Chasing Trillion-Dollar Companies

70% of France is still nuclear in terms of its power generation.

法国到今天仍有 70% 的电力来自核能。

70%?

70%?

Where are all the accidents and where are all the kerfuffles and you know nothing nothing's happened.

那事故在哪儿?那些乱子在哪儿?什么都没发生。

US is 18% and we haven't built a reactor in 40 years.

美国只有 18%,而且我们 40 年没建过一座反应堆。

We had a safety lobby in the 70s basically kill abundantly in energy for us.

70 年代那波安全游说,基本上替我们扼杀了廉价充裕的能源。

There are real outcomes where safety has hurt us and the question is where do we want the spectrum to be on AI for this stuff and there's many worlds, many scenarios, many outcomes.

确实有些结果是安全把我们害了,所以问题是:在 AI 这件事上,我们希望这根标尺停在哪儿?可能的世界有很多,场景有很多,结局也有很多。

Hi listeners, welcome back to No Priors.

各位听众好,欢迎回到 No Priors。

Today is just me and Elad talking about risk management, RSI, how many trillion-dollar companies there can really be, and the ills on regulatory capture.

今天只有我和 Elad,聊风险管理、RSI、到底能有多少家万亿美元公司,以及监管俘获的害处。

For any new founders out there, it's also time to apply to Embed Conviction's low-overhead, high-signal grant program for 10 exceptional startups building at the frontier.

如果你是刚起步的创始人,现在也是申请 Embed 的时候——这是 Conviction 面向 10 家前沿创业公司的低负担、高信号资助计划。

We hold this program twice a year and it's $250,000 in cash on an uncapped note as well as computes and services from our partners, OpenAI, Anthropic, Baseten, and others.

这个项目一年办两次,给的是 25 万美元现金、以无估值上限的可转债形式发放,外加我们合作方 OpenAI、Anthropic、Baseten 等提供的算力和服务。

Most importantly, it's about the company you keep.

但最重要的是:你身边是一群什么样的人。

Our first handful of cohorts have included companies like Cognition, Chai Discovery, Listen Labs, Physical Intelligence, and Floppy Airplanes.

前几批入选的公司包括 Cognition、Chai Discovery、Listen Labs、Physical Intelligence 和 Floppy Airplanes。

People advancing the frontier and diffusing AI into every corner of the economy.

都是在推进前沿、把 AI 渗透进经济每个角落的人。

Find the app online at embed.conviction.com.

申请入口在 embed.conviction.com。

Okay, let's get started.

好,我们开始。

Sarah G, how you doing?

Sarah G,最近怎么样?

Elad, it's good to see you.

Elad,见到你真好。

It's been a while since we just get to hang out with each other.

我们已经好久没这样单纯地坐下来聊天了。

I know, it's been too long.

确实,太久了。

What happened?

怎么回事?

Where you been?

你都跑哪儿去了?

You know, working at companies in DC.

在 DC 那边跟几家公司一起干活。

Trying to take a day off.

想抽一天休息。

You?

你呢?

Uh there's just so much going on right now in AI.

AI 这边现在实在太热闹了。

There's so much going on.

事情多得不行。

It's non-stop.

一刻不停。

It's very exciting times.

是非常刺激的时候。

You chasing the next trillion-dollar company?

你在追下一家万亿美元公司?

Yeah, it's a it's a really interesting point because basically what we had is over the last five years or so, we had three companies roughly go from close to zero to a trillion dollars in market cap, right?

是啊,这个点其实挺有意思的:过去五年左右,我们眼看着三家公司差不多从零冲到了一万亿美元市值。

Anthropic basically didn't exist 5 years ago.

Anthropic 五年前基本上还不存在。

OpenAI um was still quite early.

OpenAI 当时也还很早期。

I think GPT-3 had just come out and SpaceX was trading at 80, 100, something like that.

我记得 GPT-3 刚出来,SpaceX 的估值也就八百亿、一千亿那个量级。

And so suddenly we had this massive inflection in terms of evaluations of these companies.

然后突然之间,这些公司的估值出现了一次巨大的跃迁。

And I think a lot of people now are assuming that there's a bunch of other trillion-dollar companies that will be formed in 3 to 5 years.

我觉得现在很多人默认:未来三到五年还会冒出一批万亿美元公司。

And you know, that's unprecedented in human history.

可这在人类历史上是没有先例的。

Usually it takes 20 years, right?

通常这得花 20 年。

SpaceX actually took since the early 2000s and Google took since the '90s.

SpaceX 其实是从 2000 年代初一路走到今天的,Google 是从 90 年代开始的。

And you know, these are usually 15-20 year arcs.

这些一般都是 15 到 20 年的曲线。

And then we had this weird 5-year inflection.

然后我们碰上了这么一次诡异的五年跃迁。

And so I feel like a lot of people now are looking at different areas that are very exciting, very promising areas, robotics, materials.

所以我感觉现在很多人盯着一堆非常令人兴奋、非常有前景的方向——机器人、材料。

And everything in everybody's mind is going to be a trillion-dollar company.

在所有人心里,什么都会长成一家万亿美元公司。

And maybe some of these will over the next decade, but it's unlikely that we'll see that many more in the next 3 to 5 years.

接下来十年里也许其中一些真能做到,但未来三到五年再冒出这么多,可能性不大。

I mean, there's one I can think of that could maybe get there, but not not multiple.

我能想到有一家也许够得着,但不会是好几家。

So, yeah.

所以,就是这样。

What's the one?

哪一家?

I'm not going to say.

我不说。

A lot.

多着呢。

Where am I going to put my money?

那我该把钱往哪儿投?

I don't know.

我不知道。

It's like throwing darts.

跟闭着眼睛扔飞镖差不多。

Yeah.

是啊。

It's why we have the dartboard here.

所以我们这儿才摆了个飞镖靶。

So, you think it's actually just like a very good special point in time vintage versus, you know, the ecosystem always gets bigger.

所以你觉得这其实就是一个特别好的时间窗口、一个特别的年份,而不是生态本身在一直变大?

Well, it's more like a punctuated equilibrium, right?

更像是‘间断平衡’。

If you look at like theories of evolution, one of them is punctuated equilibrium where you have like a Cambrian explosion and you have consolidation and things are kind of steady state for a while.

你看进化论里的几种说法,其中一种就是间断平衡:先来一次寒武纪大爆发,然后是整合,接着有一段时间处在稳态。

And you have an explosion and it goes up.

然后又来一次爆发,再往上走一层。

And so that's kind of like the history of technology, right?

技术史差不多就是这个样子。

If you think about it, we had a big social wave, but there isn't like a dozen new social companies all the time right now.

你想想,我们经历过一次很大的社交浪潮,但现在并没有源源不断地冒出十几家新的社交公司。

And we had a SaaS wave and then you know, they kind of settled down.

我们经历过 SaaS 浪潮,后来也就慢慢平稳下来了。

And so we just had a giant AI wave.

而我们刚刚经历了一次巨大的 AI 浪潮。

And there's still more to come, right?

而且后面还有。

Like one could argue that internet had like four or five periods to it, right?

你甚至可以说,互联网本身就分成四五个阶段。

I had the internet of the '90s.

有 90 年代的互联网。

It had social of the early to like 2010-2012ish kind of era.

有 2010 到 2012 年前后那一波社交。

You had SaaS, you had cloud, you had, you know, big security companies.

有 SaaS,有云,有一批大的安全公司。

So you had kind of like you had crypto as a wave, so you had like all these waves happening and sometimes they had two pieces, right?

还有加密货币这一波——这些浪潮一波接一波,有时候一波里还分成两段。

Bitcoin had two a couple different cycles.

比特币就走过好几个不同的周期。

And um you know, other technologies will have that.

别的技术也会这样。

AI undoubtedly there'll be some giant breakthrough in model capability and we'll see another step and suddenly all these startups again, right?

AI 毫无疑问会在模型能力上出现某次巨大突破,我们会看到又一级台阶,然后一堆创业公司又冒出来。

But we see um these moments in time where things go from zero to a lot and then those things become consolidators.

但我们看到的是:某些时刻,东西从零变成一大堆,然后这些东西变成整合者。

And then the question is what comes after that.

接下来的问题是:再往后是什么。

And so I think we've now seen at least some of the consolidators emerge.

所以我觉得现在至少已经有一部分整合者浮出水面了。

And the question is how many more giant companies are coming in in the next handful of years and that's different from saying what happens over the next 20 years.

问题是接下来几年还会冒出多少家巨型公司——这跟问‘未来 20 年会发生什么’是两回事。

Of course there's going to be tons of interesting stuff over 20 years.

20 年里当然会有海量有意思的东西。

Over two years, three years, there's still things that will grow a lot.

两年、三年这个尺度上,也还是有东西会长得很快。

You know, there's still a lot of 100 billion dollar companies to be built, but multi-trillion dollar companies are kind of hard to get to.

千亿美元级别的公司还有很多可以做,但几万亿美元的公司就很难够得着了。

I want to talk to the investors you were talking to because I feel like I run more into um a failure of imagination of how much bigger or better something can be than the closest proxy market from a previous era.

我倒想跟你说的那些投资人聊聊,因为我碰到更多的是另一种毛病:想象力不足——想不出一个东西能比上个时代最接近的对标市场大多少、好多少。

Um and like I I think like being able to rethink market size is just still like a key underpriced investor skill right now at any stage, right?

我觉得‘重新估市场规模’这件事,到今天仍然是被严重低估的投资人核心能力,在任何阶段都是。

If you think about

你想想——

Sure.

当然。

take some of the like application companies that, you know, we have in common or that other people invested in.

拿我们共同投的、或者别人投的那些应用层公司来说。

I think there are a lot of investors who have intellectually recognized this idea of AI companies delivering services value.

我觉得有很多投资人在智识层面已经认可‘AI 公司在交付服务价值’这个说法。

They don't act like they believe it.

但他们的行为不像是真信。

They look at everything a little bit more linearly, right?

他们看什么都还是偏线性的。

So think of um if you're looking at Harvey or Abridge or something, then they think about a per seat or per yeah, per like lawyer or per doctor TAM.

比如你看 Harvey 或者 Abridge,他们脑子里算的还是按席位、或者按每个律师、每个医生去估 TAM。

And they're not actually, you know, asking the question of like what does the company look like if they can charge for outcomes.

他们并没有真的去问:如果这家公司可以按结果收费,它会长成什么样?

Um and actually thinking about like what's happening in the the coding domain, which is consumption and value, you know, 100x from here.

也没有真的去想编程这个领域正在发生什么——那是按用量算的,而且价值还能从这儿再涨 100 倍。

Coding I think is like a much, much bigger market than anyone thought.

我觉得编程是一个比所有人想象中大得多、大得多的市场。

And, you know, I think both of us were saying that a year or two ago.

而且我们俩一两年前就在这么说了。

But now the evidence is out there.

但现在证据已经摆在那儿了。

You don't have to be a genius to like take that to domains.

你不用是天才,也能把这个结论推到别的领域去。

The evidence is out there, but it's also what is a what is a trillion-dollar market and what is a hundred-billion-dollar market?

证据是摆在那儿了,但还有一个问题:什么叫万亿美元市场,什么叫千亿美元市场?

Both of those are big numbers, right?

这两个都是大数字。

I actually wrote a blog post like in 2010 or something talking about how hard it was to get to 10 billion in market cap, right?

我 2010 年前后还写过一篇博客,讲要做到 100 亿美元市值有多难。

Which is now like a seed round for some of these neolabs.

而 100 亿现在差不多是某些 neolabs 的种子轮估值了。

You know, don't get me wrong.

别误会我的意思。

I think that the reality is that there's a lot of these things that could be a hundred, but I don't think there's that many that could be a trillion.

我觉得现实是:能做到千亿的东西有不少,但能做到万亿的没那么多。

Those are just different orders of magnitude.

这是完全不同的数量级。

And so then the question is what are these things that could actually be a trillion-dollar market?

那么问题就成了:到底哪些东西真的能撑起一个万亿美元的市场?

Cuz you just think of the revenue basis that's needed for that, right?

因为你只要想想那需要多大的收入基数就知道了。

You need a hundred billion of revenue or 50 to 100 billion pretty easily.

你得有 1000 亿美元收入,或者说轻松做到 500 亿到 1000 亿。

And so then the question is with good margin, right?

而且还得有不错的利润率。

So then the question is where are the 50 to 100 billion-dollar revenue streams for single companies?

所以问题变成:单家公司 500 亿到 1000 亿美元的收入流,到底在哪儿?

That's a different question than is the TAM really, really big, right?

这跟问‘TAM 是不是特别特别大’完全是两个问题。

That's huge TAM, right?

TAM 大是一回事。

That's a very small number of markets in the world.

而全世界这样的市场少之又少。

Um there's there's a lot of them, you know, there's like, you know, a dozen plus companies that that are there-ish.

当然这样的公司也有一批,差不多十几家已经在那个位置附近了。

But um how many more will there be in the next five years?

但未来五年还会再多出几家?

That's my question.

这才是我的问题。

It's not what in the 20 years.

不是 20 年里会怎样。

It's what in the next five years will be able to get to 50 to 100 billion of revenue.

而是未来五年里,谁能做到 500 亿到 1000 亿美元的收入。

And that changes how you think about this, right?

这会改变你思考这件事的方式。

There's tons that can get to five or 10 billion of revenue and they'll be a hundred-billion-dollar company.

能做到 50 亿、100 亿美元收入的多得是,它们会是千亿美元市值的公司。

I think I'm looking at more uh both a little further out and then I'd say like I don't know that there are that many markets that are going to get to a hundred billion of revenue in the next couple of years that aren't like in France, right?

我看的更远一点,而且我不觉得未来几年会有那么多市场能做到 1000 亿美元收入——除非是法国那种量级的存在。

Tell me what else you what else you think in that timeline.

你告诉我,在那个时间尺度上你还看到什么。

Perhaps some supply chain like energy type technologies.

也许是供应链、能源这类技术。

Maybe, yeah.

也许吧。

Yeah, there's there's like a list.

确实可以列个单子。

You can make a list of like five or six areas that seem promising.

你能列出五六个看起来有戏的方向。

And then part of it too is you actually, if it's a physically it's a physical goods company, energy, robotics, etc., do you actually have the footprint to get there that fast?

但还有一层:如果它是个实体商品公司——能源、机器人这些——你真的有那个产能规模,能那么快铺到那个位置吗?

Again, I'm not I'm not doubting the size of some of these markets.

我再说一遍,我不是在质疑这些市场的规模。

I'm doubting the speed at which you can get there.

我质疑的是你到达那儿的速度。

Yeah, 100% and that's the issue.

对,百分之百,这才是症结。

And people um are, at least in my experience, uh collectively, at least investing against the fact that they believe the speed is there, which is different from the market size.

至少在我看来,大家整体上是在按‘速度就在那儿’这个信念下注,而这跟市场规模是两码事。

People are conflating the two things right now, in my opinion.

在我看来,现在大家把这两件事混为一谈了。

The other phenomena that I think is happening is almost the opposite of it, which is I see some really, really excellent founders going after niche markets because they're now scared of the neolabs.

另一个我觉得正在发生的现象几乎是它的反面:我看到一些非常非常优秀的创始人跑去做小众市场,因为他们现在害怕这些 neolabs。

And I think there's much less head-to-head competition.

结果是正面硬碰硬的竞争少了很多。

If you look at the markets that Harvey or um OpenEvidence or Decagon or any of these folks at Sierra entered you know, four, five years ago, three, four years ago even.

你看 Harvey、OpenEvidence、Decagon,或者 Sierra 那帮人当年进入的那些市场——那是四五年前,甚至三四年前。

Uh Cognition two years ago, it was big, big markets that could be in the road maps of these labs.

Cognition 是两年前——那些都是大市场,大到可能就写在这些 labs 的路线图里。

But I feel like the the the two things are happening at the same time.

但我觉得这两件事是同时在发生的。

One is for the mid-to-late stage technology markets, people are continuing to invest as if there's velocity to get to a trillion for many companies, where I don't think there's a velocity.

一是在中后期的技术市场上,大家继续按‘很多公司有速度冲到万亿’去投,而我不认为那个速度存在。

Again, I think though some of them get to 20, some of them get to 100, some of course will go to zero.

再说一次,我认为其中一些能到 200 亿,一些能到 1000 亿,当然也有一些会归零。

And then there's a separate thread of all the new stuff that's coming.

另一条线是所有正在冒出来的新东西。

How aggressive and ambitious are the founders relative to what the labs are doing?

相对于这些 labs 在做的事,创始人有多进取、多有野心?

And I think that's why you're seeing a flight to hardware companies.

我觉得这就是为什么你会看到一股往硬件公司跑的风潮。

Oh, the labs will never do this hardware thing and so we'll do that.

‘哦,labs 永远不会做这个硬件的事,那我们就做这个。’

And American dynamism and um niche applications of AI and something that should be provided by an inference cloud and etc., etc., right?

还有 American Dynamism、AI 的小众应用、本该由推理云提供的东西,等等等等。

So, there's a lot of these types of companies that I think are going to be potentially a bit more derivative and they're going to have these people doing huge, amazing things simultaneously, right?

所以有一大批这类公司,我觉得可能会更偏衍生性;与此同时,也有人在同时做着巨大而了不起的事。

It's not every startup.

不是每家创业公司都这样。

But there's more and more, at least in my perception, of people doing smaller niche things out of fear of the labs.

但至少在我的感知里,越来越多的人因为害怕 labs 而去做更小、更窄的事。

And that's also, I think, a negative.

我觉得这也是个坏信号。

And you feel like they're being too meek, like they should just take on the head-on competition because you can create a much better experience and go just compete on the product, on the distribution, any of it.

你是觉得他们太怂了?觉得他们就该正面硬刚,因为你完全可以做出体验好得多的产品,在产品上、在渠道上、在任何一点上跟它们竞争。

I think so, yeah.

我是这么觉得的。

For certain markets, of course, there's some markets where the labs will just eat it naturally, but there's a bunch of markets where they won't.

当然,有些市场 labs 自然而然就会吃掉,但也有一堆市场它们吃不掉。

But I think people are staying away from both.

而我觉得现在大家两种都在躲。

Well, we have companies in the portfolio that are going against like pretty central premises.

但我们组合里就有一些公司,是冲着相当核心的命题去的。

So, I I don't think all the founders are being too meek.

所以我不觉得所有创始人都太怂。

No, I don't think it's all.

不,我不是说所有人。

I think there's more.

我是说这样的人变多了。

My point is that it's more a trendline, and it's the newest stuff.

我的意思是这更像一条趋势线,而且发生在最新的那批东西上。

I'm not saying a thing that's a year old or 2 years old or, you know, I feel like it's a trendline that's shifting.

我说的不是一年前、两年前起的那些公司——我感觉是这条趋势线在偏移。

And again, I'm it's not all of them.

再强调一次,不是全部。

It's just a It's just enough of a subset, and it's not just a subset, it's a subset of the good founders.

只是这个子集已经够大了,而且这不只是一个普通子集,是好创始人里的一个子集。

I'm not concerned about the median founder, I'm concerned about the best founders.

我不关心中位数创始人,我关心最好的那批创始人。

What are they doing?

他们在做什么?

I am more often disappointed right now that founders are being like less ambitious than they could be.

我现在更常感到失望的是,创始人的野心比他们本可以有的要小。

So, maybe that's the trendline you're talking about.

所以也许那就是你说的那条趋势线。

We were talking about when companies when founders should sell their companies.

我们之前聊到过,创始人到底什么时候该把公司卖掉。

What is your thinking on it at this point in time or your framework for it?

在当下这个时点,你对这件事怎么想?或者说你的判断框架是什么?

There's a handful of companies that should never, ever sell, at least anytime in the near term.

有少数几家公司永远不该卖,至少近期内不该。

If you're Anthropic, you shouldn't sell.

如果你是 Anthropic,就不该卖。

If you're OpenAI, you shouldn't sell.

如果你是 OpenAI,就不该卖。

If you, you know, there's a handful of these things that should never sell.

总之有那么少数几家是永远不该卖的。

Um most companies in any given era should at least consider it, and there's usually a time-maximizing window where your best outcome is a sale within that window.

但任何时代的绝大多数公司,至少都该认真考虑一下——而且通常存在一个价值最大化的时间窗口,在那个窗口里卖掉就是你能拿到的最好结果。

It's like a 12- to 18-month period, usually where the company's worth the most it'll ever be worth.

那大概是 12 到 18 个月的一段时间,公司在那段时间里值的钱是它这辈子的最高点。

And um I I think we saw one major exit where that was probably the case uh reasonably recently.

我觉得最近就有一笔重量级退出,大概率正好落在那个窗口里。

I think there's other companies that, you know, um should really actively think about it, and from a hygiene perspective, maybe what companies should do I think Ben Horowitz wrote about this once, you know, basically do a pre-planned once-a-year board meeting where the discussion topic is, in a non-emotional way, should we consider exiting this next 6-month period?

还有一些公司真该主动去想这件事;从治理卫生的角度说,我记得 Ben Horowitz 写过一次:基本上就是预先安排一次一年一度的董事会,议题就是——不带情绪地——我们要不要考虑在接下来 6 个月里退出?

And it's pre-scheduled, so it's not the founders pushing for it, it's not the investors pushing it.

而且是提前排好的,所以既不是创始人在推,也不是投资人在推。

It's just a rational conversation.

它就是一次理性的讨论。

And the answer to the conversation may be no, we should keep going.

讨论的结论也可能是:不,我们应该继续干。

We still think we have XYZ ahead of us.

我们觉得前面还有这个那个可以做。

It's amazing.

那太好了。

But I think it's very useful for people to have that sort of conversation because I feel like in the cycle every year of AI time is like 3 to 4 years of normal cycle time.

但我觉得让大家有这样一次对话非常有用,因为我感觉在这个周期里,AI 时间的一年相当于正常周期的三到四年。

And so 3 years is like a decade, right?

所以三年就相当于十年。

Like if you think about what existed in AI 3 years ago from a model capability perspective, from a vertical app perspective, from AI roll-ups, from you name it, any any of the stuff like infrastructure, whatever, radically different world 3 years ago.

你想想三年前 AI 里有什么——从模型能力、从垂直应用、从 AI roll-up,到基础设施,随便哪一块——三年前完全是另一个世界。

And so we're on an accelerated timeline right now where everything is moving faster.

所以我们现在处在一条被加速的时间线上,一切都在更快地推进。

And that means that you should double-check your thinking more frequently because the underlying fact set is changing faster than it ever has.

这意味着你应该更频繁地复核自己的判断,因为底层事实变化的速度前所未有。

I don't know, what do you think?

我不知道,你怎么看?

What's your What's your approach to exits or not exits?

你对卖或不卖是什么思路?

I agree with you that um there uh there are a set of companies that should never sell unless they cannot finance their future, right?

我同意你说的:有一批公司永远不该卖,除非它们没法为自己的未来融到钱。

Um if I think about maybe one principle that is like new for this point in time is um I might ask at that board meeting or at that meeting once a quarter or once a year, whatever you think is the right pacing today.

如果说有哪条原则是这个时点上新出现的,那就是我在那次董事会上、或者那次每季度、每年一次的会上——节奏你自己定——会多问一个问题。

And it's more often than it was a a few years ago.

而且这个节奏比几年前要密。

Yeah, it's every 6 months.

对,现在是每 6 个月一次。

Okay, so every 6 months, great.

行,那就每 6 个月一次,很好。

Uh are you capturing value as costs fall and capabilities increase?

问题是:在成本下降、能力上升的过程中,你有没有把价值抓在手里?

Because if you're the on the wrong side of this circular change and you can't get to the other side of it, you should in fact like sell.

因为如果你站在这轮循环变化的错误一侧,又跨不到另一侧去,那你其实就该卖。

If you don't have good ideas about how to be on the right side of history.

前提是你对‘怎么站到历史正确的一边’没有好想法。

Um so I I I I think that's the question people should ask them ask themselves.

所以我觉得这是大家应该问自己的问题。

And then if you think about our our friends um at Cursor uh is the way you want to compete capital compute access and is it perhaps a maximally valuable point in time?

然后你想想我们那些朋友——比如 Cursor——你想用来竞争的手段是资本、是算力获取吗?现在是不是恰好是价值最高的时点?

That's an interesting question.

这是个很有意思的问题。

Um But I think that like you know, more broadly it's a it's a I feel like it's a very personal and very interesting risk management question.

但我觉得更广义地说,这是一个非常个人化、也非常有意思的风险管理问题。

I do think people should ask themselves, right?

我确实认为大家应该问问自己。

Like the idea that there is pride around like never considering this is nonsense.

那种‘以从不考虑卖公司为荣’的想法,是胡扯。

The situational awareness situation is a good reminder that everyone has to stay alive to profit as well.

Situational Awareness 那件事是个很好的提醒:所有人都得先活着,才谈得上赚钱。

Hedge funds are different than companies, they have to survive to compound, but I I think even just the firmness of like you need to match your financing structure to your thesis horizon and then be able to like continually finance the company to the promised land of whatever you're trying to do.

对冲基金和公司不一样,它们得先活下来才能复利;但我觉得哪怕只是这条硬道理——你得让融资结构匹配你的论点周期,然后能持续把公司融到你想抵达的那片应许之地。

Yeah, I think the financing part though is going to be there because basically what's happening because of this rapid rise of three trillion dollar plus companies in a short time frame, an enormous amount of venture capital is starting to get returned.

不过我觉得融资这一环是不成问题的,因为短时间里冒出三家万亿美元以上的公司,意味着大量风险资本开始被返还。

And that means people are raising bigger and bigger funds and they need to put it somewhere and they're going to put it against trillion dollar companies of the future.

这意味着大家在募越来越大的基金,而这些钱总得放到什么地方去,于是就会押在未来的万亿美元公司上。

And so I do think we're going to see a ongoing rise in valuations most likely over the next year or two, much more than we've seen to date.

所以我确实认为未来一两年估值大概率会持续上涨,而且比我们迄今看到的涨得更猛。

And obviously there'll be some great things in there and there'll be a bunch of stuff that doesn't deserve it, but I actually think financing is going to get easier not harder.

里面显然会有一些非常好的东西,也会有一堆配不上那个价的东西,但我其实认为融资会变得更容易,而不是更难。

And so I view it less as financing and more um what do you think is the true likely expected outcome of your company?

所以我不太把它当成融资问题,而更看成:你觉得你公司真正的、大概率的预期结果是什么?

Not what investors are telling you, not what the press is telling you, not what Twitter is telling you.

不是投资人告诉你的,不是媒体告诉你的,也不是 Twitter 告诉你的。

Like just sit down and run the math and then remember that at some point you'll probably trade trade it like 10x or something.

就是坐下来把账算一遍,然后记住:到某个时点你大概会按 10 倍左右的倍数交易。

You know?

你懂吧?

Maybe 15x.

也许 15 倍。

And so then the question is what is your thing going to be worth?

那么问题就是:你这东西最后能值多少钱?

Right?

对吧?

And remember eventually things slow down in terms of compounding, too.

还要记住,复利最终也会慢下来。

And you can decide where that slow down happens, but you kind of do that math, you do future dilution, you look at your potential outcome, you look at years of work it'll take to get there.

你可以自己判断慢下来的点在哪儿,但你得把这笔账算了:算未来的稀释,看你可能的结果,看要花多少年才能走到那儿。

And then you can come to a conclusion because there's two types of opportunity cost that or risk management.

然后你就能得出结论了,因为机会成本、或者说风险管理有两种。

There's risk management against the value of the thing you're doing.

一种是针对你手上这件事本身价值的风险管理。

But the biggest opportunity cost is your time.

但最大的机会成本是你的时间。

Your most productive years of your life are on the line right now.

你这辈子最有产出的那些年,现在正押在牌桌上。

And you could either walk away with a good amount of money, go to the next giant thing, now having done it before, they're going to work with you again, etc., etc., or you can roll the dice, and you can decide to roll the dice.

你可以拿着一笔不错的钱走人,去做下一件大事——而且这次是做过一轮的人了,大家还会愿意跟你干;你也可以选择掷骰子搏一把。

That may be the right answer.

搏一把也可能是对的答案。

And again, for some companies, absolutely you should do that.

再说一次,对某些公司来说,你绝对应该那么干。

But for others, it may be, "Hey, actually now is maybe the time to go."

但对另一些公司,答案可能是:‘嘿,现在也许就是该走的时候。’

Secondary is an intermediate option, which I actually don't think is always that great, because it solves for some short-term needs, but it doesn't actually um create a solution.

老股转让是个折中选项,但我其实不觉得它总是那么好,因为它只解决了一些短期需求,并没有真正给出一个解法。

And you see a lot of people from 2020, 2021 still running companies 5 years later that aren't working.

你能看到很多 2020、2021 年那批人,五年后还在经营一家跑不起来的公司。

And think of that or 6-year period where they've been locked up when all the AI change happened.

想想那五六年——AI 的巨变全部发生的那段时间,他们被锁死在里面。

What is the cost of that to a great founder?

对一个优秀的创始人来说,这个代价是多少?

So, I think I think there's that kind of cost that people don't really talk about as much, which I think is the real cost.

所以我觉得,这种代价是大家谈得不够多的,而我认为那才是真正的代价。

That's your lifetime cost, right?

那是你的人生成本。

And you only live once, and it's a short life, and so do you want to eventually be working on something that's going to continue to struggle, that's overcapitalized, that has runway for the next 10 years?

人只活一次,而且很短;所以你愿不愿意最后一直在做一件持续挣扎、融资过度、但账上还有十年跑道的事?

Or not.

还是不愿意。

And that's where you end up.

而那就是你最后的处境。

That's what happened with the 2020, 2021 cohort.

2020、2021 那一批就是这么回事。

There's tons of people running these companies that aren't working still.

有一大堆人还在经营那些跑不起来的公司。

And we forgot about them, cuz we're talking about AI all the time.

而我们把他们忘了,因为我们一天到晚在聊 AI。

That is a huge waste.

那是巨大的浪费。

Um my uh I think my point was really that even if there are lots of dollars still rotating into venture or being produced by these huge outcomes over, you know, now and over the next year, too, private markets don't have to be rational or right for long periods of time, right?

我想说的其实是:哪怕现在和未来一年里还有大量资金轮转进风投、或者被这些巨大的退出创造出来,私募市场也完全可以在很长一段时间里既不理性也不正确。

And so, being smart about your ability to finance a company is the equivalent of avoiding margin calls, right?

所以,想清楚自己有没有能力持续融资,等价于避免被追加保证金爆掉。

And I I I you know, some founders who are working on something that requires a like a technical point of view, for example, or even a a structural point of view about how the market resolves, can get very frustrated, because investors will believe something that they think is wrong or stupid for a long time.

有些创始人做的事需要一个技术判断、甚至是对市场最终如何收敛的结构性判断,他们会非常沮丧,因为投资人可能长期相信一些他们认为错误或者愚蠢的东西。

And it's just the job of the founders to go navigate that narrative or that set of beliefs.

而创始人的工作就是去驾驭那套叙事、那套信念。

And if if they think it's untenable, or if they think they're like down some um wasteful path with their time as you described, then they should sell the company.

如果他们觉得这局面撑不住,或者像你说的觉得自己走在一条浪费时间的路上,那就该把公司卖掉。

But it could be worse.

但情况还可能更糟。

I mean, founders, they have the concentration risk, but they could be hedge fund managers facing retail um irrational acts in the market and redemptions next quarter.

创始人至少只是承担集中度风险;他们还可能是对冲基金经理,要面对市场上散户的非理性行为和下个季度的赎回。

So, it's just different um uh different environment.

所以只是环境不同而已。

But I don't think it's as simple as like financing is now free.

但我不觉得事情简单到‘融资现在免费了’。

I do think it is going to skew, as you said, toward scale of opportunity, naturally.

我确实认为,正如你说的,钱会自然地向机会的量级倾斜。

Or perceived scale.

或者说,向被感知到的量级。

Perceived scale, yeah.

被感知到的量级,对。

Perceived scale is important not I think all um So, the other thing a lot of people out here are working on or talking about is um if you talk to people at the labs, there's this enormous manic energy right now.

被感知到的量级很重要。所以,这边很多人在做、在聊的另一件事是:如果你去跟 labs 里的人聊,现在有一种巨大的躁狂能量。

We're 6 months-ish or towards the end of the year to be completely done with code, like it's a solved problem.

他们说大概再有 6 个月、或者到今年年底,代码这件事就彻底解决了,变成一个已解问题。

And then we'll probably have some form of, you know, light RSI by end of next year.

然后到明年年底,我们大概会有某种轻量版的 RSI。

And at that point you'll have models training big chunks of the models themselves.

到那时候,模型会自己训练模型的很大一部分。

I think it it's probably more pre-training initially, maybe it could impact pre-training over time more quickly as well.

我觉得初期可能更多是在预训练那一侧,也可能随着时间推移更快地波及预训练。

And because of that, many people believe, "Hey, you know, if I have a year, year and a half left of productive work in my career, I should be working 16 hours a day because every week is a you know, 2% of all the time I have left to be productive before I get displaced by AI."

正因为这样,很多人相信:‘嘿,如果我职业生涯里只剩一年、一年半的有效产出时间,那我就该一天工作 16 小时,因为在被 AI 取代之前,每一周都占我剩余有效时间的 2%。’

Um what do you think about that?

你怎么看这个?

Like do I believe it or what happens if it's true?

你是问我信不信,还是问如果它是真的会怎样?

Do you believe it?

你信不信?

I think the idea that the models can

我觉得,‘模型能够

improve their own training if the leading scientists working on this believe it, and it's an extension of what we are already seeing in code and math.

改进自己的训练’这个想法——如果做这件事的顶尖科学家都相信,而且它是我们已经在代码和数学上看到的东西的延伸。

Of course, you should believe it.

那你当然应该信。

Right.

对。

Yeah, but on that timeline.

对,但关键是那个时间表。

The data I have is that a number of very smart and even very self-aware research scientists have uh felt that, you know, there was some knee in the curve on recursive self-improvement or ASI 18 months away every 18 months for the last 5 years.

我手上的数据是:一批非常聪明、甚至非常有自省能力的研究科学家,过去五年里每隔 18 个月就觉得‘递归自我改进、或者说 ASI 的曲线拐点还有 18 个月’。

So, how good of it is a predictor?

所以这个预测有多准?

It's not clear.

并不清楚。

I think the like the extension from code to training code to data pipeline work um is way easier to believe.

我觉得从写代码延伸到训练代码、再延伸到数据管线的工作,这条线要好信得多。

The question of like how you are going to go gather that data for less verifiable, more complex domains.

问题在于:在那些更难验证、更复杂的领域里,你要怎么去把数据搞到手。

Or do you run into the actual constraints on the like physical compute accessibility side?

或者说,你会不会撞上物理算力可得性这一侧的真实约束?

Like I think that's probably more of a limiter than um this being algorithmically possible.

我觉得那大概比‘算法上能不能做到’更像是瓶颈。

Yeah, I mean the physical compute basically um reinforces an oligopoly market because what it does is it creates a ceiling on the rate of progress any single lab can get if effectively assume the compute is roughly prorated across the ecosystem to the big labs.

对,物理算力基本上强化了一个寡头市场,因为它给任何单一 lab 的进步速度设了一个天花板——如果我们大致假设算力是按比例分摊给生态里这些大 labs 的话。

And so in the absence of a lack of compute constraints, you are almost having an enforced oligopoly market up to a point.

所以在算力约束存在的情况下,你几乎等于有了一个被强制的寡头市场,至少在某个程度上是。

Or at least you force closer competition between the players than would exist otherwise, which I think is an interesting odd effect of this moment in time.

或者至少,它逼着这些玩家之间的竞争比原本更贴身——我觉得这是这个时点上一个很有意思的怪效应。

And the question is when does that lift and what does that look like?

问题是这个约束什么时候解除,解除之后又是什么样子。

Say yeah, it's kind of it's a very exciting time.

对,这确实是个非常令人兴奋的时候。

I mean, I was wondering about second-order effects of that belief of it's 18 months away because that does suggest there could be a burnout cycle in 18 months.

我在想的是‘还有 18 个月’这个信念的二阶效应,因为它确实意味着 18 个月后可能会有一轮集体倦怠。

Like I know some people that one of them is at your labs who for a while um brought up with me, should they get married?

比如我认识一些人,其中一个在你们那边的 labs,有阵子来问我:他们该不该结婚?

Like two people want to get married.

就是两个人想结婚。

Should they get married because I don't know what happens in 18 months to the world.

他们该不该结,因为我不知道 18 个月后世界会变成什么样。

It's like I you should get married.

我的反应是:你该结婚。

You should go ahead.

去结吧。

It'll be okay.

会没事的。

And so, I do think we're living through this very manic, very exciting, very intense work period.

所以我确实觉得,我们正活在一段非常躁狂、非常兴奋、非常紧绷的工作期里。

Um so yeah, it's really fun stuff.

所以是啊,挺有意思的。

I think it's kind of tragic, man.

我倒觉得这有点悲哀,老兄。

Really?

真的?

Why?

为什么?

I think the reactions of some really extraordinary research friends to it is it feels a little bit tragic.

我看到一些非常出色的研究员朋友对这件事的反应,感觉是有点悲哀的。

I think it like I I feel like it's psychologically most similar to if people think they're going to die.

我觉得从心理上看,这跟人以为自己快死了最像。

Right?

对吧?

Like how would you spend the last 2 years of your life?

你会怎么过生命最后的两年?

Would you spend it the way you are today?

你会像今天这样过吗?

Or would you spend it in a very different way?

还是会过得完全不一样?

Is it like a a not unrelated philosophical question?

这算不算一个并非无关的哲学问题?

And so you do have people who are like, "Ah, like my contribution is a bit irrelevant given ASI in the next 18 months.

于是你确实会看到有人说:‘啊,反正未来 18 个月就ASI了,我这点贡献有点无所谓了。

So, should I get married?

那我该结婚吗?

Should I bother to work?

我还值得工作吗?

Should I travel?

我该去旅行吗?

Should I only work?"

还是应该只工作?’

Um and you know, I I I just actually think it's a much more stable and satisfying state if people act as if they have.

而我其实觉得,如果人按‘自己还有时间’来行动,那会是一个稳定得多、也令人满足得多的状态。

But maybe you think that's blind.

不过也许你觉得那是自欺欺人。

Yeah, I just think there's a lot of um second order effects that are happening or going to happen.

对,我只是觉得有很多二阶效应正在发生、或者将要发生。

And part of them are driven by this belief system and potential burnout over time.

其中一部分是被这套信念体系驱动的,还有随时间累积的潜在倦怠。

Uh part of it is going to be um you know, one thing that I've noticed that is happening at some of the labs is that you know, as compute becomes really the scarce resource, it turns out that there's a few dozen researchers that drive a lot of like 80% of the results at any given place.

另一部分是——我注意到某些 labs 里正在发生的一件事:当算力真正成为稀缺资源之后,你会发现任何一个地方大约 80% 的成果,是由几十个研究员做出来的。

Which is a really interesting human power law, right?

这是一条非常有意思的‘人的幂律’。

If you actually look at it, in any field there's at most a few dozen people who drive the field.

你真去看的话,任何领域里,推动这个领域的人最多也就几十个。

You look at breast cancer research, you look at certain subfields of mathematics, you look at subfields of physics, you look at the entrepreneurial ecosystem and founders.

你看乳腺癌研究,看数学的某些子领域,看物理的子领域,看创业生态和创始人。

Like there's a handful of people, dozens of people who drive most progress.

都是少数几个人、几十个人在推动绝大部分进展。

Um and that also happens in AI research.

AI 研究里也一样。

And you know, increasingly compute is differentially provided to those people, right?

而且算力正越来越倾斜地配给这些人。

And so I know some labs have slowed down on their hiring of researchers unless they're above a very, very high bar because the cost isn't the researchers, the compute associated with the person.

所以我知道有些 labs 已经放慢了研究员招聘,除非候选人高过一条非常非常高的线——因为成本不在研究员本身,而在跟这个人绑定的算力。

That's real where the real bottleneck is.

真正的瓶颈在那儿。

I think there's this broader concept of like return on invested tokens, like an ROIT kind of metric, which is if you have a certain token budget, who do you give it to and why?

我觉得这里有一个更大的概念,叫‘token 投入回报率’,类似一个 ROIT 指标:如果你有一定的 token 预算,你把它给谁、为什么给?

This is kind of like engineering back in the day, right?

这有点像当年的工程团队。

The internal tools teams at companies were always starved for resources cuz many, at least tech companies, would rather use the same engineers to build product than to build internal tools that would make other functions more productive.

公司里的内部工具团队永远资源紧张,因为很多公司——至少是科技公司——宁愿把同一批工程师用来做产品,而不是做那些能提升其他职能生产力的内部工具。

That's why I think the death of SaaS is a little bit overstated.

所以我觉得‘SaaS 已死’这个说法有点夸张了。

Because why would you use tokens on a bunch of SaaS stuff that you're not actually paying that much for per year relative to the outcome of those same tokens being invested against a core product or against some massive margin lift or some other thing, right?

因为你为什么要把 token 花在一堆你一年其实也没付多少钱的 SaaS 上?同样这些 token 投到核心产品、或者投到某个巨大的利润率提升、或者别的什么上面,产出会大得多。

And so, I think increasingly we've shifted from a world where people said, "Hey, everybody use AI and do whatever you want." to hey, we have to like measure spend and move more things to open source and then I think the next wave is, what are the projects and people that should actually get outsized pieces of a token budget.

所以我觉得我们正越来越多地从‘嘿,所有人都用 AI,想干嘛干嘛’那个世界,转向‘我们得开始计量开销、把更多东西挪到开源上’;而下一波我觉得是:到底哪些项目、哪些人应该拿到超额的 token 预算。

And what is that return on investment?

这笔投入的回报又是多少?

Is this sort of next shift that's coming.

这是接下来要发生的转变。

It'll take some time though.

不过还需要一些时间。

I mean, many people are still at like, "Hey, everybody try AI or whatever."

我是说,很多人现在还停在‘嘿,大家都去试试 AI 吧’这个阶段。

You know, at big enterprises.

尤其是大型企业里。

What do you think is the appropriate compute like token budget for a business or a human being 3 to 5 years from now?

你觉得三到五年后,一家企业、或者一个人的算力、token 预算应该是多少才合适?

Should I look at it like rent?

我该像看房租一样看它吗?

No, I mean, it depends on what the budget is for what.

不,我是说,这取决于这笔预算是用来干嘛的。

I mean, people forget too, Minecraft was like, what was it?

而且大家也忘了,Minecraft 当年是多少人来着?

Five people, 10 people when it was bought for billions of dollars by Microsoft.

被微软以几十亿美元买下的时候,也就五个人、十个人。

People keep talking about someday there'll be like a multi-billion dollar single person company.

大家总在说,总有一天会出现一家几十亿美元估值的单人公司。

That was basically Minecraft.

Minecraft 基本上就是那个。

Roughly.

差不多吧。

It already happened like 15 years ago or whenever that was.

这事十五年前、或者不管多久前,就已经发生过了。

So, there are always people who can take outsized advantages of technology.

所以永远都有人能从技术里拿到超额的杠杆。

And AI has accelerated that radically.

而 AI 把这件事极大地加速了。

And so, at some point it's like, why give tokens to people who can't do that on a relative basis unless you just run out of those people.

所以到某个时点问题就变成:相对而言,为什么要把 token 给那些做不到这一点的人?除非你把能做到的人用完了。

And you may run out of them.

而你确实可能会把他们用完。

This is back to like, will all the engineers get laid off?

这又回到那个问题:工程师是不是都会被裁掉?

Probably not anytime soon.

短期内大概不会。

But, you could argue that at some companies, even before AI, there was a bunch of engineers that weren't that productive.

但你也可以说,在某些公司,哪怕在 AI 之前,就有一批工程师产出并不高。

They could be like, "Oh."

他们可能就是‘哦’那种状态。

Especially some of the big tech companies.

尤其是某些大型科技公司。

And I think a lot of those folks will be very coveted by GE or PG&E or Hershey's.

而我觉得这些人里有很多会被 GE、PG&E 或者好时这样的公司抢着要。

Or so, even if there is some displacement of engineers at some point in the future, I don't know when that is or if it happens.

所以哪怕未来某个时点真的出现工程师被替代的情况——我不知道那是什么时候、会不会发生。

But, if it does happen, there's lots and lots of homes for them.

但如果真发生了,他们能去的地方多得是。

Because there's tons of enterprises that never had the capability set or ability to recruit these people, and they want the capabilities they bring.

因为有大量企业从来没有过这种能力储备、也从来招不到这些人,而它们想要这些人带来的能力。

Even if they're mediocre in the context of a Google or Meta or whatever, they may be exceptional in the context of a of a certain subset of old-school enterprises.

哪怕在 Google 或 Meta 的语境里他们只是中等水平,在某一类老派企业的语境里,他们可能就是顶尖的。

And so, you know, I do think there's going to be this permeation through um the enterprise landscape of engineering talent in an unexpected way.

所以我确实认为,工程人才会以一种出人意料的方式渗透进整个企业版图。

This is probably many years away.

这大概还要很多年。

I'm just saying I think that's probably a likely outcome.

我只是说,我觉得这大概率会是结局之一。

I think relatedly, if, you know, if you are researcher #800 and you have not been allocated an outsized number of tokens to work with at one of the major labs.

我觉得与此相关的是:如果你是第 800 号研究员,在某个主要 lab 里没有被分到超额的 token。

Yeah.

嗯。

I think the opportunity to go spend your energy on, you know, something where you have like comparative advantage and understanding and should benefit from all of this.

我觉得,去把你的精力花在一个你有比较优势和理解、并且理应从这一切里受益的地方,是有机会的。

The supply chain bottlenecks, um domains that should accelerate like bio, diffusion into other valuable fields.

比如供应链瓶颈,比如生物这种本该被加速的领域,比如向其他有价值领域的扩散。

Like, to me that uh that seems a lot more exciting than being concerned about the downfall of mathematics and going on, you know, vacation until the world ends.

对我来说,那比担心数学的没落、然后休假等世界末日,要令人兴奋得多。

Oh, yeah.

哦,是啊。

Lots of places to go do stuff.

可以去做事的地方多得是。

And I I do think that's where a subset of the research community will end up over time, right?

而且我确实认为,随着时间推移,研究圈里的一部分人最后就会去那些地方。

And that that's an interesting question is how many researchers do you need if you're a top AI lab?

这里有个有意思的问题:如果你是一家顶级 AI lab,你到底需要多少研究员?

And what how does that number relate to the number that you have now?

这个数字跟你现在的人数是什么关系?

And is it you have the right number?

是你现在人数刚好?

Is it you need five times as many people?

还是你需要五倍的人?

Is it need you need half as many, but they only need to be above a certain bar cuz it's compute constrained?

还是因为受算力约束,你只需要一半的人,但他们都得高过某条线?

Um and you want to map it against the best ideas, and best ideas come from a subset of people on average.

而且你想让人力对得上最好的想法,而最好的想法平均而言来自其中一小部分人。

Not always, but on average.

不总是这样,但平均而言是。

And so, it's a really interesting question of like, what is How does all this stuff fall out?

所以这是个非常有意思的问题:这一切最后会怎么落定?

And then, where does the n plus one person go?

然后,第 n+1 个人该去哪儿?

And there's lots and lots and lots of places for the n plus one person to go.

而第 n+1 个人能去的地方多得是、多得是。

They're still exceptional.

他们依然出类拔萃。

They're still top of the bell curve.

他们依然在钟形曲线的顶端。

You know, again, I don't want to have that misinterpreted as the person is not being amazing.

再说一次,我不希望这被误解成那个人不厉害。

It's just at some point, people will do some cut-off on their power law.

只是到某个时点,大家会在自己的幂律上画一条线。

I think you will appreciate this.

我觉得你会喜欢这个。

Maybe you've heard it, cuz it's an old uh ex-Googler joke.

你也许听过,因为这是个 Google 老员工的老段子。

But, um when Google was like, I don't know, 50,000 people or something, the question was, how many people does it take to run Google?

当年 Google 大概有五万人吧,有人问:要多少人才能把 Google 运转起来?

And if you ask somebody within search and ads, they're like, oh, like 20% of people in search and ads.

你问搜索和广告部门里的人,他们会说:大概搜索和广告里的 20% 的人。

And if you ask somebody outside of, you know, search and ads, they'd say like, 50,000 people or whatever Google is.

你问搜索和广告之外的人,他们会说:五万人,或者说 Google 有多少人就是多少人。

Um, so, I think this is probably some very different perspectives on the uh concentration of contribution.

所以我觉得,这大概是关于‘贡献集中度’的几种不同视角。

I don't know.

我也说不好。

I think I've never worked at Google.

毕竟我没在 Google 待过。

Yeah, I mean, we definitely know that uh Yeah, I mean, I worked at Google.

对,我们当然知道——对,我在 Google 待过。

Um, I thought it was a wonderful place.

我觉得那是个很棒的地方。

In search?

在搜索?

I worked on mobile search a bit, and I worked on um I worked on ads a bit.

我做过一阵移动搜索,也做过一阵广告。

Uh I mean, I worked on a bunch of mobile stuff, and then I worked on a bunch of like AI ads-related stuff.

我是说,我做过一堆移动相关的东西,后来又做过一堆 AI 广告相关的东西。

May I ask you um a very different question, which is like, can you think of anything that could disrupt this all right now?

我能问你一个很不一样的问题吗:你能想到有什么东西可能把眼下这一切打断?

You could have like investors, like a number of different players collapse in their commitment on the capex side, because the market's heated.

比如投资人、比如一批不同的玩家在 capex 上的承诺集体崩掉,因为市场太热了。

There's some sort of freak out about the debt um and the returns profile.

比如出现某种对债务和回报结构的恐慌。

You see like minor indication of that, but not not real pressure yet.

你能看到一点苗头,但还没有真正的压力。

Um, and the last one is, do you think there's a technological disruption that's possible?

最后一个是:你觉得有没有可能出现技术层面的颠覆?

Like alternatives to transformers?

比如 transformer 的替代方案?

Does that still matter at all?

这件事现在还重要吗?

Is there anything that would make the landscape look really different technically?

有没有什么东西会让整个格局在技术上变得很不一样?

I think there's always technology unknowns and then I I think the idea of attempting to restrict model usage of models we already have or open-source to dramatically constrain like pace of progress, I think it's the other.

我觉得技术上的未知永远都在;而另一条,我觉得是试图限制我们已有模型的使用、或者限制开源,以此大幅压住进步速度。

Yeah, and I agree with the regulatory angle.

对,监管这条我同意。

What do you think is going to happen in California?

你觉得加州会发生什么?

So, they passed the billionaire tax.

他们通过了亿万富翁税。

And then I mean the Democratic Party in California came out in in favor of it.

而且加州民主党还公开表示支持。

Um you're a founder of one of these companies that you've backed that's now worth 10 billion plus.

假设你是你投过的某家公司的创始人,公司现在值 100 亿美元以上。

Is the founder going to have a forced asset sale now next year?

这位创始人明年会被迫卖资产吗?

Assuming it passes, will dozens of founders have to sell big chunks of their companies?

假设它通过了,是不是会有几十位创始人不得不卖掉自己公司的一大块?

It's not clear the regulators have thought through the execution and compliance of this.

监管者有没有想清楚这件事怎么执行、怎么合规,这不好说。

But I I think the immediate effect is that um a huge number of people that are attempting to create value or do new things in California choose to leave.

但我觉得直接效果是:大量想在加州创造价值、做新东西的人会选择离开。

It's It's already happening.

这已经在发生了。

It's hard to move an entire ecosystem very quickly.

要让整个生态快速搬走是很难的。

This is the fastest way I can think of to chase the entire ecosystem out.

但这是我能想到的、把整个生态赶走的最快方式。

What do you think happens?

你觉得会怎样?

Yeah, I mean the way that that law is written is um my sense is it's reasonably broad in terms of once it passes, they can re-implement it, they can lower the bar in future years, etc. And my sense is in '28, there's increasing talk about also trying to um add a exit tax in California.

我的感觉是,这条法律的写法相当宽:一旦通过,他们可以再实施、可以在往后的年份里下调门槛,等等。而且我的感觉是,到 28 年,关于在加州再加一道离境税的讨论会越来越多。

Um so, if you actually try and leave, they'll they'll try and take a big chunk as sort of a penalty for that.

所以如果你真的想走,他们会试着拿走一大块,算是对你走的一种惩罚。

So, is your prediction mass migration in '27 to Miami?

所以你的预测是 27 年大规模迁往迈阿密?

Miami finally happens?

迈阿密终于成了?

I think it will um take some time.

我觉得这会需要一段时间。

Um I think the people who wrote the bill want the flight to happen.

我觉得写这个法案的人就是想让人跑。

Um I think they want people to leave and I think the two really negative signs for California was the was this bill and then um this sort of ballot harvesting harvesting initiatives.

我觉得他们希望大家离开;而我觉得加州最负面的两个信号,一个是这个法案,另一个是这类代收选票(ballot harvesting)的动议。

I think those are the two things that kind of make a a potentially worse future for the state in different ways.

我觉得这两件事以不同的方式,让这个州的未来可能变得更糟。

So I'm hopeful like as usual that California figures it out but I think if there was any alternative that was um, easy to do a lot of people would even more people would be leaving.

所以我跟往常一样,还是希望加州能想明白;但我觉得如果有任何一个容易落脚的替代选项,离开的人会更多。

I do think a lot of people like I know quite a few who are starting to go now or planning to go uh, by you know, the fall next month or so.

我确实觉得有不少人——我认识好些现在就开始走、或者计划在今年秋天、下个月左右走的人。

What is your second choice ecosystem?

你的第二选择生态是哪里?

I think that there's a few different places that a lot of people are considering and the question is um like what is critical mass look like in two years at each one of those spots.

我觉得有几个地方是很多人在考虑的,问题在于两年后这几个地方各自的临界规模会是什么样。

So I think I think a lot of these things kind of self-assemble and people talk about weather and they talk about all these other things but the reality is um, you know, Boston used to be one of the main startup hubs.

我觉得这类事情很大程度上是自组装的;大家会聊天气、聊其他一堆东西,但现实是——波士顿当年也是主要的创业中心之一。

It still is for biotech, right?

它在生物科技上现在依然是。

Um, they sort of lost competitively in the early 90s, right?

但它大概在 90 年代初的竞争中输掉了。

In the 80s Boston was sort of the counterweight to Silicon Valley.

80 年代的波士顿算是硅谷的对手。

Um, and the weather there was awful, you know?

而那儿的天气很糟糕。

And so I think um, it's more about where do you have enough smart people aggregated working on common things and then that's where these renaissances tend to happen.

所以我觉得,关键更在于:哪里聚集了足够多聪明人在做共同的事——这类文艺复兴往往就发生在那儿。

I think it's been really exciting to see the amount of um, like great technology migration and innovation in um, in Texas around energy.

我觉得特别令人兴奋的是,德州围绕能源出现了这么大量的技术迁移和创新。

Mhm, mhm.

嗯,嗯。

Because that is really like a reaction to regulatory environment and demand where um, I've seen a lot of people either move from Silicon Valley or move from other places because it is a place where you can experiment um, and uh, the there is actually an ecosystem now that that's super exciting.

因为那真的是对监管环境和需求的一种回应:我看到很多人从硅谷、或者从别的地方搬过去,因为那是一个你可以做实验的地方;而且现在那儿是真的有一个生态了,这非常令人兴奋。

Energy and hardware actually, there's a really growing hardware corridor there well, which is you know, it was originally all around El Segundo cuz that's where SpaceX was and then Anduril.

其实是能源和硬件——那边有一条正在快速成长的硬件走廊。这东西最早都聚在 El Segundo,因为 SpaceX 在那儿,后来是 Anduril。

And now, you know, SpaceX and I think part of Tesla and stuff moved to Texas.

而现在,SpaceX、还有我觉得特斯拉的一部分之类的,都搬去了德州。

And so, there's like this new ecosystem kind of emerging around sort of

所以有这么一个新生态正在成形,围绕着——

a part of Texas as well in addition to Austin.

德州的另一片区域,而不只是奥斯汀。

So, I do think we are seeing these shifts and these shifts are purely driven by regulation.

所以我确实认为我们正在看到这些迁移,而且这些迁移纯粹是由监管驱动的。

They're not driven by is Texas a better or worse place to live.

不是因为德州是个更好或更差的居住地。

I mean, it impacts things, right? it's regulatory shifts driving people out.

当然那也有影响,但真正把人赶出去的是监管的变化。

You didn't take my bait on um architecture and psych technology.

你没接我在架构和psych技术上抛的钩子。

What do you think about architectures?

那你觉得架构怎么样?

I think we are going to like as an industry consume all of the compute and power available.

我觉得作为一个行业,我们会把所有可用的算力和电力全部吃掉。

Whatever the uh underlying architectures.

不管底层架构是什么。

Uh so, the idea that you are going to have a lot of pressure to find more memory or power efficient um architectures is like more interesting than ever.

所以‘会有很大压力去找更省内存、更省电的架构’这个判断,比以往任何时候都更有意思。

Uh but catching up to transformers in scale and match for hardware remains pretty tough.

但要在规模上追上 transformer、并且跟硬件匹配得一样好,仍然相当难。

But I think people make that they will make that bet as they get more desperate in terms of more experimentation.

不过我觉得随着大家在实验上越来越急迫,还是会有人下这个注。

I don't think it changes the direction of the industry.

我不认为它会改变行业的方向。

Think it whatever it is gets copied and then the labs do it and they have all the compute anyway.

我觉得不管那是什么,都会被抄走,然后 labs 也去做——反正算力都在它们手里。

You know, that's that's the high probability outcome.

那是高概率的结果。

It's not the only outcome.

但不是唯一的结果。

There could be some lower probability thing where some neo lab comes up with something.

也可能出现一个低概率情况:某个 neolab 搞出了什么东西。

They keep it super super secret.

他们把它捂得死死的。

They scale on it and suddenly their model is better than anyone else's by far.

他们在上面做规模化,然后突然之间他们的模型远远好过所有人。

And then they can afford all that extra compute and everything else and everybody rallies around them.

于是他们就能负担得起那些额外算力和其他一切,所有人都围到他们那边去。

You know, you could always imagine a scenario like that.

这种情形你总是可以想象的。

But you could also imagine a scenario where just one person from that team leaves for Anthropic or OpenAI and the knowledge spreads and the next thing you know, everybody has it.

但你也可以想象另一种:那个团队里只要有一个人跳槽去了 Anthropic 或 OpenAI,知识就扩散了,然后你会发现所有人都有了。

Which is what's been happening so far in terms of these models.

而在这些模型上,迄今为止发生的就是后者。

Do you think if the dominant thing is access to compute then and it's an oligopoly because of it.

如果主导因素是算力获取,并且正因如此形成了寡头格局。

Um, do you see the labs using that uh access to compute to control other verticals that they want to be in?

你觉得这些 labs 会用这种算力优势去控制它们想进入的其他垂直领域吗?

Or or what is the safety that's really needed that's actually protective of people, right?

或者说,真正需要的、真正保护到人的‘安全’是什么?

What is the risk?

风险是什么?

What is the outcome?

结果是什么?

Um, that's kind of the, you know, Janssen from Janssen Pharmaceutica.

这有点像 Janssen Pharmaceutica 的 Janssen。

You know, uh he's he's considered one of the best drug developers of all times.

他被认为是史上最好的药物研发者之一。

He has these great videos on YouTube where he's interviewed 30 40 years ago talking about regulatory capture in pharma.

YouTube 上有他三四十年前受访的很棒的视频,讲的就是制药业的监管俘获。

And the reason things got so expensive and so slow is number one, uh regulatory capture, and the second is risk-reward scenarios where the FDA, in his mind, I'm not saying this is correct or incorrect, in his mind the FDA um focuses too much on safety and risk and not enough on benefit.

而事情之所以变得这么贵、这么慢,第一是监管俘获,第二是风险-收益的取舍:在他看来——我不评价这个说法对不对——FDA 太看重安全和风险,而不够看重收益。

He says no risk-reward, there's only risk.

他说的是:根本没有风险-收益,只有风险。

So, that slows everything down cuz you're only looking at one side of the equation.

所以一切都被拖慢了,因为你只看等式的一边。

One could imagine a scenario where in the labs a version of that is created as well.

你可以想象在这些 labs 里也造出一个类似的版本。

Right?

对吧?

Where the safety burden is so high, even if the outcome is even higher, even if the positive outcome is dramatically higher relative to the risk.

安全负担被抬得极高,哪怕收益更高,哪怕正面结果相对于风险要高出一大截。

And so, this is back to if you only focus on one side of the equation, you'll always constrain things.

所以这又回到那句话:如果你只盯着等式的一边,你永远会把事情束住。

And if you constrain things, but then push progress forward internally on an exponent, and a year is worth three or four years in normal time, then you're a year ahead internally.

而如果你把外面束住了,内部却按指数往前推进——而且一年抵得上正常时间的三四年——那你在内部就领先了一年。

That's a massive advantage.

那是巨大的优势。

And so, this is very interesting question of like, where do we as society feel comfortable on the risk-reward spectrum for different things?

所以这是个非常有意思的问题:在不同的事情上,我们作为社会觉得自己在风险-收益这根标尺的哪个位置上是舒服的?

Like, if my email gets hacked, is that so terrible relative to better healthcare through AI models sooner?

比如,我的邮箱被黑,跟更早通过 AI 模型拿到更好的医疗相比,真有那么可怕吗?

Right?

对吧?

And so, that's kind of the trade-off.

这大概就是那个取舍。

So, yeah, these are all things we'll have to work through from a societal perspective.

所以,这些都是我们要从社会层面去消化的问题。

I think part of the challenge here is um it's not a comfortable stance for the regulators to for many policy makers to hear from technologists that you have to see what happens with the technology versus

我觉得这里有一部分难点在于:对监管者、对很多政策制定者来说,听技术人说‘你得先看看这项技术会发生什么’,是个不舒服的立场——而不是——

control.

——控制。

We've always said that.

我们一直都这么说。

We've It's always been a tech thing.

这一直是科技这边的说法。

It's always been throughout history, hey like of course this technology could be used in negative ways.

整个历史上都是这样:‘嘿,这项技术当然可能被用在坏的方面。’

Right?

对吧?

Every technology has both positive and negative applications.

每项技术都有正面和负面的用法。

Biotech, you could create a virus but you can cure cancer.

生物技术,你可以造出病毒,但也可以治愈癌症。

Nuclear, you could have free, cheap, abundant energy.

核能,你可以有免费、便宜、充裕的能源。

You can also create weapons.

你也可以造武器。

And if you actually look at it, you know, 70% of France is still nuclear in terms of its power generation, right?

而你真去看的话,法国到今天仍有 70% 的电力来自核能。

70% Where are all the accidents and where are all the kerfuffles and you know, nothing.

70%。那事故在哪儿?那些乱子在哪儿?什么都没有。

Nothing's happened.

什么都没发生。

US is 18% and we haven't built a reactor in 40 years.

美国只有 18%,而且我们 40 年没建过一座反应堆。

Japan is 25%.

日本是 25%。

Very safe, very abundant, but we had a safety lobby in the 70s basically kill abundant clean energy for us, right?

非常安全、非常充裕——但 70 年代那波安全游说,基本上替我们扼杀了廉价充裕的清洁能源。

Well, we're making them now.

可我们现在在建了。

We just need to make a lot of them.

我们只是需要建很多座。

We're not making much.

我们建得并不多。

We're not making much.

我们建得并不多。

So, I think um we've seen there there are real outcomes where safety has hurt us.

所以我觉得,我们确实看到有些结果是安全把我们害了。

And that's hurt us in power and energy production.

它在电力和能源生产上害了我们。

It's hurt us in aspects of medicine.

它在医学的某些方面害了我们。

It's hurt us in lots of places.

它在很多地方都害了我们。

And the question is where do we want this spectrum to be on AI for this stuff?

所以问题是:在 AI 这件事上,我们希望这根标尺停在哪儿?

And there's many worlds, many scenarios, many outcomes.

可能的世界有很多,场景有很多,结局也有很多。

And societally, we kind of get to choose where do we want to where do we want to place that needle on that on the wheel of safety versus risk versus outcome.

而从社会层面说,这个指针放在安全、风险、结果这个转盘的哪个位置上,是我们自己可以选的。

Elad, before we go, um what is something that you're just excited about that is on the positive end of that wheel?

Elad,在结束之前——有什么是你单纯感到兴奋的、落在这个转盘正面那一端的东西?

I mean, there's so much stuff I'm excited about there.

那边让我兴奋的东西太多了。

Like I think there's so much we can do from a human productivity perspective, from an education perspective, from a health care perspective, from a daily life and benefit to life perspective, self-driving and elderly, everything.

我觉得在人的生产力上、在教育上、在医疗上、在日常生活和生活质量上,我们能做的太多了——自动驾驶、养老,什么都算。

You know, like there's so much good that can come of all this.

这一切能带来的好处实在太多了。

So, I'm optimistic about a lot of applications, and that's why I'm cautious about where we should end up on that spectrum because um I do think it's always good to make sure that we have the proper safeguards societally, but I think that historically for big industries, we've gone too far.

所以我对很多应用是乐观的;也正因为如此,我对我们最终该落在那根标尺的哪个位置上很谨慎——因为我确实觉得,从社会层面确保有恰当的防护始终是好事,但历史上在大产业上,我们都走过头了。

And the reason tech has been so successful so quickly and has had so much human impact is because it's been lightly regulated.

而科技之所以这么快就取得成功、对人类产生这么大影响,正是因为它一直处在轻监管之下。

And I think it's better to keep it that way than not, and we'll lose optimism, we'll lose momentum, we'll lose progress.

我觉得保持这样比不保持要好;否则我们会失去乐观、失去势头、失去进步。

And that's what happened in biotech, and that's what's happened in a variety of areas over time.

生物科技就是这么被搞掉的,还有其他一系列领域也是。

So, happened in energy for a long time.

能源被这样搞了很长一段时间。

A call to arms against regulatory capture.

这是一份反监管俘获的檄文。

All right, we'll see you guys.

好了,我们下次见。

Find us on Twitter at No Priors Pod.

在 Twitter 上关注 @NoPriorsPod。

Subscribe to our YouTube channel if you want to see our faces.

想看我们的脸就订阅我们的 YouTube 频道。

Follow the show on Apple Podcasts, Spotify, or wherever you listen.

在 Apple Podcasts、Spotify 或者你常用的任何地方关注这档节目。

That way you get a new episode every week.

这样你每周都能收到新一期。

And sign up for emails or find transcripts for every episode at no-priors.com.

在 no-priors.com 可以订阅邮件,或者找到每一期的文字稿。