Anthropic 指控阿里巴巴对 Claude 实施蒸馏攻击

今天这期播客，Meta改变了主意，不再把数千名员工重新分配去训练AI了；同时，Anthropic指控阿里巴巴对Claude发动了一次价值2800万美元的查询蒸馏攻击——就是你向Claude提2800万次问题，把回答全收回来，再用这些问答去训练自己的模型，这叫蒸馏。

Today on the podcast, Meta is changing its mind on reassigning thousands of its employees to AI training, and we also have Anthropic, which is accusing Alibaba of a $28,000,000 query distillation attack on Claude, basically where you ask Claude, you know, 28,000,000 questions, you get them back, you take the questions and answers, and it's called distillation, but you use them to train your model.

General Intuition刚以23亿美元的估值融了3.2亿美元，专门用来训练AI代理玩电子游戏；而Claude的付费用户自一月以来增长了75%。

General Intuition has just raised $320,000,000 at a $2,300,000,000 valuation to train AI agents on gameplay, and Claude's paying customers are up 75% since January.

这事儿挺关键的，因为它们正在严重蚕食ChatGPT的领先优势。

This is pretty big because they're starting to eat severely into ChatGPT's lead.

还有Accenture，他们开始限制AI的使用配额了，因为员工们把算力都花在了一些不重要的基础任务上。

We also have Accenture who is rationing AI tokens because their employees are burning them on basic tasks that are not super important.

Naveen Rao的非传统AI方案声称能把推理能耗降低一千倍——说实话，每次听说能大幅降低AI的水电消耗，我都特别兴奋，因为这才是让AI真正规模化、释放更多潜力的关键。

And Naveen Rao's unconventional AI claims a 1,000 x cut in inference power, which, again, I'm super excited anytime we can cut the water and the power used for AI because that is basically how we make this scalable, and we unlock a lot of really cool potential.

你有没有烦过，得为好几个AI模型分别买订阅，想跟所有主流AI聊一聊，却只能被困在Claude、ChatGPT或者Gemini其中一个里？

If you've ever been sick of having to pay for multiple subscriptions to multiple AI models and you wanna talk to all of the top AI models in one chat, this is something that always annoyed me when I was stuck with just Claude or just chat GPT or just Gemini.

我老想，要是能同时看到其他模型的回复就好了，比如Claude不会生成图片，或者Grok在音频生成上可能还不如Eleven Labs。

You know, I wish I'd get the responses from others in there, especially when you have something like Claude, which doesn't generate images, or you have something like Grok, which might not be as good at generating audio as eleven Labs.

我做了一个叫AI Box的实验平台，算是我自己的创业项目，你可以在一个地方直接跟前80个主流AI模型对话。

I've built the AI Box playground, my very own startup, where you can talk to all of the top 80 models in one place.

你可以问Grok那个Gemini拒绝回答的问题，再让11 Labs帮你生成一段音频。

So you can ask Grok that question that Gemini refuses to answer, and then you can get 11 Labs to create an audio file for you.

让ChatGPT帮你做深度研究，再让Claude帮你分析推理。

You can get chat GBT to do deep research, and Claude to help you with reasoning.

所有这些都能在一个聊天窗口里完成。

You can do it all in the same chat.

如果你想试试，上面有80多个AI模型。

So if you wanna go check it out, there's 80 plus AI models.

里面包含图像、音频、视频，甚至音乐生成模型，你可以在同一个对话里和所有模型交流，还能同时开启多个对话，直接对比结果。

There's image, audio, video, even music models that we have in there, and you can talk with all of them in the same conversation, start multiple conversations, and you can compare the results side by side.

我特别喜欢让一个模型生成图片，再让ChatGPT也生成一张，然后并排看看效果。

Something I love is asking, you know, one model to generate an image and then asking ChatGPT to generate the image, and you can see the image results side by side.

这样你就能判断哪个模型更适合你手头的任务。

Pick what model is better for which task you're working on.

想试试的话，访问 aibox.ai/playground 就行。

If you wanna try it out, it's aibox.ai/playground.

每个月只要8.99美元，我会在描述里放个链接，你们可以去看看。

It's only $8.99 a month, and I will leave a link in the description to go check it out.

好的。

Okay.

咱们聊聊Meta最近的情况吧。

Let's talk about what's going on with Meta.

他们推翻了上个月决定重新调配数千名工程师的计划，原本是要把这些人调去各种AI训练岗位。

They have reversed the decision that they made last month to redeploy thousands of their engineers, and they were gonna do this to a whole bunch of different AI training roles.

有一份内部备忘录泄露了。

They had this internal memo that leaked.

被《商业内幕》看到后，迅速火了。

It was seen by Business Insider, and it it went very viral.

简单说，这份备忘录表示，他们会尊重每个人自己的选择，让员工自己决定是否参与AI项目。

But, basically, it said that they are going to, quote, defer to each individual's choice when they're talking about if they want, you know, people to go and work on the AI effort.

之前他们还说，好吧，

So in the you know, before they were like, okay.

你瞧。

Look.

我们有很多员工其实都被迫要去参与AI训练的工作。

We're we're making a whole bunch of our employees are gonna be forced to kinda go and work on this, AI training.

他们之前在重新调整整个公司的组织结构，我觉得公司内部对此反应很大。

They were reshuffling how they were organizing their whole company, and I think they got a lot of backlash from inside of the company.

据说他们的首席技术官安德鲁·博斯沃斯在最近一次采访中说，这可能是他任职以来公司士气最差的时候，主要是因为之前进行了大规模裁员。

Apparently, their CTO, Andrew Bosworth, said in a recent interview, he said that it was, quote, probably one of the worst, it's ever been when he was talking about the morale over at Meta just because I think they'd done a bunch of layoffs.

他们正在做一系列调整。

They're making a bunch of shifts.

所以，最终引发了强烈的反弹。

And so, anyways, they had this big backlash.

公司内部有很多人甚至联名反对这项重大决定。

Had a whole bunch of people inside the company that were, like, signing up titans against this big move.

因此，现在他们决定让员工自己选择是否参与AI项目。

And so now they're going to let people decide if they wanna work on the AI effort or not.

我只想说一点，当时大约有七千名员工在五月被动员起来，负责数据标注和其他一些强制性的AI相关工作。

My only thing that I will say here, because this was about 7,000 employees that were kind of recruited in May that were going to be doing like data labeling and a bunch of other mandatory AI stuff.

他们现在可以选择做或不做，但我猜如果有人拒绝，很可能就会被辞退之类的。

You know, they have the option to do it or not, but I'm sure there's some percentage if they say no, they just will be let go or something.

所以我很好奇，这到底会产生多大的影响。

So I'm curious to see how much of an impact this is.

不过我觉得，至少给员工选择权，对整体士气肯定是件好事，但也不至于让人觉得特别振奋。

However, I think just giving people the choice is definitely something that will have a good, like, it's it's good overall for the morale, but I don't know if it's like a really like, oh, man.

你瞧。

Look.

管理层现在就是任由他们自己决定。

Men is just letting them do whatever they want.

我觉得这里面多少也有点这个意思。

So I think there's a little bit of that.

明白了。

Okay.

Anthropic指控阿里巴巴在四月到六月期间，对Claude进行了两千八百八十万次提问，实施了一次蒸馏攻击。

Anthropic is accusing Alibaba of doing 28,800,000 questions in a distillation attack on Claude between April and June.

他们表示，发现了大约两万五千个虚假账户。

They said they had about 25,000 fraudulent accounts.

这是迄今为止已知的最大规模针对该公司的蒸馏攻击。

This is the largest known distillation attack on the company to date.

他们提到，这一点是在致参议院银行委员会的一封信中披露的，称存在一个重大漏洞。

They said, and they kind of disclosed this in a letter to the Senate Banking Committee, that there's a major loophole.

他们指出，外国实验室只需抓取模型的输出结果，就能绕过美国的芯片出口管制来训练自己的AI系统。

They said that foreign labs can bypass US chip export controls by simply scraping model's outputs to train their own AI.

明白了。

Okay.

这里头能聊的东西太多了。

There is so much to unpack here.

首先，他们确实是通过这种方式绕过了美国的芯片出口管制。

The first thing is they're like, they're bypassing US chip export controls.

明白了。

Okay.

别管什么美国芯片出口管制了。

Forget US chip export controls.

他们真正想说的是，任何人都能这么做，比如阿里巴巴，或者别的任何人，都可以拿去训练自己的模型，甚至把模型开源。但问题是，如果阿里巴巴只要跑两千八百万次查询就能把他们的模型蒸馏出来，那Anthropic的竞争优势到底在哪？

What they're really saying here is anybody can do this, like Alibaba, whatever, anyone, and could go train a model, could go make that model open source, But, like, what is Anthropic's moat if Alibaba can go and run 28,000,000 queries and distill their model?

别提什么美国芯片出口管制了。

Forget about The US chip export controls.

Anthropic真是在说，他们的模型训练这么脆弱，光靠这么几次查询就能被复制？

Like, is Anthropic really saying that their model training is so fragile it could just be copied by doing that?

他们花的几十亿美金，居然可能被两千八百万次查询就复制了。

You know, the billions and billions they spend could just be copied with 28,000,000 queries.

你想想，两万个虚假账户，假设每个账户每月花20美元，虽然实际上他们更可能是按API调用付费的。

And if you think about it, you know, 20,000 fraudulent accounts let's say they're they're paying $20 for each of these fraudulent accounts, although likely they're they're paying for the API.

但我们就当是每月20美元的账户吧。

But let's just say it was the $20, you know, a month account.

我们说的是五十万美元啊。

We're talking about, you know, $500,000.

真的吗？

Like, really?

所以花五十万美元，他们就能搞出这样的事来。

So for $500,000, they could they could pull something like this off.

现在我觉得可能更多，但咱们就按十倍算吧。

Now I think it's probably more, but, I mean, let's just let's go 10 x.

五千万美元。

$50,000,000.

对吧？

Right?

假设他们花了五千万美元在API调用上。

Let's say they spent $50,000,000 in API tokens.

这样的话，Anthropic就得去找政府了。

Like, Anthropic is is going to the government.

这件事要提交给参议院银行委员会了，他们一听就慌了。

It's going to the senate banking committee, and it's like, oh, no.

这些外国势力居然在干这种事。

Like, these foreign actors are doing this.

我觉得就连马斯克在最近与OpenAI的听证会上也承认，Grok和xAI确实对OpenAI做过模型蒸馏攻击。

And I think even Elon Musk admitted in his recent hearing with OpenAI that they that Grok and and xAI had done some distillation attack like, distillation model distillation on OpenAI.

说白了，现在大家都这么干。

So everybody's basically doing this.

我觉得特别有意思的是，他们这么紧张，明明花几十亿训练模型，结果别人花五千万做一次蒸馏，就能做出几乎一样好的模型。

And, anyways, it's fascinating to me that they're so concerned about this because they're spending billions to train these models, and someone could come in and spend 50,000,000, let's say, do distillation and have a model almost as good.

这简直太离谱了。

That's pretty crazy.

据称，这次的攻击规模远超今年二月DeepSeek、月之暗面和MiniMax对阿里巴巴发起的那些蒸馏攻击。

Apparently, this particular attack was way bigger than DeepSeek, Moonshot, Moonshot, and MiniMax who all did distillation attacks allegedly to Alibaba, back in February.

所以这应该是目前为止最大的一起。

So this is kind of the the biggest one.

看起来这些中国公司都在想，嘿。

It seems like all these Chinese companies are like, hey.

别管美国的芯片出口了。

Well, forget the The US chip exports.

我们直接去蒸馏Anthropic的模型就行。

We're just gonna go distill Anthropic.

特朗普政府曾单独要求Anthropic暂停向所有外国人员提供Fable Five和Mythos Five，理由是国家安全，但他们却说，你看。

The Trump administration separately ordered Anthropic to suspend Fable five and Mythos five to to all foreign nationals because of national security, but they're saying, like, look.

我们知道Fable Five确实暂停了，但假设他们没停呢？

You know, we we had to suspend Fable five, but, like, let's say they didn't suspend it.

或者在没暂停的那段时间里，如果阿里巴巴跑去对Fable Five做了一次蒸馏攻击，后来Anthropic被政府逼着停了，可那时候阿里巴巴已经把模型蒸馏出来了，能力也拿到了，而我们却没有。

Or even in the amount of time where it wasn't suspended, what if Alibaba went and did a distillation attack on Fable five, and then Anthropic, you know, the the Trump administration forces Anthropic to suspend it, but now, you know, Alibaba has it distilled, and now they have that capability and we don't.

这一切都让我对开源模型的未来非常看好，毕竟Anthropic的模型也许会一直持续地变得越来越强。

All of this makes me quite bullish on open source models in the future because, I mean, perhaps Anthropic's model is just gonna get, you know, exponentially better and better for all eternity.

但只要有一天它达到瓶颈，迟早会被蒸馏，然后放进开源模型里。

But assuming there's ever a plateau, it's gonna get distilled and put into an open source model.

如果人们有合适的硬件，就不必向Anthropic付费了。

And people, if they have the right hardware, won't have to pay Anthropic.

不过，现在内存和其他部件的价格涨得厉害，你看苹果不得不把几乎所有笔记本电脑的价格至少上调了100美元，专业款更是涨了300美元，因为零件太贵，供应也严重不足。

Although, with the price of, you know, memory and everything going up like crazy, if you saw Apple had to boost the prices of basically every single Apple laptop by at least a $100, and for the pros it was like $300 because parts are so much more expensive and there's such a shortage.

所以，就算我们以后都用本地运行的开源模型，最终可能还是逃不过被涨价坑一把的命运。

So maybe even if we were all using open source models on our own devices, we're all gonna get hosed in the future anyways.

时间会给出答案。

Only time will tell.

General Intuition 刚刚以23亿美元的估值融资了3.2亿美元，专门用于用游戏实况来训练智能体。

General Intuition has just raised $320,000,000 at a $2,300,000,000 valuation to train agents on gameplay, so specifically video game footage.

这家公司总部在纽约。

This is a company out of New York.

他们周四表示，Coastal Adventures 领投了这轮融资。

They said on Thursday that Coastal Adventures led their round.

General Catalyst、杰夫·贝佐斯、埃里克·施密特、前F1冠军尼科·罗森伯格，还有谷歌、DeepMind和麻省理工的研究人员，都纷纷投了钱。

They had General Catalyst, Jeff Bezos, Eric Schmidt, a former f one champion, Nico Rosenberg, and researchers at Google, DeepMind, and MIT were all writing checks into this.

我的意思是，如果连深脑和麻省理工的研究人员都参与了，那这项目肯定有真本事。

I mean, if you got the researchers at DeepMind and MIT, you probably know you're onto something pretty solid.

这轮融资规模不小啊。

This is a pretty big round.

他们去年十月其实已经融过一轮，金额是1.34亿美元。

They actually did a round of funding back in October of last year for a 134,000,000.

加起来，他们总共筹到了4.54亿美元。

So total, they've raised $454,000,000.

这家公司挺有名的，因为他们用八分钟的现实世界数据微调了一个机器人，再结合他们公司Metal的游戏视频做预训练，结果那机器人真能走路、能动了。

They're kind of a famous company because General Intuition fine tuned a robot using eight minutes of real world data, and then after pretraining on gameplay from their company, Metal, they're you know, the the robot was actually able to walk and move.

他们用电子游戏的数据来帮助现实中的机器人，也正是靠这个，他们才慢慢出名的。

But they're using this data from, like, video games to help robots in the real world, and this is where they're kind of becoming, you know, this is what they're getting famous for.

这次特别让人兴奋的地方在于，如果这套方法行得通，那General Intuition很可能成为驱动大量机器人、模拟和游戏初创公司的底层模型。

The reason why this one in particular is exciting is because if this works, then general intuition is gonna be kind of the the model layer that runs a whole bunch of robotics and simulation and gaming startups.

就像现在Anthropic、OpenAI这些通用模型正在支撑着所有软件类初创公司一样。

So the same way that Anthropic and OpenAI and, you know, all of these kind of general models are running all the startups today that do software.

这将会支撑起所有的机器人创业公司，所以才这么让人兴奋。

This will be doing all of the robotics startups, so that's why it's exciting.

自一月以来，Claude的企业客户增长了75%。

Claude's pain customers are up 75% since January.

这个增幅非常大。

This is a huge jump.

数据来自两千八百万美国用户的交易记录，这是首次出现明显突破，可以说是对ChatGPT订阅主导地位的真正挑战。

This is according to a bunch of transaction data from 28,000,000 US consumers, and this is the first serious jump, and I guess you could say competition to ChatGPT's subscription dominance.

因为现在大家都清楚，嗯——

Because right now everyone knows, okay.

你看，

Look.

Claude在企业市场表现非常出色。

Claude's doing really well with with enterprise.

很多大公司都在给Anthropic付大笔费用，我觉得他们在这方面已经稍微领先于OpenAI了。

There's lot of enterprise players paying Anthropic a lot of money, and I think it, you know, slightly is beating out OpenAI there.

但在消费者市场，OpenAI 一直是最强的。

But OpenAI has always been the king when it comes to consumers.

对吧？

Right?

他们的模型价格最便宜。

They're they're the ones that they're they have the cheapest model.

他们也是最早推出的。

They were came out first.

他们那边的使用量算是最大的，不过Claude在那个人群里也越来越受欢迎了。

They have kind of the most usage, but Claude is getting more and more popular with that demographic as well.

你要是去查查，看人们讨论这些指标时的一个参考，就会发现，嗯，其实有不少相关的课程。

If you go look at, one of the indicators when people are talking about these metrics is there's, I guess, a bunch of, like, courses.

这些基本上都属于自学类的课程。

And so these are kind of self directed learning courses.

你看那些数据的话，Claude课程的需求量远远高于ChatGPT。

And if you look at those, the Claude course demand is way higher than ChatGPT.

实际上是三比一。

It's actually three to one.

所以变化很大，也许你会说，这是因为Claude比较新，而很多人已经熟悉怎么用ChatGPT了，现在得重新学Claude。

So there's a huge shift, and and I guess you could say, well, maybe it's because it's newer and so many people already know how to use ChatGPT and now they need to use Claude.

说实话，我觉得Claude对普通人来说功能更丰富。

And to be fair, I think Claude has probably more to it for the average person.

对吧？

Right?

我们都习惯了和ChatGPT聊天。

We're all used to talking with Chad GPT.

我可不会专门去学它，虽然可能有人会去。

I'm not gonna go take a course on that, although maybe some people will.

但说到Claude，就有Claude Work、Claude Code，还有各种方式能把它整合到工具开发里，我觉得很多人会想报这些课程。

But when it comes to Claude and it's like, well, there's Claude co work and there's Claude code and there's, like, all these different ways you can integrate it into the the, you know, building tools, There's a lot of courses that I think people would wanna take.

所以我觉得这大概就是它突然火爆的原因。

And so I think this is probably why it's spiking so much.

据数据科学平台DataCamp称，他们的用户搜索‘Claude’的次数已经超过了搜索‘AI’的次数，过去三十天里，关于Claude的课程需求增长了18倍。

Apparently, if you go to data camp, they said that their users have searched for the keyword Claude more than the keyword AI, and they said that the demand for Claude courses is up 18 x in the last thirty days.

这确实是个不小的趋势。

So this is kind of a really big trend.

再看企业端的使用情况，我们也看到用量大幅上升。

And if we look at even the usage of, you know, enterprise, we're seeing just a huge upswing.

虽然ChatGPT在付费用户总数上依然领先，但Claude的增长势头非常强劲。

So Chateapati is still the leader in kind of the absolute amount of paying customers that are, you know, using their their model, but there is some really big growth coming out of Claude.

埃森哲现在开始限制员工使用AI的令牌配额，因为不少员工把预算全花在把PDF转成幻灯片，还有各种低价值任务上，全靠这些昂贵的模型来处理。

Accenture is now rationing employees' AI token access because apparently a whole bunch of their staff burned through all of their budgets converting PDFs to slides and running a lot of other really low value tasks through these super expensive, models.

有意思的是，这正是我节目里经常提到的一点：能用AI做的事，不等于该用AI去做。

Now it's interesting, and this is something I talk about a lot on my show, just because you can do something with AI doesn't mean you should do something with AI.

很多工作其实早就有很成熟的普通软件了，用起来完全够用。

There's a lot of tasks where we just have very basic software that we've had for a long time, and it it, you know, gets the job done.

根本没必要非得用AI来处理这些事。

You do not need to use AI to do it.

这正是其中一个例子。

This is one of the cases.

对吧？

Right?

把PDF转成幻灯片。

Converting a PDF to slides.

其实有很多更便宜的方法可以做这事，但Accenture是吃了亏才明白的。

There's probably a lot of other ways to do this that are using the the most expensive frontier models, and Accenture learned this the hard way.

这事儿其实是在反转一个月前的政策，当时他们还威胁员工说：

This is something that is, I guess, kind of reversing a month's old mandate where they were threatening people and saying, hey.

听着。

Look.

不使用AI就别想升职，结果人人都开始拿AI干所有事。

You can't get a promotion if you don't adopt AI, and now everyone uses uses AI for everything.

现在突然发现，糟了。

And now all of a sudden, like, oh, shoot.

这成本可不低啊。

Like, this is pretty expensive.

有些任务的投入产出比实在不高，所以他们正在慢慢调整，改变主意了。

There's not a lot of ROI on some of these tasks that people are doing, so they're they're kinda walking it back and changing their mind there.

负责代理式AI战略的贾斯汀·夸克告诉高管们，每月AI支出波动太大，让财务团队根本没法准确预测成本，引发了CFO、COO和CIO级别的关注。

Justice Quack, who is the agentic AI strategy lead, told executives that unpredictable month to month AI spend makes it nearly impossible for finance teams to model costs triggering CFO, COO, and CIO level scrutiny.

今年早些时候，埃森哲内部搞过一套排行榜，把员工的AI使用量作为指标来推动普及。

Earlier the earlier this year, Accenture built internal leaderboards where they're basically ranking their employees on AI usage to drive adoption.

现在回头看，这做法明显适得其反——谁用的token最多，哪怕是在做价值很低的工作，也不该因此得到最多奖励或晋升。所以现在他们正在转向新的方向。

This obviously looks super counterproductive today, just saying whoever burns the most tokens on you know, and maybe that could be on really low value work, shouldn't be the people that are are being, I guess, compensated the most or or promoted the So right now, they're now switching gears.

他们打算开始限制使用量，这正好赶上一波持续数日的AI概念股抛售潮，不少内存芯片厂商都受了冲击。

They're gonna start rationing, and this is also coinciding with a multi day AI stock sell off, which is hitting a lot of memory chip makers.

我觉得这说明很多企业其实很难为他们的推理硬件支出找到合理依据。

I think this shows that a lot of enterprises are unable to justify some of their inference hardware spending.

我特别想聊的一个故事是，纳文·拉奥，他以前是Databricks的AI负责人。

A story I'm super excited about is that Naveen Rao so he is the ex AI chief at Databricks.

他离职后创办了自己的公司，名叫Unconventional AI。

He left, and he launched his own company, which is called Unconventional AI.

他说，基于振荡器的芯片架构能把推理功耗降低多达一千倍，相比GPU。

And he said that an oscillator based chip architecture can cut inference power by up to 1,000 x versus a GPU.

他这项新技术，据称在能效上比GPU高出一千倍以上。

So he has this new technology, and he believes it is 1,000 times more product or more, you know, efficient power efficient than a GPU.

他们刚发布了首款产品，一个可运行的图像生成模型。

They just released their first uno, which is a working image model.

画质基本能达到Stable Diffusion的水平。

It's basically matching stable diffusion's quality.

这简直来得正是时候——现在我们正面临能源和算力的双重瓶颈，各大云服务商都吃不消了。

This could be built at the perfect time when we're basically facing a massive crunch of not just energy, but just, you know, having enough compute power with all of these hyperscalers.

这听起来有点技术性，但振荡器计算是通过耦合的物理振荡器自动进入低能耗状态，而不是像GPU那样每秒数十亿次地在逻辑门之间搬运数据。

Sort of this sounds a little bit technical, but oscillator computing uses coupled physical oscillators to settle into low energy states, and they're doing this instead of what we're we're we're typically used to with GPUs where they're shuffling bits through logic gates billions of times a second.

我们这是在彻底改变计算架构的底层逻辑。

So we're completely changing the way that we're doing the architecture here.

我觉得我们真的迫切需要解决这个问题，因为大型云服务商正面临巨大的能耗瓶颈。

And I think we have a real need because hyperscalers have this huge power wall.

对吧？

Right?

所有顶尖的前沿实验室，比如OpenAI、Anthropic、谷歌，都在签下吉瓦级别的核能长期购电协议。

All of the big frontier labs, OpenAI, Anthropic, Google, they're all signing, you know, nuclear PPAs for gigawatts of generation.

他们都在努力从这些新建的能源设施获取电力，但无论是建数据中心，还是建发电设施，都需要很长的时间。

And all of these you know, they're they're trying to get the the energy generation from all of these new energy facilities, but it it takes so much time to build not just the data centers, but also a lot of the energy generation.

对吧？

Right?

微软签下的那些核能协议，要到2020年代末才能真正投入使用，所以今天我们能实现的任何能效提升都至关重要，它们能帮我们撑到这些新电厂和数据中心陆续上线。

The the nuclear deals that Microsoft is signing, all of that isn't gonna arrive till the late, you know, the late twenty twenties, and that makes, you know, basically any real efficiency gains that we can make today huge because that ties us over until a lot of these new power plants come online, a lot of new data centers come online.

未来会有大量基础设施要建设，但我们可能很快就会耗尽现有的资源。

We have a lot of, you know, infrastructure that's gonna get built out, and we like, we're gonna run out very quickly.

我觉得现在大家都开始意识到这一点了。

I think everyone's realizing that.

所以这些能效提升对整个行业来说意义重大。

And so when we have these efficiency gains, it means a lot for the industry.

这真的让人非常振奋。

So very exciting.

各位，非常感谢收听本期播客。

Guys, thank you so much for tuning into the podcast.

别忘了给个评分和评论。

Make sure to leave a rating and review.

如果你在使用苹果设备，就在 Spotify 上点几颗星吧。

So if you're on Apple, dropping some stars on Spotify.

这对我们帮助真的很大。

It really helps a show out.

它能直接提升节目在算法中的排名。

It it basically boosts it up in the algorithm.

让更多人发现这个节目，也让我受益匪浅。

It helps more people find it, and it helps me out a ton.

另外，如果你喜欢我们聊过的某个话题，或者想听更多相关内容，也欢迎在评论里告诉我们。

Also, if you have topics that you really like that we cover or you wanna hear more about something, make sure to leave it in a review.

我每一条评论都会看。

I read all of the reviews.

所有留言我都会认真读，非常感谢大家的反馈，这些意见真的帮助我不断改进节目的质量。

I read all the comments, and I really appreciate them, and I help I try to let those help guide the show to to improve the quality on it.

如果你们还有什么想法或建议，随时在评论里说，我们下一期再见。

So if there's anything else you guys want, let me know in the comments, and I will catch you guys all in the next episode.

别忘了去 aibox.ai 看看，每月 8.99 美元，就能体验 80 多个 AI 模型。

Make sure to check out aibox.ai as always if you wanna try 80 different AI models in one place for $8.99 a month.

好了。

Alright.

下期见。

Catch you all in the next episode.