为什么a16z的Martin Casado认为人工智能热潮还将持续数年

是的。

Yeah.

是的。

Yeah.

我的意思是，没错。

That's I mean, that's the yeah.

这就是工作的一部分，这是一个欠定系统。

That's the that's the part of the job where it's an it's an underdetermined system.

变量远多于方程，我们只能尽力而为。

There's way more variables than equations than we just do our best.

我们的分析是多方面的。

And our analysis is multifarious.

我知道作为投资者，可能受到像X这样的因素推动，我们喜欢把风投简化为这五件事、我们的基本观点等等。

I know as investors and probably fueled by things like X, we like to reduce VC to like here are these five things, here's our basic thesis, and you know.

事实上，大多数投资决策都需要大量工作，你会考虑很多因素，最后才做出判断。

The reality is most investment decisions take a lot of work, you consider an awful lot of things, and then at the very end you kind of look at it and you make a judgment on that.

那么我们关注哪些方面呢？

So what are the things we look at?

正如我提到的，创始人与市场的契合度非常重要。

Like I mentioned, founder market fit is very important.

战术方法也非常重要。

Tactical approach is very important.

对我而言，市场本身极其重要。

The market itself to me is incredibly important.

我刚刚了解到，如果你进入的是一个萎缩的市场，生活会很艰难。

I've just learned that if you're selling into a market that's shrinking, life sucks.

即使是一个巨大的市场，哪怕是一个非常、非常大的市场，比如交换和路由市场，但如果它只增长3%、停滞或萎缩，你面对的就是预算缩减、人员裁员，所有现有企业都在为生存而战。

Even if it's a huge market, even if it's a huge, huge market, let's say switching and routing is this huge market, but if it's only growing 3% or is flat or is shrinking, you're dealing with budgets that are contracting, people that are losing their jobs, all of the incumbents are going to be fighting for their lives.

因此，我对市场是增长还是萎缩、招聘能力、融资能力都非常敏感。

So I'm very sensitive to markets that are growing versus shrinking, ability to hire, ability to fundraise.

我的意思是，所有这些因素都会影响决策。

I mean, all of these things go.

我的意思是，投资的最终备忘录通常相当全面，因此所有这些都必然要求我们在公司融资之前做大量工作。

I mean, the final memos for investments tend to be fairly comprehensive, and so all of this also necessarily requires us to do a lot of work before companies are fundraising.

因此，这个过程中的一个必要部分是，你必须不断列举出市场上存在的公司，然后进行分析，以确定谁处于领先地位，谁不是。

And so there's kind of a necessary part of this motion, which is you're constantly trying to enumerate the companies that are out there and then doing the analysis to determine who is in the lead and who is not.

然后你说得对，归根结底，你会想：好吧，我们做了这么多工作，认为可以在这里提出这个论点，但我们经常出错。

And then you're right, at the end of the day you're kind of going be like, Ah, okay, we did all of this work and we think that you can make this argument here, and we get it wrong a lot.

对吧？

Right?

没人能预测未来。

Nobody can predict the future.

是的，这正是这个资产类别的魅力所在，对吧？

Yeah, that's the beauty of this asset class, right?

没错，100%。

Yeah, 100%.

我的意思是，你必须接受这样一个事实：即使一家公司现在看起来是领先者，任何事情都可能发生。

I mean, you you just have to be comfortable knowing that even if a company looks like the leader now, anything can happen.

他们可能第二天就被收购了，而收购方是他们自己决定的；也可能突然冒出一家以前不存在的新公司，任何事情都可能发生。

They can get acquired the next day for an acquire that they decide to do, a new company can show up that didn't exist before, anything can happen.

可能会出现平台变革等等。

There could be a platform shift, etcetera.

因此，整个目标是：在一系列投资中，能否超越其他风险投资公司的上四分之一表现？

And so the entire goal is can you over a set of investments beat the upper quartile of the other venture capital firms?

这就是目标，同时你要接受沿途的亏损。

That is the goal, and you take the losses along the way.

我们在讨论企业家或高管的重要性。

We're talking about the importance of the entrepreneur or the executive.

在X上，我看到你提到你觉得哈克·谭——博通的首席执行官——是过去十多年最伟大的CEO之一。

On X, I saw you mentioned that you thought Hak Tan, the the broad cam Broadcom CEO, was, you know, one of the great CEOs of, you know, the past decade plus.

但这个名字我通常不会在这样的讨论中听到。

And that's not a name that I usually hear discussed in those in that debate.

你能告诉我，你为什么这么认为吗？

Like, can you tell me where that comes from and why you think that?

我来更强烈地表达一下。

I'll make a stronger form to say.

我认为，除了詹森和少数几个人之外，哈克坦可能是基础设施行业有史以来最出色的CEO。

I think Hakten may be the best outside of maybe Jensen and a handful of others, he may be the best CEO the industry has ever seen in infrastructure.

他简直不可思议。

He's just unbelievable.

有人应该写一本关于哈克坦的书，或者做个概述、投资组合分析或专题报道，他的员工留存率简直惊人。

Somebody should do the hot tan book or overview or portfolio or focus piece or whatever, the employee retention is unbelievable.

他成功完成了这些极其复杂的收购。

He's managed to do these incredibly complex acquisitions.

我要说，通常当你收购一家公司时，无论哪家公司，你都会把被收购的团队整合进自己的团队，你知道，会有大量律师、企业并购、业务发展和人力资源人员忙前忙后，整个整合过程都有一整套委员会负责。

And I will say, so normally when you buy a company, any company at all, like the team that you integrate the acquired company into, you know when you've got all these kind of lawyers and corp dev and biz dev and HR people running around, you've got this entire committee for integration.

当哈克坦收购一家公司时，即使是像VMware这样规模的公司，并购委员会和整合委员会就只有哈克坦一个人。

When Hawktown acquires a company, even something like the size of a VMware, the M and A committee is The integration committee is Haktan.

这家伙在工作强度和会议管理方面简直传奇。

Mean, the guy is just legendary on how hard he works, how he runs his meetings.

他对自己业务的方方面面都了如指掌。

He knows everything about his business.

他熟知所有数据。

He knows all of the numbers.

有趣的是，他是个商业专家。

And what's interesting, he's a business guy.

他不是技术专家，也不是产品专家，但他一直远离聚光灯，值得称赞的是，他只专注于业务，我们所有人都能从他的所作所为和未来规划中学到很多。

He's not a technologist nor a product You know, but he has stayed away from the limelight, and to his credit, he just focuses on the business, but there's a lot we can all learn from what he has done and what he's gonna do.

我真的认为他可能是目前最具标志性的首席执行官。

I really do think he is probably the most iconic CEO right now.

你给我编辑日历上提了个好建议，我一定会多做些研究，并让我朋友帮忙。

Well, you've put good a good marker on my editorial calendar there, so I'm gonna I'm gonna make sure to do some more research and ask my buddy to

开始着手吧。

get started.

不知道他是否曾经有过正式的职位。

Don't know if he's ever had a form.

你应该。

You should.

我的意思是，是啊。

I mean Yeah.

为什么不呢？

Why not?

是啊。

Yeah.

这是个很好的想法。

That's a great thought.

你还有另一条推文，我觉得非常有趣，也在风投圈引起了一些热议，真有意思，到底是什么样的话题会引发这些讨论的波澜。

You had another tweet that I thought was really interesting and and caused a little bit of a stir in VC world, which it's it's so it's so fun what things happen to cause a stir or not in these discussions.

对于没看到那条推文的人，它的观点是：非共识投资在早期阶段实际上相当危险。

The tweet, for folks that didn't see it, is the idea that nonconsensus investing is where the alpha is is actually quite dangerous in the early stage.

后面还有一点，但这就是核心内容。

There's a little bit after that, but that's sort of the the meat of it.

你认为为什么这条推文引起了如此强烈的共鸣，并引发了如此多的讨论，而不是愤怒？

Why do you think that struck such a chord and, you know, caused such, not outrage, but discussion?

我觉得它只是惹恼了每一个人。

Well, think it just managed to piss everybody off.

我觉得每个利益群体都找到了讨厌它的理由，对吧？

I think there was like every constituency found a reason to hate it, right?

完美的推文。

The ideal tweet.

是的，没错。

Yeah, that's right.

这简直就是所有战争测试的母版，对吧？

It's like the mother of all Warshock tests, right?

你知道，外界有一种观点认为，风投只是在做模式匹配，毫无价值，所以对这些人来说，这证实了他们的看法。

And You know, I think for you know, there's this sense outside of VC that VCs are just pattern matching and add no value, and so for those people it was a confirmation.

于是他们就说：哦，我就知道。

And so they're like, Oh, I know it.

风投不过是我们的共识，而玛蒂娜只是承认了这一点，我之前完全没意识到，但我们稍后可以深入讨论。

VC is just consensus of us, you know, and now Martina is just acknowledging it, which I totally wasn't, but we can get into that.

对于投资者来说，这就像对他们原创性的攻击，仿佛在说：我可不会那样做。

And then for the investors, it was like an attack on their originality, which is like, I don't do that.

我就是一个共识。

I'm a consensus.

你知道，有很多初级投资者根本不知道自己在说什么，他们说了些随机的话，但也有几位资深投资者说：哦，我做的都是非共识投资，等等等等。

It's like that all you know, you had many junior investors who don't know what they're talking about that They kind of said a bunch of random stuff, but you had some very senior investors that were like, Oh, I do all these non consensus bets and whatever, whatever.

所以每个人都找到了反对厄姆布里奇的理由，顺便说一句，我根本没深思过这个观点，因为这本来是一件相当无害的事情。

So everybody found some reason to take Umbridge, by the way, which was like, I hadn't even thought deeply about the take, because this is a fairly innocuous thing.

我觉得这太明显了。

Thought it was just so obvious.

我当时想，周日早上随便说点显而易见的话，结果却成了导火索。

I was like, I'll say some obvious thing on a Sunday morning, and it just turns out to have been a lightning rod.

是什么促使你

What prompted you

说出来，你当时想传达的是什么？也许很多人其实都错过了真正的重点？

to say it, and what were you sort of trying to communicate that probably a lot of people maybe talked past to the actual point I think?

我与一个庞大的投资者团队合作，经常需要提供指导。

Well I work with a large team of investors and I'm often in the position of providing guidance.

如果你不考虑后续融资，那就没有全面评估潜在机会。

And if you're not considering follow on capital, then you're not fully evaluating the opportunity set.

我发现，那种陈腐的风投格言手册里，总说一切都要追求阿尔法收益，如此这般。

And I've found that the cliche VC aphorism rule book is like everything must be alpha and this and that.

所以我觉得，已经有很多人在谈论寻找璞玉了。

So I just thought there's plenty of people talking about fighting the diamond in the rough.

有很多人谈论开拓蓝海，但另一方面却很少被提及：随着投资阶段越来越靠后，风投变得越来越追求共识，而这正是因为他们投入的资金更多，需要更高的可预测性。

There's plenty of people that are talking about fighting the white space, but there's this another side to it that isn't as represented, which is as you go later in later stages, VCs become more and more consensus driven, and that's exactly because they're putting more money in and they need more predictability.

这自然源于整个系统。

It follows naturally out of the system.

所以，从某种意义上说，这可能是你所能想象的最平淡无奇的推文了。

So in a way, this is the most banal tweet you could ever imagine.

这实际上完全显而易见。

It's actually totally obvious.

我不是说我能感知到最好的东西。

I'm not saying I can sense the best.

我做过大量非共识的事情。

I've done tons of non consensus stuff.

我只是说，如果你不考虑这一点，那就很危险，而我们经常忽视这一点。

I'm just saying that if you don't consider this, it's dangerous, and so often we don't talk about that.

所以，这正是起源——一条来自显而易见之处的极其平庸的推文。

So that was the genesis, is a totally banal tweet from a very obvious place.

嗯，时不时制造一点小风波总是好的。

Well, it's always good to to cause a little bit

尤其是针对一些最终无害的事情。

of a stir every once in a while, especially over something that is ultimately benign.

我只是觉得，我的意思是，我觉得X简直太混乱了。

I I just feel like I mean, I feel like, like, X is, like it's just it's just totally chaotic.

有些推文我觉得特别深刻又精辟，但我却忽略了另一条，那条纯粹是毫无意义的东西。

There's some tweets I'm like, This is so deep and pithy, and I ignores another one who's like this kind of pointless thing.

所以，从某种意义上说，再次着眼于市场而非公司，我认为这些推文更能反映接收者的心态，而不是发推者本人。

And so in a way, again, just looking at the market as opposed to the company, I think that tweets are much more indicative of the people receiving it than the person actually tweeting it.

说到这个，目前相当一致的领域，让我们

Speaking of, well, quite consensus sectors at the moment, let's

进入人工智能以及我们当下所处的这个疯狂世界，你最近花了很多时间在上面。

get into AI and this wild world we're living in at the moment, which you're spending a lot of time on.

我知道你提到过，你所看到的AI领域的某些能量让你想起了九十年代的互联网泡沫。

I I know that you have mentioned that some of the energy that you're seeing in AI really reminds you of the nineties.com boom.

比如，你注意到哪些象征性的现象让你联想到那种蓬勃发展的景象？

Like, what are those sort of symbols of that effervescence that you spotted that that that did bring that to mind?

是的，让我想想，我1996年满20岁，可能从1997年或1998年开始在劳伦斯利弗莫尔实验室实习，我不太记得具体是哪一年了。那几年我来回奔波，2000年时我全职在利弗莫尔工作，我只记得那段时间有一种缓慢酝酿、最终爆发的感觉。

Yeah, so let's see, I turned 20 in 'ninety six, and I was interning at Livermore probably starting, I don't remember if it was 'ninety seven or 'ninety So I was going back and forth for a few years, I worked full time in Livermore in 2000, and I just remember this kind of slow boil that erupted during that time.

当我上世纪90年代中期，比如1995年，刚开始学计算机科学时，这还是一门有点古怪的学科。

When I started computer science as an undergrad, let's say 'ninety five, it was kind of this wonky discipline.

当时我其实有点低迷，但互联网刚刚兴起，你能感受到那种兴奋。

Was actually kind of in a little bit of a slump, but the web was just starting and you could feel this excitement.

等我毕业时，我去了北亚利桑那大学，那是一所我父亲当教授的学校，位于菲格斯塔夫，即使在这个小山城，也有学生毕业就能拿到全美各地的疯狂编程工作，而且他们正被积极招揽，到处都充满这种激情。

And then by the time I graduated, I I went to Northern Arizona University, it was a school my father was a professor in Feigstaff, Arizona, and even in this small mountain town school we had students that were graduating, getting these crazy jobs as programmers all over the nation, and they were being actively recruited, so there was just kind of all of this excitement.

当我去湾区时，就会被那里层出不穷的创业者氛围所感染，什么都有。

And then when I would go to the Bay Area, I would kind of get caught up in all the founder it is that was going on, and you had everything.

你有各种派对，我记得第一次降落在硅谷，开车沿着101号公路行驶时，心想：这些广告牌全都在跟我对话，对吧？

You had all the parties, you had all I remember the first time I landed in Silicon Valley and I drove down the 101, I'm like, All these billboards are talking to me, right?

当时就有一种能量，弥漫在街头巷尾，有Linux大会、Python大会，人人都来参加，无数公司不断涌现，每个领域都充满乐观与混乱。

So there was just this energy, and it was in the streets, and you'd have Linux conference, the Python conference was going on, and everybody would show up, and all these companies getting created, and it was just optimism and chaos in every sector that you look at.

然后我觉得，过去二十年里，事情变得有点制度化了，就像只是日常做生意的普通一天。

And then it feels to me that things got a bit institutionalized, which is it's just kind of like another day to do business for the last twenty years.

但我觉得现在，你又看到了类似的能量，比如那些广告牌我们已经看了很久了。

And I feel like now you have a lot of the same type of energy, which is like, mean, you know, the billboards we've had for a very long time.

但同样，你又看到了随之而来的文化运动，以及无数创业者和投资活动。

But again, you've got these kind of cultural movements that follow it and all the founders and all the investing going on.

所以我感觉现在的能量和九十年代末期一样强烈。

So I just feel like it has the same level of energy that we had in the late '90s.

你觉得我们现在是接近1996年，还是更接近1999年或2000年初？

Do you think we're circa 'ninety '6 or closer to circa 'ninety nine, early 2000s?

1996年，没错。

'ninety six, yeah.

真的吗？你觉得还有很大的发展空间？

Really, you think you got some room to run?

我觉得人们已经忘了泡沫是什么样子了。

I think people forget what a bubble looks like.

每次估值上涨，人们就说这是泡沫。

Every time valuations go up, people say bubble.

但听好了，泡沫就是当你上车后，出租车司机开始给你推荐股票的时候。

But listen, a bubble is like when you get into a car and the taxi driver's giving you stock tips.

那才是泡沫。

That's a bubble.

还记得那些疯狂的过度行为和各种崩盘吗？

Remember all of the crazy excesses and all the crazy blowups?

这完全、完全不一样。

It's totally, totally different.

所以这感觉很像九六年初期，最大的区别是，当时那些公司甚至还没盈利，而且持续了更长时间。

So this feels a lot like early 'ninety six, and the big difference is then the companies weren't even making money, and it lasted so much longer.

顺便说一句，人们在九七年、九八年、九九年和2000年都在抱怨泡沫。

By the way, people were decrying bubble in 'ninety seven and 'ninety eight and 'ninety nine and 2000.

你最终会是对的，没错。

You'll be right eventually, yeah.

人们当时都在这么说，对吧？

People were saying it, right?

而且他们确实有非常合理的担忧。

And they actually had really legitimate concerns.

你有世通公司，它有400亿美元的债务，杠杆极高，几乎是所有这些东西背后唯一的供应商。

You had WorldCom, which had $40,000,000,000 in debt, is super levered, was like a single supplier that was underlying all of this stuff.

你甚至可以为一家几乎没有收入、收入极少的公司进行首次公开募股。

You could IPO a company with basically no revenue, very little revenue.

这些疯狂的价值型公司中的许多根本没有资金。

Many of these companies, these crazy value businesses had no money.

它们什么都没赚，对吧？

They were making nothing, right?

因此，当时确实存在这些非常合理的担忧，但这些担忧如今几乎都不复存在了。

And so there was these very legitimate concerns, and none of those really exist today.

我的意思是，为许多基础设施提供资金的公司，比如谷歌、Meta、微软，资产负债表上都有数千亿美元的资金。

I mean, the companies that are bankrolling a lot of the infrastructure have hundreds of billions of dollars on the balance sheet, Google, Meta, Microsoft.

OpenAI 拥有真实的收入。

OpenAI has real revenue.

Cursor 也有真实收入，而且其估值并未与收入严重脱节。

Cursor has real revenue, and the valuations aren't totally out of whack with the revenue.

所以，是的，市场肯定会波动，会上涨也会下跌，会出现回调之类的，但我认为我们离九十年代末那种泡沫还远得很。

So yes, markets will oscillate for sure, and so they'll go up and down and you'll have pullbacks or whatever, I but don't think we're anywhere close to like a late nineties level bubble.

不。

No.

我认为这可能会发生。

I think that could come.

而且，你知道，当像

And, you know, listen, when like

而且很可能就会，对吧？

And probably will, right?

它很可能就会，但我觉得我们还远没到那个地步。

And it probably will, but like I don't think we're anywhere close.

我只是觉得人们忘记了什么是真正的泡沫。

I just think people forgot what a good bubble looks like.

有很多有趣的事，伙计。

There are a lot of fun, man.

所以

So

开心的时候音乐还在响

The fun fun while the music is

呢。

on.

我保证。

I promise.

我觉得，你知道，我不确定我们现在处于周期的哪个阶段，我没在那个时期当过成年人，所以我没法比较，但我认为现在连出租车司机都对这些事了如指掌。

I think, you know, I I don't know where we are in the cycle, you know, I didn't live through that period as an adult, so I I can't I can't compare, but I think we're at the stage of taxi drivers knowing these things very well.

我觉得，你知道，某些领域的估值确实变得相当火热了。

I do think, you know, the the the valuations are certainly getting spicy in some levels.

也许还没到顶峰。

Maybe they're not quite at the at the peak.

是的。

Yeah.

但我认真问你一个问题。

But I I mean, honest question for you.

你觉得现在和2021年相比失衡了吗？

Do you think right now it's out of whack with 2021?

我觉得我不认为

I it's I don't think

这不是2021年。

it's 2021.

不是。

Nope.

我同意。

I agree.

我觉得我们还没到那个地步。

I think we're not there yet.

但我也不确定。

But I don't know.

这感觉像是2019年，或者2018年中期吗？

Does it feel like 2019, mid twenty eighteen to me?

对，没错。

Like, yeah.

是啊，看起来确实如此，所以也许我们还得再等十八个月到两年，但我不确定，如果我在2019年签了大额支票，我不知道在2022年会有多少笔投资会让我感到兴奋，对吧？

That's that seems and about so, yeah, maybe we got another eighteen months or two years, but I don't know if I'd love like, how many if I was writing big checks in, let's say 2019, I don't know how many of those I would have been thrilled about in 2022, right?

当然，估值会起伏不定。

For sure you have valuations waxing and waning.

我觉得把这一点应用到2021年很有意义。

I think it's great to actually apply it to 2021.

我的意思是，2021年确实有很多热情，但这种热情并不是由真实的业务使用驱动的，对吧？

I mean, 2021 there was a lot of excitement, but it wasn't actually driven by real business usage, right?

那只是因为疫情导致人们转向线上，再加上大量私人资本涌入市场。

It was like COVID, the flight to online, and then just a bunch of private capital flooded in the market.

记得Tiger、Cotu、Insight这些机构当时都在大力投资，所以某种程度上，这种热情和狂热并非源于任何可持续的商业原因。

Remember Tiger, Cotu, Insight, all of these were deploying very heavily, and so in a way there was kind of this excitement and exuberance, but not for any sustainable business reason.

这其实只是资本大量涌入，以及当时一种不可持续的宏观环境造成的奇特现象。

It was really like an influx of capital and then this kind of quirk of the macro that wasn't sustainable.

但就人工智能而言，我们已经走了三四年，看起来是可持续的，我们理解用户留存、增长和利润率。

But with AI, I mean, we were three, four years in, it looks sustainable, we understand retention, we understand growth, we understand margins.

是的。

Yeah.

而且技术突破的成分少多了。

And much less of a tech revelation.

知道吗，当时真的

Know, there was really

是的。

Yeah.

没错。

That's right.

所以我们实际上有一个坚实的基础支撑着它。

So we actually have a foundation underlying it.

所以我会说，是的，它有点像2019年，但这次是真实的。

So I would say, yeah, I mean, it kinda feels a little bit 2,019, but it's but it's real.

所以，你知道，与2021到2022年的崩盘不同，你可以说我们还处于周期的早期阶段，而且这种波动还会持续下去。

And so, you know, unlike, you know, the the twenty twenty one, twenty two collapse, I mean, you could argue that we're still early in cycle, and, it's gonna continue to oscillate.

但我认为我们离顶峰还远得很。

But I I don't think we're anywhere near near the top.

有意思。

Interesting.

是的。

Yeah.

我觉得，我的意思是，我得再好好想想，但你确实提出了很多非常有力的观点。

I think that's I mean, I I wanna think about it more, but I think you make a lot of very good cases there.

我们没有看到老虎基金涌入，但确实有不少主权财富基金的资金正在进入，还有大量的资金。

We don't have the tigers coming in, but we do have a lot of sort of sovereign wealth fund money perhaps coming in and a lot of Totally.

大型企业的现金。

Big corporate cash.

对吧？

Right?

完全正确。

Totally.

完全正确。

Totally.

完全正确。

Totally.

这是不同的层次。

That's different level.

这实际上是非常有趣的，我的意思是，也许在这个播客里，我们没有时间深入探讨。

This this is actually a very I mean, maybe on this podcast, we're not gonna have the time to dig into it.

关于当前的技术浪潮，有一个非常有趣的观点：你实际上可以投入资本，并在另一端获得收入，而这些是非常资本密集型的业务，对吧？

This is very interesting construction about the current technology wave is you can actually deploy capital and you can get revenue on the other side of it, and these are very capital intensive businesses, right?

我认为这就是市场正在试图调整的地方。

And I think that is what the market is trying to normalize.

例如，对于这些基础模型，你甚至没有十亿美元根本无法进入这个赌场。

You can't even really enter the casino without a billion dollars for these foundation models for example.

所以我同意，当我们长期筹集了这么多资金后，对于资本结构的理解确实处于一种未知领域，但我们知道的是，你可以将其转化为收入和用户。

So I agree, we're in a bit of like terra incognito as far as understanding what the capital structure is long time after you've raised this much money, but what we do know is you can actually convert it into revenue and into users.

因此，我认为这正是市场将出现大量理性化和正常化的地方。

And so I think this is where we're gonna see a lot of rationalization and normalization in the market.

但再次强调，别以为这是基础的，它并不是投机行为，对吧？

But again, don't think it's basic, it isn't speculative, right?

它只是试图理解市场正在做什么。

It is just trying to understand what the market is doing.

我最终认为市场是非常有效的。

I ultimately think markets are very efficient.

所以我认为，我们会走向理性，但人工智能领域确实创造了真实的价值，我认为，如果资金不投入，就会错失过去二十年里最大的超级周期。

And so I think, you know, we'll rationalize, but there's a true value being created in this AI, and I think that if, you know, money's not falling, it's gonna miss the greatest super cycle in the last twenty years.

是的。

Yeah.

那另一方面，你真的可能会错失良机。

That's the that's the you know, the other side of it is, you could you could really miss out.

你知道，你提到显然正在创造一些真正有价值的东西，我完全同意。

You know, you you mentioned that there's really something obviously valuable being created, and I fully agree.

但我对你们看到这些研究感到感兴趣。

But I I was interested in the fact that you see these studies.

你知道，麻省理工学院不久前发布了一项研究，说是什么来着？

You know, MIT had their study not long ago that, you know, said that, what was it?

95%的企业部署都没有带来价值。

95% of these enterprise deployments are not delivering value.

我们看到的这种差距为什么会存在？

Why is there that gap in what we're seeing?

这是一个测量问题吗？

Is that like a measurement problem?

这是一个部署问题吗？

Is it a, you know, a deployment problem?

我认为其中一个

I think one of

AI的问题在于它已经存在了很久，因此我们对它有着各种先入为主的观念。

the problems with AI is that it's been around forever and so we have all these presuppositions on what it is.

这是我对AI的看法。

So here's my view on AI.

目前的AI本质上是一种与个人行为紧密相关的个体产消者型技术。

Right now AI as it is is very much an individual prosumer type technology that's attached to individual behavior.

就像我使用ChatGPT、我使用Cursor、我使用Ideogram、我使用Midjourney，企业获得的价值在于他们的用户在使用ChatGPT。

It's like me using ChatGPT, me using Cursor, me using Ideogram, me using Midjourney, and the value that organizations get is that their users are using ChatGPT.

他们的用户在使用各种工具，这就是全部真相。

Their users are using whatever, that's what it is.

然而，企业内部有平台团队，他们的董事会说：我们需要更多AI，去实施一些项目。

However, there are platform teams within the enterprise and their boards are like, We need more AI, go implement stuff.

于是他们匆忙开展这些AI项目，而这些项目当然都失败了。

And so they're scrambling to do these AI projects, and of course those are failing.

这是一种完全不同的技术和不同的变革。

This is such a different technology and a different shift.

所以，如果你衡量的是内部自行开展这些工作的努力，那么我可以说失败率当然会很高，但这与如今数千万用户正在使用这些技术、从中获得价值，并将这种价值带入各自工作场所的事实毫无关系。

So if you measure some internal effort to go ahead and do stuff by yourself, then I would say the failure rate of course is gonna be very high, but that has nothing to do with the fact that now many tens of millions of users are using the technologies, getting value from them, and driving that value into whatever their workplace is.

因此，我认为在谈到这波人工智能时，我们必须意识到它是一个全新的事物。

And so I just think that when it comes to this wave of AI we have to realize it's a very new thing.

它的采用周期将完全不同。

It's gonna have a totally different adoption cycle.

我们尚未真正攻克直接销售的企业市场。

We've not yet cracked like the direct sales enterprise.

我想对那些正在聆听的企业的建议是：目前与其自己做这类项目，不如与真正从事这些工作的供应商或产品公司合作。

I would say for those enterprises that are listening, than doing your own kind of project for now, it's probably better to work with a vendor or a product company that's actually doing these things.

随着时间推移，就像互联网一样，互联网当初也是如此。

Then over time, just like the Internet by the way, the Internet was the same way.

就像互联网一样，它最终将以我们所有人都能理解的方式进入企业。

Just like the Internet, it will make its way into the enterprise in a way that we all understand.

但它目前还尚未到来。

But it's just not there yet.

你所说的

What are the

你最常将它融入生活的方面有哪些呢？

ways that you've ended up incorporating it into your life most, would you say?

另一方面，有没有哪些领域是你特别保护、不愿使用它的？

And on the other side, are there areas in which you're, you know, especially protective of not using it sort

为了保持你的独立思考？

of to preserve your your thinking?

我的意思是，就像我提到的，我用AI来写代码。

I mean, like like I mentioned, so I I code with AI.

我停止编程的原因就是我不想再学新的框架，对吧？

So so the reason I stopped coding is I just didn't wanna learn the next framework, right?

我的意思是，九十年代末开发时，你坐到电脑前就开始写代码，所有东西都摆在那儿，你不需要学太多东西。

I mean, the thing with developing the late nineties is you'd sit down to your computer and you'd write code, you know, and it was all kind of there and you didn't have to learn a lot of stuff.

你主要就是直接写代码。

You'd mostly just write in code.

然后在2000年代，我完成了我的博士学业。

Then through the 2000s, I did my PhD.

所以我花了不少时间去理解所有的框架和其他东西。

So then I kind of invested enough time to understand all the frameworks and whatever.

但我后来退出了，因为我开始创业或转做投资者。

But I step away because I'm building a business or I'm becoming an investor.

当我重新回来时，我不得不重新学习所有这些新东西，尤其是那些网页相关的内容。

When I go back to it, I just have to learn all of these new things, especially with all this web stuff.

你学的并不是计算机科学的基础知识，也不是任何有普遍价值的核心内容。

And you're not learning anything fundamental to computer science or anything foundational or anything that's useful outside of that context.

你只是在学习某个随机创建了这个框架的人做的愚蠢设计决策。

You're learning whatever stupid design decision some random person that created the framework did.

这正是让我远离编程的主要原因，而有了AI，我就不用再应付这些了。

And so that's really what slowed me down from coding, and with AI I don't have to deal with any of that.

我只是说，随便给我一个应用的样板代码，让我能写个视频游戏就行，所有这些决策都由AI来完成。

I'm like, whatever, give me boilerplate for an app so I can write a video game, and all of those decisions are made by AI.

所以我大量使用AI编程。

So I use AI coding very heavily.

我几乎每天晚上都这么做，它是

I do it almost every night it's

真的吗？

really Really?

是啊，

Just been Yeah,

是的。

yeah.

这算是我的放松时间，但能再次专注于编程真的非常愉快。

It's kind of my relaxing time, but it's really just lovely to be able to just kind of focus on code again.

我还有一件个人喜欢的事，我喜欢阅读与创新或经济学密切相关的著作，通常我会有很多问题，所以现在我会读一章，然后散步遛狗时——这听起来有点傻——我会用Grok的音频模式，和它讨论这一章的内容。

Another kind of just personal thing I like, so I love reading kind of figures that are closely tied to innovation or economics, and often I have a lot of questions, and so these days what I'll do is I'll read a chapter, and then when I walk my dog this is silly when I walk my dog, I use Grock audio mode, and I actually have conversations about chapter.

某种程度上，我提出的问题并不是事实型问题，而是分析和综合型问题。

In a way I don't even Like the questions I have are kind of analysis synthesis questions, not fact based questions.

比如，论证14世纪西班牙萨拉曼卡学派是奥地利经济学派的先驱。

Like make an argument of why the school of Salamanca in Spain in the 1300s was a progenitor to the Austrian School of Economics.

所以我确实会就我读过的内容进行这些对话，我发现这样做能让我更深入地思考，而且我觉得这很有趣，也让我更全面。

So I actually have these conversations about what I read, and I just find that I think more deeply about it when I do it, and I actually find it interesting and it's kind of more well rounded.

所以这对我个人来说非常棒。

So that's personally been great.

我在写作方面有点强迫症，所以我不用AI来写作。

I think I'm a bit OCD when it comes to writing, so I will not use AI for writing.

我认为写作就是思考，我用写作来思考，如果有什么东西替我做了这件事，我就无法思考了。对我来说，写作一直是一种终身工具，所以我从没用AI写过任何东西——当然，也许这么说太绝对了，但我确实从不使用AI来写作，这是我唯一真正想保护的领域。

I think writing is thinking, and I use writing to think, and so if something did that for me, I wouldn't be thinking, and I think this for me has just been a lifelong tool, and so I don't think I've ever used AI to write a single, I mean, maybe that's not true, don't want me to be too categorical, but I never, never use AI for writing, so that's the one area that I've really tried to protect.

非常有趣。

Really interesting.

我忍不住想问，你曾经喜欢的一些历史书籍有哪些呢？

I can't help but ask, you know, what what some of those historical books that you have enjoyed might be.

是的。

Yeah.

嗯。

Yeah.

最近我一直在读艾森豪威尔的书。

So I've I've just been in in Eisenhower lately.

我一直在看一堆关于艾森豪威尔的书。

I've been going through a bunch of Eisenhower books.

艾森豪威尔有什么有趣的地方呢？

What's interesting about Eisenhower, right?

他是一位保守派总统，算是温和派，但在他任内，沃伦法院成立了。

He's a conservative president, he was a moderate, but he was also, you know, it was under his watch that the Warren Court was created.

沃伦法院众所周知，是民权运动的先锋，通过推翻政策废除了大量吉姆·克劳法等等。

And the Warren Court, you know, I mean, very famously, you know, was the vanguard of the civil rights movement as far as overturning policy and kind of getting rid of a lot of the Jim Crow laws, etcetera.

因此，我最近一直在思考一些问题：如今人们批评法院体系，有很多关于法院被‘塞人’的言论，于是我经常和格罗克讨论，把对沃伦法院的批评与对当前最高法院的批评进行对比，发现这些批评竟然如此相似。

And so a lot of kind of my questions have been, today people criticize the court system and there's a lot of rhetoric on how the court's being stacked, etcetera, and so I've actually been kind of having conversations with the Grock comparing and contrasting the rhetoric around the Warren Court with the rhetoric around the current Supreme Court, and it's so interesting how similar the criticisms actually are.

当然，时代背景和环境已经不同了。

And so, of course, it's a different environment and a different era.

但对我来说，我觉得现在发生的事更像是一个历史趋势的一部分，而不是某种完全的异常。

But I for me, I I just feel much, much closer to what's going on now as being part of a historical like a historical trend than some total aberration.

我喜欢融入更宏大的叙事之中。

I like to be part of, like, the broader narrative.

也许从新冠疫情开始，甚至更早，我们就非常清楚地意识到，我们正生活在历史之中，而这种感觉在几年前可能并不那么明显。

Perhaps starting in COVID or perhaps you would start earlier even, it feels very, very clear that we're living in history in a way that, you know, maybe wasn't as as obvious, you know, a couple

几十年前相比，真正酷的是倒回时间，听听越战时期的言论，听听互联网泡沫时期的言论，听听关于战争和法院的言论，你会发现其实并没有太大不同。

decades ago versus something like What's cool is to actually rewind the clock and listen to the rhetoric during the Vietnam War and listen to the rhetoric during the .com boom, and listen to the rhetoric around the war and court and realize I don't think it's actually that much You know, we always have this kind

差不多。

of Not so different.

是的，我们总是把自己的故事讲成：‘这是前所未有的时代，我们以前从没做过这种事’，诸如此类的话。

Yeah, we always tell our stories like, Oh, these are unprecedented times, we've never done this before, and blah blah blah blah blah.

但那时候的人也说过同样的话。

But they said those words back then too.

他们确实说过。

They did.

他们说：‘这前所未有。’

They were like, Oh, this is unprecedented.

我们从未这样做过。

We've never done this.

这是国家的终结。

This is the end of the nation.

真的吗？

Really?

哇。

Wow.

对我而言，意识到这是一个连续体很好。

For me anyways, it's nice to realize that this is a continuum.

它已经持续了很长时间。

It's been going on for a long time.

美国具有反脆弱性。

The US is anti fragile.

这是世界上最好的国家，你知道的。

It is the best country you know, on the planet.

你知道，我们一直都有挑战要应对，而且我们很好地处理了这些挑战。

You know, like, we have always have challenges to deal with, and and we do a good job of dealing with them.

现在人工智能领域中，有没有哪些部分你觉得几乎是海市蜃楼？

Are there parts of the AI world right now that you consider almost a mirage?

你知道，就是那种看起来好像会有什么成果，但由于某些根本原因，不太可能持久的东西。

You know, something that looks like it could be something, but for some fundamental reason, it's unlikely to last.

你知道，我觉得人们讨论过，比如提示工程，这可能只是一种过渡状态，我想知道，你有没有发现类似性质的东西。

You know, I think folks have talked about, you know, prompt engineering, for example, is something that's maybe more of a transitory state of affairs, and I wonder, yeah, how you you know, what you might point to that has a similar quality.

所以我认为，我们现在看到的是，这些人工智能模型公司正在走两条截然不同的路径。

So I think that what we're seeing is we're seeing two pretty distinct paths that these AI model companies take.

其中一条路径是，模型的能力越来越强、越来越多。

So one of those paths, the model just does more and more and more.

所以你基本上只有一个模型，它的能力越来越强、越来越多。

And so you basically have one model that does more and more and more.

这就像这些代码命令行工具，比如 Codecs。

It's like these code CLI tools, like Codecs, for example.

所以你可以选择把事情搞得特别复杂，做大量的提示工程，开发一堆软件，或者直接把模型暴露给用户。

So you could be like, I'm gonna make it super complicated and do all this prompt engineering and have all this software, or I could just expose the model to the user.

在这种情况下，你只需把模型直接交给用户，它反而表现得更好，因为模型比你写的任何代码都更聪明，而且你根本很难预测模型会说什么。

And it seems in these situations you just expose the model to the user, and it just does better because the model is smarter than whatever code you're gonna write, and then it's just so hard to kind of interpret what the model's gonna say anyway.

所以这是一种路径。

So that's one path.

我们在像素空间也看到了类似的情况：不再为图像、3D、音乐、角色分别设计模型，而是用一个视频模型来完成所有任务。

We're also seeing this in the pixel space, which is instead of having a model for image and a model for three d and a model for music and a model for characters, I'm gonna have one video model that does everything.

顺便说一句，我会让它具有交互性。

And oh, by the way, I'm gonna make it interactive.

这就是 Genie 三号。

This is Genie three.

所以它什么都能做，对吧？

So it just does everything, right?

有一种路径，就是上帝模型路径，支持这一路径的理由是苦涩的教训论。

There's like one path, which is the God model path, and the argument for that is the bitter lesson argument.

也就是说，你拥有所有数据，模型很聪明，等等。

That's, you you have all the data, the model is smart, etcetera.

这显然是一个可行的路径。

And that's clearly a viable path.

另一种路径是模型组合路径，我们再来看像素案例。

The other path is the composition of models' paths, which is let's take the Pixel case again.

今天早上我在X上刚看到一个很棒的视频，有人制作了一个视频，说他用Midjourney生成图像。

Actually I just saw this amazing video this morning on X where somebody made a video and they're like, I used Midjourney to make the images.

他用World Labs生成三维场景。

I use World Labs for the three d scenes.

他用Suno生成音乐。

I use Suno for the music.

这是不同模型的组合，你一看，就会发现他们制作出了这段精美的视频。

And it's this composition of different models, and you look and they have this beautiful video that was created.

因此，组合路径的论点是，如果你对输出结果有自己的看法，你就能拥有更多的控制权。

And so the argument for the composition is if you have an opinion on what comes out, you'll just have a lot more control.

如果我想要精细的摄像机运动和视角，我就需要三维模型。

If I want fine grained camera movements and field of view, I'll need three d.

也许我想要非常特定的图像，并且希望这些图像保持一致，所以我需要一个图像模型。

Maybe I want very specific images and I want consistency across those, so I'll need an image model.

我想让音乐以某种特定的方式呈现。

I want the music a certain way.

我可能希望随着时间推移改变音乐，所以我需要一个独立的音乐模块。

I may want to change it over time, so I want a separate thing for music.

因此，我真诚地相信我们会看到这两种路径，而最大的错误是人们认为只会有一种路径。

And so I honestly believe we're gonna see both of these paths, and I think the biggest mistake is people assume it's gonna be one or the other.

一切都会变成一个模型，但问题是，组合确实是真实存在的。

Everything's gonna be one model, but the problem with that is composition is just real.

我们已经有一套现成的工具和现有的工具链，它们使用的是你想要整合的组件输出，所以我认为这种想法是错误的。

We've got an existing you know, set of tool we've got existing tool chains that use components of outputs that you're gonna wanna use, and so I think that's a mistake.

另一个观点是，这些大型模型没什么用。

And the other one is like, oh, these big models aren't gonna be useful.

你需要一组小型模型。

You need a collection of small models.

这显然不对，因为更好的方法将继续让这些单一模型变得更强大。

Clearly that's not true because the better lesson will continue to make these single models much more powerful.

这真的很有趣。

That's really interesting.

你提到了Codex，还说过晚上会用很多这些工具，这让我想到了Cursor，我知道你深度参与其中。

You mentioned Codex there, and you've talked about using a lot of these tools in your evenings, which brings me on to Cursor, which I know you're very involved with.

当你在分析AI协同创作，思考‘谁是这里的领导者’时，

When you were doing the analysis on, you know, AI cogeneration and thinking about, okay, who's the leader here?

Cursor是否明显就是那个领跑者？

Was it just blazingly obvious that it was Cursor?

这个过程是怎么发生的？

How how did that sort of come about?

是的。

Yeah.

我的意思是，这取决于你在做什么。

I mean, listen, it depends on what you're doing.

我是个开发者。

I'm a developer.

我们当时在研究开发者工具，而开发者工具就是IDE。

We were looking at developer tools, and the developer tool is the IDE.

现在听好了，编码的空间非常广阔。

Now listen, coding space is enormous.

对吧？

Right?

有代码仓库，有测试，有PR管理等等，对吧？

There's repos, there's testing, there's PR management, etcetera, right?

但在编码的情况下，Copilot让我们看到了AI如何强大地融入开发过程。

But in the case of coding, mean, the Copilot had given us a glimmer of how powerful AI could be integrated in the development process.

Cursor 团队执行得极其出色。

The Cursor team executed so exceptionally well.

我们一半的公司都在使用它，当时他们只是非常专注于打造 IDE，并成为领导者。

Half of our companies were using it, and they were just, at the time, just very, very focused on building out the IDE and being the leader.

因此，仅凭这一赌注，就已经非常明确了。

So it's just for that bet, that was very clear.

这并不意味着我们认为 CLI 不是一个好的方向，只是有所不同。

That didn't mean that we didn't think CLIs were a good bet, so it was just different.

这是不同的。

That's different.

当时，真正利用 PR 作为开发者接口的方法并不多，比如把 GitHub 作为开发者的界面。

And at the time, there really wasn't as many approaches that were using the PR for an interface to the developer, right, like using GitHub as interface for the developer.

但也很清楚的是，Cursor 的野心是改变整个代码世界。

But it was also pretty clear that Cursor's ambitions was to change all of code.

同时也很清楚的是，Coder 也在不断演进，因此从我们的角度来看，一个专注于开发者工具、拥有这种宏大愿景的极度产品导向的团队，对我们而言是开发者工具领域的正确选择，当然，这最终取得了非常好的成果。

It was also very clear that Coder was evolving, and so from our perspective, a very, very product focused team working on tools for developers that has this kind of broad vision was the right bet for developer tooling for us, and of course that's worked out quite well.

这太重要了，我之前说过，现在还想再强调一遍，这并不意味着其他所有领域都没有巨大价值，比如编码模型，价值巨大；CLI工具，价值巨大。

It's just so important, I said it before, I wanna say it again, that that doesn't mean that all of these there isn't tons of value in all of these other areas, like coding models, tons of value, CLI tools, tons of value.

我的意思是，这个领域非常庞大。

I mean, this space is enormous.

如果你粗略算一下，假设有三千万名开发者。

If you just do rough math, let's say there's 30,000,000 developers.

实际人数更多，但我们就假设是三千万吧。

There's more, but let's say it's 30,000,000.

假设他们每年平均产生十万行代码。

Let's say they get on average 100 ks a year.

我的意思是，这个数字对吗？

I mean, is that?

所以这可能是一个三万亿美元的市场？

So it's like a $30,000,000,000,000 market or something?

是的。

Yeah.

假设你拿到10%，那就是3万亿美元，我的意思是，我们讨论的是一个无限大的市场。

Let's say you get 10%, that's $3,000,000,000,000 I mean, we're talking about an infinitely sized market.

如果问我，人工智能最让你惊讶的领域是什么？

And if you ask me, What is the one area that AI has surprised you?

是编程。

It's in coding.

听我说，我一直在搞开发，我根本没想到会这么好。

Listen, I've been developing and my whole I would never have guessed it'd be this good.

所以你面对的是一个AI非常擅长切入的无限大市场，因此我认为我们会看到一大批极其成功的公司。

And so you've got an infinitely sized market that AI is very effective at going, and so I think we're gonna see a bunch of super successful companies.

你觉得什么会决定胜负，并在这里建立真正的护城河？

What do

因为显然，鉴于这个市场的规模，你会看到大量大型公司和初创企业都想分一杯羹。

you think will sort of dictate winners and produce real defensibility here?

因为显然，鉴于这个市场的规模，你会看到大量大型公司和初创企业都想分一杯羹。

Because obviously it's given the size of that market, you see lots of interest from from large companies and in insurgents sort of take a piece of that space.

所以我的一般经验法则是：当市场加速增长时，它们会变得碎片化，这是一种自然的物理规律。

So my my my general rule of thumbs is while markets are accelerating in their growth, they will fragment, and that's a natural law of physics.

因此，每个人在第一天就担心可防御性，这在我看来简直是愚蠢的。

And so everybody worries about defensibility on day zero, which is just dumb in my opinion.

直到市场放缓或整合时，这才变得重要，对吧？

It doesn't matter until markets slow down or they consolidate, right?

实际上，这就像：听好了，如果我是一家公司的负责人，必须花一美元，那我为什么不去花在没有竞争的领域，而不是有竞争的领域呢？

And it literally just kind of falls out of the like, listen, if I have to spend a dollar if I'm in a company and I've gotta spend a dollar, why don't I spend a dollar in an area where I don't have competition or I do have competition?

当然，你肯定会选择那些没有竞争、而你又是领导者的领域。

Well, of course, you're gonna do it where you don't and where you're the leader anyways.

我们基本上已经看到这些领域中出现了碎片化现象。

That's we've seen basically fragmentation in most of these domains.

在这些大多数领域中，都有公司在不断成长。

We've got companies growing in most of these domains.

所以我认为，从长远来看，是什么让代码领域的公司保持可防御性？

And so I think when it comes to code, long term what keeps them defensible?

所以，我目前的观点是，我认为人工智能本身没有任何固有的护城河。

So here's my current view is I don't think there's any inherent defensibility in AI.

我认为这种护城河并不存在。

I don't think that exists.

我认为人工智能克服了启动难题，因此它解决了你的客户获取问题，因为它太神奇了，但这种情况不会永远持续下去；而现在，就像有人发明了冷聚变，人们蜂拥而至来获取电力，所以它解决了你的客户获取问题。

I think that AI overcomes the bootstrap problem, so it kind of solves your customer acquisition problem because it's so magic, and that won't be the case forever, but right now it's like somebody invented cold fusion and people show up for electricity, so it solves your customer acquisition problem.

但从护城河的角度来看，你必须回归传统的护城河。

But from a defensibility standpoint, have to go to traditional moats.

我们知道如何构建护城河——比如双边市场、整合护城河、工作流护城河，无论是什么，作为一家公司，你仍然需要去建立它。

We know how to do moats, that's a two sided marketplace, it's an integration moat, it's a workflow moat, whatever it is, you still as a company have to build that.

相对于现有企业而言，好消息是，当你面对新行为时， incumbents 很难执行，而我们这里显然出现了新行为。

The good news relative to incumbents, last point on this is when you have new behaviors, incumbents have a tough time executing, and we clearly have new behavior here.

这是一种个人行为，一种新的关系。

It's an individual behavior, it's a new relationship.

实际上，往往还伴随着情感因素，比如从 TPT 四到五的转变，我们就看到了这一点。

There's actually often an emotional component too, like the shift between TPT four and five, we saw that.

所以我认为，新的行为、优势和挑战者，我们正在见证这一切的发生，因此我对现有巨头的担忧少了很多。

And so I think new behaviors, advantage, challengers, I think we're seeing this play out, and so I worry much less about the incumbents.

我认为，如果你是一位创业者正在听这段话，并且你在做一家AI公司，那么首要任务是找到那个空白领域，而不是担心防御性，我觉得是这样。

I think if you're a founder listening to this and you're doing an AI company, priority zero is finding that white space, not worrying about defensibility in my opinion, you know?

一旦你找到了那个空白领域，就依靠传统的护城河来在市场放缓时保护它。

And then once you find that white space, rely on traditional emotes to protect it when the market slows down.

还有你旗下的另一家公司，我一直从外部非常感兴趣，真想听听你和他们的故事，也让那些可能还没接触过他们的人了解他们在做什么。

There's another of your companies that I have been so interested in from, you know, only from the outside that I'd love to hear, you know, your story with them and and for folks that maybe haven't come across them yet for them to understand what they're building.

这就是World Labs，我不知道，但只要你上X平台，现在总能看到一些非常有趣的东西，某种三维模型，都是World Labs做的。

And this is World Labs, which, I don't know, reliably anytime you go on x, there is now something really interesting, some sort of three d model that, you know, WorldLabs is responsible for that

这真的很神奇。

is is quite It's magic.

这是最神奇、最令人着迷的。

It's the most magical mesmerizing.

那它是怎么诞生的呢？

So how did that come about?

是的。

Yeah.

这甚至可以追溯到我为视频游戏编写3D引擎的经历。

This even goes back to my experience for writing three d engines for video games.

这只是一个特别的兴趣。

It's just a particular interest.

所以，WorldLabs是由3D领域的真正先驱创立的。

So listen, WorldLabs was created by the true pioneers in three d.

它是李飞飞，就是那个创造了ImageNet的著名人物李飞飞。

It was Fei Fei Li, who did ImageNet, you know, super famous Fei Fei.

它是本·米伦哈尔，他创造了NERF，也就是邻域辐射场。

It's Ben Mildenhall who created NERF, which is the near Radiance Field.

它是克里斯托夫·拉瑟纳，他在高斯点云流行之前就已涉足此领域。

It's Kristoff Lassner who was like Gaussian splats before they were cool.

它是贾斯汀·约翰逊，也就是风格迁移领域的专家。

It's Justin Johnson who's the style transfer guy.

我的意思是，他们真的拥有最顶尖的团队。

I mean, like, they've just got the most epic team.

描述他们所做事情的最简单方式是：他们想从一个二维表示（比如一张图片）中创建出一个三维表示，比如一个场景或世界。

The easiest way to articulate what they're doing is they wanna take a two d representation, like an image, and create a three d representation from it, like a scene or a world.

这是一个非常非常困难的问题，因为如果你只有图片，就看不到所有东西。

And it's a very, very tough problem because if you just have an image, you can't see everything.

你看不到桌子后面，看不到你身后等等。

You can't see in the back of the table, can't see behind you, etcetera.

所以这其中涉及大量的生成式组件。

So there's a ton of generative components to it.

这还是个难题，因为训练模型需要数据，而三维数据本身就非常少，因此这可以说是一个尚未解决的问题，但它却是一个非常通用的问题。

And it's also a tough problem because the way that you train models is with data, and there just isn't a lot of three d data, so it's kind of this unsolved problem, but it turns out to be a very horizontal problem.

举个例子，你为什么需要一个三维场景？

So for example, why would you need a three d scene?

嗯，你可以用它来创建一个纯粹的虚拟环境，与之互动，放置角色，改变视角，进行增强，或者像在VR中一样步入其中。

Well, you could use it because you wanted to create a pure virtual environment that you wanna interact with, wanna place a character with, you wanna change the angle with it, you wanna augment it, you wanna step into it like VR.

你也可以用它来做各种设计，比如建筑，或者用于增强现实（AR）。

You could also use it for any sort of kind of design, like architecture, you could use it for AR.

实际上，我刚看到一个很酷的例子，一个叫伊恩·柯蒂斯的人，他在Oculus中创建了自己客厅的三维模型，并用World Labs叠加了他所做的修改。

Actually I just saw this great, this guy named Ian Curtis did this cool thing where he had a three d representation of his living room in his Oculus, and then he was overlaying changes on it that he made with World Labs.

因此，他可以在虚拟重建和真实场景之间切换，更改家具和其他物品。

And so he could switch between the virtual labs recreation and the real recreation, he could change furniture and change things like that.

而实际上，这在机器人领域也非常相关。

And then actually ultimately this is very relevant in robotics.

所以，仅靠二维视频的问题在于，你实际上没有深度信息。

So the problem with just two d video is you actually don't have depth.

你看不到物体后面的东西，因此如果你想让一个传统程序——比如机器人大脑——来判断距离、推测物体另一侧的样貌，或规划如何绕过这些障碍，你就需要创建某种三维表示。

You can't see behind things, and so you need to create some three d representation if you want a traditional program, let's say like a robotics brain, to decide things like how far away is this, what might it look like on the other side, how do I plan around these things?

因此，你能创建的三维表示越多，具身AI就会越智能。

So the more three d representation you can create, the smarter an embodied AI would be.

所以他们真正试图攻克这个终极难题：我只有世界上一个二维视角——就像我们的大脑那样——我该如何重建出三维信息，以便对其进行处理？

And so they're really trying to tackle this holy grail of problems of I just have one view in the world that's two d, which is what our brain does, and then how do I kind of recreate that in three d so that I can kind of process it?

机器人这块我觉得特别有意思。

The robotics piece was you know, is the piece that I think is so interesting.

显然，VR应用在某种程度上看起来很直观，但另一方面，目前这个市场还不算庞大。

Obviously, you know, the VR applications feel very obvious in a way, but, obviously, on the other hand, still not a market that is massive at this point.

但机器人这块感觉可能真正会带来颠覆性改变，尤其是结合了过去几年该行业出现的其他一些进展。

But the robotics piece feels like that could be, I I don't know, truly a game changer when you're combining some of these other developments we've seen in that industry over the past couple years.

它真的像是在解决一个主要的限制因素。

Like, it really feels like it's addressing one of the major sort of limiting factors.

我们还是先回到市场规模吧，因为我感觉我们在AI领域一直犯一个错误：没人会说‘二维图像’是一个市场。

Well, let's go back to the let's go back to the market size first just because I I feel this is a mistake we keep making in the AI world, which is nobody would have said two d image is a market.

没人会这么说。

Nobody.

过去有一大类公司，它们只是些小规模的收购对象，从未真正盈利，一直在试图构建二维图像。

There's a whole class of companies which were like, in the past, were small acquisitions, they were never really profitable that were trying to build two d images.

但如今，我们有了像Midjourney这样的公司，它仅靠自身努力就实现了数亿美元的年经常性收入。

And yet now we have companies like Midjourney, which Famous do has bootstrapped hundreds of millions of ARR.

我们有BFL，有Ideogram，都是非常成功的公司，等等。

We've got BFL, we've got Ideogram, very successful companies, etcetera.

所以我认为，一般来说，当创作的边际成本降为零时，市场规模就会爆炸性增长，对吧？

And so I think in general, when you bring the marginal cost of creation to zero, the market size explodes, right?

图像创作、视频创作、音乐创作等的边际成本。

Marginal cost of image creation, of video creation, of music creation, etcetera.

而且，虽然这不是你问题的重点，但我认为非常重要的一点是：如果你把三维内容创作的边际成本降到零，我认为这个市场将是无限大的。

And then again, know that this wasn't the point of your question, but I think it's very important to touch on is if you bring the marginal cost of three d content creation to zero, I think that that market is infinitely large.

我的意思是，VR体验糟糕的原因之一就是内容匮乏。

I mean, one of the reasons VR sucks is because there's no content.

我的意思是，我有一台Quest III，我很喜欢它，但我花24美元，却只能买到一些最无聊的小玩意儿。

I mean, I don't know if I've got a Quest III, I love it, but like I go, I spend $24, and I get the stupidest little thing.

所以我会说，很多元宇宙——我讨厌这个词——但为了让我们达成共识，VR、在线游戏等等，其发展实际上受限于内容。

And so I would say a lot of metaverse, I hate the term, but I mean, just to get us all on the same page, VR, online gaming, etcetera, is really gated on content.

创建三维场景太难了。

It is so hard to build three d scenes.

这太贵了。

It is so expensive.

因此，我认为以前不是市场的领域现在可以成为市场。

And so I think that markets that weren't markets before can become markets.

话虽如此，我同意你的观点：从长远来看，如果我们真的要实现具身化的AGI（我不是AGI专家，但我想说，如果我们真要实现具身化的AI或AGI，它能观察世界、构建对世界的表征，并决定如何与世界互动，那么无论如何，你都必须在三维空间中重建这个世界。

That said, I agree with you that long term, if we're gonna have embodied AGI'm not an AGI guy, but I'm gonna say if we're gonna have embodied AI, embodied AGI that looks at the world and then creates a re representation of that world and decides how to interact with that world, somewhere, somehow you're gonna need to recreate that world in three d.

你无法用链接来实现这一点。

You can't do it with links.

我喜欢这样描述：假设我蒙上你的眼睛，把你放进一个黑暗的房间，然后试图用语言描述这个房间，让你能导航或完成任务，但语言根本不够准确。

The description I like to say is like, let's say I blindfold you and I put you in a room, the lights are off, and I try to describe the room so you can navigate it or do any task, the words are just not gonna be accurate enough.

我会说：‘你面前有个杯子，离你大约三英尺远’——这根本行不通。

I'll be like, There's a cup in front of you, it's about three feet, that won't work.

但另一方面，如果我给你一个摄像头，让你能重建出三维场景，并让你置身于这个三维环境中，那你当然就能自如地在房间里导航了。

On the other hand, if I give you a camera and then you could recreate the three d and you're positioned in that three d, of course you can now navigate the room.

因此，对于具身化AI来说，这种解决方案具有某种根本性的意义。

And so there's something very fundamental to this solution space for embodied embodied AI.

你刚才说的几件事让我觉得非常有趣。

You said a few things there that were super interesting to me.

其中一件事，我真的很想深入探讨一下VR部分。

One of them I I I do wanna dig into the VR piece.

确实，内容还不够多，但真正的限制难道不是硬件吗？

It's true that there's not enough content, but isn't the real constraint there the hardware?

从功能上讲，现有的内容已经足够让我们过上近乎无限多样的生活了。

Like, functionally, there's more than enough content for us to live almost infinity lives for us to be in it.

但只要它还不够逼真，就谈不上足够令人享受。

But until it actually feels sufficiently high fidelity, it's just not enjoyable enough.

对吧？

Right?

也许对某些人来说是这样。

May maybe for some people.

我的意思是，听我说。

I mean, listen.

我非常喜欢VR。

I love VR.

每次有新的VR设备推出，我都会买下来。

I have I every time a new VR thing comes, I buy it.

我的问题是，与视频游戏不同，视频游戏深度沉浸且内容丰富，而我走上木板后就结束了。

And the my problem is is is unlike a video game, which are deeply immersive and, you've got a ton of content, I walk a plank and then I'm done.

我打了一个僵尸，然后就觉得你缺乏足够的沉浸式内容。

I shoot a zombie and then I'm I just feel like you don't have enough immersive content.

所以可能中间 somewhere 有个平衡点。

So there's probably somewhere in the middle.

如果你看看许多纯在线的虚拟体验，瓶颈在于如何构建这些非常庞大的世界？

If you look at a lot of online purely virtual experiences, the gating factor is like how do you build these very, very large worlds?

这需要很多年。

It takes years.

需要团队花费数年时间来构建这些关卡、世界和三维环境。

It takes teams of people years to build kind of these levels and these worlds and these three d environments.

非常有趣的是，我认为这是一个非常重要的观点，我与World Labs合作得非常紧密。

And what's very interesting, I think this is such an important point, what's very interesting is I work very close with World Labs.

我每周三去那里，和一个团队一起写代码。

I go in on Wednesdays, I work with a team, I write code.

你知道的，我有点傻里傻气的。

It's you know, I'm all silly.

这太棒了。

That's awesome.

你知道，

You know,

我就像是个测试用户，对吧？

like, I'm like a beta user, right?

做一些奇怪的事情，但我跟他们非常亲近。

Do some kind of silly things, but I'm very, very close.

他们和很多艺术家合作，这些是真正有3D背景的传统艺术家，他们制作出美丽的世界，花大量时间在上面。

And they work with a lot of artists, and these are traditional true artists that have backgrounds in 3Ds, and they make these beautiful worlds and whatever, they spend a ton of time on it.

他们会花几十个小时来制作它。

They'll spend tens of hours making it.

因此，你最终得到的是一个非常细致、非常丰富的虚拟世界，如果由一群人来完成，可能需要一年时间。

And so what you end up with is you end up with a very detailed, very rich virtual world that would've taken maybe a year if you had a team of humans doing it.

一个人可以用更少的时间完成，但这仍然需要艺术家投入大量的技艺和努力。

One person can do it with less time, but it still requires a ton of craft and a ton of work from an artist.

因此，我认为像这样的技术将增加我们能够观看和探索的虚拟场景和世界的数量，我认为因此，任何需要这些内容的市场都将增长，因为你能够更快速、更高质量地生产出来。

And so I think that technology like this is going to increase the amount of virtual scenes and worlds that are there for us to of view and explore, and I think as a result, any market that requires these is just gonna grow because you can produce more, better quality, and faster.

非常有趣。

Really interesting.

然后你说你不是搞通用人工智能的。

And then you said you're not an AGI guy.

我认为，从最根本的层面来说，我们还没有弄清楚人类大脑是如何工作的，我认为语言模型之类的东西可能只是其中的一小部分，我倾向于同意杨立昆的观点：我们终将实现通用人工智能，我们会一点一点地攻克它，但从我们目前所处的位置到那里并没有一条直路，不是简单地增加算力和数据，现有模型就能变成通用人工智能。

I think Tell us at the very foundations, I don't think we have figured out how the human brain works, and I don't think, I think maybe a language model or something is a small subset of it, but I tend to agree with Jan Lecun, which is we'll get to AGI at some point in time and we keep chipping off pieces of it, but there isn't a straight path from where we are now, it's not like you just add compute and data to the existing models and then we have AGI.

我认为我们只是在一点一点地解决每个小问题，因此，对我来说，把通用人工智能当作目标、衡量标准或终点，只会助长模糊的思维，因为它最终成了你寄托所有期望和恐惧的地方，而事实上，它现在甚至不是一个真实存在的地方。

I think that we just keep chipping off each pieces of the problem, and so for me using AGI as some goal or measuring stick or destination, all it does is encourage very sloppy thinking because it ends up becoming the place that you put all of your expectations and all of your fears, and right now it's not even a real place.

所以我真的努力让人们不要使用'AGI'这个术语，不要被它误导，因为这个词承载了太多魔法和恐惧，让对话变得极其困难。

And so I really try and force people not to use the term AGI, not to trick in terms of it because it's very hard to have a conversation because it's such a holding place for magic and magic fears.

我喜欢讨论具体的问题、解决方案、产品、技术、技术趋势和技术方向，也许在某个时候，我们会找到能够实现人类水平智能、具备同等灵活性并能像人类一样快速学习的架构，那时我们再讨论AGI。但在那之前，这个词只会削弱对话的质量。

And so I like to talk about concrete problems, solutions, products, technologies, technology trends, technology directions, and then hey, maybe at some point in time we will know the architecture that will provide human level intelligence with all the flexibility, that can learn just as fast, etcetera, and then we can start talking about AGI, But until that time, it just erodes conversational quality.

它并不能提升对话质量。

It does not enhance it.

我完全同意。

I I fully agree.

是的，我也有这种感觉。

It feels like it yeah.

它掩盖了意义，远多于它揭示的内容。

It obscures meaning much more than it reveals anything.

是的。

Yeah.

它在对话中毫无帮助。

It just doesn't help in a conversation.

对吧？

Right?

它确实助长了懒惰的思维方式。

It It really encourages lazy thinking.

它很快就会变成几乎完全是语义上的争论，你会想：‘等等，你所说的AGI到底是什么意思？’

It quickly becomes almost entirely semantic, where you're like, Well, actually, what do you mean by AGI?

哦，我的意思是这个。

Oh, well, this is what I mean.

好吧，那这就是，我们通常如何理解

Okay, well then, this is And how we sort of

此外，它成了一个无需真正提供理由的万能借口。

also, it becomes the universal justification without having to actually have a justification.

为什么人工智能的边际风险比传统计算机系统更高？

Well, why is the marginal risk for AI greater than traditional computer systems?

哦，AGI。

Oh, AGI.

这没有任何意义。

That doesn't mean anything.

这根本不是一种陈述，对吧？

It's not a statement, right?

为什么这会导致人们失业？

Why is this going to put end people out of a job?

哦，AGI。

Oh, AGI.

这两个都是很好的问题。

Now both of these are great questions.

劳动力问题是一个重要的问题。

The labor question is an important question.

边际风险问题也是一个重要的问题。

The marginal risk question is an important question.

我们应该进行这些讨论。

We should have those discussions.

但我们讨论这些问题时不应以AGI为框架，因为AGI并不存在，我们应当基于当前真正发生的事情来讨论。

But we have those discussions not in terms of AGI because that's not a thing, we should do it in terms of what's actually happening now.

依我经验，每次你提到AGI，人们都会用它来为自己的恐惧、担忧或最乐观的期望找借口。

And in my experience, every time you say AGI, this is what people use to justify whatever their fear is, whatever their concern is, or whatever their most optimistic hope is.

问题是，当你深入探究时，会发现这其实是一种信念，认为存在某种神奇的东西能带来这一切。

And the problem is when you dig into it, it's this kind of belief that there's this magic thing that will provide it.

因此，对我来说，问题不在于某天我们终将拥有最聪明的计算机团队，而在于‘AGI’这个词本身导致了对话和讨论质量的下降。

And so for me it's conversational and discourse quality that's the problem with the term AGI, not the fact that someday we will have computers on the smartest teams.

当然，我们会拥有这样的计算机，但眼下这并无帮助。

Of course, we will, but right now, that's not helpful.

你提到，计算能力和数据不足以让我们直接通向AGI——无论我们如何称呼它，比如在各方面都等同于人类大脑的系统。

You you mentioned that, you know, compute and data is not gonna be enough for us to sort of have a straight shot to, you know, AGI, whatever we might call it, let's say, a brain equivalent in in every way to a human.

这对你从投资角度思考这一领域的进展有何影响？

How does that impact how you think about, you know, the progression of this from an investment perspective?

你是否预期在未来几年里，这些模型的能力会持续出现巨大的飞跃？

Do you expect, you know, continuous large leaps in the capability of these models over the next few years?