本周最重要的AI新闻正被忽视

这周我们看到了很多值得关注的模型发布。

We got a lot of notable model releases this week.

Muse带着Wonder和Wonder Full重新回到了前沿。

Muse jumped back to the frontier with Wonder and Wonder Full.

谷歌实际上发布了Gemini 4 Argon，但看起来周四公布的那两个不起眼的小模型，也就是今天节目的主题，在所有发布中受到的关注最少，但可能实际上是最重要的，至少从在企业中实际使用AI的现实角度来看是这样。

Google actually released Gemini 4 Argon, and yet it seems like the two little unassuming models that were announced on Thursday, that are the subject of today's show, the ones that are getting the least fanfare of all of these, might actually be the most significant, at least when it comes to the practical reality of using AI inside businesses.

《AI每日简报》是一个每日播客和视频节目，涵盖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.

好了，朋友们，在我们深入之前，先快速说几个通知。

All right, friends, quick announcements before we dive in.

首先，感谢今天的赞助商：KPMG、Harbor、Granola和Blitzy。

First of all, thank you to today's sponsors, KPMG, Harbor, Granola, and Blitzy.

想要收听无广告版本的节目，请访问patreon.com/aidailybrief，或者你也可以在Apple Podcasts上订阅。

To get an ad-free version of the show, go to patreon.com/aidailybrief, or you can subscribe on Apple Podcasts.

如果想了解如何赞助这个节目，请发送邮件到sponsors@aidailybrief.ai。

And to learn more about sponsoring the show, send us a note at sponsors@aidailybrief.ai.

再提一句，我们刚刚宣布了最新的超级智能高级培训项目。

One more shout-out, we have just announced the latest super intelligent premium training program.

我们很快就要启动第一期AI冠军训练营了。

We are soon kicking off our first AI Champions boot camp.

这个项目不仅能提升你使用AI的技能，还能帮助你在整个企业中推动AI的应用。

This builds not only your skills to use AI, but to support adoption across your enterprise.

项目包含六次直播课程，内容非常丰富，你可以在AI Daily Brief网站顶部找到报名链接。

It's anchored by six live sessions and has a ton of value built in, and you can find a link to it from the top of the AI Daily Brief website.

Anthropic已经开始将第三方安全评估人员嵌入到他们的组织中。

Anthropic has begun the process of embedding third-party safety evaluators into their organization.

上个月，为了响应加速前沿发展的呼吁，Dario Amodei承诺将第三方评估人员整合进Anthropic，让他们拥有员工级别的权限来监督安全流程和报告事件。

Last month, as part of calls to pace the frontier, Dario Amodei pledged to integrate third-party evaluators into Anthropic, giving them staff-level access to monitor safety procedures and reports on incidents.

在周四发布的最新消息中，Anthropic表示他们已经整合了第一批评估人员，也就是来自Meter的六人团队。

In an update published on Thursday, Anthropic said that they've now integrated the first batch, which is, of course, a six-person team from Meter.

他们说Meter的成员拥有类似员工的访问权限，这是他们的说法，包括使用公司电脑、办公空间和内部工具。

They said the Meter folks have employee-like access, which is their term, including use of company computers, office space, and use of internal tooling.

Anthropic还提到，他们正在与Secure Bio和Accenture合作，作为评估方，并且已经和其他组织，包括Hugging Face，进行了初步的讨论。

Anthropic added that they're working with Secure Bio and Accenture as evaluators and are in early discussions with other organizations, including Hugging Face.

在解释为什么支持这种安全监控模式时，Anthropic写道：“我们希望对自己所做的决定公开负责，但我们不能公开分享所有与安全工作相关的信息。

Explaining why they support this model of safety monitoring, Anthropic wrote, "We want to be publicly accountable for the decisions we're making, but we can't publicly share all the information relevant to our safety work.

其中一些信息可能对恶意行为者有用，或者属于商业敏感信息，比如计算机训练信息。

Some of it could be useful to malicious actors or is commercially sensitive, like computer training information.

相反，我们让第三方评估人员深入了解我们的信息，并让他们分享关于构建模型所涉及风险的证据和合理结论。

Instead, we're giving third-party evaluators deep access to our information and are having them share evidence and reasoned conclusions about the risks involved in building our models.

我们的目标是缩小评估人员对前沿AI公司系统的理解与公司自身理解之间的差距。”

We aim to close the gap between what evaluators understand about a frontier AI company's systems and what the company itself understands."

随着这一安全制度的推行，有两个主要的讨论领域，坦白说也是批评的焦点，值得我们去了解。

Now, as this safety regime gets rolled out, there are two big areas of discussion and frankly, criticism that are worth understanding.

第一个是外部安全评估人员的独立性。

The first is the independence of the outside safety evaluators.

与一些人的说法不同，Anthropic声称他们并不资助Meter，而且坦白说，这些关系网有点超出了本期头条要讨论的范围。

Unlike what some have said, Anthropic claims they do not fund Meter, and frankly, the web of connections is a bit beyond the scope of this particular headlines to unpack.

不过，Meter和Anthropic确实有共同的大金主，都是围绕有效利他主义这个圈子的。

However, Meter does share major backers with Anthropic aligned around the effective altruism space.

这倒不是说他们有什么阴谋，只是他们来自同一个哲学根源，所以最后很多出资方都一样。

It's not necessarily full tinfoil hat, it's just that they come from the same philosophical roots and end up having a lot of the same funders because of that.

对于这种安排，大家有各种不同的看法也正常，但这解释了为什么Anthropic会这么快推进这个合作，而OpenAI看起来就有点犹豫。

Now, there are a range of valid opinions that could be held about the arrangement, but it's context for why Anthropic is quickly moving forward with that particular partnership, while OpenAI appears a little more hesitant.

另一个大问题是，实验室对安全审计报告的发布有多大的控制权。

The other big issue is how much control the labs have over the publication of safety audits.

这周早些时候，路透社报道说，Anthropic和OpenAI都在争取对报告有严格的编辑控制权。

Earlier this week, Reuters published a report claiming that both Anthropic and OpenAI were seeking strict editorial control over reports.

报道的时候，这个谈判看起来还在变动中，双方似乎正在朝着一个妥协的方向走，就是删掉机密信息，并且注明哪些地方做了重大删减。

This seemed to be a fluid negotiation at the time of reporting, and the parties seemed to be moving toward a compromise that involved redacting confidential information and noting when major redactions had occurred.

对于Meter，Anthropic写道，他们“不能否决发布，但如果信息属于少数预先定义的类别，我们可以要求他们删减。”

With Meter, Anthropic wrote that they, quote, "Can't veto publication, but we can require them to redact information if it falls into narrow, predefined categories.

Meter可以公开评论我们的删减，也可以说他们认为这些删减是否合理，只要不泄露被删掉的内容就行。”

Meter is free to characterize our redactions publicly and say whether it believes the redactions were justified as long as it does so without revealing the redacted information."

提到的类别包括客户数据、法律特权信息以及模型架构信息。

The categories mentioned were data about customers, legally privileged information, and information about model architectures.

Anthropic 还为嵌入式审计员定义了三个层级，Meter 是唯一处于最高层级的，享有员工级别的访问权限。

Anthropic has also defined three tiers for embedded auditors, with Meter the only one in the top tier, defined by employee-level access.

Secure Bio 目前处于最低层级，他们刚开始审查 Anthropic 八月风险报告中已经公开的化学和生物武器风险。

Secure Bio is currently in the bottom tier as they begin initial work reviewing existing chemical and biological weapons risk already publicly disclosed in Anthropic's August risk report.

在这个层级，Anthropic 要求审计员在发布前必须获得批准。

In that tier, Anthropic requires auditors to seek approval before publishing.

Anthropic 表示，他们会在 Secure Bio 完成第一份报告后，评估是否将其提升到第二层级，这样他们就能获得更多信息，并且无需事先批准。

Anthropic said that they will evaluate monitoring Secure Bio to the second tier, which would get them more access to information and no approval required following the completion of their first report.

Accenture 尚未开始工作，但 Anthropic 表示他们计划与其建立合作伙伴关系，包括红队测试模型、护栏测试和对齐评估。

Accenture has not begun work, but Anthropic says they are aiming to develop a partnership with them that involves red teaming models, guardrail testing, and alignment assessments.

围绕这一事件的大部分讨论都集中在 Meter 的独立性上，Embroidery 创始人 Zach Corman 总结了许多人的看法，他写道：“对于所有推动像 Meter 和 AISI 这样的独立评估机构的 AI 安全人士来说，其实早就有一个词来形容这种情况：安全表演。”

Most of the chatter coming out of this is about the independence of Meter, with Embroidery founder Zach Corman summing up a big strand of thought when he writes, "For all the AI safety people pushing for independent evaluators like Meter and AISI, there's already a term for this: security theater."

对很多人来说，另一个关键点不在于起点，而在于终点。

And the other thing for a lot of folks, it's less about where it starts and more about where it ends.

换句话说，如果Meter是唯一的评估者，那就有点像摩根大通挑几个它喜欢的银行评估机构，然后说这就是够格的独立审计了。

In other words, if Meter is the only evaluator, it's kind of like JPMorgan picking a handful of bank evaluators that it likes and saying that's a good enough independent audit.

看起来最终不太可能是这种情况。

It seems like that's unlikely to ultimately be the scenario.

就连Meter团队的一些人，比如Parv Mahajan，也提醒说：“我对这个作为风险评估的原型感到非常兴奋，但重要的是要记住，这不能替代正式的监督。”

Even some from the Meter team, like Parv Mahajan, reminded, "I'm pretty excited about this as a prototype for risk assessments, but it's important to remember that this is not a substitute for formal oversight.

Meter是在自愿协议下工作的评估者，而不是审计员，因为没有既定的安全标准可供审计。

Meter is an assessor working under a voluntary agreement, not an auditor, because there are no set safety standards to audit."

我觉得最终这必须走向某种更外部的审计机构。

I think ultimately this is going to have to end up in some sort of much more external auditor.

事实上，这可能是四大会计师事务所的一项业务。

In fact, this is probably a line of business for the Big Four.

当然，现在的限制是，真正有资格做这种评估的人并不多，而且几乎没有人供职于那些真正独立的组织。

Right now, of course, the constraint is that there just aren't that many people qualified to actually do this sort of assessment, and pretty much none of them work for those truly independent organizations.

至于前沿实验室对这一举措的态度，许多人注意到，埃隆·马斯克在X上回应了Anthropic关于这一举措的帖子，说SpaceX可能也会这么做。

Now, in terms of frontier lab alignment around this as a move, many noted that Elon Musk responded to Anthropic's post about the move on X, saying SpaceX will probably do the same.

接下来，Cognition 推出了一套新系统，帮助组织更高效地使用 token。

Next up, Cognition is launching a new system to help organizations token max more efficiently.

今年到目前为止，我们在 token 花费上大致看到了两种模式。

So far this year, we've broadly seen two settings for token spend.

第二季度是 token 大量使用的盛宴期，预算无上限，要求尽可能多用。

There was the feast of the token maxing era in Q2, with unlimited token budgets and a mandate to use as much as possible.

然后是 token 紧缩的饥荒期，组织开始对每月使用量设置上限。

Then there was the famine of the token austerity era, when organizations started placing caps on monthly usage.

实际上，大多数组织在 token 上限上仍然相当慷慨，但这传递了一个信号：token 预算是未来的方向。Cognition 的客户工程副总裁 Calvin Giroud 认为这两种做法都不对，他评论说：“用户应该有的限制不应该是静态的。”

Realistically, most organizations were still pretty generous with their token caps, but it sent the message that token budgeting was the way of the future Calvin Giroud, Cognition's VP of customer engineering, believes both approaches are wrong-headed, commenting, "The limits that a user should have should not be static.

它们应该根据你的使用效率和从中获得的价值动态调整。”

They should be dynamic based on the efficiency of your usage and the value that you're getting out of it."

Cognition 想出了一个有趣的方法，为使用其编程代理 Devin 的工程师动态调整预算。

Cognition has come up with an interesting way to dynamically adjust budgets for engineers using their coding agent, Devin.

Cognition 表示，他们可以用 AI 检查一个会话，并估算出人类完成同样工作量所需的大致小时数。

Cognition says that they can use AI to examine a session and estimate the equivalent number of hours it would have taken a human to achieve the same amount of work.

当工程师达到他们的限额时，系统会评估他们使用令牌预算的效率。

When engineers hit their limits, the system scores how efficiently they're using their token budget.

如果他们的评分令人满意，限额就会自动提高。

If they have a satisfactory grade, the limits automatically increase.

吉鲁说：“这就像信用卡的工作原理一样。

Giroud said, "This is how credit cards work, for example.

你有一个固定的信用额度，当你证明自己是个好的信用卡用户，并且按时还款，你的额度就会越来越高。”

You have a set credit limit, and as you prove you are a good steward of the credit card and you're paying back things on time, you get higher and higher and higher limits."

Cognition 认为，在代理编码竞争激烈的环境中，给客户提供效率工具来建立信任，比鼓励最大消费更有回报。

Cognition believes that in the hyper-competitive landscape of agentic coding, giving customers efficiency tools to build trust is going to have a better payoff than encouraging maximum spend.

吉鲁评论说，无节制的令牌最大化，引用他的话，“不是一种可持续的长期方法，而且我们的许多客户是大型上市公司，他们必须在公开市场上对股东负责，说明他们使用的工具的效率。”

Giroud commented that unrestrained token maxing is, quote, "not a sustainable long-term approach, and many of our customers are large public companies that have to answer to their shareholders in public markets around the efficiency of the tools they're using."

Cognition 还推出了一项名为 Devin Coach 的新功能，这是一个 AI 系统，用于反馈工程师如何使用他们的代理。

Cognition is also rolling out a new feature called Devin Coach, an AI system to give feedback on how engineers are using their agent.

这个教练旨在识别几十种最常见的错误，比如提出模糊的请求，或者为简单的任务选择昂贵的模型。

The coach is designed to pick up a few dozen of the most common errors, like making ambiguous requests or selecting an expensive model for a mundane task.

开发者的主要反应是：“去你的吧。”

The main response from developers was, "Screw this."

软件工程师杰西写道：“软件工程现在成了工厂流水线的活儿。”而阿米莎写道：“这有点苏联的味道。”

Software engineer Jesse wrote, "Software engineering is now a factory role," while Amisha wrote, "There's something very Soviet about this."

布莱恩·尚写道：“用人类需要多少小时来构建来衡量，这本身就是在衡量错误的东西。”

Brian Shang wrote, "How many hours would a human have needed to build this is measuring the wrong thing.

这是在激励人们去构建最复杂、最令人困惑的东西，而不是最有价值的东西。”

It's incentivizing folks to build the most complex and bewildering thing possible, not the most valuable thing."

但也有人明白，即使这种特定的方法不对，这类系统也只会成为我们现在生活的世界的一个特征。

And yet others got that even if this particular methodology is wrong, this type of system is just going to be a feature of the world that we live in now.

Weave的创始人亚当·科恩写道：“代币分配将是未来十年最重要的技能。”

Wrote Weave founder Adam Cohen, "Token allocation will be the most important skill in the next decade."

在政府领域，白宫人工智能特别工作组的首次会议于周四举行，为政府内部的人工智能监管设定了议程。

Over in the government realm, the first meeting of the White House's AI taskforce was held on Thursday, setting the agenda for AI oversight within the administration.

根据总统的行政命令，主要任务是监督一份将在120天内发布的人工智能安全报告。

The primary task is overseeing an AI safety report to be published within 120 days, according to the president's executive order.

总统科技顾问委员会已经受命准备这份报告，报告将涵盖网络安全和关键基础设施中的AI风险、生物安全和双重用途应用，以及模型控制和问责。

The President's Council of Advisors on Science and Technology has been commissioned to prepare the report, which will cover AI risk in cybersecurity and critical infrastructure, biosecurity and dual-purpose applications, as well as model control and accountability.

工作组还同意设立一个工作组来落实白宫协议，该协议由各前沿实验室和英伟达签署。

The taskforce also agreed to set up a working group to implement the White House accord signed by each of the frontier labs and NVIDIA.

该协议要求实验室在董事会层面的监督下实施安全控制，并与外部审计机构合作。

The accord commits the labs to implement safety controls with board-level oversight, as well as partnering with external auditors.

国防部副部长埃米尔·迈克尔将牵头一项倡议，建立报告协议，使实验室在发生安全事件时能够通知政府。

Defense Undersecretary Emil Michael will lead an initiative to establish reporting protocols to enable the labs to inform the government when security incidents occur.

与此同时，联邦贸易委员会主席安德鲁·弗格森的任务是在司法部和联邦贸易委员会之间建立一个执法和消费者保护协调委员会。

FTC Chair Andrew Ferguson, meanwhile, is tasked with creating a law enforcement and consumer protection coordination committee across the DOJ and FTC.

最后，人事管理办公室主任斯科特·卡波拉将与公共和私营部门合作，加强关键系统的防护，并协调劳工部的工作，研究AI对美国劳动力和经济的影响。

Finally, Office of Personnel Management Director Scott Capora will work with the public and private sector on the hardening of critical systems, as well as coordinating work at the Department of Labor to study the impact of AI on the American workforce and economy.

一个悬而未决的问题是政府将如何强制执行合规，因为白宫协议只是自愿的，或者正如总统所说，是道德约束。

One thing that's still up in the air is how the government will enforce compliance, given that the White House accord is only voluntary or, as the president put it, morally binding.

一位白宫官员告诉Semafor，一个可能的惩罚手段是威胁实施出口管制，今年早些时候曾对Anthropic使用过这一手段。

A White House official told Semafor that one stick could be the threat of export controls, which was used against Anthropic earlier this year.

与此同时，现在X上大家都在讨论新款特斯拉Roadster，它基本上能飞，多亏了SpaceX的推进器。

Meanwhile, everyone on X right now is talking about the new Tesla Roadster, which basically kind of flies thanks to some SpaceX thrusters.

但这并不是本周特斯拉和SpaceX合作的唯一大事。

But that's not the only big Tesla-SpaceX collaboration this week.

巴克莱银行的一份研究报告称，期待已久的合并可能在年底前实现。

A research note from Barclays claims that a long-awaited merger could be in the cards before the end of the year.

分析师丹·莱维在周三的一份报告中写道，合并可能会在12月9日特斯拉年度股东大会之前宣布。

Analyst Dan Levy wrote in a Wednesday note that a merger could be announced ahead of Tesla's annual shareholder meeting on December 9th.

两家公司都没有宣布任何消息，也没有向美国证券交易委员会提交任何文件。

Neither company has announced anything, and nothing has been filed with the SEC.

尽管如此，莱维写道：“虽然合并的产业逻辑曾受到质疑，但我们认为最近的经理人评论表明，他们看到了合并的关键好处。”

Still, Levy wrote, "While the industrial logic of a combination had been in question, we believe recent managerial comments indicate they see key benefits of a combination."

莱维表示，两家公司都没有聘请银行家或成立董事会委员会来开始谈判。

Levy said that neither company has hired bankers or created a board committee to begin talks.

不过，他认为联合的Terrafab芯片工厂将是一个关键推动因素。

However, he believes the joint Terrafab chip plant would be a key driver.

他建议这次合并可以采用全股票交易的方式，用特斯拉的股票换SpaceX的股票，比例大概在0.7到1之间。

He suggested the merger could be structured as an all-stock deal, swapping Tesla shares for SpaceX stock at around 0.7 to 1.

巴克莱银行认为，这样马斯克将拥有合并后公司37%的股份，以及大约56%的投票权。

Barclays suggests that this would leave Musk with 37% ownership of the combined company and roughly 56% voting power.

到目前为止，预测市场并不相信这件事，年底前合并的几率仍然只有24%。

So far, the prediction markets don't believe it, with odds of a merger before the end of the year still at 24%.

然而，尽管看起来没什么准备，这又是一笔埃隆和埃隆之间的交易，而埃隆已经同意了。

And yet, despite the lack of obvious preparation, this is another deal between Elon and Elon, and Elon is already in.

在上个月的采访中，他说：“想象一下，当两家公司合作如此紧密时，你会采取什么行动。”

During an interview last month, he said, "Imagine what action one might take when there's so much close collaboration."

最后今天，我们昨天报道了一个很有意思的后续，就是最新这一代模型开始展现出类似Opus 4.5的影响力和不断扩展的能力。

Lastly today, a really cool follow-up to our story yesterday about how this latest generation of models is starting to turn out to look more like Opus 4.5 in terms of their impact and expanding capabilities.

YouTuber Max Blade和他的主播朋友们用vibe coding做了一个叫Kong Royale的大逃杀游戏。

YouTuber Max Blade and his streamer buddies have vibe coded a battle royale game called Kong Royale.

Blade称这是全球首款AAA级vibe coding大逃杀游戏，从截图来看，确实挺不错的。

Blade called it the world's first AAA vibe-coded battle royale, and looking at the screenshots, it seems pretty good.

这个游戏有一个巨大的岛屿场景，有载具，玩家角色是一只大猩猩。

The game has a huge island setting, vehicles, and a giant gorilla as the player character.

Blade在早期尝试过这个想法，但他说当时的模型还不足以支持。

Now, Blade worked on this idea during the early days of vibe coding but said the models just weren't ready for it.

但有了Opus 5.5和几个Claude Max账号，他花了几周时间就把它做出来了。

But armed with Opus 5.5 and a couple of Claude Max accounts, he was able to put it together with a few weeks' work.

Blade称他们使用的方法为“蜂群编码”，基本上就是把一堆AI代理扔进他的日常直播聊天中，让人类开发者参与进来。

Blade calls the methodology that they used hive coding to do the work, which is basically throwing a bunch of agents into his daily live stream chat with human developers who tune in.

到了晚上，这些代理会回顾聊天记录，挑选出最好的改动，然后发布到早上的版本中。

Overnight, the agents would review the chat, select the best changes, and ship them for a morning build.

Blade基本上是通过观众的想法来众包开发这个游戏的。

Blade essentially crowdsourced the development of the game using ideas from his audience.

距离Peter Level的Plane Game已经过去大约一年半了，那个游戏开启了用AI编码制作完整游戏的概念。

We're now about a year and a half out from Peter Level's Plane Game, which kicked off the idea of vibe coding full games.

现在一个独立开发者只要有毅力和一个AI订阅，就能做出这么厉害的东西，真是令人惊叹。

It is pretty amazing what a solo dev can now achieve with a bit of perseverance and an AI subscription.

Riley Brown认为这种社交创作显然是下一步的发展方向。

Riley Brown thinks that this social creation is the obvious next step.

他写道：“这显然是电子游戏的未来，主播和他们的社区一起构建游戏，然后一起玩。”

He wrote, "This is obviously the future of video games, streamers and their community building the game together, then playing it.

随着模型每个月都在进步，这些游戏会看起来越来越专业。”

Every month that goes by as models improve, these games will look more and more professional."

Alex Hanson写道：“这是兄弟之夜的新时代。”

Alex Hanson wrote, "This is the new era of bros nights.

与其在《堡垒之夜》里骂那些比你小一半的孩子，不如去买点激浪、披萨，再拿块白板。

Instead of cussing at kids half your age on 'Fortnite,' go get Mountain Dew, pizza, and a whiteboard.

你规划自己的游戏，或者想想你青少年时期玩过的游戏的翻新或模组。

You map out your own game or think of a refresh or mod of a game from your teenage years.

让Opus 5.5每晚循环处理，再一起分摊一个Claude Max订阅。

Give it Opus 5.5 to loop on it every night and a Claude Max sub you all split.

兄弟之夜又回来了。”

Bros nights are so back."

不过，这就是今天的头条新闻了。

That, however, is going to do it for the headlines.

接下来，是本期的主要内容。

Next up, the main episode.

毕马威和德克萨斯大学奥斯汀分校的一项新研究发现，当人们与人工智能合作时，相似的技能并不能保证相似的成果。

A new study from KPMG and the University of Texas at Austin found that when people work with AI, similar skills don't guarantee similar outcomes.

研究人员研究了500多名早期职业专业人士，发现表现最好的人通过指导、评估和完善人工智能的输出，持续放大了人工智能的价值。

Researchers studied more than 500 early career professionals and found that the best performers consistently amplified the value of AI by guiding, evaluating, and refining its outputs.

这些顶尖表现者被称为“人工智能放大器”，他们的定义不仅仅在于他们知道什么，还在于他们如何与人工智能合作。

These top performers, called AI amplifiers, weren't defined by what they knew alone, but by how they worked with AI.

想了解更多关于人工智能放大器与其他人的区别，请访问kpmg.com/us/aiamplifiers。

Learn more about what separates AI amplifiers from everyone else at kpmg.com/us/aiamplifiers.

每期节目，我都会讨论OpenAI、Anthropic、SpaceX AI、谷歌和Meta之间的竞争。

Every episode, I talk about the competition between OpenAI, Anthropic, SpaceX AI, Google, and Meta.

如果你一直在听，你可能已经有了最喜欢的公司。

And if you've been listening for a while, you might have a favorite.

也许你觉得OpenAI和Anthropic能保持领先，或者Meta的开源策略能胜出。

Maybe you think OpenAI and Anthropic can stay ahead, or perhaps Meta's open source strategy can win out.

不管你怎么看，每个AI实验室都创造了不同的投资机会。

Whatever your view, every AI lab creates a different investment opportunity.

Harbor Capital Advisors的AI实验室生态系统ETF系列，让你投资于你相信的AI实验室背后的生态系统。

Harbor Capital Advisors AI Lab Ecosystem ETF suite lets you invest in the ecosystem behind the AI lab you believe in.

在你投资的任何地方搜索Harbor AI实验室生态系统ETF，或者在X上关注@HarborCapital了解更多。

Search Harbor AI Lab Ecosystem ETFs wherever you invest or follow @HarborCapital on X to learn more.

访问harborcapital.com获取包含投资目标、风险、费用、支出及其他重要信息的招股说明书。

Visit harborcapital.com for a prospectus containing investment objectives, risks, fees, expenses, and other important information.

投资前请仔细阅读并考虑这些信息。

Read and consider it carefully before investing.

风险包括本金损失和与人工智能相关的风险。

Risks include principal loss and artificial intelligence-related risks.

Harbor ETFs由Foresight Fund Services, LLC分销。

Harbor ETFs are distributed by Foresight Fund Services, LLC.

Harbor与AI Daily Brief没有关联，这些基金也不隶属于任何AI实验室，也没有得到它们的赞助或认可。

Harbor is not affiliated with AI Daily Brief, and the funds are not affiliated with, sponsored by, or endorsed by any AI lab.

这是一则付费广告，并非个性化的投资建议。

This is a paid advertisement and not personalized investment advice.

投资有风险，包括可能损失本金。

Investing involves risk, including possible loss of principal.

开会的时候，我会全身心投入。

When I'm in a meeting, I'm fully in it.

我思考的是如何迭代，以及推动目标前进所需的创意交流。

I'm thinking about the iteration and creative back and forth it takes to actually push a goal forward.

我不会去想记录要点、跟踪待办事项这些事，而这就是Granola发挥作用的地方。

What I'm not thinking about is capturing takeaways, tracking to-dos, or any of that, and that's where Granola comes in.

Granola是一款AI驱动的记事本，它能捕捉会议内容，并将其转化为清晰、结构化的笔记，突出决策和行动项，方便查找。

Granola is an AI-powered notepad that captures what happens in your meetings and turns it into clean, structured notes with the decisions and action items pulled out and easy to find.

无需设置，也无需配置。

There's no setup and no configuration.

它正好融入你现有的工作方式。

It just fits into how you already work.

对我来说，这意味着我可以一直保持创意状态，而Granola确保这些创意真正变成行动。

For me, it means I get to stay in idea mode, and Granola makes sure those ideas actually become action.

一旦你在第一次会议上试用Granola，就很难再离开它了。

Once you try Granola on a first meeting, it is hard to go without.

你可以在granola.ai/brief完全免费试用。

You can try it totally free at granola.ai/brief.

访问granola.ai/brief，把时间拿回来。

That's granola.ai/brief to get your time back.

Blitzy对庞大代码库的理解，能够实现自主安全修复、现代化改造和新功能开发。

Blitzy's understanding of massive code bases unlocks autonomous security fixes, modernization, and new features.

那么，当完全没有遗留代码时，会发生什么？

So what happens when there's no legacy code at all?

绿地项目本该是容易的部分。

Greenfield is supposed to be the easy part.

全新开始，没有技术债务。

Clean slate, no technical debt.

但即便是全新项目，也只能按人类的速度推进，一次一个冲刺。

But even Greenfield moves at human speed, one sprint at a time.

Blitzy 把工作单元从开发者变成了整个项目，自主规划、构建、测试和验证整个应用，从零开始。

Blitzy changes the unit of work from the developer to the project, autonomously planning, building, testing, and validating entire applications from scratch.

数十万行可直接上线的代码。

Hundreds of thousands of lines of production-ready code.

有一位 Blitzy 客户搭建了一个全新的应用，53.4 万行代码，把原本 65 周的路线图压缩到了两周。

One Blitzy customer stood up a brand-new application, 534,000 lines of code, compressing a 65-week roadmap into two weeks.

另一位客户在没有前端工程师的情况下，交付了整个应用。

Another shipped an entire application with no front-end engineer.

无论是旧系统还是全新项目，答案都一样：软件以计算的速度运行。

Legacy or greenfield, the answer is the same: software at the speed of compute.

在 blitzy.com 上构建你的下一个项目。

Build what's next at blitzy.com.

网址是 B-L-I-T-Z-Y.com。

That's B-L-I-T-Z-Y.com.

欢迎回到《AI每日简报》。

Welcome back to the AI Daily Brief.

嘿，你信吗？

Well, would you believe it?

这周最重要的AI模型发布，反而是大家讨论最少的那些。

The most important AI model announcements of this week are the ones that people are talking the least about.

周四，Meta和微软都发布了新模型。

On Thursday, we got model announcements from Meta and Microsoft.

这些模型里，你看不到任何能改变行业标杆的基准测试成绩。

Nowhere among them are you going to see benchmarks that change the state of the art.

Artificial Analysis 也不会花大篇幅去报道这些。

Artificial analysis isn't going to dedicate pages and pages to these.

然而，尽管这些发布在X上没引起什么轰动，但它们对AI在商业中的应用方式影响深远，所以微软和Meta的最新模型发布，真的值得我们多加关注。

And yet, for reasons that have nothing to do with the excited hype discourse on X and everything to do with how business usage of AI is changing, Microsoft and Meta's latest model announcements deserve a lot more attention.

我们先来说说微软。

Let's talk about Microsoft first.

周四，微软的AI首席执行官穆斯塔法·苏莱曼发布了一款新的最先进的基础模型，叫做MAI Base 2。

On Thursday, Microsoft's AI CEO, Mustafa Suleyman, announced a new state-of-the-art base model called MAI Base 2.

这是一个2.8万亿参数的模型，和Kimi K3的规模一样大。

It's a 2.8 trillion parameter model, making it the same size as Kimi K3.

它可能比Sol或Opus稍大一点，但不如Fable或Astra那么大，不过我们并不清楚这些模型的具体规模。

It's likely a little larger than Sol or Opus, but not quite as large as Fable or Astra, although we don't have good information on how large any of those models are.

它采用了混合专家架构，有800亿个活跃参数，这再次和Kimi K3的架构相当接近。

It uses a mixture of experts architecture with 80 billion active parameters, which is again, fairly similar to the architecture of Kimi K3.

而且这是一次全新的预训练，完全没有从其他模型进行蒸馏。

This is also a completely fresh pre-training run with zero distillation from other models.

现在，这只是一个基础模型，也就是说它还没有经过强化学习。

Now, this is just a base model, meaning that it hasn't gone through reinforcement learning.

但微软对他们目前看到的结果感到非常兴奋。

But Microsoft is pretty excited about what they've seen so far.

在发布前的全员大会上，苏莱曼告诉员工：“我们现在拥有的预训练基础模型，比目前驱动GPT-5.6 Sol的模型还要强。”

During a town hall prior to the announcement, Suleyman told staff, "We now have a stronger pre-trained base model than the model that currently powers GPT-5.6 Sol.

而且实际上，我们只用了OpenAI训练他们模型的四分之一的计算量。”

And in fact, we used one quarter of the compute that OpenAI used to train their model."

这个与OpenAI模型的对比，是微软凭借其投资协议获得专有信息后进行的，并没有公开分享。

Now, the comparison against OpenAI's models was done using proprietary information Microsoft has access to as part of their investment deal and wasn't shared publicly.

不过，在公开场合，微软将MAI 2的性能与所有超过一万亿参数的开源模型进行了比较。

However, in public, Microsoft compared the MAI 2's performance against all open models over a trillion parameters.

在预训练基准测试中，MAI 2在每个基准上都超过了所有模型，包括Kimi K3，有些情况下还大幅领先。

Across pre-training benchmarks, MAI 2 outperformed on each benchmark against every model, including Kimi K3, and by a wide margin in some cases.

微软表示，他们实现这一结果的预训练效率是DeepSeek V4 Pro的12倍以上。

Microsoft said that they managed the result with more than 12x the pre-training efficiency as DeepSeek V4 Pro.

这个模型是在16,000块GB200 Blackwell GPU上训练了38天。

The model was trained on 16,000 GB200 Blackwell GPUs across 38 days.

接下来，公告帖子变得有点技术性，但重点在于，微软在前沿规模上成功完成了一次非常出色的预训练。

From there, the announcement post gets a little technical, but the big takeaway is that Microsoft pulled off a very successful pre-training run at frontier scale.

Prime Intellect的Elie Boukaus写道：“哇。

Prime Intellect's Elie Boukaus writes, "Wow.

这篇关于微软新3T预训练的技术博客文章非常精彩，里面关于如何训练这种规模模型的见解非常罕见且清晰。”

Amazing technical blog post about Microsoft's new 3T pre-training with very rare and clear insights on how to train models of this size."

现在，关于MAI 2为什么重要，对微软和客户来说各有答案。

Now, in terms of why MAI 2 matters, there's the answer for Microsoft and the answer for customers.

在所有训练AI模型的公司中，可以说微软通往前沿的旅程最为曲折，或者至少是最复杂的之一。

Among all the company's training AI models, one could argue that Microsoft has had the most torturous journey to the frontier, or at least one of the most complicated.

起初，他们决定买而不是自己造，在2019年到2022年底ChatGPT发布后不久之间，向OpenAI投资了130亿美元。

Initially, they decided to buy rather than build, investing $13 billion into OpenAI between 2019 and shortly after the launch of ChatGPT in late 2022.

那笔投资看起来很有远见，当其他大公司还在摸索时，微软已经利用OpenAI的势头，一开始就推出了高质量的AI。

That investment looked prophetic, and as the other big companies bumbled around, Microsoft was able to take advantage of OpenAI's momentum and jump right out of the gate with quality AI.

然而，没过多久，路上就出现了一些相当大的坎坷。

However, it wasn't long before there were some fairly significant bumps in the road.

最引人注目的可能是2023年底，萨姆·奥尔特曼被罢免后又重新担任OpenAI CEO，期间有一段时间看起来基本上整个OpenAI团队都要搬到微软去。

Probably the most notable came in late 2023 when Sam Altman was ousted and then reinstated as OpenAI CEO, with a short period where it looked like basically the entire OpenAI team would go move into Microsoft.

我想那时候微软的领导层已经看得很清楚了，不管他们多喜欢OpenAI，多喜欢用OpenAI的模型，单从受托责任的角度来看，他们也需要更多的独立性。

I think at that point it became pretty clear to Microsoft's leadership that no matter how much they liked OpenAI and liked working with OpenAI models, simply from a fiduciary responsibility standpoint, they were going to need to have a bit more independence.

2024年3月，微软聘请了DeepMind的联合创始人穆斯塔法·苏莱曼，让他担任新AI部门的CEO。

In March of 2024, Microsoft hired DeepMind co-founder Mustafa Suleyman to become CEO of a new AI division.

接下来的一年里，那个内部部门并没有拿出什么成果。

For the next year, we didn't get much from that internal division.

我们听到的只有各种不和的传闻，比如2024年10月《纽约时报》报道说，苏莱曼在一次视频通话中对OpenAI的员工大喊大叫，基本上是在指责OpenAI故意隐瞒技术突破。

All we got was reports of acrimony, like an October 2024 report from "The New York Times" that had Suleyman yelling at an OpenAI employee on a video call, essentially accusing OpenAI of sandbagging technological breakthroughs.

尽管微软能大量使用OpenAI的资源，但在2025年这一年里，微软还是过得很艰难。

Even with all of its big access to OpenAI, Microsoft struggled across 2025.

今年情况开始有所转变，或者说微软的新策略开始变得清晰，那就是微软发布了他们的MAI-1系列模型。

Things started to shift this year, or at least Microsoft's new strategy started to become clear when Microsoft released their MAI-1 family of models.

从表面上看，这些模型并不怎么出彩。

Now on the surface, these models were nothing to write home about.

基础模型比当时的前沿模型要小，在基准测试上的表现也只是平平。

The base model was smaller than frontier models of the era, and it was only okay on the benchmarks.

这些模型作为处理日常工作的通用模型是够用的，微软也开始把它们部署到整个Copilot生态系统中用于这个目的。

The models were adequate as commodity models for doing routine work, and Microsoft started to deploy them across the Copilot ecosystem for that purpose.

但它们显然无法与Anthropic在近半年前发布的Opus 5.5相提并论。

But they were certainly a shadow of what Anthropic had achieved with Opus 5.5, which had been released almost a half year earlier.

然而，对于任何仔细观察的人来说，这些模型只是开胃菜，主菜是新的商业模式。

However, to anyone watching closely, the models were just the appetizer to the entrée, which was the new business model.

这个模式叫做Frontier Tuning，它允许客户在微软的基础模型之上微调自己的模型。

That model was called Frontier Tuning, and it allowed customers to fine-tune their own models on top of a Microsoft base.

为什么这很重要呢？因为与此同时，企业通过实际应用发现，仅仅订阅最先进的公司并试图通过顶级模型满足所有智能需求，根本不可能做到成本效益。

Now, why this mattered is that this was happening at exactly the same time that enterprises were realizing, by seeing it in practice, that simply subscribing to the most state-of-the-art company and trying to run all of their intelligence needs through the top models had no chance of being anywhere near cost efficient.

事实证明，你可以用AI做很多事情，尤其是当我们进入智能体时代，开始用代码和软件解决各种问题，甚至包括工程领域之外的问题时，总的token账单变得相当可观。

Turns out that you can use AI for a lot of things, and especially as we entered the agentic era, where we began to use code and software to solve all sorts of problems, even outside of engineering, the total token bill that was coming due was quite significant.

而且这还是在真正大规模采用这些新的智能体方法之前的情况。

And this was even before you had true mass-scale adoption of these new agentic methods across the whole enterprise.

从那以后，在商业方面，大问题一直是如何解决这个问题。

Since then, on the business side, the big questions have been about how you address that.

其中一部分是政策层面的，比如新的令牌上限，规定不同的人能用多少，以及他们需要什么条件才能获得更多访问权限。

Some of it's policy in terms of new token caps on how much different people can use and what it takes for them to get more access.

另外，企业现在开始变成模型杂食者，希望根据手头具体任务的需求，在不同模型之间切换，而不是被锁定在某个供应商的生态里。

Then there's the fact that enterprises are becoming model omnivores now, wanting the ability to move in and out of different models based on the needs of the particular task at hand, rather than be locked into one vendor ecosystem.

这样一来，Frontier Tuning 模式就显得更有意思了。

And that's where the Frontier Tuning model starts to look a little more interesting.

微软首席执行官萨蒂亚·纳德拉一直在强调，企业必须掌握智能模型之上的智能层，不能把这层交给合作的供应商。

Microsoft CEO Satya Nadella has been hammering the drum that it is imperative for enterprises to own the intelligence layer around the intelligence models and to not give that layer to the companies they work with.

这需要构建更复杂的基础设施，可能包括模型路由器、开源模型或本地托管模型的组合，而微软正是在这样的环境中推出了它的 MAI 模型和新的 Frontier Tuning 商业模式。

That requires building a more sophisticated infrastructure that involves some combination of model routers or open models or locally hosted models, and it's into that environment that Microsoft dropped its MAI models and its new Frontier Tuning business model.

然而，如果没有最先进的基座模型，这个计划能走多远也是有限的。

And yet, without a state-of-the-art base model, there's only so far that program could go.

而这正是 MAI 2 开始变得重要的地方。

And that's where MAI 2 starts to matter.

如果后训练模型真的能接近最先进水平，我们可能会看到六家美国公司争夺前沿领域，而几个月前还只有两家。

If the post-train model actually brings it up close to the state of the art, we could have six US companies competing for the frontier, up from just two a few months ago.

而且更重要的是，微软通过其 Frontier Tuning 产品，针对它认为企业面临的一个关键且日益严峻的挑战，给出了自己的答案。

And what's more, of those models, Microsoft, with its Frontier Tuning product, is offering an answer to what it perceives to be a key and growing enterprise challenge.

虽然其他实验室，比如 Anthropic 和 OpenAI，只关注成本方面，通过提供更快、能力稍弱但更高效的模型来解决问题，但微软的 Frontier Tuning 模式也隐含地回应了主权问题。

While the other labs, like Anthropic and OpenAI, focus simply on the cost side of the equation by offering faster, less capable, but more efficient models, Microsoft's Frontier Tuning model has an implicit answer to the sovereignty question as well.

当然，对于主权和成本问题，开放模型也有其他的解决方案。

Now, of course, there are other answers to both the sovereignty and the cost question in open models.

但考虑到微软已经深深嵌入商业生态系统，你可以想象，很多公司会更愿意和微软合作来微调模型，而不是自己去搭建基础设施，托管一个基于中国基础模型的本地部署。

But given how embedded Microsoft already is in the business ecosystem, you have to imagine that there are a lot of companies that are going to be much more comfortable working with Microsoft to fine-tune a model than going out and figuring out the infrastructure to host an on-premise model built on a Chinese base.

不过，中国开放模型可能并不是那些公司的唯一选择。

Except that Chinese open models may not be those companies' only option.

Meta 这周已经发布了重大消息，推出了 Muse Wonder 和 Muse Wonderful。

Meta already made a big announcement this week with Muse Wonder and Muse Wonderful.

虽然这些模型确实让 Meta 重新回到了美国实验室的前沿竞争行列，并大大增强了其面向消费者的 Muse 代理的能力，但精明的人都在关注他们周四的公告，认为 Muse Spark Open 可能是一个更大的事件。

And while those models absolutely put them back in the frontier conversation among US labs and significantly increase what their consumer-focused Muse agent can do, the smart money is looking at their announcement from Thursday and thinking that Muse Spark Open might be an even bigger deal.

Meta 称 Muse Spark Open 是目前美国最强大的开放权重模型。

Meta is calling Muse Spark Open the most capable US open-weight model to date.

这个模型总共有8160亿个参数，其中420亿个是活跃的，上下文窗口达到一百万个token。

The model has 816 billion total parameters, 42 billion active in a million token context window.

相比之下，GLM 53是一个7530亿参数的模型，所以规模差不多。

GLM 53, by way of comparison, is a 753 billion parameter model, so on a similar scale.

AI首席执行官亚历山大·王表示，这个模型是为编码和智能体工作设计的，比如长周期任务、复杂的软件工程和多步推理。

AI CEO Alexander Wang said the model is built for coding and agentic work like long horizon tasks, hard software engineering, and multi-step reasoning.

在基准测试方面，Spark Open介于美国顶级开源模型（如Inkling和Nemotron-3 Ultra）和中国开源前沿模型（如GLM 53和Kimi K3）之间。

In terms of benchmarks, Spark Open is somewhere between the top US open models like Inkling and Nemotron-3 Ultra and the Chinese open frontier represented by GLM 53 and Kimi K3.

在编码基准上，它的分数比Kimi K3略低，但Meta声称在智能体基准（如JobBench和MCP-Atlas）上取得了开源模型的最优结果。

Its scores are a bit behind Kimi K3 on the coding benchmarks, but Meta claimed open state-of-the-art results for agentic benchmarks like JobBench and MCP-Atlas.

Artificial Analysis给Muse Open打了39分，这使它和DeepSeek V41 Flash并列第17位，在开源模型中排第7。

Artificial Analysis scored Muse Open at 39, which put it tied with DeepSeek V41 Flash in 17th place overall and seventh place among open models.

它比开源模型的最优水平低大约7分，但比Nemotron和Inkling高出整整13分。

It's about seven points off the state of the art for open models, but a solid 13 points ahead of Nemotron and Inkling.

Artificial Analysis还没有完成对Reflection AI的Beam的测试，所以很难进行直接的对比。

Artificial Analysis hasn't completed their testing of Beam from Reflection AI, so it's difficult to get an apples-to-apples comparison there.

但在Reflection公布的少数几个基准测试中，Muse Open看起来不相上下，甚至可能稍微领先一点。

But on the few benchmarks that Reflection has published, Muse Open looks on par, if not ever so slightly ahead.

而且和Reflection的模型不同，Muse Open在各大推理提供商那里都能完整使用，权重也都可以下载。

And unlike Reflections model, Muse Open is fully available across major inference providers with full weights available to download.

对Meta来说，这标志着他们重返开放权重模型训练的胜利。

Now for Meta, this is a triumphant return to open-weight model training.

在AI发展初期，Meta就是开放权重模型的代名词，发布了曾经备受喜爱的Llama系列。

Early on in the AI cycle, Meta was synonymous with open-weight models, releasing the once beloved Llama family.

尤其是在2024年的Llama 3时代，可以说Meta在前沿领域紧追OpenAI。

Particularly during the Llama 3 era in 2024, Meta was, one could say, nipping at OpenAI's heels on the frontier.

当时，很多人认为开放模型对Meta来说是最佳策略。

At the time, many viewed open models as the optimal strategy for Meta.

他们不需要向消费者收取每月20美元的聊天机器人订阅费，因为他们已经拥有科技行业最赚钱的商业模式之一。

They didn't need to charge consumers $20 for a chatbot subscription, given that they already had one of the most profitable business models in tech.

Meta只需要通过免费提供AI来搅局，阻止OpenAI获胜。

All Meta had to do was play spoiler by giving away AI for free, preventing OpenAI from winning.

然后Llama 4就来了。

Then came Llama 4.

这个模型在2025年4月发布，当时就让人们大失所望，而一周后OpenAI发布o3，情况就更糟了。

Released in April of 2025, the model was disappointing to people right away, and started looking a whole lot worse when OpenAI released o3 the following week.

Meta从未发布Llama 4旗舰版“巨兽”那个两万亿参数的版本，到了那年夏天，马克·扎克伯格已经开始对Meta的AI部门进行彻底的重组。

Meta never released the flagship two trillion parameter version of Llama 4, dubbed Behemoth, and by the summer of that year, Mark Zuckerberg had embarked on a full tear-down rebuild of Meta's AI division.

今年四月，Meta开始重新发布模型，推出了MuSpark 1.0，当时有些人以为他们不再做开源模型了。

When Meta started releasing models again in April of this year with MuSpark 1.0, some assumed that they were done with open models.

那时，Meta已经投入了数千亿美元用于AI建设，显然不可能全部免费开放。

By then, Meta had committed hundreds of billions of dollars to the AI build-out, and they clearly couldn't give it all away for free.

亚历山大·王承诺会回归开源模型，但有些人对此持怀疑态度，觉得他们未必会兑现。

Alexander Wang pledged to return to open models, but some were skeptical that they would get around to it.

不过现在看来，他们并不是在空口说白话，Meta确实已经重新回到了开源模型排行榜的榜首，至少在美国是这样。

It seems, however, they weren't just blowing smoke, and Meta has now returned to the top of the open model ranking, at least in the US.

开源模型研究员内森·兰伯特写道：“各位，这对我来说就像超级碗一样重要。

Open model researcher Nathan Lambert writes, "Guys, this is like my Super Bowl.

我等了好久，终于等到美国开放权重模型重新变得重要了。

I waited so long for US open-weight models to be relevant again.

在今天之前，美国模型还落后于2026年2月就出现的中国模型。

Before today, the US models were still behind Chinese models from as early as February 2026.

还有更多工作要做，但这次至少把差距缩小了一半。

Still more work to do, but this cut the gap by at least half."

他后来补充说：“如果它的表现大致符合分数所显示的，这个模型的采用率会飙升。

He later added, "If it performs roughly like what the scores indicate, adoption of this model is going to go gangbusters.

这是美国第一个接近前沿智能体能力的开放权重模型。”

First open-weight model in the US with near frontier agentic capabilities."

Nathan提醒说，GLM-52和Kimi K3是中国那边的重大突破，但他补充说，美国模型会获得疯狂的采用。

Nathan reminded that GLM-52 and Kimi K3 was that step change from China, but added that US models get crazy adoption.

Box的Aaron Levie写道：“这将进一步加速AI的普及，因为企业可以在编排和高阶任务上混合使用前沿性能，同时为可重复的工作流程拥有低成本且定制化的后训练模型。

Box's Aaron Levie wrote, "This is going to further accelerate AI diffusion because enterprises will be able to mix and match frontier performance for orchestration and high-end tasks while being able to have low-cost and tailored post-trained models for repeatable workflows.

总的来说，这会大幅增加整个技术栈的token消耗和工作负载。”

Overall, this dramatically increases token consumption and workloads across the stack."

所以现在我们又回到了之前关于微软的那个话题，那些不太愿意把所有数据和依赖都交给OpenAI或Anthropic的企业，现在比几个月前有了多得多的选择。

And so now we are right back again to the same conversation we were having about Microsoft, where enterprises who are not so keen to just hand over all their data and all their dependencies to OpenAI or Anthropic are starting to have a heck of a lot more choices than they did just a few months ago.

现在，关于MuSpark Open还有一个有趣的细节。

Now, one more interesting nugget on the MuSpark Open story.

这个模型，信不信由你，得到了苹果的第一方支持。

The model, believe it or not, has first-party support from Apple.

苹果和Meta不仅不是朋友，我觉得连“亦敌亦友”都算不上。

Now, Apple and Meta not only aren't friends, I don't even think you could call them frenemies.

马克·扎克伯格作为CEO的相当大一部分时间，都在努力摆脱苹果的控制，因为苹果掌控着移动用户和App Store。

A fairly big chunk of Mark Zuckerberg's life as CEO has been spent trying to get out from under the thumb of Apple based on its ownership of mobile users and the App Store.

然而，这个新模型却带来了一个新的合作。

And yet, this new model comes with a new partnership.

作为合作的一部分，苹果更新了他们的MLX软件工具包，确保Spark Open能在苹果硬件上流畅运行，而且两家公司的CEO都花时间互相称赞了对方。

As part of that partnership, Apple has updated their MLX software toolkit to ensure Spark Open will run smoothly on Apple hardware, and both CEOs of these companies took time to talk about each other.

扎克伯格发布了一段Muse在Mac Studio上运行的视频，并说：“MuSpark Open，权重今天发布，在新款Mac Studio上运行得很好，这要感谢与John Ternus和苹果的合作。”

Zuckerberg posted a video of Muse running on Mac Studios and said, "MuSpark Open, weights releasing today, runs great on new Mac Studios thanks to work with John Ternus and Apple."

特纳斯回应说：“恭喜团队。

Ternus responded, "Congrats to the team.

能在你的办公桌上如此高效地运行像MuSpark Open这样规模和级别的模型，真是太了不起了。

Amazing to run a model of MuSpark Open's caliber and scale this efficiently right from your desk.

Mac Studio不断突破端侧AI的极限。”

Mac Studio keeps pushing the boundaries of on-device AI."

这真是宣布你作为苹果新任CEO将以不同方式做事的好方法。

Talk about a way to announce that you are going to do things differently as the new CEO of Apple.

不过，尽管在苹果硬件上运行这个模型是可能的，但这并不意味着容易。

Now still, just because running this model on Apple hardware is possible doesn't mean it's easy.

Spark Open的完整版需要一TB的内存，所以你得考虑两台顶配的M5 Ultra Mac Studio。

The full version of Spark Open requires a terabyte of memory, so you're talking about a pair of tricked-out M5 Ultra Mac Studios.

那个型号的最高内存配置还没发货，两台加起来大概要花四万美元。

The top memory spec for that model hasn't shipped yet, and a pair is going to cost around $40,000.

所以，虽然这对最狂热的专业消费者来说也不太现实，但对于关心私有推理的小初创公司来说，并不是高不可攀的价格。

So while that's not really in reach for even the most devoted of prosumers, it's not unobtainably expensive for small startups that care about running private inference.

而且因为这是个开源模型，你当然可以跑一个量化版本，把它塞进Mac Mini里，只要你接受那些取舍。

And because it's an open model, you can of course run a quantized version and cram it into a Mac Mini, as long as you're happy with the trade-offs.

这两家公司对过去的问题并不避讳。

The companies weren't coy about their past issues.

事实上，在接受Axios采访时，亚历山大·王感谢了前苹果高管丹尼尔·格罗斯——现在在Meta超级智能实验室——帮忙化解了过去的恩怨，让双方能好好合作。

In fact, in an interview with Axios, Alexander Wang credited former Apple executive Daniel Gross, now at Meta Superintelligence Labs, for smoothing over the past grievances to work nicely together.

王说：“我们很高兴看到爸爸妈妈又能好好相处了。”

Said Wang, "We're excited for Mommy and Daddy to be playing nice again."

现在，大多数企业会放弃那种简单的、直接订阅Claude或ChatGPT就能解决问题的方案，这种想法其实不太现实，至少短期内是这样。

Now, the idea that most enterprises are going to opt away from the easy, hand-me-a-solution sort of approaches that just getting a Claude or a ChatGPT subscription offer is unrealistic, at least in the short term.

但我确实认为，越来越多的公司，尤其是那些有一定技术能力的，会开始设计更复杂的模型架构，让它们能根据需求和限制在不同模型之间切换，这些需求可能是性能和能力方面的，也可能是成本和效率方面的。

But I do think that a growing portion of companies, especially those that do have some technical capabilities, are going to be looking to design more complex model architectures that allow them to move in and out of different models based on their needs and constraints, which could be performance and capability needs, or they could be cost and efficiency needs.

这周，Meta和Muse都推出了大幅升级的模型加入竞争，我觉得这对美国的企业来说是件好事。

This week, both Meta and Muse threw majorly upgraded models into that fray, and I think businesses in America are better off for it.

这个趋势我们当然会继续关注，但今天就到这里，以上就是今天的AI每日简报。

This is a trend we will, of course, continue to watch, but for now, that's going to do it for today's AI Daily Brief.

感谢你的收听或观看，一如既往。下次再见，祝一切安好。

Appreciate you listening or watching, as always, and until next time, peace.