萨姆·阿尔特曼：OpenAI 如何取胜？AI 建设逻辑？2026 年上市？

因为如果你有一个能够协调多个不同机器人的人类，那么你可能希望留下他们。

Because if you have a human who's gonna be able to sort of orchestrate a bunch of different bots, then you might wanna keep them.

我不知道。

I don't know.

你怎么看这个问题？

How do you think about this?

所以，我同意你的观点，很明显，每个人都会负责管理大量执行不同任务的AI。

So I I agree with you that it's clear to see how everyone's gonna be managing, like, lot of AIs doing different stuff.

最终，就像任何一位优秀的管理者一样，希望团队会变得越来越好，但你承担的范围和责任也会越来越大。

Eventually, like any good manager, hopefully, team gets better and better, but you just take on more scope and more responsibility.

我不是一个就业悲观主义者。

I am not I am not a jobs doomer.

短期内，我有些担忧。

Short term, I have some worry.

我认为在某些情况下，过渡过程可能会很艰难。

I think the transition is likely to be rough in some cases.

但我们天生就非常在意他人在做什么。

But we are so deeply wired to care about other people, what other people do.

我们似乎如此关注相对地位，总是渴望更多，渴望有所作为、服务他人，表达创造力——无论是什么驱动我们走到今天，我认为这些都不会消失。

We are so we seem to be so focused on relative status and always wanting more and to be of use and service to express creative spirit, whatever whatever whatever has driven us this long, I don't think that's going away.

现在我认为，未来的工作，或者说，我甚至不确定‘工作’这个词是否合适，到2050年我们所有人每天要做的事情，很可能与今天大不相同。

Now I do think the jobs of the future, or I don't even know if jobs is the right word, whatever we're all gonna do all day in 2050 probably looks very different than it does today.

但我并不认为生活会失去意义，经济会彻底崩溃。

But I but I I don't have any of this like, oh, life is gonna be without meaning, and the economy is gonna totally break.

我相信，我们会找到更多意义，而经济我认为也会发生巨大变化。

Like, we will find, I hope, much more meaning, and the economy, I think, will significantly change.

但我觉得，你不能跟进化生物学对着干。

But I I think you just don't bet against evolutionary biology.

我经常思考如何自动化OpenAI的所有职能，甚至更进一步，我会思考拥有一个AI首席执行官的OpenAI意味着什么。

You know, I think a lot about how we can automate all the functions at OpenAI, and then even more than that, I think about, like, what it means to have an AI CEO of OpenAI.

这并不会让我感到困扰。

It doesn't bother me.

我对此感到非常兴奋。

I'm, like, thrilled for it.

我不会反对它。

I won't fight it.

我不想成为那个固守手工方式、觉得这样能做得更好的人。

Like, I don't wanna be I don't wanna be the person hanging on being like, can do this better the the handmade way.

AI首席执行官会做出大量决策，以指导我们把所有资源都用于给AI提供更多能量和权力。

AI CEO just make a bunch of decisions to sort of, like, direct all of our resources to giving AI more energy and power.

这就像是

It's like

我的意思是，不会。

I mean, no.

你会

You would

真的会设置严格的护栏。

really put a guard guardrail on.

是的。

Yeah.

显然，你并不希望一个不受人类监管的AI首席执行官。

Like, obviously, you don't want an AI CEO that is not governed by humans.

但如果你想想，也许这个类比有点疯狂，但我还是说一下。

But if you think about if if you think about maybe, like this is a crazy analogy, but I'll give it anyway.

如果你想象一种情况，世界上每个人实际上都是某家AI公司的董事会成员，都能告诉AI首席执行官该做什么，如果它表现不好就解雇它，并对决策进行监管。

If you think about a version where, like, every person in the world was effectively on the board of directors of an AI company and got to, you know, tell the AI CEO what to do and fire them if they weren't doing a good job at that and, you know, got governance on the decisions.

但AI首席执行官则负责努力执行董事会的意愿。

But the AI CEO got to try to, like, execute the wishes of the board.

我认为，对未来的很多人来说，这看起来会相当合理。

I think to people of the future, might seem like quite a reasonable thing.

好吧。

Okay.

我们马上转向基础设施的话题。

So we're gonna move to infrastructure in a minute.

但在我们离开关于模型和能力的这部分之前，GPT-6什么时候发布？

But before we leave this section on models and capabilities, when's g p GPT six coming?

我不确定我们会什么时候称一个模型为GPT-6，但我预计在明年第一季度会出现比5.2有显著提升的新模型。

I expect I don't know when we'll call a model GPT six, but I would expect new models that are significant gains from 5.2 in the first quarter of next year.

“显著提升”是什么意思？

What does significant gains mean?

我目前还没有一个具体的评估分数可以告诉你，但是，

I don't have, like, an eval score in mind for you yet, but,

更多是企业方面的应用

more enterprise side of things

或者说是两者都有。

or Definitely both.

对于消费者来说，模型会有很多改进。

The there will be a lot of improvements to the model for consumers.

目前消费者最想要的并不是更高的智商。

The main thing consumers want right now is not more IQ.

企业仍然希望提升智力水平。

Enterprises still do want more IQ.

因此，我们会以不同方式改进模型，以适应不同用途，但我们的目标是打造一个每个人都更喜欢的模型。

So we'll improve the model in different ways for the kind of for different uses, but, I our goal is a model that everybody likes much better.

基础设施。

So infrastructure.

你们已经承诺了大约1.4万亿美元用于基础设施建设。

You have 1,400,000,000,000.0 thereabouts in commitments, to build infrastructure.

我听了你们关于基础设施的很多说法。

I've listened to a lot of what you've said about infrastructure.

以下是你们提到的一些内容。

Here are some of the things you said.

如果人们了解我们能用算力做些什么，他们会想要多得多的算力。

If people knew what we could do with compute, they would want way, way more.

你说过，今天我们能提供的能力与10倍算力、100倍算力之间的差距是巨大的。

You said the gap between what we could offer today versus 10x compute and a 100 k x compute is substantial.

你能帮我们把这一点再展开一点吗？

What what can you help flush that out a little bit?

你打算如何利用这么多的计算能力？

What are you gonna do with so much more compute?

我之前稍微提过这一点。

Well, I mentioned this earlier a little bit.

我个人最兴奋的是利用人工智能和大量计算能力来发现新的科学。

The thing I'm personally more excited most excited about is to use AI and lots of compute to discover new science.

我相信，科学发现是让世界对每个人变得更好的关键所在。

I am a believer that scientific discovery is the high order bit of how the world gets better for everybody.

如果我们能将海量计算资源投入到科学问题中，发现新的知识——这一点现在才刚刚开始发生。

And if we can throw huge amounts of compute at scientific problems and discover new knowledge, which the tiniest bit is starting to happen now.

这还处于非常早期的阶段。

It's very early.

这些都是一些非常小的事情。

These are very small things.

但你知道，根据我对这个领域的了解和历史经验，一旦曲线开始上升并略微脱离x轴，我们就知道如何让它变得越来越好。

But, you know, my learning and history of this field is once the squiggle start and it lifts off the x axis a little bit, we know how to make that better and better.

但这需要巨大的计算资源。

But that takes huge amounts of compute to do.

因此，这是其中一个领域——利用大量AI来发现新科学、治愈疾病，以及其他许多事情。

So that's one area where, like, throwing lots of AI at discovering new science, curing disease, lots of other things.

一个最近的酷炫例子是，我们用Codecs在不到一个月的时间内开发了Sora安卓应用。

A kind of recent cool example here is we built the Sora Android app using Codecs, and they did it in, less than a month.

他们使用了海量的资源。

They used a huge amount.

在OpenAI工作的一个好处是，你不会受到Codex使用量的限制。

One of the nice things about working at OpenAI is you don't get any limits on Codex.

他们使用了海量的token，但却能完成通常需要很多人花更长时间才能完成的工作，而Codex基本上替我们完成了大部分任务。

They used a huge amount of tokens, but they were able to do what would normally have taken a lot of people much longer, and Codex kinda mostly did it for us.

你可以想象，未来整个公司都可以利用大量计算资源来构建他们的产品。

And you can imagine that going much further, where entire companies can build their products using lots of compute.

人们已经广泛讨论过视频模型将如何指向这些生成的实时用户界面。

People have talked a lot about how video models are gonna point towards these generated real time generated user user interfaces.

这将需要大量的计算资源。

That will take a lot of compute.

希望转型的企业将使用大量的计算资源。

Enterprises that want to transform their business will use a lot of compute.

想要提供优质个性化医疗的医生，会持续测量每位患者所能提供的所有体征数据。

Doctors that want to offer good personalized healthcare that are like constantly measuring every sign they can get from each individual patient.

你可以想象这会消耗大量的计算资源。

You can imagine that using a lot of compute.

很难准确描述我们目前为生成AI输出所使用的计算量有多大。

It's hard to frame how much compute we're already using to generate AI output in the world.

但这些数字都非常粗略。

But these are horribly rough numbers.

所以，以这种方式谈论可能显得不够严谨，但我总觉得这类思维实验有点用处。

So and I think it's like undisciplined to talk this way, but I I always find these like mental thought experiments a little bit useful.

所以，请原谅我的不严谨。

So forgive me for the sloppiness.

假设今天一家AI公司每天可能从前沿模型中生成大约一万亿个token。

Let's say that an AI company today might be generating something on the order of 10,000,000,000,000 tokens a day out of frontier models.

你知道，可能更多，但我认为没有人会达到千万亿token的量级。

You know, more, but not it's not like a a quadrillion tokens for anybody, I don't think.

假设世界上有80亿人，假设平均每个人——我认为这些数字完全错了。

Let's say there's 8,000,000,000 people in the world, and let's say on average someone's these are, I think, totally wrong.

但假设一个人每天输出的token平均为两万个。

But let's say someone the average number of tokens outputted by a person per day is like 20,000.

然后你可以开始比较，公平地说，我们需要比较的是今天模型提供商的输出token，而不是所有消耗的token。

You can then start to and the token you could to be fair, then we have to compare the output tokens of a model provider today, not not all the tokens consumed.

但你可以开始观察并说，一家公司的这些模型每天输出的token将超过全人类的总和。

But you can start to look at this and you can say, we're gonna have these models at a company be outputting more tokens per day than all of humanity put together.

然后是十倍，再然后是一百倍。

And then 10 times that, and then a 100 times that.

而且，某种程度上，这种比较真的很荒谬。

And, you know, in some sense, it's like a really silly comparison.

但某种程度上，它揭示了地球上智力运算的规模——人类大脑与AI大脑之间的对比。

But in some sense, it gives a magnitude for, like, how much of the intellectual crunching on the planet is, like, human brains versus AI brains.

那种相对增长率确实很有趣。

That's kind of the relative growth rates there are are interesting.

所以我在想，你知道吗，是否存在对这种算力的需求？

And so I'm wondering, are do you know that there is this demand to use this compute?

比如说，如果OpenAI将算力翻倍用于科学或医学研究，我们会不会取得明确的科学突破？

Potentially, like so for instance, would we have sure fires like scientific breakthroughs if OpenAI were to put double the compute towards science or with medicine like are Yeah.

我们会不会有明确的能力来辅助医生？

Would we have that clear ability to assist doctors?

这些在多大程度上是基于推测，而不是基于你今天所见的清晰理解？

Like, how much of this is sort of supposition of what's to happen versus clear understanding based off of what you see today that it

根据我们今天所看到的一切，这一切都必将发生。

Everything will based off what we see today is that it will happen.

这并不意味着未来不会发生什么疯狂的事情。

It does not mean some crazy thing can't happen in the future.

有人可能会发现一种全新的架构，从而带来一万倍的效率提升，那时我们可能就过度建设了。

Someone could discover some completely new architecture and there could be a 10,000 times, you know, efficiency gain, then we would have really probably overbuilt for a while.

但我们现在看到的一切——模型在每个新层级上进步的速度、人们多么想使用它们、每次成本下降后人们多么更想使用它们——

But everything we see right now about how quickly the models are getting better at each new level, how much more people wanna use them, each time we can bring the cost down, how much more people really wanna use them.

所有这些都向我表明，需求将会增加，人们会用它们来做各种美好的事情，也会用来做些无聊的事情。

Everything about that indicates to me that there will be increasing demand and people using these for wonderful things, for silly things.

但这看起来分明就是未来的模样。

But it it just so seems like this is the shape of the future.

这不仅仅是每天能处理多少个标记的问题。

It's not just like it's not just, you know, how many tokens we can do per day.

而是我们能多快地完成它们。

It's how fast we can do them.

随着这些编码模型变得越来越好，它们可以长时间思考，但你并不想等那么久。

As these coding models have gotten better, they can think for a really long time, but you don't wanna wait for a really long time.

所以还会存在其他维度。

So there will be other dimensions.

不仅仅是我们可以处理的令牌数量，还有在少数几个关键维度上对智能的需求以及我们能用它们做什么。

It will not just be the number of tokens that that we can do, but the demand for intelligence across a small number of axes and what we can do with those.

如果你面临一个非常困难的医疗问题，你是想用5.2，还是愿意用5.2 Pro，即使它需要多得多的令牌？

You know, if you're like, if you have like a really difficult health care problem, do you wanna use 5.2 or do you wanna use 5.2 pro even if it takes dramatically more tokens?

我会选择更好的模型。

I'll go with the better model.

我觉得你会的。

I think you will.

让我们再深入一层。

Let's just try to go one level deeper.

谈到科学发现。

Going to the scientific discovery.

你能举一个科学家的例子吗？

Can you give an example of, a scientist?

它不一定非得是将来的事，也许就是你现在就知道的一个。

It doesn't have to well, maybe it's one that you know today.

就像是，解决某个问题 x。

That's like, have problem x.

如果我投入计算资源 y，我就能解决它，但我现在还做不到。

And if I put, you know, compute y towards it, I will solve it, but I'm not able to today.

今天早上在 Twitter 上有一件事，一群数学家正在互相回复推文。

There was a thing this morning on Twitter where a bunch of mathematicians were saying they were all, like, replying to each other's tweets.

他们说，我之前真的怀疑语言模型能否变得优秀。

They're like, I was really skeptical that LMs were ever gonna be good.

5.2 是那个让我突破界限的版本。

5.2 is the one that crossed the boundary for me.

它做到了，你知道的，在一些帮助下，它完成了这个小证明，发现了一个小成果，但这真的改变了我的工作流程。

It did it, you know, figured out this, it with some help, it did this small proof, it it discovered this small thing, but it's this is actually changing my workflow.

然后其他人也纷纷附和说：‘我也是。’

And then people were piling on saying, yeah, me too.

我的意思是，有些人说5.1已经达到了，但没多少人这么认为。

I mean, some people were saying 5.1 was already there, not many.

但那是一个非常近期的例子。

But that that was like that's a very recent example.

这个模型才发布五天左右。

This model's only been out for five days or something.

人们都说，好吧。

Where people are like, alright.

你知道的，数学家们，没错。

You know, the mathematic Yeah.

数学研究界似乎在说：好吧，重要的事情刚刚发生了。

The mathematics research community seems to say like, okay, something important just happened.

我看到格雷格·布罗克曼在他的动态中一直在强调这些不同的数学和科学应用，我认为5.2在这些群体中引起了某种共鸣。

I've seen Greg Brockman has been highlighting all these different mathematical scientific uses in his feed and something has clicked, I think, with 5.2 among these communities.

所以随着事情的发展，看看会发生什么会很有趣。

So it'll be interesting to see what happens as as things progress.

我们不喜欢，计算方面的一个难点是，你必须提前很久做准备。

We don't like, one of the hard parts about compute Mhmm.

在规模上，你得提前很久规划。

At the scale is you have to do it so far in advance.

所以，你提到的那1.4万亿，我们会花很长一段时间来投入。

So, you know, that 1,400,000,000,000.0 you mentioned, we'll spend it over a very long period of time.

我希望我们能更快一点。

I wish we could do it faster.

我认为如果我们能更快完成，会有需求，但建造这些项目、为数据中心供电、制造芯片和系统、搭建网络等等，都需要极其漫长的时间。

I think there would be demand if we could do it faster, but it just takes an enormously long time to build these projects and the energy to run the data centers and the chips and the systems and the networking and everything else.

这会持续一段时间，但从去年到现在，我们的算力大概增长了三倍。

So that will be over a while, but, you know, we, from a year ago to now, we probably about tripled our compute.

明年我们希望再次将算力翻三倍，之后再翻三倍。

We'll triple our compute again next year, hopefully, again after that.

收入增长可能比算力还快一点，但大致还是跟我们的算力规模同步。

Revenue grows even a little bit faster than that, but it does roughly track our compute fleet.

所以我们从未遇到过无法充分变现所有计算资源的情况。

So we we have never yet found a situation where we can't really well monetize all the compute we have.

如果我们有，我想如果我们有双倍的计算资源，我们的收入也会翻倍，对吧？

If we had I think if we had, you know, double the compute, we'd be at double the revenue right

好的。

Okay.

让我们来谈谈数字吧，既然你提到了。

Let's let's talk about numbers since you brought it up.

收入正在增长。

Revenue is growing.

计算支出在增长，但计算支出的增长速度仍然超过收入增长。

Compute spend is growing, but compute spend still outpaces revenue growth.

我认为报道的数字是，OpenAI预计从现在到2028或2029年1月之间将亏损约1200亿美元，之后才会实现盈利。

I think the numbers that have been reported are OpenAI is supposed to lose something like 120,000,000,000 between now and January twenty twenty eight, twenty nine where you're gonna become profitable.

所以你能谈谈这会如何变化吗？

So talk a little bit about like, how does that change?

转折点在哪里？

Where does the turn happen?

我的意思是，随着收入增长，推理在整体算力中的占比越来越大，最终会超过训练成本。

I mean, as revenue grows and as inference becomes a larger and larger part of the fleet, it eventually subsumes the training expense.

这就是我们的计划。

So that's the plan.

在训练上花很多钱，但赚得越来越多。

Spend a lot of money training, but make more and more.

如果我们没有持续如此大幅地增加训练成本，我们早就盈利了。

If we if we weren't continuing to grow our training costs by so much, we would be profitable way, way earlier.

但我们所押注的是，要非常积极地投资于这些大型模型的训练。

But the bet we're making is to invest very aggressively in training these big models.

全世界都在关注你们的收入如何与支出相匹配。

The whole world is wondering how your revenue will line up with the spend.

有人问过，今年收入能否达到200亿美元，而支出承诺高达1.4万亿美元。

The question's been asked if the trajectory is to hit $20,000,000,000 in revenue this year And the spend commitment is 1,400,000,000,000.0.

所以我认为最好明确地说明一下

So I think it would be great just to lay out

很长一段时间。

very long period.

是的。

Yeah.

总体而言。

Overall.

这就是我想跟你提这件事的原因。

And that's why I wanted to bring it up to you.

我认为最好一次性向所有人清楚地说明这些数字将如何运作。

I think it would be great to just lay it out for everyone once and for all how those numbers are gonna work.

这非常困难，说实话，我肯定做不到，我认识的几乎没人能做到。

It's very hard to like really I find that one thing I certainly can't do it and very few people I've ever met can do it.

你知道，你可能对很多数学问题有很好的直觉，但指数增长通常很难让人在脑海中快速建立一个清晰的框架。

You know, you can like you have good intuition for a lot of mathematical things in your head, but exponential growth is usually very hard for people to do a good, quick mental framework on.

不知为何，进化让我们必须在头脑中擅长处理很多数学问题。

Like, for whatever reason, there were a lot of things that evolution needed us to be able to do well with math in our heads.

但建模指数增长似乎并不在其中。

Modeling exponential growth doesn't seem to be one of them.

因此，我们相信，我们可以在相当长的一段时间内保持收入的陡峭增长曲线。

So, the thing we believe is that we can stay on a very steep growth curve of revenue for quite a while.

而我们目前看到的一切都继续印证了这一点。

And everything we see right now continues to indicate that.

如果我们没有足够的计算能力，就无法做到这一点。

We cannot do it if we don't have the compute.

再说一遍，我们目前严重受限于计算能力，这直接影响了收入。我认为，如果我们有一天拥有了大量无法以盈利方式每单位计算能力变现的算力，那时就很有必要问一句：这一切到底该怎么运作？

Again, we're so compute constrained, and it hits the revenue line so hard that I think if we get to a point where we have like a lot of compute sitting around that we can't monetize on a, you know, profitable per unit of compute basis, be very reasonable to say, okay, this is like a little, how's this all going to work?

但我们已经用多种方式推演过这个问题。

But we've penciled this out a bunch of ways.

当然，随着我们一直在做的降低计算成本的工作逐步见效，我们在每美元算力性能（flops per dollar）上也会变得更加高效。

We will, of course, also get more efficient on, like, a flops per dollar basis as, you know, all of the work we've been doing to make compute cheaper comes to pass.

但我们看到了消费者增长，也看到了企业增长，还有很多我们尚未推出但即将推出的新业务类型。

But we see this consumer growth, we see this enterprise growth, there's a whole bunch of new kinds of businesses that we haven't even launched yet, but will.

但计算能力是推动这一切的生命线。

But compute is really the lifeblood that enables all of this.

所以，你知道，途中会有一些关键节点。

So we you know, there's like checkpoints along the way.

如果我们对时间或计算的预测稍有偏差，我们还是有一些灵活性的。

If we're a little bit wrong about our timing or math, we can we have some flexibility.

但我们一直面临计算能力不足的问题。

But we have always been in a compute deficit.

我们能做的事情始终受到限制。

It is always constrained what we're able to do.

不幸的是，我认为这种情况将一直存在，但我希望它能逐渐减轻，因为我认为我们可以推出许多优秀的产品和服务，这将是一个非常出色的业务。

I unfortunately think it will always be the case, but I wish it were less the case and I'd like to get it to be less of the case over time, because I think there's so many great products and services that we can deliver and it'll be a great business.

好的。

Okay.

所以训练成本实际上降低了。

So it's effectively training costs go down.

以百分比来看。

As a percentage.

但总体上成本是上升的。

They go up essentially overall.

但你的预期是，通过这种企业推广，以及人们愿意为ChatGPT的API付费，OpenAI能够通过收入增长来覆盖这些成本。

But And then your expectation is through things like this this enterprise push, through things like people being willing to pay for ChatGPT through the API, OpenAI will be able to grow revenue enough to pay for it with revenue.

是的，这就是计划。

Yeah, that is the plan.

现在我认为，最近市场在这方面有点失控了。

Now I think the thing, so the market's been kind of losing its mind over this recently.

我认为让市场感到不安的是债务进入了这个等式。

I think the thing that has spooked the market has been the debt has entered into this equation.

而关于债务的概念是，当你有可预测的东西时，才会去借债。

And the idea around debt is you take debt out when there's something that's predictable.

然后公司会借入债务，进行建设，并获得可预测的收入。

And then companies will take the debt out, they'll build and they'll have predictable revenue.

但这是一种全新的类别。

But it's it's the this is a new category.

它是不可预测的。

It's it is unpredictable.

那么，你怎么看待债务在这里出现这一事实呢？

Is is that how do you think about the fact that debt has entered the picture here?

首先，我认为今年早些时候，市场更加疯狂了，比如我们会见一些公司，第二天它们的股价就上涨了20%或15%。

So first of all, I think the market more lost its mind when earlier this year, you know, we would like meet with some company and that company's stock would go up 20% or 15% the next day.

那太疯狂了。

That was crazy.

这感觉非常不健康。

That felt really unhealthy.

实际上，我现在很高兴市场中多了一点怀疑和理性，因为在我看来，我们当时正完全走向一个极其不稳定的泡沫。

I'm actually happy that there's like a little bit more skepticism and rationality in the market now because it felt to me like we were just totally heading towards a very unstable bubble.

现在我认为人们有了一定的自律性。

And now I think people are some degree of discipline.

所以我认为事情是这样的：之前人们疯狂了，而现在人们变得更加理性了。

So I actually think things are I think people went crazy earlier and now people are being more rational.

在债务方面，我认为我们确实知道，如果我们建设基础设施，这个行业就会有人从中获得价值。

On the debt front, I think we do kind of we know that if we build infrastructure, we the industry, someone's going to get value out of it.

而且这仍然处于非常早期的阶段。

And it's still, it's still totally early.

我同意你的观点，但我认为没有人再质疑AI基础设施不会创造价值。

I agree with you, but I don't think anyone's still questioning there's not going be value from like AI infrastructure.

因此，我认为债务进入这个市场是合理的。

And so I think it is reasonable for debt to enter this market.

我认为还会有其他类型的金融工具。

I think there will also be other kinds of financial instruments.

我猜测我们会看到一些不合理的金融工具，因为人们真的在不断创新融资方式来支持这类事物。

I suspect we'll see some unreasonable ones as people really, you know, innovate about how to finance this sort of stuff.

但你知道，像贷款给公司建设数据中心，我觉得这没问题。

But, you know, like lending companies money to build data centers, that seems fine to me.

我认为担忧在于，如果事情不能持续快速推进，这里有一种可能的情景，你可能不同意，但模型的进步会达到饱和，那么基础设施的价值就会低于预期价值。

I think the fear is that if things don't continue at pace, like here's one scenario, and you'll probably disagree with this, but like the model progress saturates, Then the infrastructure becomes worth less than the anticipated value was.

然后，是的，这些数据中心对某些人来说仍然有价值，但它们可能会被清算，然后有人以折扣价购入。

And then yes, those data centers will be worth something to someone, but it could be that they get liquidated and someone buys them at a discount.

是的。

Yeah.

而且我怀疑，在此过程中，会经历一些繁荣与萧条。

And I do suspect, by the way, there will be some like booms and busts along the way.

这些事情从来都不是一条完美的直线。

These things are never a perfectly smooth line.

首先，对我来说这非常清楚，我甚至愿意拿公司做赌注，模型将会变得好得多、好得多。

First of all, it seems very clear to me and this is like a thing I'd happily bet the company on that the models are going to get much, much better.

我们对这一点已经有了相当好的洞察。

We have like a pretty good window into this.

我们对这一点非常有信心。

We're very confident about that.

即使模型没有继续进步，我认为世界上也存在大量惯性，需要很长时间才能适应这些变化。

Even if they did not, I think the there's like a lot of inertia in the world that takes a while to figure out how to adapt to things.

我认为5.2版本所代表的经济价值，相对于世界迄今为止从中获取的成果，是如此巨大，即使你把模型冻结在5.2版本，还能创造出多少更多的价值，从而带来多少收入呢？

The overhang of the economic value that I believe 5.2 represents relative to what the world has figured out how to get out of it so far is so huge that even if you froze the model at 5.2, how much more, like, value can you create and thus revenue can you drive?

我敢赌上一大笔钱。

I bet a huge amount.

事实上，你没问这个，但容我稍微发几句牢骚。

In fact, you didn't ask this, but if I can go on a rant for a second.

我们过去经常讨论一个二维矩阵：短周期 vs 长周期，慢起飞 vs 快起飞，不同时间点我们觉得各种可能性在变化，而根据你在该矩阵中的位置，可以大致理解世界应该优化哪些决策和战略。

We used to talk a lot about this two by two matrix of short timelines, long timelines, slow takeoff, fast takeoff, and where we felt at different times the kind of probability was shifting and that that was going to be, you could kind of understand a lot of the decisions and strategy that the world should optimize for based off of where you were going to be on that two by two matrix.

在我的脑海里，关于这个问题的图景中，出现了一个Z轴，那就是小滞后 vs 大滞后。

There's like a z axis in my head, in my picture of this that's emerged, which is small overhang, big overhang.

我之前大概没怎么认真思考过这一点，但现在回过头来看，我一定是默认了滞后不会那么巨大。

And I kind of thought that I guess I didn't think about that hard, but like my retro on this is I must have assumed that the overhang was not going to be that massive.

如果这些模型具有巨大价值，世界本应很快学会如何利用它们。

That if the models had a lot of value in them, the world was pretty quickly going to figure out how to deploy that.

但在我看来，如今大多数地区的潜力尚未被充分释放，这种滞后将非常巨大。

But it looks to me now like the overhang is going be massive in most of the world.

你会看到一些特定领域，比如某些程序员，通过采用这些工具，生产力将大幅提升。

You'll have these like areas like, you know, some set of coders that'll get massively more productive by adopting these tools.

但总体而言，你拥有一个如此聪明的模型，而坦白说，大多数人仍在问着与GPT-4时代类似的问题。

But on the whole, you have this crazy smart model that to be perfectly honest, most people are still asking the similar questions they did in the GPT-four realm.

科学家不同，程序员不同，也许知识型工作将会发生变化。

Scientists different, coders different, Maybe knowledge work is going to get different.

但存在巨大的滞后效应，这给世界带来了一系列非常奇特的后果。

But there is a huge overhang and that has a bunch of very strange consequences for the world.

我们尚未完全理解这一切将如何展开，但这与我几年前的预期截然不同。

We have not wrapped our head around all the ways that's going to play out yet, but it's very much not what I would have expected a few years ago.

我有

I have

关于这个能力过剩，我有个问题想问你。

a question for you about this capability overhang.

基本上，模型能做的事情远比它们目前所做的要多。

Basically, models can do a lot more than they've been doing.

我想弄清楚，为什么这些模型的性能远超当前的使用水平，但许多企业在尝试应用时，却没能获得投资回报。

I'm trying to figure out how the models can be that much better than they're being used for, but a lot of businesses, when they try to implement them, they're not getting a return on their investment.

或者至少，他们是这么告诉麻省理工学院的。

Or at least that's what they tell MIT.

我不太确定该怎么看待这个问题，因为我们不断听到企业说，如果你把GPT 5.2的价格提高十倍，我们依然愿意支付。

I'm not sure quite how to think about that because we hear all these businesses saying, you know, if you 10x the price of GPT 5.2, we would still pay for it.

你们的价格严重低估了它的价值。

You're hugely underpricing this.

我们从它身上获得了巨大的价值。

We're getting all this value out of it.

所以我觉得这种说法不太对。

So I don't that doesn't seem right to me.

当然，如果你听听程序员怎么说，他们会说，这东西我愿意付一百倍的价格之类的。

Certainly, if you talk about, like, what coders say, they're like, this is, you know, I'd pay a 100 times the price or whatever.

只是官僚主义把事情搞砸了。

Just be bureaucracy that's messing things up.

假设你相信GDP估值数据，也许你有充分的理由不相信，因为它们可能是错的，但我们就假设它们是对的。

Let's say you believe the GDP valve numbers and maybe you don't for good reason, maybe they're wrong, but let's say it were true.

对于这些定义明确、并非持续时间很长的知识型任务，有七成的时间，人们会对5.2的输出感到满意甚至更满意。

And for kind of these well specified, not super long duration knowledge work tasks, seven out of 10 times, would be as happy or happier with the 5.2 output.

那你就应该大量使用它。

You should then be using that a lot.

但人们改变工作流程却需要很长时间。

And yet it takes people so long to change their workflow.

他们太习惯让初级分析师做演示文稿之类的事了，结果就是，这种惯性比我想象的还要顽固。

They're so used to asking the junior analyst to make a deck or whatever that they're gonna like, it just that's stickier than I thought it was.

你知道，我到现在还是用几乎同样的方式管理我的工作流程，尽管我知道自己本可以更多地使用AI。

You know, I still kind of run my workflow in very much the same way, although I know that I could be using AI much more than I am.

嗯。

Yep.

好的。

Alright.

我们还剩十分钟。

We got ten minutes left.

我还有四个，哇。

I got four Wow.

真快啊。

That was quick.

我有四个问题。

I got four questions.

我们来看看能不能快速过一遍。

Let's see if we can lightning round, through them.

所以，你们正在开发的设备。

So, the device that you're working on.

我们将在广告后回归，邀请OpenAI首席执行官萨姆·阿尔特曼。

We'll be back with OpenAI CEO Sam Altman right after this.

Capital One的技术团队不仅仅在谈论多智能体AI。

Capital One's tech team isn't just talking about multiegenthic AI.

他们已经部署了一个。

They already deployed one.

它被称为聊天礼宾服务，正在简化汽车购买流程。

It's called chat concierge, and it's simplifying car shopping.

通过自我反思、分层推理和实时API检查，它不仅能帮助买家找到心仪的汽车。

Using self reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love.

还能帮助安排试驾、获得贷款预批，并估算旧车置换价值。

It helps schedule a test drive, get preapproved for financing, and estimate trade in value.

先进、直观且已部署。

Advanced, intuitive, and deployed.

这就是他们的优势所在。

That's how they stack.

这就是Capital One的技术。

That's technology at Capital One.

如今，每一分钱都该更努力地工作，但要弄清楚把钱放在哪里可能会让人困惑。

These days, it feels like every dollar should be working a little harder, but figuring out where to put your cash can be confusing.

这就是Wealthfront的用武之地。

That's where Wealthfront comes in.

Wealthfront是一个以技术为驱动的金融平台，旨在帮助你将储蓄转化为长期财富。

Wealthfront is a tech driven financial platform built to help you grow your savings into long term wealth.

他们的高收益现金账户通过合作银行，截至2025年11月7日，为未投资资金提供3.5%的年化收益率。

Their high yield cash account through program banks offers a 3.5% APY on your uninvested cash as of 11/07/2025.

而且没有月费，无需最低或最高余额即可享受该利率，你甚至可以随时通过24/7即时免费提现至符合条件的账户，让你的资金始终触手可及。

And there are no monthly fees, no minimum or maximum balance to earn that rate, and you can even make free instant withdrawals to eligible accounts in just minutes, twenty four seven, so your money can always be within reach.

目前，Wealthfront为新客户在前三个月内，对最高15万美元的余额提供比基础利率高出0.65%的额外年化收益率。

Right now, Wealthfront is offering new clients an extra point 65% APY over the base rate for three months on up to a $150,000 balance.

当你开设首个现金账户时，总年化收益率可达4.15%的浮动利率。

That's a total of 4.15% variable APY when you open your first cash account.

立即前往 wealthfront.com/bigtech 注册。

Go to wealthfront.com/bigtech to sign up today.

这是对 Wealthfront 的付费推荐。

This is a paid testimonial for Wealthfront.

此内容可能无法反映其他人的体验，且不保证未来表现或成功。

It may not reflect the experience of others, and there's no guarantee of future performance or success.

Wealthfront 证券业务并非一家大型机构。

Wealthfront brokerage is not a big.

利率可能随时调整。

Rate is subject to change.

促销条款和条件适用。

Promo terms and conditions apply.

如需更多信息，请参见本集描述。

For more information, see the episode description.

这里是迈克尔·刘易斯。

Michael Lewis here.

我最畅销的书《大空头》讲述了2008年美国房地产市场泡沫形成与破裂的故事。

My best selling book, The Big Short, tells the story of the buildup and birth of The US housing market back in 2008.

十年前，《大空头》被拍成了获得奥斯卡奖的电影，现在我首次将其制作成有声书，并由我亲自朗读。

A decade ago, The Big Short was made into an Academy Award winning movie, and now I'm bringing it to you for the first time as an audiobook narrated by yours truly.

《大空头》所讲述的做空市场意味着什么，以及谁真正为失控的金融体系买单，如今比以往任何时候都更具有现实意义。

The Big Short story, what it means to bet against the market, and who really pays for an unchecked financial system, is as relevant today as it's ever been.

现在请前往 pushkin.fm/audiobooks 或任何有声书销售平台获取《大空头》。

Get The Big Short now at pushkin.fm/audiobooks or wherever audiobooks are sold.

我听说的是，手机尺寸，没有屏幕。

What I've heard, phone size, no screen.

如果手机本来就没有屏幕，那为什么不能只是一个应用呢？

Why couldn't it be an app if it's the phone if it's the phone without a screen?

首先，我们将推出一系列小型设备。

First, we're gonna do a fam a small family of devices.

它不会是一个单一的设备。

It will not be a single device.

随着时间的推移，这并不是猜测，所以我尽量不要完全错。

There will be over time a this is this is not speculation, so I may try not to be totally wrong.

但我认为，随着时间的推移，人们使用计算机的方式会发生转变：从一种迟钝、被动的工具，变成一种非常智能、主动的系统，它能理解你生活的方方面面、你所处的环境，以及你身边的人——无论是物理上靠近你，还是通过你正在使用的计算机与你关联的人。

But I think there will be a shift over time to the way people use computers, where they go from a sort of dumb reactive thing to a very smart, proactive thing that is understanding your whole life, your context, everything going on around you, aware of that people around you physically or close to you via a computer that you're working with.

我认为，当前的设备并不适合这种未来的世界。

And I don't think current devices are well suited to that kind of world.

我深信，我们总是处于设备能力的极限边缘，你知道，你有一台电脑，它有各种设计选择。

And I am a big believer that we like, we work at the limit of our devices, you know, you have that computer and it has a bunch of design choices.

比如，它可以是开放的或封闭的，但不可能——注意这个访谈——它既封闭，又能在你忘记问萨姆问题时，悄悄在你耳边提醒你。

Like, could be open or closed, but it can't be you know, there's not like a, pay attention to this interview, but be closed and like whisper in my ear if I forget to ask Sam a question or whatever.

也许那样会很有帮助。

Maybe that would be helpful.

而且，比如，它有一个屏幕，这限制了你能做的事情，就像图形用户界面几十年来一直如此。

And there's, like, you know, there's, like, a screen and that, limits you to the kind of same way we've had graphical user interfaces working for many decades.

还有，比如，键盘的设计初衷是为了减缓你输入信息的速度。

And there's, you know, a keyboard that was built to like slow down how fast you could get information into it.

这些长期以来一直是不容置疑的假设，但它们确实有效。

And these have just been unquestioned assumptions for a long time, but they worked.

然后，一种全新的事物出现了，它打开了一个可能性空间，但我认为当前设备的形态并不是最优的匹配。

And then this totally new thing came along and it opens up a possibility space, but I don't think the current form factor of devices is the optimal fit.

如果这种全新的可用性出现时，设备形态却恰好是最优的，那反而会很奇怪。

It'd be very odd if it were for this like incredible new affordance we have.

天啊，关于这个我们可以聊上一小时，但我们还是继续下一个话题吧。

Oh man, we could talk for an hour about this, but let's move on to the next one.

云。

Cloud.

我们已经讨论过构建云了。

We've talked about building a cloud.

这是我们收到的一封听众来信。

Here's an email we got from a listener.

在我们公司，我们正在从Azure迁移出来，直接与OpenAI集成，以驱动产品中的AI体验。

At my company, we're moving off Azure and directly integrating with OpenAI to power our AI experiences in the product.

重点是通过整个系统注入数万亿个令牌来驱动AI体验。

The focus is to insert a stream of trillions of tokens powering AI experiences through the stack.

这就是计划以这种方式打造一个庞大的云业务吗？

Is is that the plan to build a big big cloud business in that in that way?

首先，数万亿个令牌，真是大量的令牌。

First of all, trillions of tokens, a lot of tokens.

如果你问到计算需求和我们的企业战略，我们的企业客户已经明确表示了。

And if, you know, you asked about the need for compute and our enterprise strategy, like, our enterprises have been clear Yeah.

他们告诉我们，希望从我们这里购买多少令牌。

With us about how many tokens they'd like to buy from us.

但到2026年，我们仍将无法满足需求。

And we are going to again fail in 2026 to meet demand.

但我们的策略是，大多数公司都希望找像我们这样的公司，说：我希望我的公司名字与AI联系在一起。

But the strategy is companies, most companies seem to want to come to a company like us and say, I like the name of my company with AI.

我需要一个为我公司定制的API。

I need an API customized for my company.

我需要为我的公司定制ChachiPTO企业版。

I need ChachiPTO Enterprise customized for my company.

我需要一个平台，能够运行所有这些代理，并且让我放心地使用我的数据。

I need a platform that can like run all these agents that I can trust my data on.

我需要有能力将数万亿的令牌集成到我的产品中。

I need the ability to get trillions of tokens into my product.

我需要让我的所有内部流程变得更加高效。

I need the ability to have all my internal processes be more efficient.

而我们目前还没有一个出色的全方位解决方案来满足他们。

And we don't currently have like a great all in one offering for them.

我们希望打造这样一个方案。

And we'd like to make that.

你的雄心是将它打造成与AWS和Azure齐名的产品吗？

Is your ambition to put it up there with the AWS and Azure of the world?

我认为它和那些属于不同类型的東西。

I think it's I think it's a different kind of thing than those.

我的意思是，我其实并没有野心去提供所有那些托管网站的服务，我甚至都不太清楚那些是什么。

Like, don't I don't really have an ambition to go offer whatever all the services you have to offer to host a website or I don't even know.

但我觉得这个概念没错。

But but I I I think the concept yeah.

我猜人们会继续使用他们所谓的网络云。

My my guess is that people will continue to have their call it web cloud.

然后我认为还会出现另一种情况，即一家公司会说：我需要一个AI平台，来处理我想做的所有内部事务、我想提供的服务等等。

And then I think there will be this other thing where, like, a company will be like, I need an AI platform for everything that I wanna do internally, service I wanna offer, whatever.

而且，你知道，从某种意义上说，它确实运行在物理硬件上，但我认为它会是一个相当不同的产品形态。

And, you know, like it does kind of live on the physical hardware in some sense, but I think it'll be a fairly different product offering.

我们快速聊聊发现吧。

Let's talk about discovery quickly.

你提到过一件事让我很感兴趣，你说模型明年会有一些小的发现，五年内会有大的突破。

You've said something that's been really interesting to me, you that you think that the models or maybe it's people working with models or the models make small discoveries next year and big ones within five.

是指模型本身吗？

Is that the models?

是那些与它们一起工作的人吗？

Is it people working alongside them?

是什么让你确信这会发生？

And what makes you confident that that's going to happen?

是的。

Yeah.

人们使用模型，比如那些能自己找出要问什么问题的模型，这感觉还很遥远。

People using the models like the models that can like figure out their own questions to ask that does feel further off.

但如果世界正在从新知识中受益，我们应当感到非常兴奋。

But if the world is benefiting from new knowledge, like we should be very thrilled.

而且，我觉得人类进步的整个历程就是我们不断制造出更好的工具，然后人们用它们去做更多的事情。

And, you know, like, I think the whole course of human progress has been that we build these better tools and then people use them to do more things.

在这个过程中，他们又制造出更多的工具。

Then out of that process, they build more tools.

这就像一层层的脚手架，我们一代代、一发现一发现地攀爬上去。

And it's this like scaffolding that we climb, like layer by layer, generation by generation, discovery by discovery.

人类提出问题这一事实，我认为丝毫不会削弱这个工具的价值。

And the fact that human's asking the question, I think in no way diminishes the value of the tool.

所以我觉得这很棒。

So I think it's great.

我非常开心。

I'm all happy.

今年年初，我以为小发现会从2026年开始。

At the beginning of this year, I thought the small discoveries were going start in 2026.

它们在2025年，也就是2025年底就开始了。

They started in 2025, in late twenty twenty five.

这些发现都还非常微小。

Again, these are very small.

我真的不想夸大它们。

I really don't want to overstate them.

但任何一点进展，对我来说都与毫无进展有着本质的不同。

But anything feels qualitatively to me very different than nothing.

当然，当我们三年前推出ChatGPT时，那个模型并不会为人类知识总量做出任何新贡献。

And certainly in the when we launched ChatGPT three years ago, that model was not going to make any new contribution to the total of human knowledge.

从现在看未来五年，这条通往重大发现的旅程，我怀疑这就像AI正常的爬坡过程。

What it looks like from here to five years from now, this journey to big discoveries, I suspect it's just like the normal hill climb of AI.

它每季度只是变得稍微好一点。

It just gets like a little bit better every quarter.

然后突然间，我们发现，由这些模型增强的人类正在做五年前人类绝对做不到的事情。

And then all of a sudden we're like, woah, humans augmented by these models are doing things that humans five years ago just absolutely couldn't do.

而且，不管我们主要将这些成就归功于更聪明的人类还是更聪明的模型，只要我们能获得科学发现，我就

And, you know, whether we mostly attribute that to smarter humans or smarter models, as long as we get the scientific discoveries, I'm

无论如何都感到非常高兴。

very happy either way.

明年上市吗？

IPO next year?

我不知道。

I don't know.

你希望成为一家上市公司吗？

Do you want to be a public company?

你看起来可以长期保持私有状态。

You seem like you can operate private for a long time.

你在资金需求之前就上市了。

When you go before you needed to, in terms of funding.

这里涉及很多因素。

There's like a whole bunch of things at play here.

我认为公众市场能够参与价值创造，这很酷。

I do think it's cool that public markets get to participate in value creation.

如果你看看任何之前的公司，我们上市的时间将会非常晚。

And, in some sense, we will be very late to go public, if you look at any previous company.

做一家私有公司真好。

It's wonderful to be a private company.

我们需要大量资金。

We need lots of capital.

我们迟早会突破所有的股东人数限制之类的问题。

We're gonna, you know, cross all of the sort of shareholder limits and stuff at some point.

那么，我作为一家上市公司的CEO会感到兴奋吗？

So am I excited to be a public company CEO?

0%。

0%.

我对OpenAI成为一家上市公司感到兴奋吗？

Am I excited for OpenAI to be a public company?

在某些方面，是的。

In some ways, I am.

但在某些方面，我觉得这会非常烦人。

And in some ways, I think it'd be really annoying.

我仔细听了你的Theo Vaughn访谈。

I listened to your Theo Vaughn interview very closely.

很棒的访谈。

Great interview.

他真的很酷。

He was really cool.

Theo 真的很懂行。

Theo really knows what he's talking about.

是的。

Yeah.

他太棒了。

He's awesome.

决定让约书亚·本吉奥。

Deciding Joshua Bengio.

他做了充分的准备。

He's he did his homework.

你告诉他，就在 GPT-5 发布之前，GPT-5 在几乎所有方面都比我们聪明。

You told him this was right before GPT five came out that GPT five is smarter than us in almost every way.

我认为这正是通用人工智能的定义。

I I thought that that was the definition of AGI.

这不就是AGI吗？

Does is that isn't that AGI?

如果不是，那么这个术语是否已经变得有些无意义了？

And and if not, has the term become somewhat meaningless?

这些模型显然在某种

These models are clearly extremely smart on a sort

纯粹的算力基础上极其聪明。

of raw horsepower basis.

你知道，最近几天到处都在说GPT 5.2的智商是147、144、151之类的。

You know, there's all this stuff out in the last couple of days about GPT 5.2 has an IQ of 147 or 144 or 151 or whatever it is.

这取决于是谁的测试，总之是个很高的数字，而且很多领域专家都说它能完成这些惊人的任务，它正在让事情变得更高效。

It's like, you know, depending on whose test it's like, it's some high number and you have like a lot of experts in their field saying it can do these amazing things and it's like contributing, it's making it more effective.

你提到了我们讨论过的所有事物的GDP。

You have the GDP of all things we talked about.

但你缺少的一点是，模型无法在今天意识到自己做不到某件事，然后去主动学习如何掌握它、理解它。

One thing you don't have is the ability for the model to not be able to do something today, realize it can't, go off and figure out how to learn to get good at that thing, learn to understand it.

当你第二天回来时，它就能做对了。

And when you come back the next day, it gets it right.

这种持续学习的能力，就像幼儿能做到的那样。

And that kind of continuous learning like toddlers can do it.

在我看来，这似乎是我们需要构建的重要部分。

It does seem to me like an important part of what we need to build.

那么，没有这种能力，你还能拥有大多数人会认为是AGI的东西吗？

Now, can you have something that most people would consider an AGI without that?

我会说，显然不能。

I would say clear.

我的意思是，很多人会认为我们现在的模型就已经是AGI了。

I mean, there's a lot of people that would say we're at AGI with our current models.

我认为几乎所有人都会同意，如果我们现在的智能水平再加上这种能力，那显然会非常接近AGI。

I think almost everyone would agree that if we were at the current level of intelligence and had that other thing, it would clearly be very AGI like.

但也许世界上大多数人会说：好吧，就算没有这个，它也能完成大多数重要的知识性任务。

But maybe most of the world will say: Okay, fine even without that, like it's doing most knowledge tasks that matter.

比我们更聪明，在大多数方面都超越了我们，我们已经达到了AGI。

Smarter than us and most of us in most ways, we're at AGI.

你知道，它正在发现新的科学小发现，我们已经达到了AGI。

You know, it's discovering small pieces of new science, we're at AGI.

我认为这意味着，尽管我们所有人都很难停止使用这个词，但它定义得非常模糊。

What I think this means is that the term, although it's been very hard for all of us to stop using, is very under defined.

我有一个建议：既然我们在AGI上搞错了，从未明确定义过它，那么现在大家关注的新术语是当我们达到超级智能时。

I have a candidate like, one thing I would love since we got it wrong with AGI, we never defined that, that, you know, the new term everyone's focused about is when we get to super intelligence.

所以，我的建议是我们达成共识：AGI就像一阵风一样过去了，它并没有对世界产生太大影响，或者至少长期来看会，但好吧，我们确实在某个时刻建造了AGI，我们现在正处于一个模糊阶段，有些人认为我们已经有了，有些人认为我们有了，更多人认为我们有了，然后我们会问：接下来是什么？

So, my proposal is that we agree that, you know, AGI kind of went wooshing by, it was didn't change the world that much, or it will in the long term, but okay fine, we've built AGIs at some point, you know, we're in this like fuzzy period where some people think we have and some people think we have and more people think we have and then we'll say, okay, what's next?

超级智能的一个候选定义是：当一个系统能够比任何人类（即使借助AI辅助）更好地担任美国总统、大型公司的首席执行官，或管理一个大型科学实验室时。

A candidate definition for superintelligence is when a system can do a better job being president of The United States, CEO of a major company, you know, running a very large scientific lab than any person can, even with the assistance of AI.

好的。

Okay.

我认为关于国际象棋发生的事很有趣，那时国际象棋已经超越了人类。

I think this was an interesting thing about what happened with chess, where chess got it could be humans.

我对此记忆犹新，就是深蓝那件事。

I remember this very vividly, that deep blue thing.

然后有一段时间，人类和AI合作比单独的AI表现更好。

And then there was a period of time where a human and the AI together were better than an AI by itself.

后来，人类反而拖了后腿。

And then the person was just making it worse.

最聪明的做法是那个没有人类参与的纯AI，它根本不理解自己的强大智能。

And the smartest thing was the unaided AI that didn't have the human, like, not understanding its, its great intelligence.

我觉得类似这样的情况，可以作为一个关于超级智能的有趣框架。

I something like that is like an interesting framework for superintelligence.

我认为这还远得很，但我希望这次能有一个更清晰的定义。

I think it's like a long way off, but I would love to have like a cleaner definition this time around.

好吧，萨姆，我用了三年时间每天都在使用你的产品。

Well, Sam, look, I have been in your products using them daily for three years.

谢谢

Thank you

非常感谢。

very much.

确实进步了很多。

Definitely gotten a lot better.

简直无法想象它们接下来会发展到什么地步。

Can't even imagine where they go from here.

我们会努力尽快让它们变得更好。

We'll we'll try to keep getting them better fast.

好的。

Okay.

这是我们第二次交谈，我非常感谢你两次都如此坦诚。

And this is our second time speaking, and I appreciate how open you've been both times.

所以谢谢你。

So thank you

感谢你的时间。

for your time.

感谢大家收听和观看。

Thank you everybody for listening and watching.

如果你是第一次来，请点击关注或订阅。

If you're here for the first time, please hit follow or subscribe.

我们的频道上有许多精彩的访谈，而且还有更多即将上线。

We have lots of great interviews on the feed and more on the way.

过去一年，我们两次邀请了谷歌DeepMind的首席执行官德米斯·哈萨比斯，其中一次还邀请了谷歌创始人谢尔盖·布林。

This past year, we've had Google DeepMind CEO, Demis Esabes on twice, including one with Google founder Sergey Brin.

我们还邀请了Anthropic的首席执行官达里奥·阿莫迪，2026年还将有更多重磅访谈即将推出。

We've also had Dario Amode, the CEO of Anthropic, and we have plenty of big interviews coming up in 2026.

再次感谢，我们下次再见于《科技大观》播客。

Thanks again, and we'll see you next time on big technology podcast.