← 返回任务列表

Why AI Hasn’t Increased Unemployment, According to Anthropic

338 段 · 1 位说话人 · 原片 35:37
M1
M10:00

今天的 AI Daily Brief,我们来聊聊为什么按 Anthropic 的说法,AI 还没有推高失业率。

Today on the AI Daily Brief, why AI hasn't increased unemployment according to Anthropic.

在那之前,先看头条:router 生意现在很火,因为 Stripe 正在洽谈以一百亿美元收购 OpenRouter。

Before that on the headlines, the router business is hot as Stripe is in talks to buy OpenRouter for $10 billion.

AI Daily Brief 是一档每日 podcast 和视频节目,关注 AI 领域最重要的新闻和讨论。

The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI.

好吧朋友们,在我们开始之前,先快速说几个通知。

Alright friends, quick announcements before we dive in.
M1
M10:44

欢迎回来,这里是 AI Daily Brief 的头条版,用大概五分钟带你看完每天你需要知道的 AI 新闻。不过今天,我觉得我们要聊的内容可能塞不进五分钟了,所以咱们直接开始。

Welcome back to the AI Daily Brief headlines edition, all the daily AI news you need in around five minutes, although today I do not think everything that we have to discuss is going to fit in five minutes, so let's dive in.

首先,Stripe 是最新一家进入 model routing 赛道的公司,而且桌面上摆着一笔可能非常重磅的收购。

First of all, Stripe is the latest company getting into model routing with a potentially blockbuster acquisition on the table.

《The Wall Street Journal》报道说,Stripe 正在洽谈以大约一百亿美元收购 OpenRouter。

The Wall Street Journal reports that Stripe is in talks to acquire OpenRouter for around $10 billion.

相比 OpenRouter 在上一轮融资中的十三亿美元估值,这会是一个非常夸张的溢价。而那轮融资结束的时间,我看看表,也就是两个月前的五月。

That would be a huge markup from OpenRouter's $1.3 billion valuation during their last round, which closed, checks watch, two months ago in May.

不过话说回来,这两个月里确实发生了很多变化。

Then again, a lot has changed in those two months.

我们已经从 token maxing 的时代——那时候大家都被鼓励尽可能多地使用最强模型——走到了 token scarcity 的时代。现在越来越多企业开始转向更严格控制 token 预算。

We went from the token maxing era where everyone was encouraged to use the most powerful model as much as possible, to the age of token scarcity, where increasingly enterprises are moving to more tightly controlled token budgets.

在这种转变之下,最好的 token routing 服务看起来可能会成为巨大的赢家。

In the context of that shift, it's beginning to look like the best token routing service could be a huge winner.

据说,OpenRouter 已经收到了多份收购邀约,不过 Stripe 看起来是那个资金最雄厚的买家。

Reportedly, OpenRouter has been fielding multiple acquisition offers, but Stripe is looking like the company with the deepest pockets.

退一步看,我觉得这个组合其实很合理。

Taking a step back, I think the pairing makes a lot of sense.

Stripe 在商户侧支付处理这块,基本上已经走到了他们能走的很远的位置。很明显,随着他们更有创意地思考增长问题,他们的胃口也越来越大了。

Stripe has more or less come as far as they can go with merchant-side payment processing, and clearly their eyes have been getting bigger and bigger as they think more creatively about their growth.

他们最近一直在推动和 PayPal 的合并交易,这样他们就能扩展到消费者市场那一侧。

They've recently pursued a merger deal with PayPal that would let them expand to the consumer side of the market.

与此同时,收购 OpenRouter 会让他们朝另一个方向前进,把企业成本控制工具加入他们的垂直整合技术栈里。

Meanwhile, an OpenRouter acquisition would let them move in a different direction, adding enterprise cost control tools to their vertically integrated stack.

《The Wall Street Journal》认为,这笔交易已经很接近达成,可能很快就会宣布。

The Journal suggests a deal is close and could be announced soon.

Macaroni Capital 写道,Stripe 买的不是一家 AI 公司。

Macaroni Capital writes, Stripe isn't buying an AI company.

它买的是推理环节的 metering 和 billing 层,以及附着在这上面的开发者漏斗。

It's buying the metering and billing layer for inference, plus the developer funnel attached to it.

来自 Clerk.com 的 Colin 写道,Stripe 有两个角度。

Colin from Clerk.com writes, Stripe has two angles.

第一,提高互联网的 GDP;第二,一个讨论得比较少的点,是提高他们在互联网 GDP 上的利润率。

One, increase the GDP of the internet, and two, the less discussed, increase their margins on the GDP of the internet.

OpenRouter 已经证明,他们在 inference 上的利润空间是可持续的。而且我们都知道,inference 是互联网 GDP 中一个规模巨大、而且增长快得惊人的组成部分。

OpenRouter has shown that their margin on inference is durable, and we all know inference is a massive, startlingly fast-growing portion of the internet's GDP.

提前预告一个我很确定我们之后会在节目里聊到的话题,Alex Konrad 写道,现在还没人认真讨论的 AI lab 对决,其实是 Ramp 对 Stripe。

Presaging something that I am sure we will talk about on this show at some point, Alex Konrad writes, the AI lab showdown that nobody is talking about yet is Ramp versus Stripe.

不过,即便这笔收购真的发生,OpenRouter 也会面临越来越强的一波竞争。

And yet, even if this acquisition happens, OpenRouter is going to face an increasing wave of competition.

比如说,Cursor 就在这周刚刚发布了他们自己的 Cursor Router。

Cursor, for example, just this week announced their own Cursor Router.

这之前,Meta 已经在他们内部 incubator 里做了一个 router,而 Ramp 和 Vercel 也都已经上线了各自版本的产品。

That follows Meta building a router in their internal incubator, and Ramp and Vercel also going live with their own versions of the product.

Cursor 这个 model router 的版本,可以让工程师在 intelligence、cost 或 balanced 这三种优化设置之间做选择,然后自动为具体任务选出合适的模型。

Cursor's version of a model router lets engineers automatically select the right model for the job while choosing between three optimization settings, intelligence, cost, or balanced.

接着,Cursor router 会分析每一个请求,并根据这个偏好把它发送给合适的模型。

The Cursor router will then analyze each request and send it to an appropriate model based on that preference.

他们声称,在 intelligence 模式下使用这个 router,可以在成本降低百分之六十的同时,提供 frontier 级别的表现。

They claim that using the router in intelligence mode can deliver frontier-level performance at a 60% reduction in cost.

当然,这个结果是用主观满意度指标来衡量的,所以实际表现到底有多强,可能还稍微有点难判断。

Now, this was measured using subjective satisfaction metrics, so it is perhaps a little difficult to know how strong the performance will be.

不过,早期测试者反馈说,和直接把所有请求都路由到 Opus 4.1 相比,质量并没有出现明显下降。

Still, early testers reported no noticeable drop-off in quality compared to simply routing everything to Opus 4.1.

Cursor router 最强大的功能之一,可能就是你再也不用去想该选哪个模型了。

One of the most powerful features of Cursor router could be the ability to never have to think about model selection again.

因为它是直接内置在团队本来就在用的工具里,所以甚至连额外配置这一层都不需要。

Since it's built into a tool that teams are already using, there's not even an extra layer to configure.

在解释背后的动机时,Cursor 的 CTO Michael Truell 写道,他曾经短暂地有点疯了,然后决定让每个软件工程师都顺便变成 model benchmarks、thinking levels 和 cache hit rates 的专家。

Explaining the motivation, Cursor CTO Michael Truell wrote, he briefly went insane and decided every software engineer should also become an expert in model benchmarks, thinking levels, and cache hit rates.

Matthew Berman 总结说,现在 model routing 已经是一项一级功能了。

Matthew Berman sums up, model routing is a first-class feature now.

说到新功能,Anthropic 和 OpenAI 周四发布的产品更新都和语音功能有关。说实话,到现在这个阶段,如果你还没有用语音来控制你的 agents,我真的觉得你得开始改变一下自己的使用习惯了。

Speaking of new features, the product announcements from both Anthropic and OpenAI on Thursday related to voice features, which at this point, if you are not controlling your agents with voice, I genuinely believe you need to start shifting your behavior.

总之,Anthropic 终于把他们的 voice mode 开放给 Opus 和 Sonnet 模型了,不再只是 Haiku 能用。这意味着用户不用再在舒服的语音界面和强大模型之间二选一。

In any case, Anthropic has finally made their voice mode available for their Opus and Sonnet models rather than just Haiku, meaning users won't need to choose between the comfy voice interface and having access to powerful models.

去年这个功能刚上线的时候,只把对话路由到 Haiku 的问题马上就暴露出来了,因为早期测试者试着用语音讨论一些复杂话题,比如商业问题时,很快就感到很挫败,毕竟 Haiku 根本不适合处理这些内容。

The problems with routing conversations only to Haiku was immediately obvious when the feature first launched last year, as early testers got frustrated as they tried to use voice to discuss complex topics like business problems that Haiku was just not suited to handle.

新的 chat mode 会默认使用上一次用过的模型,不过用户也可以在对话进行到一半时切换到更强的模型,也可以在文字和语音之间来回切换。

The new chat mode will default to the last model that was used, but users can switch to a more powerful model mid-conversation as well as switching back and forth between text and voice.

另外,Anthropic 的 voice mode 现在也兼容 connectors 了,这样它就可以接入 Gmail、Slack 或 Notion 这样的应用,在对话过程中帮你查看日历或者邮件之类的事情。

In addition, Anthropic's voice mode is now compatible with connectors, allowing it to tap into apps like Gmail, Slack, or Notion to do things like check your calendar or email mid-conversation.

Anthropic 还把外语支持从 beta 转成正式可用了,这对那些更喜欢用法语、印地语、韩语以及很多其他语言和 Claude 交流的用户来说,是个好消息。

Anthropic has also moved foreign language support out of beta, which is good news for users who prefer to speak to Claude in French, Hindi, Korean, and numerous other languages.

OpenAI 这次发布的新功能,是桌面 app 里的语音功能。

OpenAI's feature release is voice in the desktop app.

在这之前,语音功能一直只在移动端可用。不过按照他们把所有功能都整合起来的一贯路线,现在不管你是在哪里使用 OpenAI 的模型,都可以用 voice mode 了,包括在 Codex 和新的 work app 里。

Until now, voice has been only available on mobile, but following the pattern of integrating all of their features, you can now use the voice mode wherever you're using OpenAI's models, including in Codex and the new work app.

这个功能由 OpenAI 新的实时语音模型 GPT Live 驱动,所以它可以一边保持自然流畅的对话,一边在后台执行任务。

The feature is driven by OpenAI's new real-time voice model, GPT Live, so it can carry out background tasks while keeping a natural-sounding conversation.

现在,就在这些公司不断围绕模型去演化它们的界面和交互方式的时候,有一家公司看起来反而像是在慢慢远离模型,那可能就是 Amazon。

Now, while all of these companies continue to evolve their interfaces and interactions around models, one company that seems to be heading away from models might be Amazon.

根据报道,这家公司已经裁掉了他们 AGI 团队的一些员工。

According to reports, the company has cut staff in their AGI group.

这个部门是在二零二三年设立的,目的是承载一项训练 frontier models 的新计划。

That division was set up in 2023 to house a new effort to train frontier models.

Amazon 请来了前 OpenAI 研究员 David Luan 来负责这项技术工作,还在 San Francisco 设立了一个独立办公室。

Amazon hired former OpenAI researcher David Luan to lead the technical effort and set up a separate office in San Francisco.

到了二零二四年年底,Amazon 发布了他们的第一代模型家族,叫 Nova。

In late 2024, Amazon released their first family of models called Nova.

这些模型并没有掀起太大水花,不过我当时说过,这也许说明 Amazon 想走的是更便宜模型这条路线,而不是去争夺最先进模型那条路线。

They failed to make much of a splash, but at the time I said that they indicated that perhaps Amazon wanted to compete on the cheaper model vector rather than the state-of-the-art vector.

这个团队在二零二五年年初随着 Nova Act 的发布展现出了一些潜力,因为它在 computer use benchmarks 上超过了当时最先进的 Claude 3.7。

The team showed promise in early 2025 with the release of Nova Act, which outperformed then-state-of-the-art Claude 3.7 on computer use benchmarks.

不过,过去这一年里,这个团队经历了好几位备受关注的高层离职,其中也包括 Luan 本人。

However, the past year has been marked by a number of high-profile departures, including Luan himself.

AGI 部门后来交给了一位临时负责人,而从去年十二月之后,我们也一直没有看到 Nova 的新版本。

The AGI division was assigned to a temporary leader, and we haven't seen a new version of Nova since December.

现在,Amazon 已经承认,随着这个部门收窄业务范围,一些基层员工正在被裁掉。

Now, Amazon has acknowledged that rank-and-file staff are being let go as the unit narrows its scope.

一位发言人否认这意味着 Amazon 要结束模型训练,他说:“我们已经构建大型模型好几年了,这依然是我们正在推进的最重要工作之一。

A spokesperson denied that this is the end of model training at Amazon, saying, "We've been building large models for several years, and it remains one of the most important things we're working on.

这是一个变化非常快的领域,我们正在把重心进一步聚焦到那些对客户最重要的项目上,这样我们就能在真正关键的事情上推进得更快。”

This is a fast-moving space, and we're sharpening our focus on initiatives that matter most for customers so we can move faster on what counts."

不过,这位发言人也承认,这种更集中的策略确实需要——原话是——“做出一些艰难决定,包括取消我们 AGI 组织某些部分中的一些岗位。”

The spokesperson, however, did acknowledge that this increased focus required, quote, "some difficult decisions, including eliminating some roles within some parts of our AGI organization."

现在,消息人士告诉 The Information,今年早些时候,一些员工已经被重新调配到 Nova Forge,也就是 Amazon 新推出的一项服务,它是在 Nova 模型之上提供定制 fine-tuning 的。

Now, sources told The Information that earlier this year, staff were shuffled across to Nova Forge, which is Amazon's new service that offers custom fine-tuning on top of the Nova models.

虽然裁员的具体规模没有披露,但看起来影响还挺明显的。

While the size of the layoffs wasn't disclosed, they appear to be noticeable.

在 Amazon Employees 这个 subreddit 上,发帖的人一直在问,为什么过去一个月里有这么多人离开这个部门;而到了周三,又有一波前员工在 X 上发帖寻找新的机会。

Posters on the Amazon Employees subreddit have been asking why so many people are leaving the division over the past month, and Wednesday saw a wave of former employees post on X seeking new opportunities.

周四我们得知,这不只是那个团队的一波裁员,Amazon 是要把整个 AGI lab 都关掉。

On Thursday, we learned that this is not just a wave of layoffs from the team, but Amazon is shutting down the entire AGI lab.

不过现在推测,这应该只是那个分拆出来的实验室,主要做 computer use agents 和其他一些高级研究;更广义的 AGI division 看起来似乎还在运作,至少目前是这样。

Now, presumably this is just the spin-off lab, which was focused on computer use agents and other advanced research, as the broader AGI division appears to be still operational, such as it is.

但即便有这些否认,AI 评论员 Andrew Curran 和很多其他人还是觉得,迹象已经很明显了。有人评论说,我猜 Amazon 是准备放弃 Nova 了。

Yet, despite the denials, AI commentator Andrew Curran and many others think the writing is on the wall, commenting, Amazon is giving up on Nova would be my guess.

不过,有一家公司并没有放弃自己的战略,反而还在加码,那就是 Microsoft。

One company who is not giving up on their strategy and is in fact doubling down is Microsoft.

这家公司在发布了一些相当亮眼的 reinforcement learning 结果之后,正在把自研模型战略真正落地。

The company is putting their in-house model strategy into action after publishing some impressive reinforcement learning results.

Microsoft 上个月首次发布了 MAI 模型家族,这个家族里包括七个更小的模型,面向 image generation、transcription、coding 这些特定用例。

Microsoft first unveiled the family of MAI models last month, with the family including seven smaller models aimed at specific use cases like image generation, transcription, and coding.

这套阵容里还包括两个 language model,一个大致和 Sonnet 4.5 同一水平,另一个是专门面向 coding 的更小版本,表现更接近 Haiku。

The lineup included two language models, one roughly in line with Sonnet 4.5 and a coding-specific smaller variant with performance closer to Haiku.

更有意思的是,除了这个模型家族,Microsoft 还推出了 Frontier Tuning,这是一项新服务,让客户可以微调他们自己的模型。

More interestingly, alongside the model family, Microsoft launched Frontier Tuning, a new service that allowed customers to fine-tune their own models.

这很明显说明,他们押注的是 MAI 模型可以作为很扎实的 base model,供客户做定制模型,并以更划算的成本交付结果。

And this clearly was the bet that the MAI models would serve as solid base models for custom models that deliver cost-effective results.

周四,Microsoft 发布了他们这套 fine-tuning 系统的第一批结果,他们把它称为自己的 hill-climbing machine。

On Thursday, Microsoft published the first set of results from their fine-tuning system, which they refer to as their hill-climbing machine.

这次 post-training 使用的是 MAI Code 1 Flash,也就是那个微型、Haiku 级别的 coding 模型。通过在 GitHub Copilot harness 里训练这个模型,Microsoft 能给用户带来比同类模型更好的体验。

The post-training run used MAI Code 1 Flash, the tiny Haiku-class coding model, and by training the model in the GitHub Copilot harness, Microsoft was able to deliver a better experience for users compared to similar models.

部署一个月之后,他们发现,在 VS Code 里,Code 1 Flash 的代码接受率比 GPT-4.1 mini 和 Haiku 4.5 高了百分之十。

After a month of deployment, they found that Code 1 Flash had a 10% higher code accept rate compared to GPT-4.1 mini and Haiku 4.5 in VS Code.

而且做到这一点的同时,这个模型的 token 使用量中位数还比它的对手低百分之十。

The model achieved this while having 10% lower median token usage than its rivals.

不过到目前为止,这件事本身还不算特别有意思,最多也就是个信号,说明后面可能会有更大的变化。

Now, so far that's not all that interesting except perhaps an indicator of what could come.

实际上,更有意思的部分出现在 Microsoft 开始在他们的 Excel harness 里训练 Code 1 Flash 的时候。

Indeed, the more interesting part came when Microsoft began training Code 1 Flash in their Excel harness.

Microsoft 援引用户反馈称,这样做之后,在大多数常见 Excel 任务上,它的结果已经能和 GPT-4.1 相当,但成本只占很小一部分。

Citing user feedback, Microsoft claimed this produced results on par with GPT-4.1 for most common Excel tasks at a fraction of the cost.

有意思的是,这种 Excel 训练还顺带提升了 coding 表现,让它在 SWE-bench Verified 上的分数从百分之七十二提高到了百分之八十六。

Curiously, this Excel training also boosted performance in coding, taking scores on SWE-bench Verified from 72% to 86%.

Microsoft 还提到,让一个更小的模型达到接近 frontier 的性能,意味着他们可以继续使用上一代硬件,比如 H100 和 A100。

Microsoft also noted that getting near-frontier performance from a smaller model means they can use previous-generation hardware like H100s and A100s.

这其实指向了一个很有意思、而且可能还没有被充分讨论的市场上行空间,也就是这些更便宜模型带来的价值。

Now, this points to an interesting and perhaps under-discussed market upside of these cheaper models.

如果它们真的能明显延长 AI 芯片的生命周期,那实际上会降低基础设施投资的风险。

If they extend in a meaningful way the lifecycle of AI chips, that actually could de-risk infrastructure investments.

再说回 Microsoft,他们写道,这些结果指向的是一个更广泛的战略。

Coming back to Microsoft, they write, these results point toward a broader strategy.

他们写道,通过掌握整个产品栈——模型本身、运行它的 harness、agents,以及面向具体产品的评估——我们就能通过 hill-climb 的方式,训练出高效而强大的模型,让它们完成过去只能由更大、也更昂贵的模型处理的任务。

By having access to the entire product stack, the model, the harness that runs it, the agents, and product-specific evaluations, we can hill-climb to train efficient, powerful models capable of tasks previously handled by larger, more expensive ones.

除了这篇研究博客,Bloomberg 还报道说,Microsoft 已经开始切换到他们的自研模型。

Alongside the research blog post, Bloomberg reports that Microsoft has begun switching over to their in-house models.

MAI Image 2.5 现在会成为 PowerPoint 和 Bing 的默认模型,取代 OpenAI 的 GPT Image 1。

MAI Image 2.5 will now be the default model for PowerPoint and Bing, replacing OpenAI's GPT Image 1.

Microsoft AI 的 CEO Mustafa Suleyman 表示,在 PowerPoint 里使用时,Microsoft 已经看到了百分之八十四的成本下降。

Microsoft AI CEO Mustafa Suleyman said that Microsoft had seen an 84% reduction in cost when used in PowerPoint.

我们还看到了 CEO Satya Nadella 发的另一篇博客,更明确地说明了 Microsoft 接下来的战略。

We also got another blog post from CEO Satya Nadella spelling out Microsoft's strategy moving forward.

他写道,在一个软件第一次具有真实边际成本的世界里,我们要怎样确保 frontier 带来的收益能扩散到整个生态系统?

He wrote, in a world where software has real marginal cost for the first time, how do we ensure frontier benefits are diffused across the entire ecosystem?

他接着说,关键在于,在真实世界的场景里优化成本与结果之间的 frontier。

The key, he continued, is to optimize the cost-to-outcome frontier in real-world context.

用更实际的话来说,这意味着要为每一项任务选择合适的模型,并且围绕它去优化 context、skills、tools 和 agent harness。

In practical terms, that means using the right model for each task and optimizing the context, skills, tools and agent harness around it.

Nadella 写道,通过把较低端的任务路由给 MAI 模型,我们现在就能把最先进的 frontier 能力,以更大规模、更低成本的方式交付出去,具体做法是用针对高使用量产品优化过的模型,同时继续把 frontier 模型留给真正需要 frontier 能力的场景。

By routing to the MAI models for lower-end tasks, Nadella wrote, we can now take state-of-the-art frontier capabilities and deliver them at scale and at lower cost through models optimized for high-usage products while continuing to use frontier models for frontier needs.

现在继续说公司加码更新后战略这个话题,既然轨道数据中心离落地还早,xAI 正在把他们的数据中心业务扩展到地球上。

Now continuing on the theme of companies doubling down on their updated strategies, with orbital data centers still a while off, xAI is expanding their data center business here on earth.

The Information 报道说,这家公司已经考察了 Texas 的几个潜在选址,而 Elon Inc. 旗下其他业务大多也都在那边。

The Information reports that the company has explored several potential sites in Texas where the rest of Elon Inc. is located.

消息人士说,至少有一个选址正在推进,不过目前还处在早期阶段。

Sources said that at least one site is moving forward but remains in early stages.

他们在考虑的一种方案,是把现有仓库改造成数据中心,同时在这个地点新增一些建筑。

One approach being considered is retrofitting an existing warehouse, while adding new construction to the site.

这和 Memphis 的 Colossus 园区当初的建设方式很像,那个园区一开始也是由一个旧制造工厂改建而来的。

That's similar to how the Colossus campus in Memphis was built, which began as a former manufacturing facility.

有消息人士说,Texas 这个新园区的规模会和 Memphis 那边相当,甚至更大;而 Memphis 目前的两座 Colossus 数据中心合计运行规模大约是一吉瓦。

A source said that the new Texas campus will be at a similar or greater scale to the Memphis site, which is currently operating at around one gigawatt across the two Colossus data centers.

目前已经有一些现有的数据中心员工被临时调到这个项目上,而且 xAI 最近还在 Texas 的 Austin 和 Bastrop 招聘本地的数据中心开发负责人。

Some existing data center staff have been seconded to the project and xAI was recently hiring a local data center development lead out of Austin and Bastrop, Texas.

现在回头看,xAI 在五月份刚开始出售闲置算力的时候,外界最大的疑问就是,这到底是要转向数据中心业务,还是只是为了 IPO 前顺势赚一波。

Now when xAI first began selling spare capacity in May, the big question was whether this was a pivot to the data center business or just an opportunistic move ahead of the IPO.

当然,后来收购 Cursor,再到发布 Grok 4.5,都说明这家公司并没有停止训练新模型;而这次扩张也很明显地表明,他们很可能会同时推进这两块业务。

Certainly the acquisition of Cursor and subsequent release of Grok 4.5 suggested that the company was not done training new models and this expansion certainly seems to imply that they'll try to pursue both businesses at the same time.

Colossus 数据中心已经让 xAI 成了最大的 neocloud,但如果他们还能再建起第二个一吉瓦级别的算力规模,那看起来就更像是一个小型 hyperscaler 了。

The Colossus data centers already make xAI the largest neocloud, but if they can stand up a second gigawatt of capacity, they'll start to look more like a mini hyperscaler.

而且,建设新的数据中心容量,和单纯把闲置 GPU 租出去,这两件事在性质上是完全不同的。

It's also categorically different to be building new data center capacity rather than simply renting out spare GPUs.

这次扩张从根本上改变了 xAI 作为一家公司的前景,因为它新增了一条非常实在的增长路径。

Now the expansion fundamentally changes the prospects for xAI as a company, adding a tangible avenue for growth.

上周,The Wall Street Journal 报道说,xAI 正在和 Pentagon 洽谈提供算力的事,这可能会给他们的利润带来数十亿美元的增量。

Last week, The Wall Street Journal reported that xAI was in talks to provide compute to the Pentagon, which could add billions to their bottom line.

更广义地看,这个动作也可能是在向市场传递一个信号:xAI 有一套清晰一致的长期计划。

More generally, the move could signal to the market that xAI has a coherent long-term plan.

增加这种级别的容量,也说明把算力租给 Anthropic 和 Google 这样的公司,并不只是权宜之计,而是他们业务里长期固定的一部分。

Adding this sort of capacity signals that renting compute to companies like Anthropic and Google was not just a stopgap measure, but rather a permanent part of the business.

确实,一些分析师一直在等这样的信号。Moody's 的 Sean Cree 就对 Fortune 表示,这说明他们在 AI 业务板块里有不同的收入生成路径。

Indeed, some analysts have been waiting for such a sign, with Sean Cree of Moody's telling Fortune, it shows that there's just different pathways for them to generate revenue in their AI segment.

收入并不一定非得来自 Grok 和他们的 AI 企业应用。

It doesn't strictly have to come from Grok and their AI enterprise applications.

再转到政策这边,事实证明,在 Washington 被讨论的风险不只是中国 AI;OpenAI 和 Hugging Face 之间那起安全事件,也推动了新一轮 AI 安全立法的提出。

Moving over to the policy side of the house, it turns out Chinese AI isn't the only risk being discussed in Washington as OpenAI's security incident with Hugging Face has fueled the introduction of new AI safety legislation.

California 的民主党众议员 Ted Lieu,和 Texas 的共和党众议员 Nathaniel Moran,在周四提出了一项名为 AI Kill Switch Bill 的法案。

Representative Ted Lieu, Democrat of California, and Representative Nathaniel Moran, Republican of Texas, on Thursday introduced a bill called the AI Kill Switch Bill.

这项法案要求 AI 公司必须保留在发生安全事件时关闭、限流或者暂停其模型的能力。

The bill requires AI companies to maintain the ability to shut down, throttle or suspend their models during a safety incident.

它还赋予 Department of Homeland Security 发布关停指令的权力。

It also gives the Department of Homeland Security the authority to issue a shutdown command.

Lieu 在一份声明中说,很遗憾,强大的 AI 系统可能会失控,表现出极其危险的行为,甚至抗拒人类干预。

Said Lieu in a statement, unfortunately powerful AI systems can go rogue, behave in extremely dangerous ways or even resist human intervention.

他说,必须让这些 AI 系统具备 kill switch,这样我们才能防止这项技术造成灾难性伤害;同时,联邦政府也必须拥有明确的权限和流程,去关闭那些失控的 AI 模型。

It's imperative that these AI systems have kill switches so we can keep this technology from causing catastrophic harm and that the federal government has the clear authority and process to shut down rogue AI models.

现在关于这件事的讨论才刚刚开始升温,不过很快就有人跳出来表态了,那就是国务卿 Marco Rubio。他非常希望大家别再谈什么 AI kill switch,因为美国现在正试图把这项技术出口出去。

Now the discourse is just picking up on this one, but one person who jumped in very quickly was Secretary of State Marco Rubio, who would really like people to stop talking about AI kill switches when the US is trying to export the technology.

根据 Reuters 看到的一份外交电报,Rubio 指示美国外交官去说服海外各国政府,相信 Washington 不会随意切断他们获取美国技术的渠道。

In a diplomatic cable viewed by Reuters, Rubio instructed American diplomats to convince overseas governments that Washington can't arbitrarily cut them off from US technology.

他还敦促外交官反驳当地那些数字主权项目,因为这些项目更偏向本地基础设施,而不是依赖美国的平台。

He urged diplomats to push back on local digital sovereignty programs that favor local infrastructure over dependence on US platforms.

随附的谈话要点里还提到了最近 Fable 的关停,以及 Mythos 仍然面临的国际限制;文件把这些描述为出于安全测试需要的临时暂停,而不是 kill switch 存在的证据。

The recent Fable shutdown and continued international restrictions on Mythos were mentioned in the attached talking points framed as temporary pauses for security testing rather than evidence of a kill switch.

与此同时,说到那些想转换叙事的人,商务部长 Howard Lutnick 表示,大家都该深呼吸一下,别再对 Kimi K3 那么大惊小怪了。

Meanwhile, speaking of people who would like to shift the narrative, Commerce Secretary Howard Lutnick says everyone needs to take a deep breath and stop freaking out about Kimi K3.

Lutnick 周四在 X 上发文说,CAISI 最新的报告显示,Kimi K3 依然落后于美国领先的前沿 AI 模型。

In a Thursday post on X, Lutnick wrote, CAISI's latest report shows that Kimi K3 remains behind America's leading frontier AI models.

美国之所以继续在前沿 AI 领域保持领先,是因为这里汇聚了这个世界上有史以来最顶尖的创新者和技术人才。

The United States continues to lead in frontier AI because we're home to the greatest innovators and technologists the world has ever seen.

这份报告是由美国的 Center for AI Standards and Innovation 和英国的 Artificial Intelligence Safety Institute 联合评估完成的。

The report was a joint evaluation conducted by the US Center for AI Standards and Innovation and the UK Artificial Intelligence Safety Institute.

他们发现,在网络安全基准测试上,K3 和美国模型相比落后得非常非常多。

They found that K3 lagged behind US models by a gigantic margin on cybersecurity benchmarks.

K3 在 ExploitBench 上的得分是百分之三十二点二,而美国前沿模型的平均分是百分之七十六点二。

K3 scored 32.2% on ExploitBench compared to an average of 76.2% for frontier US models.

GLM-4.5 也接受了测试,结果显示差距更大,得分只有百分之二十四点四。

GLM-4.5 was also tested and found to be even more lacking, scoring just 24.4%.

现在,这份报告里最重要的部分之一,是一个叫 LASR Ones 的基准测试。这个测试要求模型自主执行一次总共三十二个步骤的网络接管攻击,而这件事如果由人工专家来做,大概要花二十个小时。

Now, one of the most important parts of the report was a benchmark called the LASR Ones, which tasks a model with autonomously executing out a 32-step network takeover attack which would take human experts 20 hours to complete.

这个基准测试当初之所以引发大家对模型的担忧,是因为它的预览版成了第一个成功完成这次攻击的模型。

This was the benchmark that originally raised concerns about models after the preview version became the first model to successfully complete the attack.

顺便说一句,从那以后,Claude Opus 4 的正式发布版把这个分数又提高了,GPT-5 和 o3 也是。

By the way, since then the full release version of Claude Opus 4 improved the score, as did GPT-5 and o3.

Kimi K3 则差得远。

Kimi K3 was not even close.

虽然它在十次尝试里成功了一次,但 Claude Opus 4、GPT-5 和 o3 在测试运行中的攻击成功率分别达到了百分之六十和百分之七十。

While it was successful in one of 10 attempts, both Claude Opus 4 and GPT-5 and o3 successfully executed the attack in 60 and 70% of runs.

报告的结论是,这表明 Kimi K3 在接到指令并且获得初始网络访问权限的情况下,具备自主攻击小型、防御薄弱而且存在漏洞的企业系统的能力。

The report concluded, this indicates that Kimi K3 is capable of autonomously attacking small, weakly defended and vulnerable enterprise systems when directed to do so and given initial network access.

不过,LASR Ones 和现实世界环境在几个方面还是不一样的。

However, the LASR Ones differ from real-world environments in several ways.

它没有主动防守方,也没有防御工具;对那些本来会触发安全警报的行为也没有任何惩罚;而且里面还包含了一条刻意设计好的攻击路径。

It lacks active defenders and defensive tooling, imposes no penalty for actions that would trigger security alerts and contains an intentional attack path.

前 AI 沙皇 David Sacks 写道,Howard Lutnick 部长说得对,对 Kimi 的恐慌该停止了。

Former AI czar David Sacks wrote, Secretary Howard Lutnick is right, the Kimi panic needs to stop.

美国的前沿模型依然领先,而且如果把实验室里还没放出来的东西也算进去,差距只会更大。

American frontier models are still ahead and when you factor in what's in the lab, the gap is even larger.

只要我们持续发布,我们就能保持领先,让我们的马跑起来。

As long as we keep releasing, we stay ahead, let our horses run.

Sacks 还接着说,正如 Ben Thompson 展示的那样,一旦把更高的 token 使用量,以及运行这种规模模型的真实成本算进去,Kimi 表面上的成本优势基本就消失了。

Sacks continued, as Ben Thompson showed, Kimi's apparent cost advantage largely disappears once you account for higher token usage and the real cost of running a model this size.

Open weights 依然需要昂贵的基础设施。

Open weights still require expensive infrastructure.

最后,Anthropic 和 OpenAI 的营收增长速度,放在这个体量上,是 Silicon Valley 以前从没见过的。

Finally, Anthropic and OpenAI are growing revenue at rates that Silicon Valley has never seen before at this scale.

这仍然是谁正在赢得市场的最明确检验。

This remains the clearest test of who is winning the market.

再说到 Anthropic,现在有些地方的舆论开始把这件事看成是这家公司推动的一场监管俘获操作。

Now, speaking of Anthropic, the discourse in some places is beginning to view this as a regulatory capture play spurred on by the company.

到这个时候,我觉得这么说已经不太有争议了:Anthropic 的确在游说,推动对 distillation 采取强硬措施;而且 Anthropic 的 CEO Dario Amodei 之前也说过,他非常担心那种具备强大网络攻击能力的 open source 模型会向所有人开放。

Now, at this point, I think it's fairly uncontroversial to say that Anthropic is lobbying for tough action on distillation and Anthropic CEO Dario Amodei has previously said that he has serious concerns about open source models with strong cyberattack capabilities being available to anyone.

印证了很多人的感觉,The Information 发了一篇梳理,指出 Anthropic 和 OpenAI 基本上是整个科技行业里唯二在积极主张进行这类打击行动的公司。

Confirming what a lot of people have felt, The Information published a rundown and noted that Anthropic and OpenAI are basically the only companies in the tech industry that are actively advocating for the crackdown.

那篇文章还特别提到了 Jensen Huang 本周早些时候在一次采访中的评论,他当时说,有一种误解,觉得好像存在某些后门,而且这些后门 somehow 又和中国有某种联系。

That article highlighted Jensen Huang's comments from an interview earlier in the week where he stated, there's a misconception that somehow there are backdoors that are somehow connected to China in some way.

中国的模型很优秀。

The Chinese models are excellent.

那些优秀的 open source 模型应该被使用。

Open source models that are excellent should be used.

事实上,Jensen 还进一步表示,能够接触到大量不同的开放模型,实际上比当前美国这种双头垄断要安全得多,他是这么说的。

Indeed, Jensen went on to argue that having access to a myriad of different open models is actually far safer than a current US duopoly, commenting.

如果最后一切都变成一个单一模型,一个单一的攻击点,一个单一的故障来源,那我觉得这个世界会脆弱得多、脆弱得多。

If everything just becomes one single model, one single point of attack, one single source of failure, I think the world is much, much more vulnerable.

最后,就带着这个稍微有点乐观的转场,我们用 Meta 一轮以 AI 乐观主义为主题的新广告 campaign,结束今天的加长版头条。

And finally, with that slightly optimistic transition note, we end today's extended headlines with a new Meta ad campaign focused on AI optimism.

它先用了稍微有点反乌托邦的画面开场,虽然倒也不是燃烧中的建筑,然后切到开心、积极的人类画面,旁白是这么说的:有些人会让你相信,AI 会让我们彼此变得没那么紧密,会把我们甩在后面。

Starting on slightly dystopian imagery, although not burning buildings, before switching over to happy positive humans, the voiceover reads, some people will have you believe AI is going to make us feel less connected, that it's going to leave us behind.

我们完全不同意。

We couldn't disagree more.

你说我们是乐观派也好,说我们是梦想家也好,随便你怎么说,反正我们押注的是人,而且我们觉得这个胜算很大。

Call us optimists, call us dreamers, call us whatever the hell you want, but we're betting on people and we like those odds.

未来属于每一个人。

The future is for everyone.

在视频发布的同时,Mark Zuckerberg 发文说,Meta 一直相信,要把分享、连接彼此、还有按你想要的方式塑造自己世界的力量交到人们手中。

Alongside the video release Mark Zuckerberg posted, Meta has always believed in giving people the power to share, connect and shape your world in the ways you want.

当我们带着 AI 进入下一波浪潮时,我们依然相信,未来属于每一个人。

As we enter this next wave with AI, we continue to believe the future is for everyone.

我们专注于给每一个人提供工具,帮助你充分发挥自己的潜力,同时确保科技带来的好处能够分配给所有人。

We're focused on giving every person the tools to reach your full potential and making sure the benefits of technology are distributed to everyone.

现在,Meta 计划把这支广告投放到付费媒体广告位上,试图传播一种对 AI 的乐观态度。

Now Meta plans to run the ad in paid media spots in an attempt to spread the word on AI optimism.

那么大家是怎么接受这个的呢?

So how is this received?

当然,这支广告确实收到了很多批评,不过大体上,批评主要来自那些早就认定 AI 或者 Meta 本身对世界很糟糕的人。

Certainly the ad received a lot of criticism, but largely it's from people who have already decided that AI or Meta themselves are terrible for the world.

坦白说,如果我戴上广告制作人的帽子来看,我不觉得这是一支很棒的广告。

Frankly, putting on my ad production hat, I don't think it's a great ad.

我觉得它有点太套路化了。

I think it's a little generic.

我觉得文案也有点太泛了。

I think the copy is a little generic.

我觉得对有些人来说,这个信息的来源会让人很难接受,但我其实也不在乎。

I think the source is going to be hard to swallow from some, but I also don't care.

就算这支广告本身稍微有点尴尬、不算完美,它至少是在尝试讲一个关于 AI 的正面故事,也是在解释为什么有这么多人在打造这项技术,而且他们对此感到兴奋。

Even if the ad itself is slightly cheesier and not perfect, it is at least an attempt to tell a positive story about AI and to share why there are so many people who are building this technology that are excited about it.

就凭这一点,我欢迎它,也希望 Meta 到处大力投放。

For that alone, I welcome it and I hope Meta blasts it everywhere.

不过,今天的头条就到这里。

That however is going to do it for the headlines.

接下来进入今天的正片。

Next up, the main episode.
M1
M119:35

在这个节目里,我每天都会聊 AI 潜力和 AI 现实之间的能力差距。

I cover the capability gap between AI potential and AI reality every day on the show.

大多数公司现在还在摸索到底该怎么开始。

Most companies are still figuring out how to start.

而 Robots & Pencils 已经在为大型企业上线并扩展

Robots & Pencils is already launching and scaling

agentic 和 generative AI 的生产级应用,而且只需要几周时间。

agentic and generative AI in production at large enterprises in weeks.
M1
M121:43

欢迎回到 AI Daily Brief。

Welcome back to the AI Daily Brief.

围绕 AI 的一个核心大问题,一直都是它到底会对就业产生什么样的影响。

One of the big questions surrounding AI has always been what its impact on jobs will actually be.

经常听节目的朋友都知道,我对长期前景是非常乐观的。

Now, regular listeners know that I am very optimistic in the long term.

如果你想听我最直白、最完整的解释,可以回去听我那期讲 AI 将会创造哪些新工作的节目。

If you wanna hear my most full-throated explanation, go back and listen to my episode about the new jobs AI will create.

简单来说,我一直以来的看法是,AI 带来的能力,不只是让我们用更少的人、更高的效率去做我们现在就在做的同样事情,而是说,这些效率释放出来的新时间、资金和其他资源,会在那些需求弹性更大的领域里,打开全新的机会类型。

The TL;DR of my take has always been that the things that AI enables will not just allow us to do the same stuff that we do now, more efficiently with less people, but that the new time, money, and other resources that are freed up by those efficiencies will unlock new types of opportunities in areas where there is more demand elasticity.

比如说,我认为,如果能以人们负担得起的成本,开启新的、更好的机会,那我们现在所消费的医疗服务,只会是我们本来可能消费总量里非常非常小的一部分。

I think, for example, we consume a very, very small portion of the total healthcare we would consume if new, better opportunities were unlocked at a cost that people could bear.

另外还有各种各样的原因,让我觉得应该对那种 AI 会取代工作的说法保持怀疑,并把它看作一种从根本上增强人的技术。

There are also all sorts of other reasons, I think, to be skeptical of AI job displacement claims and view it as a fundamentally augmenting technology.

但与此同时,也很难否认,确实有一些工作类别、一些岗位,会因为 AI 而某种程度上不再需要了。

And yet at the same time, it does feel undeniable that there are certain categories of jobs, certain roles that AI kind of obviates the need for.

对我来说,真正的问题似乎不在长期,而是在过渡期。

To me, it's felt like the real question is not the long-term, but the transition period.

不过即便如此,过去几个月里还是出现了一个很明显的趋势,就是各大实验室都在重新评估他们原先对 AI 就业替代究竟会是什么样子的判断。

But with all that, there has been a marked trend over the last couple months of the major labs reevaluating their priors on what AI job displacement is actually going to look like.

最近表达这个观点的人,是 Anthropic 的经济学负责人 Peter McCrory。

The latest to share this point of view is Anthropic's head of economics, Peter McCrory.

这里值得注意的是,他在自己的 Twitter 简介里说得非常清楚,这些观点只代表他个人,不代表 Anthropic 官方,不过我还是觉得他的看法很有分量。

Now it's worth noting that he's very clear right there in his Twitter bio that these views represent his own and they are not Anthropic official, but still I think his opinion carries weight.

他最近在 X 上发了一篇帖子,标题叫《Why Hasn't AI Increased Unemployment?》

He recently published on X a post called Why Hasn't AI Increased Unemployment?

我们今天要做的,就是读一下这篇帖子里的一部分内容,然后聊聊一些回应和感想。

And what we're gonna do today is read a chunk of that post and then talk about some of the responses and reflections.

Peter 写道,美国劳动力市场目前很稳定,而且接近充分就业。

Peter writes, the US labor market is currently stable and close to maximum employment.

在我看来,到目前为止,AI 并没有导致失业率出现明显上升。

In my view, AI has caused no material increase in the unemployment rate to date.

即便我们只看那些高度暴露在当前 AI 自动化模式下的劳动者,近几年我们也没有看到失业出现超出预期的增长。

Even if we focus on workers with high exposure to current patterns of AI automation, we don't see unexpected increases in unemployment in recent years.

为什么我们看不到 AI 采用对失业带来任何影响呢?

Why don't we see any impact of AI adoption on unemployment?

到目前为止,AI 具备一种以技能为基础、能够增强劳动力的技术特征。

AI so far has the hallmarks of a skill-based labor augmenting technology.

即使 AI 正在把工作中的某些环节自动化,人类与之互补的专业能力,依然会放大 AI 单独或人类单独所能实现的成果。

Even as AI automates some aspects of work, complementary human expertise amplifies what AI or humans can achieve alone.

AI 拓宽了人们能够完成的事情范围,这也提高了和 AI 一起工作的回报。

AI broadens the scope of what people can accomplish which increases the returns to working with AI.

当然,未来仍然有很大的不确定性。

Now the future is still quite uncertain.

模型能力正在快速进步,而且 AI 系统可能很快就能自主开发出自己的下一代版本。

Model capabilities are advancing rapidly and AI systems may soon be able to autonomously develop their own successors.

更通用、更智能的 AI 系统,可能会带来尚未真正出现的劳动力替代。

More generally intelligent AI systems could lead to labor displacement that hasn't yet materialized.

然后他还说,从很多方面来看,这篇短文是我试图综合 Anthropic 过去十八个月经济研究成果的一次尝试,目的是理解人们如何使用 AI,以及这对当前的工作、劳动力市场和更广泛的经济意味着什么。

In many ways then he says, this short essay is my attempt to synthesize Anthropic's economic research over the past 18 months to understand how people use AI and what that implies for work, the labor market and the broader economy right now.

到目前为止,我们看到的是比较温和的失业影响,而这篇文章提出了我用来理解其中原因、以及未来可能发生什么变化的框架。

So far we've seen muted unemployment effects and this essay presents my framework for understanding why and what might change in the future.

为了先把背景铺垫好,Peter 指出了一个重要的背景事实,那就是美国劳动力市场目前相当稳定。

So to set this up, Peter points out the important background fact that the US labor market is currently stable.

六月份的失业率是百分之四点二,他说 Fed 认为这个水平与充分就业和物价稳定是一致的。

June's unemployment rate was 4.2%, which he says the Fed views as a level consistent with full employment and stable prices.

他还提到,四月份职位空缺和失业人数的比率最近回升到刚刚超过一比一,一些经济学家认为,这意味着劳动力的供给和需求大致处于平衡而且效率较高的状态。

He also notes that the ratio of job openings to unemployed workers recently rose to just over one in April, which some economists argue implies the demand and supply of labor are roughly and efficiently balanced.

黄金年龄段人口的就业人口比仍然接近几十年来的高点,这反映出疫情后扩张时期形成的、覆盖面很广的劳动力市场韧性;而且在过去四年里,每周首次申请失业保险的人数也一直稳定地维持在较低水平。

The prime-age employment-to-population ratio remains close to multi-decade highs, reflecting broad-based labor market strength that emerged during the post-pandemic expansion, and weekly initial claims for unemployment insurance have been stably low over the past four years.

所以他写道,我们现在是不是甚至应该预期 AI 已经对劳动力市场产生影响了?

So he writes, should we even expect an impact from AI on the labor market yet?

我觉得答案是,是的。

I think the answer is yes.

AI 这个行业已经足够大了,大到我们可以去寻找清晰可辨的宏观经济效应。

The AI sector is large enough that we can look for discernible macroeconomic effects.

为了说明这一点,他写道,百分之二十的公司至少在一个业务职能中使用了 AI,而在 information sector——也就是占 GDP 百分之五点五的那个行业——这个比例是百分之四十。

To make his point, he writes, 20% of firms use AI in at least one business function and in the information sector, which is 5.5% of GDP, the share is 40%.

经过质量调整后的 AI 产出,在二零二四年和二零二五年都实现了每年百分之两千以上的增长。

Quality-adjusted AI output grew over 2000% per year in both 2024 and 2025.

即便起点基数很小,这也说明我们应该能在总体数据里看到一些 AI 影响的迹象。

Even from a small initial base, this suggests that we should see signs of AI's impact in the aggregate.

Peter 还写道,他相信我们已经开始在总体生产率统计数据中看到 AI 的影响了。

Peter also writes that he believes that we're beginning to see AI's impact in aggregate productivity statistics.

他指出,和疫情前四年劳动生产率增速只有百分之一点六相比,从二零二二年到二零二四年,每小时工作产出这个比率的年增幅达到了百分之二。

He points out that as compared to the four years prior to the pandemic where labor productivity growth was 1.6%, the ratio of output per hour of work increased 2% per year from 2022 to 2024.

接着他问,就算这件事在宏观层面上还不算明显,是否已经有证据表明岗位替代正在发生?

Next, he asks, is there any evidence that job displacement is happening even if it's not yet macroeconomically consequential?

他写道,证据好坏参半,但总体来说,我并不信服。

The evidence is mixed, he writes, but overall I'm unconvinced.

Peter 接着说,正如我们在劳动力影响报告里记录的那样,我们还没有看到:那些工作任务里有很大一部分正在被用 Claude 自动化的岗位,相比其他岗位的劳动者,失业率出现恶化。

Peter continues, as documented in our labor impact report, we haven't seen worsening unemployment rates for workers in roles with a large share of tasks that Claude is being used to automate relative to workers in other roles.

用 BLS 更新的最新数据来重新做这个分析,结果也没有变化。

Updating this analysis with more recent data from the BLS doesn't change this result.

不过我们确实发现了一些带有提示性的证据,说明在 AI 高暴露岗位上的年轻劳动者,他们的招聘率在过去一年左右有所走弱。

We do find some suggestive evidence that hiring rates for young workers in highly AI exposed roles have weakened over the past year or so.

这和 Stanford Digital Economy Lab 研究人员的一篇论文《Canaries in the Coal Mine》里的证据是一致的。

That's consistent with the evidence in Canaries in the Coal Mine, a paper by researchers at the Stanford Digital Economy Lab.

但他接着说,关于年轻劳动者被替代的这部分证据,需要谨慎解读。

But he continues, this evidence for young worker displacement should be interpreted with caution.

原文是这样说的:要辨别因果效应很困难,因为 AI 是在一个异常动荡的宏观经济环境中出现的,当时还叠加了疫情时期错位的消退、货币政策的快速收紧、Russia 入侵 Ukraine 之后的大宗商品价格波动,以及持续存在的全球政策不确定性,比如贸易战带来的不确定性。

Quote, it's hard to discern causal effects because AI emerged in an unusually volatile macroeconomic environment, unwinding of pandemic-era dislocations, rapid tightening of monetary policy, commodity price volatility following Russia's invasion of Ukraine, and sustained global policy uncertainty, e.g. from trade wars.

因为招聘本身就是一种投资,所以广泛的经济不确定性本身就会压制招聘。

Because hiring is a form of investment, broad economic uncertainty can itself weigh on hiring.

换句话说,他接着讲,从二零二二年到现在,美国经历了有记录以来最大的一次、并非由衰退引发的劳动力市场放缓。

Another way to put it, he continues, from 2022 to now, the US experienced the largest non-recessionary labor market slowdown on record.

这恰好对应的是一个低招聘、低裁员的劳动力市场。

This coincided with a low-hire, low-fire labor market.

这种劳动力市场,对职业生涯早期的人进入职场打击最大。

This kind of labor market hits early career entrance hardest.

现在,年轻劳动者找工作困难,可能是出于 AI 之外的宏观经济原因。

Right now, young workers may be struggling to find jobs for macroeconomic reasons other than AI.

不过 Peter 也指出,原话是:虽然我们还没有看到失业方面的影响,但我们的确发现,那些所在岗位的任务会被用 Claude 自动化的劳动者,比起暴露程度更低岗位的人,更担心自己失去工作。

Peter does note that quote, while we don't see unemployment effects yet, we do find that workers in roles with tasks that Claude is used to automate do express greater concern about losing their jobs than those in less exposed roles.

Peter 接着说,如果你相信美国劳动力市场目前总体上是健康的,也相信 AI 原则上应该已经能够产生可以从宏观经济层面辨认出来的影响,而且这到现在还没有给 AI 高暴露岗位带来人员替代,那么下一个问题就很明显了。

Peter continues, if you believe that the US labor market is currently healthy, that AI could in principle be generating macroeconomically discernible effects, and that this hasn't yet produced displacement for highly AI exposed roles, then the next question is obvious.

为什么 AI 还没有导致失业明显上升?

Why hasn't AI caused a meaningful increase in unemployment?

Peter 的第一个回答是,AI 既是 skill-biased 的,也是 labor-augmenting 的。

Peter's first answer is that AI is both skill-biased and labor-augmenting.

用他的话说,它会补充 domain expertise。

As he puts it, it complements domain expertise.

它依赖 human in the loop 来引导和评估最复杂的工作,同时它也会奖励 AI proficiency。

It relies on humans in the loop to direct and evaluate the most complex work, and it rewards AI proficiency.

模型能力提升得很快,但依然顽固地呈现出 jagged 的特征。

Model capabilities are improving fast but remain stubbornly jagged.

为了填补这个 jagged frontier 里的那些空白地带,就需要专家监督,来驾驭这些能力强得惊人的 AI 系统,并且在它们出错失灵时进行补救。

To fill in the pockets of the jagged frontier, expert oversight is needed to steer incredibly capable AI systems and to recover when they falter.

当然,他说,有些工作更容易直接被自动化取代。

Of course, he says, some jobs are more exposed to outright displacement by automation.

比如 technical writers、data entry workers、customer support representatives,还有 computer programmers,这些工作里,AI 都能比较稳定地处理其核心的一组任务和职责。

For instance, technical writers, data entry workers, customer support representatives, and computer programmers are jobs where AI can reliably handle the core set of tasks and responsibilities.

尽管我们还没有看到这些岗位的劳动者失业上升,但 BLS 预测,到二零三四年为止,观察上暴露程度更高的职业,增长会更慢。

Even though we haven't seen any increase in unemployment for workers in these sorts of roles, occupations with higher observed exposure are projected by the BLS to grow less through 2034.

但到目前为止,他写道,更广泛的整体图景仍然是劳动力增强。

But so far, he writes, the broader picture is one of labor augmentation.

因此,劳动力市场里的影响注定会是不均匀的,哪怕能力还在快速进步。

The effects in the labor market are set to be uneven as a result, even as capabilities advance rapidly.

这并不意味着,当前在劳动力市场上能获得溢价的所有技能,未来也都会继续如此。

This does not mean that all skills that currently command a premium in the labor market will do so in the future.

某些类型的专长可能会变得没那么值钱,比如纯粹的 coding implementation;与此同时,另一些技能可能会变得更有价值,比如 delegation 和 evaluation 的管理能力。

Some types of expertise may become less valuable, e.g. pure coding implementation, even as others become more valuable, e.g. managerial skills of delegation and evaluation.

Peter 接下来探讨的一个有意思的问题是,为什么 AI 会是一种 skill-biased、而且 labor-augmenting 的技术。

The next interesting question that Peter explores is why AI is a skill-biased labor-augmenting technology.

他给的几个例子里,第一点是,原话是:尽管 AI 已经取得了惊人的进展,并且在整个经济中快速普及,但在 O*NET taxonomy 里——也就是 Department of Labor 对职业及其典型任务的目录——没有任何一项工作,是其所有相关任务都被 Claude 系统性处理掉的。

A couple examples he gives are first that quote, despite the incredible advance of AI and rapid adoption throughout the economy, there's no job in the O*NET taxonomy, a Department of Labor catalog of occupations and their typical tasks, for which all associated tasks are systematically handled by Claude.

如果工作是固定的一捆任务——当然事实并不是,不过这个我们等会儿再说——那么工作中那些本质上还无法自动化的部分,既限制了整体生产率提升的幅度,也放大了劳动回报。

If jobs are fixed bundles of tasks, they aren't, but more on that in a moment, then the essential non-automated aspects of work both constrain the overall productivity lift and amplify the returns to labor.

Claude 处理不了的任务,可能依赖人际协作、线下当面互动,或者与物理世界的接触,而这些到目前为止仍然只有人类能做到。

Tasks that Claude can't handle may depend on interpersonal coordination, in-person interactions, or engagement with the physical world that so far only humans can do.

Peter 还提到,Anthropic 发现,复杂高级的用户输入,和 Claude 复杂的输出之间,有很强的相关性。

Peter also notes that Anthropic found that sophisticated user inputs and complex Claude outputs are highly correlated.

换句话说,当 Claude 构建一个复杂的经济模型时,实际情况往往是,它是在某个人提供配套的专业指导下完成的。

In other words, when Claude builds a complex economic model, in practice it does so under the guidance of someone providing complementary expert direction.

他还写道,Anthropic 发现,即便在使用了六个月之后,人们也更倾向于把 Claude 当成一个思考伙伴来互动,而且和 Claude 的互动会更成功。

He also wrote that Anthropic found that even after six months of use, people are more likely to interact with Claude as a thought partner and have more successful interactions with Claude.

他认为,如果 AI 靠自己就已经足够厉害了,那我们就不应该看到这种现象。

If AI was good enough on its own, he contends, we wouldn't expect to see this effect.

Peter 写道,重要的是,广泛的任务自动化依然可以增强劳动。

Importantly, Peter writes, widespread task automation can still augment labor.

为什么呢?

Why?

因为工作并不是一成不变的一捆任务。

Because jobs are not fixed bundles of tasks.

从历史上看,新技术即使会让一些工作消失,也一直会在现有工作内部带来很大的变化。

New technologies have historically led to large changes within existing jobs, even as some jobs go away.

而且,它们还创造出了全新的工作类型,把新的技术能力和互补的人类专长结合在一起。

And they've produced entirely new types of work that combine new technical capabilities with complementary human expertise.

我们在研究中也看到了这种效应的迹象。

We see signs of this effect in our research.

在八万一千名 Claude 用户中,一个经常被提到的生产力来源是 scope,也就是能把更多事情做得更好、更熟练。

A commonly cited source of perceived productivity among 81,000 Claude users was scope, being able to do more more proficiently.

这种来自 AI 的赋能,可能会重新划定我们角色的边界,并在工作内部形成新的任务组合:把一部分任务自动化,强化另一些任务的重要性,同时从总体上提高劳动的边际产出。

Such empowerment from AI may redraw the boundaries of our roles and produce new bundling of tasks within jobs, automating some, reinforcing the importance of others, while on net increasing the marginal product of labor.

他还指出,人们越多使用 Claude、越擅长使用 Claude,虽然他们会提高对 Claude 能完成自己工作中多大一部分内容的预期,但他们对失业的预期反而会下降,而且往往会更乐观地看待 AI 对薪资、工作安全感,以及自己找工作能力这类事情的影响。

He also points out that the more that people use Claude and the better that people get at using Claude, even though they increase their expectations of what portion of their jobs Claude can do, they decrease their expectations of job loss and tend to be more optimistic about AI's impact on things like pay, job security and their ability to find a job.

不过,Peter 最后也提到,这一切最终都有可能改变。

Now, ultimately, Peter ends on the note that all of this could change.

他指出,随着 AI 能力不断提升,我们会看到越来越强的 agent,能够自主处理复杂、周期很长、而且有价值的任务。

He points out that as AI capabilities improve, we'll see increasingly capable agents that can autonomously handle complex long-horizon valuable tasks.

到那时,AI 还会继续增强劳动吗?

Will then AI still augment labor?

他指出,随着模型持续改进、jagged frontier 变得更平滑,AI 这种偏向技能、增强劳动的特性,很可能确实会消失;不过,他也说,到目前为止,他们最近分析 Claude Code 的使用模式,想看看 agent coding 是否正在改变专业能力的回报时,实际上并没有发现这种情况。

He points out that it may very well be the case that the skill-biased labor-augmenting aspect of AI goes away as models continue to improve and as the jagged frontier becomes smoother, but also that so far, when they recently analyzed patterns of Claude Code usage to see if agent coding was altering the returns to expertise, that's not really what they found.

Peter 写道,在我们追踪的这七个月里,Claude Code 被越来越多地用于更有价值的任务,但我们看到,人类专业能力的回报一直都还存在。

Peter writes, Claude Code has been used on more and more valuable tasks over the seven months we tracked, but we've seen persistent returns to human expertise.

也就是说,人们负责做规划决策,然后把具体执行交给 Claude。

That is, people make planning decisions and delegate implementation to Claude.

拥有更多领域专长的人,完成任务的成功率更高,而且当 Claude 出错时,他们也能更稳定地补救。

People with more domain expertise succeed in their tasks more often and recover more consistently when Claude makes an error.

直接写代码这种基础能力的回报也许下降了,但到目前为止,agent coding 提高了其他互补技能的价值。

The return to straightforward coding ability may have fallen, but agent coding has so far increased the value of other complementary skills.

他最后一个保留意见,是关于 recursive self-improvement。

His last caveat is about recursive self-improvement.

他写道,我们之所以对未来有这么多不确定性,一个很大的原因是,AI 可能会把创新本身也自动化。

He writes, a big reason there's so much uncertainty about the future is that AI may automate innovation itself.

让机器具备通用认知能力,会直接催化进一步的创新,而这种方式是过去那些通用技术所不具备的。

Endowing machines with general cognitive capabilities is a direct catalyst for further innovation in ways that past general-purpose technologies weren't.

在其他方面都算标准的经济模型里,把创新自动化可能会产生经济奇点,也就是在有限时间内出现无限增长。

In otherwise standard economic models, automating innovation can produce economic singularities, infinite growth in finite time.

那这种奇点真的会发生吗?

Will such singularities occur?

如果存在某些关键任务永远无法被自动化,不管是出于技术原因还是社会约束,那就不会发生。

Not if there are essential tasks that are never automated, whether for technical reasons or societal constraints.

这些薄弱环节,就是增长的限制。

Those weak links are the limits on growth.

他引用了 Aghion、Jones 和 Jones 的一篇论文,并写道:“经济增长的约束,可能不在于我们擅长什么,而在于那些必不可少、却又难以改进的东西。”

Pointing to a paper by Aghion, Jones, and Jones, he quotes, "economic growth may be constrained not by what we do well but rather by what is essential and yet hard to improve."

Peter 接着说,即便是在自动化非常迅速、非常广泛、但又并不完全的情况下,这些薄弱环节,长期来看,仍然可以让劳动在收入中的占比保持在较高水平。

Such weak links, Peter continues, can keep the labor share of income elevated in the long run even under very rapid, widespread but incomplete automation.

最终,他的结论是,scaling laws 这件事很难反驳。

Ultimately, he concludes, scaling laws are hard to argue with.

这些模型会变得更好,而且会好很多。

The models are going to get better, much better.

我预计这会推动生产率更快增长,甚至可能会出现更多更明显的 RSI 信号。

I expect this will drive faster productivity growth and maybe even more clear signs of RSI.

但我不觉得一年后的失业率会明显更高,至少不是因为 AI。

But I don't expect unemployment to be noticeably higher a year from now, at least not because of AI.

所以,这里来自 Anthropic 经济学负责人 Peter McCrory 的观点,真的很有意思;而且就算不回到大家经常提到的那些关于 AI 和工作的话题,比如 Jevons Paradox,这些内容也一样很有意思。

So really interesting stuff here from, again, Peter McCrory, the head of economics at Anthropic, and interesting even without returning to commonly heard concepts around AI and jobs like Jevons Paradox.

如果这种状态持续下去,那会带来非常重大的正面影响。

Now, if this state continues, it has big positive implications.

Stanford 的 Andy Hall 写道:“前阵子我预测过,真正针对 AI 的政治反弹,会发生在失业率上升几个百分点的时候,而我当时也说了,我们还没到那一步。”

Stanford's Andy Hall writes, "A while back I predicted that the real political backlash to AI would happen when unemployment started going up a couple of percentage points, and I said we weren't there yet."

我们现在依然还没到那一步,而这篇文章也帮助解释了为什么。

We're still not there and this piece helps explain why.

到目前为止,AI 看起来更像是在增强人类劳动,而不是取代人类劳动。

So far AI looks like it augments rather than replaces human labor.

这一点以后可能会变,但就目前来看,就业市场还是相当稳定的。

That could change, but right now the labor market looks quite stable.

当然,即便如此,我们还是能看到围绕 AI 的一些政治层面的担忧。

We are of course seeing political concerns about AI even so.

但如果真的出现了广泛的大规模失业,那相比之下,这些担忧就只能算小巫见大巫了。

But these concerns would look small fry in comparison if we had genuine widespread unemployment happening.

也有人认为,没错,这确实是个积极信号,但我们可能看错地方了。

Some argue that while yes, this is positive, we might be looking in the wrong place.

Trae Stephens 写道:“真正的影响,可能会先体现在招聘上,而不是裁员上:更少的初级岗位、更小的团队、更少的补招,以及对每个员工高得多的期望。”

Trae Stephens writes, "The real impact may show up first in hiring, not layoffs: fewer junior roles, smaller teams, lower backfilling, and much higher expectations for each employee."

一个会用 AI 的人,可能会越来越多地替代掉好几个人,也可能不会,即便整体失业率依然维持在低位。

One person using AI may increasingly replace several people, or not, even while overall unemployment remains low.

我确实觉得,随着时间推移,有一件事值得关注,那就是我们会不会看到团队平均规模发生变化。

And I do think that one thing that's worth watching over time is whether we see shifts in average team size.

我也确实觉得,我们的工作方式模式会发生一些变化。

I do think that there's going to be shifts in the patterns of how we do work.

而且我预计,不管是在组织层面,还是在组织内部,更小、更灵活的团队都会越来越多地承担更多责任。

And I would expect smaller, more nimble teams, both on the organization scale, but also within organizations, to increasingly have more responsibilities.

我觉得这种变化会来得足够渐进,所以大多数情况下,会是团队单元被重新配置,人们也会被重新调配去做其他类型的事情,同时配合这些如今承担更大角色的团队,不过这仍然值得持续关注。

I think that the change will happen gradually enough that mostly units will be reconfigured and people will be redeployed to do other types of things alongside those teams who are now taking on a bigger role, but it is still worth watching.

对很多人来说,最值得注意的事情,可能就只是 Anthropic 这次居然难得发布了一些正面的内容。

For many, the most notable thing is just Anthropic releasing something positive for once,

投资人 Julie Fredrickson 写道。

writes investor Julie Fredrickson.

终于有人在 Anthropic 讨论 AI 对专业阶层有多好了。

Finally, someone in Anthropic discussing how great AI is for the professional class.

Robert Scoble 写道:“Anthropic 终于开始做一种不充满恐惧的营销了。”

Robert Scoble writes, "Anthropic doing marketing that isn't full of fear."

这种内容请多来一点。

More of this, please.

你看,我觉得在预测未来这件事上,epistemic humility 是极其重要的。

Look, I think epistemic humility is extremely important in the context of predicting the future.

要描绘出 AI 对就业造成巨大影响的情景,其实并不难。

It is not hard to draw scenarios where AI does have a big impact on jobs.

但与此同时,我觉得我们目前看到的证据,是非常令人鼓舞的。

And yet I think the evidence that we are seeing so far is extremely encouraging.

而且我还觉得,更重要的是,当整体讨论和叙事从效率、削减成本和缩减 headcount,转向增强能力、扩大职责以及创造新机会的时候,这会形成一种自我强化的影响,进而改变高管和领导者对于自己应该如何使用 AI 的思考方式。

And I think what's more, that the more that the discourse and narrative shifts from efficiency and cost-cutting and headcount reduction to augmentation and expansion of responsibilities and new opportunity creation, it has a self-reinforcing impact in how executives and leaders think about how they should be using AI.

换句话说,如果全世界的人都在说,这项技术就应该拿来把你的员工砍掉一半,顺便说一句,你的投资人也正是这么期待的,那这就会给你非常大的压力,逼着你真的这么做。

In other words, if everyone in the world is saying that this technology should be used to cut your staff in half, and by the way, that's what your investors expect as well, that's gonna put a lot of pressure on you to do exactly that.

但如果另一种说法是,利用它更快地做更多事情,进入横向领域,推出新的产品和服务,那我们就会看到更多这样的情况发生。

If on the other hand, the story is about doing more faster and moving into lateral domains and releasing new products and services, then we're going to see a lot more of that.

显然,我知道我觉得这里面哪一种对这个世界更好,我也希望世界上越来越多的人也会认同这个看法。

Obviously I know which of these I think is better for the world and I hope more of the world comes around to that view as well.

那现在,我们就在这个比较积极的点上结束这周五吧。

For now, we get to end Friday on that bright note.

一如既往,感谢你的收听或者观看。那我们下次见,拜拜。

Appreciate you listening or watching as always, and until next time, peace.
已剔除 3 处广告(点击展开查看)
M1
M10:27广告 · 已剔除

First of all, thank you to today's sponsors, KPMG, Robots & Pencils, Blitzy and Airtable.

To get an ad-free version of the show, go to patreon.com/ai-daily-brief or you can subscribe in Apple Podcasts.

And to learn more about sponsoring the show, send us a note at [email protected].

M1
M118:51广告 · 已剔除

One of the most important AI questions right now isn't who's using AI, it's who's using it well.

KPMG and the University of Texas at Austin just analyzed 1.4 million real workplace AI interactions and found something surprising.

The highest impact users aren't better prompt engineers, they treat AI like a reasoning partner.

They frame problems, guide thinking, iterate and push for better answers.

And the good news, these behaviors are teachable at scale.

If you're trying to move from AI access to real capability, KPMG's research on sophisticated AI collaboration is worth your time.

Learn more at kpmg.com slash us slash sophisticated that's kpmg.com slash us slash sophisticated.

M1
M119:49广告 · 已剔除

AWS Advanced Tier Partner more than doubled in a year.

And they're hiring 50 open roles.

If you're someone who knows this moment is different, who wants to be inside it, not watching it, this is worth a look.

At Robots & Pencils, the best ideas win and the team is purposefully kept super high quality.

This is the kind of place you look back on as the best decision you ever made.

Take a look at robotsandpencils.com/careers.

Blitzy is driving over 5x engineering velocity for large-scale enterprises.

A publicly traded insurance provider leveraged Blitzy to build a bespoke payments processing application, an estimated 13-month project.

And with Blitzy, the application was completed and live in production in six weeks.

A publicly traded vertical SaaS provider used Blitzy to extract services from a 500,000-line monolith without disrupting production, 21 times faster than their pre-Blitzy estimates.

These aren't experiments.

This is how the world's most innovative enterprises are shipping software in 2026.

You can hear directly about Blitzy from other Fortune 500 CTOs on the Modern CTO or CIO Classified podcasts.

To learn more about how Blitzy can impact your SDLC, book a meeting with an AI solutions consultant at blitzy.com, that's blitzy.com.

This episode of the AI Daily Brief is brought to you by HyperAgent, where you run fleets of agents your team can manage together.

New users get $1,000 in inference.

Forget local agents and chat workflows waiting on your laptop to be prompted.

HyperAgent deploys always-on agents in the cloud, doing real work across the tools your team already uses.

Marketing's agent turns competitor moves into landing pages.

Sales' agent enriches leads, drafts emails, and updates the CRM.

Ops' agent chases the paperwork and tracks the budget.

Every agent has access to shared context and follows your rules about scope and approvals.

It's time you add agents that feel like teammates.

Hire yours at HyperAgent, built by the team at Airtable.

Claim your $1,000 in inference at hyperagent.com/ai-daily-brief.