, even though you have more than some of your competitors. Um, and at the same time you're driving the cost of tokens it seems like way down and you're about to release Jalapeno, your first custom chip for inference. Is there still though any part of this compute buildout that you're on and the astronomical numbers associated with this that you feel is at risk at all when you look at all of this?

我不担心我们的算力建设计划。我担心的是世界的算力建设计划。我觉得我们将能够非常盈利地使用我们计划建造的所有算力。嗯，但我看到了第一批迹象，在我看来，这像是不可持续的愚蠢行为，比如随机出现的新 Neocloud，人们声称他们明年将建造巨量算力，而我认为他们没有收入来支持或没有买家。呃，是的，我确实对世界整体正在做的事情感到一些担忧。尽管我认为我们对自己承诺的事情感觉很好。

I'm not worried about our compute buildout plans. I am worried about the world's compute buildout plans. Like I think we are going to be able to use all of the compute very profitably that we are planning to build. Um, but I am seeing the first signs of what feels to me like unsustainable silliness of, you know, random new Neocloud popping up, people claiming that they're going to build gigantic amounts of compute next year that I think they don't have the revenue to support or a buyer. Uh, yeah, I definitely feel like some fear about what the world is doing as a whole. Although I think we feel very good about what we've committed to.

但你描述的传染效应肯定会影响你。

But the contagion of what you're describing could certainly impact you.

如果整个经济崩溃，是的，那可能会影响我们，比如能否自信地支付我们承诺的算力费用。嗯，我对此感觉良好。我觉得现在人们有点不计成本。我们只是要疯狂地建设大量算力，甚至以更高的价格。如果我们能够成功大幅降低算力成本并大幅提高算力效率，那么你可以想象一个世界，有些人做出了愚蠢的财务决策，这几乎在每个繁荣时期或大多数繁荣时期都会发生。所以你知道，如果发生这种情况，也不是世界末日，也不是什么疯狂意外。

If the whole economy blows up, yes, that could impact us in terms of like being able to confidently pay for the compute we are committed to. Um, I feel good about that. Like I think people right now are kind of in a cost is no object. We're just going to build out crazy amounts of compute even higher price for it. And if we are able to succeed with our efforts to hugely drive down the cost of compute and the efficiency of compute up a lot then you know you can imagine a world where there are some people that made dumb financial decisions that happens in kind of like every boom uh or most of them. So you know not the end not not like a crazy surprise if it does.

你看到 OpenAI 成为行业算力供应商的世界吗？

Do you see a world where OpenAI becomes a supplier of compute to the industry?

呃，短期内不会。就像我们只是需要算力。

Uh not anytime soon. Like we just we need the compute.

我得到的感觉是你们内部正在讨论这个，但尚未决定。

The vibe I'm getting is you all are discussing this internally and it's not decided.

所以人们经常谈论递归自我改进

So people talk a lot about recursively self-improving

是的

yes

AI 模型。他们不太谈论在物理世界中做到这一点的能力。但如果我们的机器人项目成功，芯片项目成功，一些供应链投资成功，我们变得非常擅长以更便宜的方式建造数据中心，拥有比任何人都更好的芯片，那么我们会考虑吗？也许。我们有当前计划吗？呃，那仍然超出了——我们还没有奢侈到专注于那个。

AI models. They do not talk as much about you know the ability to do this in the physical world. But if our robotics program comes together, our chip program comes together, some of our supply chain investments come together, we get really great at building data centers way more cheaply and better chip than anybody else has, like you know, would we consider it? Maybe. Do we have any current plans? Uh that still is like outside of we don't have the luxury of focusing on that yet.

嗯。你提到了递归自我改进。我很高兴你提到了。呃，现在旧金山的人们经常谈论 RSI。嗯，有一份你发给员工的备忘录，在你们申请 IPO 时泄露了，你说潜在的 RSI 起飞看起来越快，延迟 IPO 可能越有利。

Mhm. You brought up recursive self-improvement. I'm glad you did. Uh people are talking about RSI a lot in San Francisco right now. Um there was a note you sent to employees um that leaked when you guys filed for the IPO where you said that the faster the potential RSI takeoff looks like it could be the more it could be advantageous to delay an IPO.

是的。

Yeah.

你是什么意思？

What did you mean by that?

嗯，我认为成为一家上市公司是一个困难的过渡。嗯，你知道人们会对激励做出反应，他们希望股价上涨，但他们不想错过季度业绩或其他什么。我从不希望我们——我希望我们尽可能容易地做出符合世界安全利益的决定。嗯，如果像这样，嘿，我们将不得不停止训练或停止某个点或其他什么，而且短期内会有很大的收入放缓。最好不要同时成为一家新上市公司并承受那种压力。现在，一年前我并不认为我们会走上超级智能的短期轨迹。现在我认为它可能会发生。嗯，我不确定它会发生。只是我们进展非常快，而且，你知道，我认为我们的使命比在任何特定时间框架内成为一家上市公司重要得多。嗯，所以我们会为使命做出最好的决定。

Um I think it's a difficult transition to become a public company. Um you know people respond to incentives and they want their stock price to go up but they don't want to miss a quart or whatever else. I never want us I want it to be as easy as possible for us to make a decision in the interest of safety of the world. Um and if it's like hey we're going to have to stop training or stop a point or whatever and we're going to like you know there's going to be a big revenue slowdown on in the short term. It'd be nice not to have a newly public company and that pressure at the same time. Now I did not think we were going to be on a kind of like you know like a short-term trajectory of super intelligence a year ago. Now I think it may happen. Um I'm not confident it's going to happen. It's just like we're making extremely fast progress and I, you know, I think our mission is way more important than being a public company on any particular time frame. Um, so we'll make the best decision for the mission.

你提到了机器人。我很想听听你们机器人项目的状态。你们在建造什么？是人形机器人吗？是机器人数据中心吗？两者都有？

You mentioned robotics. I'd love to hear from you the state of your robotics effort. What are you building? Is it a humanoid? Is it a robotic data center? Both.

我们肯定会做一个人形机器人。我们也会做其他形态。嗯，这个世界很大程度上是为人类设计的。所以如果你想想开门、在电脑上打字、驾驶设备、清洁厨房等等的能力，我们已经为人类建造了这个世界，我想确保我们继续为人类建造这个世界。所以匹配那种形态似乎很好。当然会有具有不同形态的数据中心机器人。我认为所有这些都不如真正弄清楚让机器人工作的“大脑”重要。

We will definitely do a humanoid. We will do other form factors as well. Um, the world is very much designed for people. So if you think about like the ability to open a door and type on a computer and drive a piece of equipment and you know clean a kitchen and whatever else, we've kind of built this world for people and I want to make sure that we keep building this world for people. So matching that form factor seems good. There will of course be data center robots that have like different form factors. I think all of that is less important than really figuring out like the the brain that makes the robot work.

所以你们在建造人形机器人。

So you are building a humanoid.

我们会。

We will.

你认为那将如何在世界上运作？你想象它会成为每个人的个人机器人吗？

How do you think that's going to work in the world? Do you imagine that being like a personal robot for everyone someday?

我不认为那是最重要的首要任务。你谈到了建造数据中心甚至建造更多机器人的能力，但，是的，总有一天。我认为每个人都应该有一个个人机器人。比如，我很想有一个个人机器人，能帮我做我不想做的任务。那太好了。

I don't think that's the most important first thing to do. And you talked about the ability to build data centers or even build more robots or whatever else, but but yes, someday. I think everyone should have a personal robot. Like I would love to have a personal robot that could like do the tasks that I don't want to do. That'd be great.

你们还有与 Johnny IV 合作的消费设备项目。我知道你不能多说，我们可能很快会在这里看到第一款设备。嗯

You also have the consumer device work with Johnny IV. I know you can't talk a lot about it and we'll probably see the first device here at some point soon. Um

快了。

Soonish.

快了。而且你经常谈到——我一直听你说，我的梦想是一个产品，它只是环境式地听我说话，吸收一切并给我上下文。我们之前讨论过计算机使用，我同意这在很多情况下似乎非常有用。但它也像是一场隐私监控噩梦，我很好奇你是否考虑过这一点，以及世界将如何反应。

Soonish. And you've talked a lot about how I've been hearing you say like my dream is a product that just is ambiently listening to me and taking everything in and giving me context. We were talking about this earlier with computer use and I agree that seems um very helpful in a lot of context. It also seems like a privacy surveillance nightmare and I'm curious if you've been thinking about that and how the world will react to that.

我们在隐私方面采取了非常强硬的立场。

We have taken a very strong stance on privacy.

我认为商业隐私也很重要，不仅仅是消费者隐私，还包括我们做出的关于不利用商业数据训练模型、零数据保留的承诺，我认为这非常重要。随着 AI 越来越深入地融入我们的生活，隐私变得极其重要。我担心的一点是，有些其他努力持不同看法，他们会推动说安全风险太大，AI 隐私不可能以同样的方式存在。我认为应该有一部 AI 特权法。我甚至认为政府不应该被允许强迫公司交出你的聊天记录之类的。如果你和医生或律师交谈，有特权这个概念。但你和 ChatGPT 交谈时没有这个特权。我认为应该有。

I think that business privacy too, not just consumer privacy, but the way we make commitments about not training on business data and about zero data retention, I think this is very important. As AI becomes more and more embedded in our lives, privacy becomes extremely important. One thing I worry about is there are other efforts that think differently and will push on, saying the safety risks are so big that AI privacy can't exist in the same kind of way. I think there should be an AI privilege law. I don't even think the government should be allowed to compel a company to give them your chat history or whatever. If you talk to a doctor or a lawyer, there's a concept of privilege. You don't have that talking to ChatGPT. I think you should.

在你刚才描述的那个情境中，律师或医生对你的数据的使用也有很多限制。不仅仅是外部共享。你认为这种监督是否应该扩展到你们如何使用……

In that context you just described, there's also a lot of limits on what a lawyer or a doctor can do with your data. It's not just sharing it externally. Do you think that kind of oversight should extend to how you use...

是的。不，我正要说到那一点。

Yeah. No, I was going to get to that.

嗯。

Yeah.

是的。所以，我认为政府能做什么应该有法律限制。我也认为公司对与 AI 共享的数据应该有很多限制，尤其是当这个东西监视你的电脑、监听你的消息、与你交谈时。我认为人们对此应该比现在更加关注。

Yeah. So, I think there should be legal limits on what the government can do. I also think companies should have a lot of restrictions on data shared with an AI, especially if you have this thing watching your computer, listening to your messages, talking to you. I think this is something people should be much more animated about than they are.

在那发生之前，你们 OpenAI 如何自我监管数据的使用？你们可能拥有世界历史上积累的最强大的个人资料数据。

Before that happens, how do you at OpenAI govern that self-govern the use of data? You probably have some of the most powerful profile data that's ever been amassed in the history of the world.

我们对于数据的使用有极其严格的内部控制，并且我们向用户做出隐私保证。随着我们接近推出这个设备，我们将讨论为这种环境计算设备构建的新隐私控制和技术。但确实，我认为我们拥有有史以来最个人化的数据库之一。

We have extremely strong internal controls about how that's used, and we make the privacy guarantees to users that we do. As we get closer to launching this device, we'll be talking about the new privacy controls and technology we're building for a device that's ambiently computing. But yeah, I think we have one of the more personal databases ever.

苹果公司公开起诉你们，指控你们窃取商业机密并雇佣他们的员工与 Johnny 一起开发这个设备，你们已经对此做出了回应。你说过这是毫无根据的，但我想知道，你担心这会减缓设备开发的进度吗？

Apple has very publicly sued you guys for allegedly stealing trade secrets and hiring their employees to work on this device with Johnny, and you've responded to it. You've said it's meritless, but I'm wondering, do you worry about this slowing down the device efforts?

不。首先，我是苹果的超级粉丝，对此我感到非常难过。当我第一次听说这件事时，我想，天哪，这听起来太恶劣了。一定是有人做了坏事。我们不想要任何公司的知识产权，当然也不想要那些会窃取公司知识产权并带给我们的员工。如果我们进行了调查并发现某人确实如此，我们当然会直接解雇他们并处理此事。但如果没有做错事，我们也会为他们辩护。我相信在我们调查之后，这是一个没有做错事的案例。我们试图解释其中一些，更多细节将在法律程序中展开。根据我的理解，我不认为这会减缓进度。

No. Look, first of all, I'm a mega Apple fanboy and I was very sad about that. When I first heard about it, I was like, man, this sounds egregious. Someone must have done something badly. We don't want any company's IP, and we certainly don't want people who are going to take a company's IP and bring it to us. If we did an investigation and found that about somebody, we would of course just terminate them and deal with it. But we're also going to defend someone if they didn't do something wrong. I believe after we looked into this that this was a case of someone not doing something wrong. We tried to explain some of that, and more of that will play out in a process. Given my understanding, I don't think this is going to slow things down.

你们如何考虑设备形态？我听说你过去说过不喜欢眼镜。你喜欢眼镜吗？

How are you thinking about form factors? I've heard you say you don't like glasses in the past. Do you like glasses?

是的。眼镜。

Yeah. Glasses.

嗯。

Yeah.

我不喜欢，因为我觉得带着摄像头和灯与人交谈非常不舒服。

I don't because I find it very uncomfortable talking to people with like a camera and a light.

是的。但还有很多其他形态。

Yeah. But there's a lot of other form factors.

有很多很好的形态。我认为我们会做少数几种形态。有适合放在桌子上的，有适合放在口袋里的，还有适合佩戴在身上的。推出所有这些产品需要一些时间。但我认为最大的调整将是适应这种主动式电脑的概念。

There's a lot of great form factors. I think we'll do a small handful of form factors. There's something that belongs on a table, something that belongs in your pocket, and something that belongs on your body. It'll take us some time to launch all of those things. But I think the big adjustment is going to be getting used to this idea of a proactive computer.

当你考虑 OpenAI 的路线图和业务时，消费设备在某种意义上是生死攸关的吗？它们纯粹是附加的吗？还有你们谈论的使命，尽管我们已经讨论过你们缩减了一些项目，但仍然有很多事情在进行中。

When you're thinking about OpenAI's roadmap and the business, are consumer devices existential in a sense? Are they purely additive? And the mission that you guys talk about, you've got a lot of things still happening even though you've whittled things down as we talked about.

我认为我们还不知道。我有强烈的直觉，有一种重大的新型电脑和新类别，历史上每隔几十年才会出现一次，改变我们使用技术的方式。但这是一个不太可能的说法，所以我认为你不应该让我这么说。你应该等着看你对这些设备的看法。

I think we don't know yet. It is my strong intuition that there is a major new kind of computer and a new category that historically has only come up every couple of decades for how we use technology. But that's an unlikely claim, so I think you shouldn't let me make it. You should just wait to see what you think of the devices.

现在人们对 OpenAI 的看法，你认为几年后会是什么样？

The way OpenAI is thought of now, how do you think it will be thought of in a couple years?

很简单。我希望人们喜欢我们推向世界的产品。这些故事我们之前提到过一些，但我能够创业，我能够为我的孩子举办一个很棒的生日派对，我能够治愈这种疾病。我最近遇到一个人，他曾经为他的狗设计 mRNA 癌症疫苗，现在创办了一家公司为其他人做这件事。我希望这些故事与几年后技术为人们所做的事情相比都显得微不足道。然后我希望很多当前对 AI 的恐惧会变成人们说：‘天哪，那是最负责任的公司，每一步都做出了非常符合我们所有人利益的决定。’我很高兴他们做得很好，因为我认为他们是这项技术的良好管理者。你看到很多其他公司采取了非常不同的方法。我认为我们在安全信念上一直相当一致，但在我们犯错时愿意调整。当我们开始这种迭代部署策略时，它被 AI 安全社区深深憎恨，但我认为事后看来，这显然是正确的。我很高兴我们有勇气做我们真正相信的事情，即使它们非常不受欢迎，而且我们大多数时候是对的，在我们犯错时我们调整了。我认为这是构建安全稳健系统的方法。所以我希望我们继续这样做，人们也能认识到这一点。

Pretty simple. I hope people love the products we put out into the world. These stories we talked about a few earlier, but I was able to start a business, I was able to do a great birthday party for my kid, I was able to get cured of this disease. I met a guy recently who used to help design an mRNA cancer vaccine for his dog and now started a company to do that for other people. I hope those stories all look small in comparison to what the technology is doing for people in a few years. And then I hope that a lot of the current AI fears people said, 'Man, that was the most responsible company at every step, they made very good calls in the interest of all of us.' And I'm glad they're doing well because I think they're being good stewards of the technology. You see lots of other companies that have taken very different approaches. I think we've been pretty consistent on our beliefs about safety, but willing to adapt when we've been wrong. When we started this strategy of iterative deployment, that was deeply hated by the AI safety community, and I think in retrospect, it was obviously correct. I'm glad we've had the courage to do the things we really believe in even when they're very unpopular, and that we've mostly been right and that we've adapted when we've been wrong. I think that is the way to build safe and robust systems. So I hope we continue to do that and people recognize it.

你们正在朝着超级智能迈进。我很好奇你个人是如何为此准备的。你对你们所构建的东西的另一端的生活有什么看法吗？

And you're building towards superintelligence. I'd be curious to know how you personally are preparing for that. Do you have a view of what life will look like on the other side of what you're building?

我认为它会看起来和现在惊人地相似。

I think it will look surprisingly similar to how it looks now.

你知道，人们会和家人在一起，谈恋爱，吵架，做自己的爱好，享受娱乐，拥有非常人性化的体验，也会感到压力、焦虑，为彼此创造价值，玩各种奇怪的游戏，并且非常关心他人。我希望它不会太不同。我希望，你知道，人类体验会更丰富。人们拥有更多的自主权、更多的自由、更多的财富，能做更多事情，能更健康，能拥有更多权力来共同定义未来，世界会变得更好更快，但人类体验仍然是非常人性化的东西。

You know, people are gonna hang out with their families and fall in love and get into fights and do their hobbies and, you know, be entertained and have a very human experience and try, you know, get stressed and get anxious and create value for each other and play all kinds of strange games and care about other people a lot. I hope it'll not be that different. I hope it'll be, you know, the human experience is richer. People have more autonomy, more freedom, more wealth, can do more, can be healthier, can kind of have like more power to collectively define the future and the world gets better faster, but that the human experience stays like a very human thing.

展望未来 12 个月，OpenAI 面临的最大风险是什么？

If you look out over the next 12 months, what is the biggest risk for OpenAI?

我认为是搞错安全、对齐和安全保障。我的意思是，我认为 12 个月后我们可能拥有极其强大的模型，如果我们能够驾驭向超级智能的过渡，在一个我们已经弄清楚如何赋能人们、如何确保权力不过度集中、如何在整个范围内提供安全保障、如何让人们感觉能很好地掌控自己未来生活的世界里，那将是巨大的成功。

I think it's like getting safety, alignment, and security wrong. I mean, I think it's possible that 12 months from now we have extremely capable models and if we are able to navigate the transition to superintelligence in a world where we have figured out how to empower people, how to make sure power is not too concentrated, how to deliver safety across the entire spectrum, how to let people feel very in control of improving their own lives in the future, that would be like a phenomenal success.

Sam Altman，谢谢你。

Sam Altman, thank you.

谢谢。

Thank you.

Granola 是我试过的最好的 AI 记事本。它适用于任何场景，无论是视频通话、电话通话、面对面还是 Apple Watch。立即在 granola.ai/sources 试用，结账时使用促销代码 sources 可享 3 个月优惠。银行业务应该像现代软件一样。在一个地方获得你需要的一切。访问 mercury.com 了解更多信息，几分钟内即可在线申请。Mercury 是一家金融科技公司，不是银行。详情请查看节目说明。Framer 是 AI 原生网站构建器，让你在不放弃控制权的情况下更快地构建。访问 framer.com/sources 可享 30% 折扣。可能适用规则和限制。Jiralassin 让你的团队和智能体在相同的上下文中工作。在 jira.com 免费试用。网址是 jir.com。

Granola is the best AI notepad I've tried. It works everywhere on a video or phone call, in person or an Apple Watch. Try it now at granola.ai/sources and use the promo code sources at checkout for 3 months off. Banking should feel like modern software. Get everything you need in one place. Visit mercury.com to learn more and apply online in minutes. Mercury is a fintech, not a bank. Check the show notes for details. Framer is the AI native website builder that lets you build faster without giving up control. Visit framer.com/sources for 30% off. Rules and restrictions may apply. Jiralassin is where your team and your agents work from the same context. Try it free at jira.com. That's jir.com.