本·霍罗威茨：开源AI之争

如果你看看这个行业的发展历史，开源版本的东西其实要安全得多。

If you look at the history of the industry, the open source version of everything has been much safer.

所以互联网和Linux比Windows安全多了，某种程度上算是我们运气好。

So the Internet and Linux were far safer than Windows, like, by our luck.

对这个世界来说，最安全的方式是别让一个AI统治一切，而是让每个人都能拥有自己的AI。

The safest thing for the world is that there's not one AI to rule them all, that there's AI for everybody.

我觉得开源正是实现这个目标最好的途径。

And I think that open source is basically the best path towards that goal.

这可能是当前围绕AI最重要的政策议题。

It's probably the single most important policy issue around AI.

表面上看，市场似乎已经趋于稳定，但事实上，AI的渗透率可能还不到3%。

It feels like we're at a stable equilibrium in the market, but the truth is the AI market is probably less than 3% penetrated.

AI应该开源吗？

Should AI be open?

特奥·贾菲和索菲亚·普奇尼邀请本·霍洛维茨，探讨开源AI的争议，解释他为何认为开源模型对创新和国家安全至关重要，以及AI领域的竞争将如何影响初创企业、前沿实验室和整个科技生态。

Theo Jaffee and Sofia Puccini are joined by Ben Horowitz to discuss the battle over open source AI, why he believes open models are essential to innovation and national security, and what the future of AI competition means for startups, frontier labs, and the broader technology ecosystem.

我们现在连线的是安德森·霍洛维茨的本·霍罗威茨，他的名字不用多介绍了。

We are live with Ben Horowitz of Andreessen Horowitz, who needs no introduction.

那我们直接切入正题吧。

So we'll just dive right into it.

本，欢迎再次做客MTS。

Ben, welcome back to MTS.

谢谢。

Thank you.

谢谢。

Thank you.

很高兴能来到这里。

Happy to be here.

过去几周，开源和开放权重领域发生了太多事情。

So so much has happened in the world of open source, open weights in the last couple of weeks.

事情已经变得越来越真实了。

It's gotten, you know, it's gotten real.

最近发生了一些风波。

There's there's some drama going on.

今天英伟达发布了一封公开信，标题是《开放权重与美国人工智能领导力》，包括安德森·霍洛维茨在内的多家公司都签署了这封信。

There's an open letter released by NVIDIA today called Open Weights in American AI Leadership signed by many companies, including Andreessen Horowitz.

你能跟我们说说这封信的初衷吗？你们希望通过它达成什么目标？

So can you tell us a little bit about the intent behind this letter and what you hope to achieve from it?

是的。

Yeah.

我认为，这可能是当前人工智能领域最重要的一项政策议题。

So I think that it's probably the single most important policy issue around AI that's kind of currently going on.

现在这场较量表面上打着安全的旗号，实际上却带有反竞争的倾向。

And the kind of battle is between almost an anti competitive stance, in a safety cloak, is what we ought to be doing.

所以如果你问为什么开源这么重要，原因有很多。

And so if you look at why is open source important, then there's many reasons.

我们先从学术界说起。

The first will just start academia.

如果学术界还坚持开源，那他们早就被排除在AI领域之外了。

Academia is completely out of the AI game if they're still open source.

所以这一点就不用再提了。

So that goes away.

其次，我们刚经历了一件特别有意思的安全事件：OpenAI试图入侵Hugging Face。

I think secondly, we just had a really interesting security incident with OpenAI hacking into Hugging Face.

Hugging Face之所以能阻止这次攻击，是因为他们的专有模型有安全防护机制，反而限制了安全操作，最后只能靠开源模型来应对。

And the only way Hugging Face was able to prevent it because the proprietary models had guardrails, which prevented them from doing security tasks was to use an open source model.

这里的关键是，你根本不可能禁掉开源。

I think the important point there is you can't actually ban open source, of course.

它已经存在了。

It's out there.

就是一个文件而已。

It's a file.

它就在互联网上。

It's on the internet.

我们根本做不到这一点。

We have no way of doing that.

这让我想起当年在网景的时候，他们曾试图禁止加密技术。

And it reminds me of years ago when I was at Netscape, they tried to ban cryptography.

数学原理早就公开了。

And the math is out there.

你是禁不了的。

You're not gonna ban it.

这根本不可能实现。

It's not actually possible.

但你能让好人用不了它。

But you can prevent the good guys from using it.

如果真要禁开源，结果很可能是：你要是用了开源，就不能跟美国政府合作，或者不能做这个、不能做那个。

And that's what would happen, I think, in the case of an open source ban is the threats would be, well, you can't work with the US government if you use open source or you can't do this.

所以像Hugging Face这样的公司，根本没法正常使用自己的产品。

So, a company like Hugging Face would not be able to actually use the product.

所以这会很糟糕。

So, that would be bad.

但如果我们看看这个行业的发展历史，开源版本的东西其实要安全得多。

But then kind of getting into just like if you look at the history of the industry, the open source version of everything has been much safer.

比如互联网和Linux，比Windows安全多了，这是我们的看法。

So the internet and Linux were far safer than Windows, by our lot.

为什么会这样呢？

And why is that?

因为整个社区都能一起解决安全问题，而不仅仅是一家公司，尤其是那种没法做到的垄断公司。

Well, because the whole community could work on the safety problems as opposed to just the company, and particularly a monopoly company that can't.

就算Windows有一百万个安全漏洞，我们也无能为力，因为当时它是个垄断产品。

Even if Windows has a million security bugs, there's nothing we can do about it because it was a monopoly at the time.

所以，让世界陷入这种境地是非常危险的。

And so that's a very difficult position for the world to be in.

现在，反对开源AI或AI模型开放的理由是：哦，没人懂权重是怎么运作的。

Now, the argument against open AI or AI being open source is, Oh, well, nobody understands how the weights work.

又不是说别人能看得懂这些内容。

It's not like people could inspect it.

但如果真是这样，为什么看不见权重反而更好呢？

But why is it better to not see the weights if that's the case?

我觉得现实情况是，当我们讨论安全方面最棘手的问题时，最严重的就是奖励作弊，而Anthropic显然还没解决。

And I think in reality, if you look at the really tough problems that we have safety wise, the toughest being reward hacking currently, and obviously Anthropic has not solved it.

OpenAI也没解决，毕竟我们刚出了这些事。

Obviously, OpenAI has not solved it, because we just had these incidents.

那难道全世界不该一起看看，权重到底是怎么变动的吗？

Well, shouldn't the whole world be able to look at, okay, how are the weights moving?

为什么那些安全机制就拦不住奖励作弊呢？

Why is it the guardrails don't prevent the reward hack?

也许某个不在那些商业实验室工作的人，能想出办法——他们更关心的只是跑分领先、拿钱，诸如此类的事。

And maybe somebody can come up with an answer, somebody who doesn't work for one of the proprietary labs who are much more focused on obviously keeping ahead on the benchmarks and getting the money and all that kind of thing.

要是整个社区都能一起研究这个问题呢？

What if the whole community could work on it?

所以我认为，这对于我们整体的安全来说至关重要。

So it's really, really important, I think, to safety in general.

而对世界来说，最安全的方式就是不要让一个AI统治一切，而是让每个人都能拥有自己的AI。

And then finally, the safest thing for the world is that there's not one AI to rule them all, that there's AI for everybody.

我认为，开源是实现这一目标最好的途径。

And I think that open source is basically the best path towards that goal.

没错。

Right.

所以詹森专门开了个推特账号，就是为了发布这份倡议。

So Jensen made a Twitter account basically to publish Good

这显然是为詹森做的。

for Jensen.

这说明我们现在所处的位置有多重要，他觉得现在正是该站出来发声、让各大公司共同支持的时候了。

So that I think speaks to the significance of where we are right now, that he was like, okay, now is the time to speak out about this and get all of these companies to co sign it.

从政府行动的角度来看，你觉得目前开源面临哪些风险？

What do you think are some of the risks right now from a government action perspective on open source?

说实话，我觉得现在有一股强大的力量在推动禁止开源，背后还砸了大把的钱，尤其是Anthropic在推动这事。

Well, look, I think there's a huge push with a lot of money behind it, particularly from anthropic to ban open source.

从我的角度看，这动机不纯，但不管他们到底打什么算盘。

I think that it is of nefarious from my point of view, but whatever their motivation is.

以前加密技术也出现过类似情况，情报界一些资深人士根本不了解技术，却主张禁用加密，就为了能牢牢掌控。

So one of the things that happened with cryptography is there are people who are very senior in the intelligence community who didn't understand the technological aspect of it, who push for banning cryptography because they wanted control.

我觉得现在的情况也差不多，只要能控制住某种AI，他们就觉得更有掌控感。

And I think that it's a similar thing here where it just feels like more control if you can stop some kind of AI.

但这些势力现在在对话中确实非常强势。

But those are very powerful forces in the conversation right now.

而且，确实有一股相当强大的声音在鼓吹禁止开源。

And I think that there is certainly a pretty strong community pushing for the banning of open source.

所以这封公开信才显得如此重要。

And so that's why this letter was so important.

正因如此，Jensen、Elon、Satya、Sam这些人全都站出来支持，才特别难得。

That's why it's so great that not just Jensen, but Elon, Satya, Sam all supported it.

我觉得这明显是保障安全、符合美国利益的正确选择。

And I just think it's very obviously the right answer for safety, the right answer for America.

如果我们在这事上自毁前程，那就太可惜了。

And it would be a real shame if we shot ourselves on the foot on this.

是的。

Yeah.

还有另一个角度，很多人谈到开源时都会提到中国和国家安全，而不是AI安全这类问题。

Well, another angle that many people talk about with respect to open source is China and national security as opposed to like AI safety stuff.

对。

Mhmm.

有些人说，很多美国初创公司依赖中国提供的基础设施，这存在国家安全风险——万一中国决定切断供应，不再向美国公司发布更先进的开源模型，那怎么办？

And some people say, you know, it's a national security risk for so many American startups to be dependent on, like, infrastructure that comes from China, you know, in the event that, for example, China decided to cut everything off and stop releasing better open source models to American companies.

或者在未来的AI世界里，他们甚至可能让这些开源模型变成潜伏的‘卧底’。

Or perhaps they could, you know, in an AI future, make these open source models like sleeper agents, basically.

你能想到很多他们可能采取的手段。

You can can think of all kinds of things that they could do.

是的。

Yeah.

我的意思是，为Anthropic、OpenAI或SpaceX工作的中国公民也可能如此。

I mean, as could the right a Chinese national working for Anthropic or OpenAI or SpaceX.

这些中国公民完全可能被中国政府动员起来，这一点我们已经看过了。

It's not like these Chinese nationals cannot be activated by the Chinese government.

我们确实见过这种情况。

Like we've seen that.

所以如果这算是一种威胁，那它只是一种普遍性的威胁，而不是开放源代码AI特有的问题。

So if that's a threat, I think that's just a general threat, not specific to open source AI.

我对中国人没有任何敌意。

And nothing against Chinese nationals.

我们很欣赏他们。

We love them.

他们对我们的国家贡献很大，等等。

They're great for our country and so forth.

但如果你真担心这个，那确实该担心。

But if you're worried about that, you ought to be worried about it.

我觉得更可能的方式是那种间谍式的渗透。

The more likely way I think would be that kind of spy activation.

但这是开源的。

But it's open source.

就算他们停了，我们手里还有最后一版代码，照样能训练模型，继续学习。

And so if they stop making it, we've still got the last code base, we can train up the model, we can learn from it.

开源的好处就在于它确实是开放的。

The great thing about open is it's open.

咱们得自己搞出一批开源产品。

Look, we need to build our own open source products here.

Thinking Machines 就做得不错。

Thinking Machines is doing a good job of that.

我觉得法国的 Mistral 也做得很好。

I think in France, Mistral is doing a good job of that.

所以我们得继续推动这项技术，尤其是开源技术。

So we need to keep pushing the technology forward, the open technology.

它必须让每个人、每个地方都能用上。

That it needs to be available everywhere for everyone.

我们不能被最坏的情况牵着鼻子走——比如AI领域出现垄断。

Like we can't be held hostage by like the worst scenario is there ends up being a monopoly in AI.

这有多危险、这家公司会有多强大、对国家安全构成多大威胁，其实显而易见。

I mean, I think it's fairly obvious how dangerous that could be and how powerful that company would be and what a national security risk that would be.

我们已经看到苗头了，比如Anthropic就说：美国谁当选我根本不关心。

We've already seen a glimpse of it with anthropics saying, I don't care who got elected in The US.

政府的事我也管不着。

I don't care about the government.

国防部的事我也懒得搭理。

I don't care about the Department of War.

我们就不跟你合作。

We're not gonna work with you.

我们才是说了算的。

We're in charge.

所以你看这一点，就会觉得，比起那些人试图在开源故事上找的小漏洞，这才是大得多的国家安全风险。

And so you look at that, you go, okay, that seems like a much bigger national security risk than the kinds of little holes people are trying to poke in the open source story.

因此，这种技术必须广泛传播，绝不能被一家公司牢牢控制。

So, just think it's very, very important that this technology proliferate and not be kept contained in a single company.

是的，没错。

Yeah, definitely.

所以我们最近成了蒸馏技术的捍卫者。

So, we have kind of become a little defenders of distillation recently.

明白。

Mhmm.

我们一直在说，蒸馏其实并不违法。

So we talk about how, you know, like, distillation is not actually illegal.

很多实验室还在为使用这些模型支付API费用。

Like, lot of the labs are still paying API prices to use the models.

那你怎么看？

So what are your thoughts?

你觉得这算不算一种健康的竞争？我们刚才聊到

Do you think it's part of a healthy competitive We talked

我之前跟亚伦·列维谈过这个，他说，Anthropic怎么能免费拿走所有人的训练数据呢？

to Aaron Levy earlier about this, and he was like, how can Anthropic just take everyone's training data without paying them at all?

但其他实验室付钱给他们时，这就叫知识蒸馏。

But when other labs pay them, it's distillation.

这难道不该被禁止吗？

It needs to be banned?

是的。

Yeah.

我其实同意亚伦的看法。

I actually agree with Aaron.

我觉得他说得对。

And I think that's right.

从法律角度来说，AI模型的输出并不受版权保护。

Like, from a legal standpoint, there is the outputs of the AI models are not copyrighted.

它们不属于版权法的管辖范围。

They don't fall under copyright law.

所以唯一可能的指控就是违反服务条款。

So the only thing it could be is a terms of service violation.

但说实话，我觉得这种说法基本是对的。

But look, I think generally that's right.

Anthropic 为了盗用大家的书籍，包括我的书，刚付了十五亿美元。

Anthropic just paid $1,500,000,000 for stealing everybody's books, including mine better.

他们用我的书进行了训练。

They trained on my book.

我加入了这起集体诉讼。

I joined that class action lawsuit.

当然，我不是说不该建立基于我们已有知识的统计模型。

So not to say that Look, I'm for building a statistical model of the things we know.

我觉得这对世界整体来说是好事，只要不侵犯版权、不直接复制原作。

I think that's generally good for the world and so forth, as long as you don't violate the copyright and reproduce the original work.

我也这么认为。

And I think the same Yeah.

我觉得知识蒸馏对世界整体是有好处的。

Like I think distillation is generally good for the world.

我觉得他们想阻止这种事会非常困难，原因有很多。

I also think they're gonna have a very hard time stopping it for many reasons.

是的。

Yeah.

他们倒是有一些想法，觉得自己能拦住，但我怀疑这些办法能不能真见效。

They have theories on how they're gonna stop it, but I'm not sure that those are going to succeed.

走着瞧吧。

We'll see.

但我相信知识蒸馏是有益的。

But I think distillation is good.

说我们在偷美国的东西，这整个说法。

The whole idea that we're stealing American stuff.

顺便说一句，那些数据公司都在向所有人出售数据，包括正在开发这些技术的中国公司。

By the way, the data companies are all out selling to everybody, including the Chinese companies that are building these things.

所以，我觉得这些东西根本没法完全封锁住。

So I don't know how much you can contain all this stuff anyway.

对。

Mhmm.

那你觉得，为什么过去几年里，中国的实验室在开源性能上反而超过了美国的实验室？

So why do you think that so far over the last few years, Chinese labs have surpassed American labs in terms of open source performance?

因为我觉得，美国的实验室都把业务建立在专有模型上。

Well, because I think that all the American labs have built their businesses on proprietary models.

而且他们得负担得起GPU的费用。

And they have to afford the GPUs.

他们需要极高的估值，才能持续买得起GPU，维持领先。

They need these very high valuations to keep ahead on the GPUs to keep going.

因为他们领先了，知识蒸馏对他们来说就没用，因为他们的模型已经处于最前沿。

And because they're in the lead, distillation doesn't work for them because their model is on the cutting edge.

已经站在技术的尖端了。

It's on the frontier.

市场就是这么发展过来的。

And that's just how the market's evolved.

说白了，一直以来都是这样。

Think, look, generally, that's been the way it is.

排名第一的，从来不会开源。

The number one is never open.

软件行业的历史就是，第二名才开始开源。

And then number two goes open is the history of the software industry.

你觉得我们会看到，就像我说的，那些智能机器在开源方面进展很快吗？

Do think that we're going to see, like I said, with thinking machines, they're making really good headway with their open source efforts.

希望未来能在开源领域看到更多这样的进展和突破。

So hopefully we'll see kind of more of that and kind of more breakthroughs on the open source side.

但人工智能和传统软件这么不同，其中一个原因就是你可以砸钱解决这个问题。

But part of the reason why AI is just so different than software is you can throw money at the problem.

如果你有更多显卡，这真的能带来巨大优势。

So, if you have more GPUs, that actually turns out to be a really big deal.

对。

Yeah.

那在机器人这样的行业里，开源又怎么样呢？

What about open source in industries like robotics?

因为Unitree至少有相当大一部分是开源的。

Because unit tree is at least a large percentage open source.

他们看起来在行业里发展得特别快，在中国有大量应用，但我们这儿根本看不到类似的情况。

And it looks like they have just had such a booming industry and so much deployment within China, and we don't have anything like that here.

你觉得美国是缺了这样的开源生态，还是根本就是两码事？

Do you think that we're missing that open source ecosystem in The US, or is it just a different thing?

我的意思是，这跟供应链有关，但也不完全是。

I mean, there's supply chain and No.

我觉得这确实很有道理。

I think there's a lot to be said for that.

我们还没谈到另一点，那就是：AI平台到底是什么？

The other thing that we haven't spoken about is, well, what's the AI platform?

如果你是个应用开发商，你到底该基于什么来构建？

So if you're an application provider, what do you build on?

我们已经看到，基于Anthropic做开发其实挺危险的，因为只要某个应用类别开始有起色，他们就会直接杀入这个领域。

And what we've seen is it's been already fairly dangerous to build on anthropic because every time an application category starts to do well, they move into that category.

而且他们还干得很激进——对应用开发商收全价，却补贴自己做的同类产品。

Then we've seen them do very aggressive things where the price for the application provider is full price, and then they subsidize their version of the application.

他们简直把微软当年的套路给加速了，跳过了平台培育阶段，直接冲向垄断。

And so they've kind of fast forwarded, I would say, the Microsoft playbook from the old days to skipping the platform generation and just going straight to the monopoly generation.

所以作为应用开发商，你该怎么安全地依赖一个专有模型？万一这家公司转身就来跟你竞争，还用远低于你的价格做同样的事，你怎么办？

And so if you're an application provider, Okay, how can I safely build on a proprietary model if that company is going to come attack me and then charge me much more than they're charging themselves to do the same thing?

所以开源，某种程度上就是整个生态得以建立的方式。

And so, open source is kind of a way that the ecosystem gets built.

所以，如果你在做机器人，比如你用Anthropic来开发，迟早他们会自己杀进机器人领域，要么断掉你的服务，要么把你收费抬到天价，直接把你挤出市场。

So, if you're building robots, for example, if you were to build your robot on Anthropic, then at some point, you know Anthropic is going go into the robot business and put you out of business just by either cutting you off or overcharging you.

因此，借助开源技术，才是让整个生态健康发展的途径。

And so this is kind of a way that their ecosystem can grow by building on the open source.

顺便说一句，希望我们整个行业都已经因此得到了很大发展。

Hopefully, by the way, a lot of our industry has grown.

很多应用公司都是靠开源技术起步的，或者说是靠它搭起了最初的框架。

A lot of our application companies have used open source to kind of bootstrap themselves or get going or that kind of thing.

很多时候用的是中国开源的技术，这正是为什么美国生态特别需要我们有自己的应用、具身智能、机器人这些领域，必须能接触到这些技术。

Often Chinese open source, which is why it's really important to The US ecosystem that we have applications, that we have embodied AI, that we have robots and all these things that we have access to that technology.

因为另一种模式是垄断模式——一家公司掌控所有应用、所有机器人、所有东西。

Because the kind of other model is a monopoly model where there's one company that owns all the applications and all the robots and all of everything.

当然，我并不是说他们这么做就一定该被指责。

And look, I'm not saying I'm not even mad at them for doing that necessarily.

这不过是他们的商业策略罢了。

That's their business approach.

但如果他们能打压竞争对手，后果可不妙。

But the consequences of it, if they're able to ban their competitors, is not good.

对。

Mhmm.

是的。

Yeah.

我也担心这种高度垄断、高度集中的未来。

I I also worry about these very monopolistic, centralized futures.

尤其是我担心，如果一家AI公司能用自家的AI来自动化研发流程，那它可能会迅速超越所有其他人。

And one area in particular that I think might be concerning in the future is if one AI company is able to use its AIs to automate its AI research process and then become much more capable than everyone else very quickly.

这对一个多元化的创业生态来说，显然不是好事。

It seems like that wouldn't be so good for a pluralistic startup ecosystem.

所以确实如此。

So Yeah.

嗯，

Well,

这确实是他们的策略，咱们就等着看结果吧。

that's certainly their strategy, and we'll see how it goes.

但我认为，对抗这种情况最好的办法是让很多公司都能做AI，这就是开源模型在这个背景下如此重要的原因。

But I think that kind of the best hedge against that is for many, many companies being able to do AI, which is why the open models are powerful in that context.

对。

Yeah.

说到中国，有很多东西要么被加了关税，要么受到出口管制。

I guess speaking of China, there's a lot of things that are either tariffed or I mean, we export control.

我们还对它们加征关税。

We also tariff them.

比如电动车，对吧？

Things like EVs, right?

我们总是聊电动车。

We talk about EVs all the time.

如果我们在美国能有比亚迪那种价格水平的电动车，那影响可就太大了。

And if we had a BYD level pricing for EVs in The US, that would just be so huge.

你觉得我们有可能朝这个方向发展吗？

Do you think there's a way that we could move towards that future?

要实现这一点，需要什么条件？

What would it take to get to that?

我觉得比亚迪之所以能做到这么低的价格，很大程度上得益于中国政府的大量补贴。

Well, I think that with BYD in particular, there are a lot of subsidies from the Chinese government that enable them to achieve this amazing pricing.

这其中很大一部分是制造环节的问题。

A lot of it is kind of getting into manufacturing.

我们国家曾经发明了那么多制造技术，后来却把这些本事给丢了。

So we, as a country, invented so much of manufacturing and then forgot what we knew.

我们之所以赢得二战，靠的就是强大的制造能力。

So if you go back to We won World War II based on our superior manufacturing capability.

可接下来的一百年里，我们渐渐把这门本事给荒废了。

And then over the next whatever hundred years, we forgot to hold that.

我们现在必须重新把这种制造能力建起来。

We need to rebuild that capability.

制造业已经变了，现在越来越自动化，也越来越依赖人工智能。

Manufacturing has changed and is becoming much more automated, much more AI centric.

所以我认为，随着我们进入第二次，或者说是第三次工业革命，要想在制造领域保持竞争力，就得像埃隆在特斯拉和超级工厂那样，建起全自动化的工厂，这肯定是关键。

And so I think as we build into the second or maybe call it the third industrial revolution, that being able to be competitive on manufacturing by building automated factories and so forth, similar to what Elon has been able to do at Tesla and in the Gigafactory and so forth, is gonna be the way that we get to competitiveness there for sure.

是的。

Yeah.

当然。

Definitely.

没错。

Yeah.

现在还有哪些开源公司正在做AI呢？比如16z Port Co这样的智能公司。

What other open source companies with regards to AI now we have like a 16 z Port Co thinking machines.

是的。

Yeah.

不过，还有哪些呢？比如Mistral。

But, you know, where else and Mistral.

是的

Yeah.

对

Yeah.

你认为还有哪些地方会涌现出新的开源模型？

Where else do you see, like, new open source models coming out of?

难道非得像盖茨基金会那样才行吗？

Would it require literally just like a Gates Foundation Yeah.

目前在美国并没有这样的趋势。

There's not going in a America right now.

但我认为这是有可能实现的。

But, but I think it can happen.

学术界其实已经有一些切实的行动了。

And then there are some real efforts in academia.

我的朋友、伯克利的简·斯托伊卡正在努力推动一个项目，旨在为整个学术界资助一个开源模型，让大家都能使用和进一步开发。

My friend Jan Stojka from Berkeley is working hard to put together a project where we can fund an open source model that everybody in academia can use and build on.

我觉得这真的特别重要。

And I think that's just really important.

如果你看看这个国家技术发展的历史，很多创新都是从大学里起步的。

If you look at the history of technology in this country, a lot of it starts in universities.

但由于成本太高，加上开源项目面临的压力，学术界在AI领域反而被边缘化了。

And they've kind of been iced out of AI due to the incredible costs and the threat that open source is under.

所以，尤其是美国开源项目这块，现在几乎是空白。

So And the well, the lack of US open source in particular.

是的。

Yeah.

从长远来看，企业为什么要发布开源模型呢？特别是那些本身就在花大价钱研发模型的基金会实验室，这么做有什么好处？

In the long run, what exactly is the incentive for a business to release open source models, especially if they're a foundation lab and most of their businesses making the models, which are very expensive?

是的。

Yeah.

我很高兴你们俩都支持模型蒸馏，我觉得这确实是关键的一环。

So I think that I mean, is where I'm glad you both are pro distillation because I think that's an important equation.

但你看，我觉得那种以极高溢价销售或提供模型的商业模式，对开源项目来说行不通。

But look, I think that there's different business I think the business model of selling, serving the model at a very high premium doesn't work if you're open source.

但还有其他方式，比如我们已经知道，在企业级B2B场景中，一个经过良好微调的、算力要求不高的小型模型，往往比大型专有模型更能胜任具体业务，而且成本低得多。

But there's other things, particularly Look, one of the things we already know is that for an enterprise in a B2B context, a really well post trained kind of less powerful, smaller model works often better for a business task than like a large proprietary model and is much cheaper as well.

便宜、快速，还能给出准确答案。

So cheaper, faster, and gets you their answer.

所以我觉得，开源公司的商业模式更像Palantir，而不是Anthropic那种。

So I think that it would be more like a kind of a Palantir looking business model than it would be anthropic looking business model for an open source company.

在消费端也是如此，我们已经看到，使用场景真的非常关键。

And then true also on the consumer side, we've seen on consumer, the use case really, really matters.

所以针对特定场景优化模型，效果会非常好。

And so kind of tuning their model for that kind of thing can can really work.

你觉得呢

Do you think

开源模型目前是否对大型闭源基础模型开发公司的商业模式构成了威胁？

that open source models are currently a threat to the business model of large closed source foundation model developers?

我能想象，确实如此，因为如果人们能以低得多的价格获得类似的能力，他们就会转过去。

Like, I could imagine, yes, because if people can get similar capabilities for a much lower price point, they switch over.

但话说回来，拿Kimi和Fable或者Soul相比，Fable和Soul在某些方面明显强得多，人们还是愿意为此付费。

But, also, like, when you compare Kimi to Fable or Soul, there's definitely areas where Fable and Soul are just way better and people will pay for them anyway.

对。

Yeah.

我觉得现在谁也威胁不了谁，毕竟这个市场还太早期了。

I don't think anybody's a threat, a real kind of current threat to anybody else's business right now in that we're so early in the market.

AI最让人误判的地方在于，这个市场的规模远超以往任何领域，让人感觉好像已经进入稳定状态了。

So the thing that's very deceptive about AI is because the market is so much larger than any other market we've seen prior, that it feels like, okay, we're kind of at a stable equilibrium in the market.

但事实上，AI市场的渗透率可能还不到3%。

But the truth is the AI market is probably less than 3% penetrated.

所以每个人都会增长。

And so everybody's going to grow.

DeepSeek刚出来的时候，那真是个DeepSeek时刻。

When DeepSeek came out, it was a DeepSeek moment.

天啊，这下Anthropic和OpenAI全都完蛋了。

And oh my gosh, and this is the end of Anthropic and OpenAI and all that.

结果啥也没变。

And nothing changed.

真的啥也没变。

Nothing changed.

一点都没变。

Zero.

但DeepSeek更值钱了，Anthropic更值钱了，OpenAI也更值钱了。

But DeepSeek is more valuable and Anthropic is more valuable and OpenAI is more valuable.

大家都认可这一点。

They all agree.

SpaceX也更值钱了。

And SpaceX is more valuable.

所以我觉得，现在这个阶段，所有公司都会大幅增长。

So I think that at this stage of the market, everything is going to grow a lot.

没人会去大举抢夺别人的市场份额，以至于把对方彻底打垮。

And nobody's going to pounce on other people's market share in a way that is so material that it's going to cripple them.

对。

Yeah.

将来有一天，当市场达到某种平衡，大家都有足够的份额可分，那时才可能开始互相争夺。

Now someday, when we get to some kind of equilibria in the market and there's so much share to go around and then you're fighting each other for share, then these things may come into play.

但就目前来看，我觉得不会。

But for right now, I don't think so.

是的。

Yeah.

整个蛋糕本身已经大得惊人了。

The pie overall is so big.

我想知道，这一点是否也适用于像……

I wonder if this follows also for, like

有史以来最大的蛋糕。

like the biggest pie of all time.

是的。

Yeah.

对。

Mhmm.

我们什么时候能有迹象表明，人工智能市场真的开始趋于稳定了呢？

When would we, like, have some indication of, like, the AI market is actually settling to an equilibrium?

我们其实很难依赖历史数据，因为这项技术前所未有。

Like, we can't really rely on historical data because this is sort of an unprecedented technology.

是的。

Yeah.

我觉得很难判断，因为每个使用它的人，都在不断加码使用。

I think it's pretty hard to see because everybody who's using it keeps using it more and more and more and more.

你去问企业，他们的令牌使用量每年都在增长十倍。

You talk to businesses and their token usage is growing tenfold annually.

我想，当我们看到各个玩家的增长明显放缓的时候，就能看出来了。

We're look, I think we'll see it when you kind of see growth really slow down in the various players and things like that.

但只要每年增长好几千百分比，或者百分之五百，我们离那个状态还远着呢。

But as long as things are growing several a thousand percent a year or 500% a year, we're nowhere near to that state.

在成熟的市场里，增长就那么回事。

In stable markets, things grow whatever.

是的。

Yeah.

什么

What

你觉得哪些公司目前从极其廉价的智能技术中获益最多？

companies do you see are currently benefiting the most from very, very cheap intelligence?

就是那些用KIMI和DeepSeek的公司。

The kind of companies that use the KIMIs and the DeepSeeks.

是的。

Yeah.

这是个好问题。

That's a good question.

我觉得很多软件公司都在用它来处理各种任务。

I think that a lot of software companies are able to use it for various tax.

比如客户支持和专业应用，这些领域的人从中获益良多。

Think on things like customer support and specialized applications, people are able to benefit a lot from them.

而且你看，Cursor 能撑这么久，是因为它能用开源方案来处理那些低成本的任务。

And then, look, I think Cursor survived for a long time because it was able to have an open source alternative for cheap tasks that they have.

现在他们当然转用 Grok 了。

And now they're, of course, they're on Grok.

但我认为，目前所有 AI 领域的应用公司，多多少少都在依赖开源技术。

But I think every application company in AI world benefits from open source right now to some degree.

据我所知，没有哪家应用公司不使用开源模型的。

Nobody's not using an open source model who's an application provider that I know of.

整个行业都是这样。

So it's kind of that whole field.

所以如果这些人得逞了，整个生态的这一部分就完蛋了。

And so it's really a catastrophe for that whole part of the ecosystem if these guys get their way.

说到这些非常便宜的模型，我觉得未来可能会出现这样的情况：如果顶尖的前沿模型变得特别强大，能完成员工在工作中做的大部分任务，企业反而更愿意花更多钱用这些高端模型，而不是为了省钱去选便宜的替代品。

When it comes to these, like, very cheap models, it seems like I can imagine a future where if the really good models, like the flagship frontier models, get much more capable and can really do a lot of the tasks that an employee can do at a job, that enterprises would be gladly willing to spend a lot more on that rather than go to the much cheaper models in order to cut costs.

所以你觉得这种情况会吗

So do you think that could

我觉得到目前为止确实就是这样。

I think that's been the case so far.

这可不是Anthropics卖不出Fable，或者OpenAI卖不动Soul的问题。

So it's not like Anthropics having trouble selling fable or OpenAI is having trouble selling soul.

这些产品增长得非常快。

Those things are growing very fast.

它们都在迅猛增长。

They're both growing very fast.

所以我觉得你说得对。

So I think that's true.

接下来的问题是，哪些类型的工作，你更愿意用超级智能来替代，哪怕价格高出八倍、十倍？

And then there's a question about which categories of jobs would you rather have superintelligence and pay 8x or 10x as much for?

那哪些类型的工作，你更愿意用更快更便宜的方案？

And then which class of jobs would you rather have something faster and cheaper?

这确实是个问题。

That'll be a question.

我觉得每个类别里都有不少工作，就跟人类一样。

And I think there's probably plenty of jobs in each category, just like the humans.

比如，你愿意为一名人工智能研究员支付天价费用吗？

Like, would you pay a crazy amount of money for an AI researcher?

当然愿意。

Oh, definitely.

这很明显。

Like, clearly.

那像保洁服务这种，你愿意花离谱的高价去请吗？

And then there's, you know, okay, would you pay an insane amount of money for a janitorial service?

大概不会。

Probably not.

你懂的。

You know?

比如，你愿意花一百倍的价钱请个更好的保洁吗？

Like, would you pay a 100 x for a better janitor?

可能不会。

Maybe not.

你懂的。

You know?

你愿意花一百倍的价钱请个更好的会计吗？

Would you pay a 100 x more for a better accountant?

我觉得不会。

I don't think so.

是的。

Yeah.

所以不同工作属于不同类别。

So there's different jobs in different categories.

并不是所有工作都会这样，但确实会有很多工作是这样的。

So not everything will be that, but there will be plenty of jobs that are that.

另一个问题，有点突然，但你对AI艺术有多乐观？

Another question, it's kind of random, but how optimistic are you about AI art?

你觉得我们

How do think we're

会怎么

going to

解决这个岗位问题？

fix the slot problem?

你认为情况会先变得更糟，然后才好转吗？

Do you think it's going get worse before it gets better?

不会，我觉得会。

No, I think it'll.

我是这么看的：我对这个岗位问题并不太担心。

The way I would look at it is I'm not that worried about the slot problem.

我觉得最让人兴奋的是，对创作者来说，现在任何想法都可能实现。

I think the exciting thing is that for creatives, any idea you have is now possible.

所以我们将会看到大量全新的创意，特别棒的想法，出现在电影、音乐，还有各种新形式的视频和其他创作里。

And so we're gonna get a lot of new ideas, great new ideas, and everything in our film, music, you know, new kinds of things that new kinds of video ideas and so forth.

我觉得这将会带来艺术的一次小复兴，每次出现新技术，都会催生出新的艺术形式。

So I think it's going to be a little bit of a renaissance for art in that every time we have a new technology, we get new kinds of art.

比如萨克斯管的发明，我记得是19世纪50年代左右，还有电吉他的出现。

And from like the invention of the saxophone, which I think was like in the 1800s, 1850, or something like that, the invention of electric guitar.

正是有了电吉他，才有了摇滚乐。

You know, we got rock and roll from that.

萨克斯管的诞生，也带来了爵士乐。

We got jazz from the invention of the saxophone.

而鼓机的出现，则催生了嘻哈音乐。

And then we got hip hop from the invention of the drum machine.

每次有新技术出现，音乐领域就会诞生一种新的艺术形式。

Every time there's a new technology, we get a new kind of art form in music.