OpenAI首席财务官为何引发AI救助计划争议

本播客由谷歌赞助播出。

This podcast is sponsored by Google.

大家好。

Hey, folks.

我是奥马尔，谷歌DeepMind的产品与设计负责人。

I'm Omar, product and design lead at Google DeepMind.

我们刚刚在AI Studio推出了全新的氛围编程体验，让你能自由组合AI功能，比以往更快地将想法变为现实。

We just launched a revamped vibe coding experience in AI Studio that lets you mix and match AI capabilities to turn your ideas into reality faster than ever.

只需描述你的应用，Gemini就会自动为你连接合适的模型和API。

Just describe your app, and Gemini will automatically wire up the right models and APIs for you.

如果需要灵感，点击'手气不错'按钮，我们会帮你启动项目。

And if you need a spark, hit I'm Feeling Lucky, and we'll help you get started.

访问a i.studio/build创建你的第一个应用。

Head to a i.studio/ build to create your first app.

今日《AI每日简报》重点关注：OpenAI首席财务官关于数据中心交易需要政府托底的争议言论。

Today in the AI Daily Brief, some extremely unfortunate comments from the OpenAI CFO talking about a government backstop of their data center deals.

头条新闻：苹果是否将每年支付10亿美元给谷歌，让Siri使用Gemini技术？

And before that in the headlines, is Apple about to pay a billion dollars a year to Google to use Gemini for Siri?

《AI每日简报》是聚焦人工智能领域重要新闻与讨论的每日播客及视频节目。

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

好的，在开始前先快速播报几则公告。

Right, quick announcements before we dive in.

首先感谢今日赞助商：Gemini、KPMG、Robots and Pencils、Blitzy和Rovo。

Firstly, thank you to today's sponsors, Gemini, KPMG, Robots and Pencils, Blitzy, and Rovo.

获取无广告版节目请访问patreon.com/aideallybrief，或通过苹果播客订阅。

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

了解更多节目信息（包括赞助机会）请访问aideallybrief.ai。

And to learn more about the show, including sponsorship opportunities and go to aideallybrief.ai.

一如既往，我对这项基准研究中涌现出如此多关于AI投资回报率的惊人数据感到无比兴奋。

And as always, I continue to be thrilled about just how much incredible information about AI ROI is coming in through this benchmarking study.

您可以在roisurvey.ai上参与调查，只需花费几分钟时间。

It's available at roisurvey.ai, and it'll only take you a couple of minutes.

相信我，您一定会想要获取这些宝贵数据。

And trust me, you are going to want to have access to this information.

我相信最终我们将收集到数千个带有投资回报率报告的实际案例，这能帮助人们更清晰地了解自身部署在行业中的相对位置。

I believe that we will have, when all is said and done, thousands of use cases with reported ROI, which I hope will give people a much better sense of where their deployments sit relative to others.

再次强调，请访问roisurvey.ai。

Again, that's roisurvey.ai.

现在让我们进入正题。

And now let's get to the show.

欢迎回到AI每日简报头条版，五分钟带您速览当日必备AI资讯。

Welcome back to the AI Daily Brief headlines edition, all the daily AI news you need in around five minutes.

经过多方猜测，谷歌似乎已赢得为苹果AI版Siri提供技术支持的合约。

After much speculation, it appears that Google has won the contract to power Apple's AI version of Siri.

根据彭博社最新报道，苹果与谷歌签署了年费十亿美元的协议，将授权使用谷歌模型作为Siri的新大脑。

According to new reporting from Bloomberg, Apple has signed a billion dollar a year deal with Google to license their models to use as Siri's new brain.

彭博社苹果内幕记者马克·古尔曼透露，谷歌将提供1.2万亿参数的定制版Gemini模型。

Bloomberg's Apple insider Mark Gurman reports that Google will provide a 1,200,000,000,000.0 parameter model which is a custom version of Gemini.

古尔曼写道，这个规模将"碾压"苹果现有仅1500亿参数的模型。

Gurman wrote that this would quote dwarf the level of Apple's current models which are only 150,000,000,000 parameters in size.

此举显然扫除了重大障碍，使苹果能够按计划在明年春季如期发布新版Siri。

Presumably this clears a major blocker and puts Apple in a strong position to release a new version of Siri as scheduled next spring.

Gemini模型将驱动Siri的摘要生成和规划功能，实现信息整合与代理任务执行。

The Gemini model will power Siri summarizer and planner functions which synthesize information and execute agentic tasks.

苹果自有模型仍将负责部分次要功能，整个流程将在苹果私有云服务器上运行，确保用户数据与谷歌完全隔离。

Apple's models will continue to drive some minor Siri features and the entire process will run on Apple's private cloud servers to ensure that user data is completely segregated from Google.

事实上，苹果已经为预期发布分配了基础设施。

Apple has in fact already allocated infrastructure in anticipation of release.

古尔曼表示，这笔交易不会广泛公开，苹果更倾向于将谷歌作为幕后合作伙伴。

Gurman said that the deal won't be widely publicized, with Apple preferring to keep Google as a behind the scenes partner.

该交易也不包括将Gemini作为聊天机器人集成到Siri中，也不会将谷歌的AI搜索添加到苹果的操作系统中。

The deal also won't extend to integrating Gemini into Siri as a chatbot, nor will it include Google's AI search being added to Apple's operating system.

古尔曼指出：苹果仍然不希望将Gemini作为长期解决方案。

Gurman noted: Apple still doesn't want to use Gemini as a long term solution.

尽管公司流失了AI人才，包括模型团队负责人，管理层仍打算继续开发新的AI技术，并希望最终用内部解决方案取代Gemini。

Despite the company bleeding AI talent, including the head of its models team, management intends to keep developing new AI technology and hopes to eventually replace Gemini with an in house solution.

为实现这一目标，苹果正在继续训练一个万亿参数模型，希望明年能用于支持消费者应用。

In service of that goal, Apple is continuing to train a trillion parameter model that they hope to use to power consumer applications by next year.

Barely AI指出了许多人在这一领域观察到的相同情况：谷歌每年支付200亿美元成为iPhone Safari浏览器的默认搜索引擎后，现在又要让苹果每年支付10亿美元用Gemini驱动Siri。

Barely AI pointed out the same thing that many noticed when it comes to the state of play in this space: Google getting Apple to pay at a billion dollars a year for Gemini to power Siri after Google had to pay Apple $20,000,000,000 a year for distribution to be the default search on iPhone's Safari browser.

与此同时，阿米特正在投资。

Amit is investing meanwhile.

刚刚指出，这再次表明谷歌今年绝对是怪兽级的一年。

Just pointed out that this is yet another indication of Google's absolutely monster year.

转向OpenAI领域，现在有百万企业正在使用ChatGPT。

Moving over to OpenAI land, a million enterprises are now using ChatGPT.

早在2022年，ChatGPT就成为历史上增长最快的消费科技产品，仅用五天就达到百万用户，五周达到1亿用户。

Back in 2022, ChatGPT became the fastest growing consumer tech product in history, reaching a million users in just five days and a 100,000,000 in five weeks.

而现在，ChatGPT似乎也成为历史上增长最快的商业软件平台。

And now ChatGPT appears to be the fastest growing business software platform in history as well.

ChatGPT工作席位在两个月内增长40%，达到700万，而ChatGPT企业席位同比增长9倍。

ChatGPT work seats are up 40% in two months to reach 7,000,000, while ChatGPT Enterprise seats are up 9x year over year.

市场负责人玛吉·霍特写道：两年半前我刚加入OpenAI时，ChatGPT企业版甚至还没建成。

Go to market leader Maggie Hot wrote, When I joined OpenAI a little over two and a half years ago, ChatGPT Enterprise wasn't even built yet.

这只是一个想法。

It was just an idea.

这个想法是：如果我们能在创新与影响力之间架起正确的桥梁，世界上最强大的技术也可以是最有用的。

An idea that the world's most powerful technology could also be the most useful if we built the right bridge between innovation and impact.

如今，全球有超过一百万家企业正在使用OpenAI的产品，使OpenAI成为历史上增长最快的商业平台。

Today, more than a million businesses around the world are using OpenAI's products, making OpenAI the fastest growing business platform in history.

与此同时，OpenAI分享了一些关于投资回报率的统计数据。

Alongside the milestone, OpenAI shared some stats around ROI.

他们表示，自8月以来，Codecs作为编码代理的使用量增长了10倍，包括思科在内的企业代码审查时间减少了50%，项目周期从数周缩短至数天。

They said that use of Codecs as a coding agent is up 10x since August, with companies including Cisco seeing a 50% reduction in code review times, cutting project timelines from weeks to days.

凯雷集团报告称，使用AgentKit后代理开发时间缩短了一半，同时准确率提高了30%。

Carlyle Group reports that agent development time has been cut in half using AgentKit, while accuracy has increased by 30%.

Indeed招聘网站报告显示，自推出AI驱动的邀请申请功能后，申请量增加了20%，同时招聘量提升了13%。

Indeed is reporting a 20% increase in applications since the introduction of their AI driven invite to apply feature alongside a 13% boost in hirings.

我认为这些公司惊人的预测并不像某些人认为的那样离谱，其中一个重要原因是企业级应用的整体潜力我们才刚刚触及皮毛。

One of the big reasons that I don't think that the enormous projections of these companies are as outlandish as they appear to some is just how much we're still scratching the surface of the eventual total surface area of enterprise usage.

在融资领域，垂直AI初创公司的淘金热仍在继续——据The Information报道，AI客服初创公司Decagon正以超40亿美元的估值进行融资谈判，最高估值可能达50亿美元。该公司去年五月融资时的估值仅为15亿美元，这意味着这家应用层AI公司在六个月内估值翻倍有余。

Over in Funding World, the gold rush for vertical AI startups continues as Decagon is rumored to be raising at a valuation north of $4,000,000,000 The Information reports that Decagon, which is an AI customer support startup, is in talks for fundraising that could see the startup valued as high as $5,000,000,000 They last raised in May achieving a $1,500,000,000 valuation, so this is yet another App Layer AI company that has seen their valuation more than double in a matter of six months.

报告指出，他们现在的年度经常性收入(ARR)已远超3000万美元，而去年仅为1000万美元。

The report stated that they're now generating significantly more than $30,000,000 in ARR, which is up from $10,000,000 last year.

这除了表明市场对高估值私募融资持续看好外，我认为也印证了AI应用层持续甚至日益增长的重要性。

Now in addition to this showing that there is continued appetite for private financing at high valuations, I think it's also part and parcel of the continued significance and maybe even growing significance of the app layer of AI.

行业另一端的融资动态：数据中心初创公司Crusoe正在推进一项二级市场交易，该公司估值将达130亿美元。Crusoe是OpenAI主要基础设施合作伙伴之一，负责德克萨斯州阿比林市Stargate设施的建设和GPU部署。

Another funding story from a slightly different part of the industry: Data center startup Crusoe is working on a secondary sale that would value that company at $13,000,000,000 Crusoe is one of OpenAI's major infrastructure partners handling the construction and GPU deployment of the Stargate facility in Abilene, Texas.

据The Information报道，此次股权转让要约将为员工提供1.2亿美元的流动性。

The Information reports the tender offer would see $120,000,000 worth of liquidity offered to employees.

知情人士透露，传闻中的130亿美元估值较数周前完成的股权融资轮上涨了30%。

The rumored $13,000,000,000 valuation is a 30% bump from an equity funding round that closed just weeks ago, according to sources familiar with the deal.

最后，让我们以一个雄心勃勃的基调结束今天的节目。

Finally today, let's end on an ambitious note.

谷歌宣布了一项新的登月计划，要将数据中心送入太空。

Google has announced a new moonshot project to put data centers in space.

CEO桑达尔·皮查伊周三发推称：我们的TPU处理器即将进军太空。

CEO Sundar Pichai tweeted on Wednesday, Our TPUs are headed to space.

受公司从量子计算到自动驾驶等登月项目的启发，SunCatcher项目正在探索如何有朝一日在太空构建可扩展的机器学习计算系统，以利用更丰富的太阳能——其能量输出超过人类总发电量的100万亿倍。

Inspired by our history of moonshots from quantum computing to autonomous driving, SunCatcher is exploring how we could one day build scalable ML compute systems in space, harnessing more of the sun's power, which emits more power than 100,000,000,000,000 times humanity's total electricity production.

与所有登月计划一样，这需要我们攻克大量复杂的工程难题。

Like any moonshot, it's going to require us to solve a lot of complex engineering challenges.

初期研究表明，我们的三色堇TPU处理器在粒子加速器模拟近地轨道辐射水平的测试中完好无损。

Early research shows our trillium generated TPUs survived without damage when tested in a particle accelerator to simulate low Earth orbit levels of radiation.

然而仍存在重大挑战，比如热管理和在轨系统可靠性问题。

However, significant challenges still remain, like thermal management and on orbit system reliability.

在倒计时2027年初发射两颗原型卫星的过程中，我们还需要更多测试和突破。

More testing and breakthroughs will be needed as we count down to launch two prototype satellites by early twenty twenty seven.

谷歌研究员特拉维斯·比尔斯在博客中写道：未来太空可能成为扩展AI算力的最佳场所。

Google researcher Travis Beals wrote in a blog post: In the future, space may be the best place to scale AI compute.

这种方案不仅具有巨大的扩展潜力，还能最大限度减少对地球资源的影响。

This approach would have tremendous potential for scale and also minimizes impact on terrestrial resources.

目前太空数据中心的发射成本仍是主要障碍，但谷歌预计到2030年代中期成本将与地面数据中心大致持平。

Now at the moment, the cost of launching a space data center is a significant blocker, but Google sees the cost converging to become roughly comparable to a terrestrial data center by the mid-2030s.

谷歌计划在2027年前与Planet公司合作开发原型卫星，以确保其硬件能承受太空的强辐射环境。

Google is planning to partner with a company called Planet on prototype satellites by 2027 to ensure their hardware can handle the increased radiation of space.

无论你认为这是AI行业彻底哗众取宠，还是将其视为最初让你对科技感到兴奋的那种宏大抱负，这个问题都见仁见智。

Now whether you think this is the AI industry completely jumping the shark, or you see it as exactly the sort of big ambition that gets you excited about technology in the first place, is a question that each person will have to answer for themselves.

以上就是本期AI每日简报的头条内容。

For now, that's going do it for the AI Daily Brief headlines edition.

接下来是正片内容。

Next up, the main episode.

如果AI不仅仅是个流行词，而是商业必需品呢？

What if AI wasn't just a buzzword, but a business imperative?

在《AI赋能》节目中，我们将带您深入全球最具前瞻性企业的董事会和战略会议现场。

On You Can With AI, we take you inside the boardrooms and strategy sessions of the world's most forward thinking enterprises.

本节目由我——贝瑟妮·惠蒂莫尔主持，毕马威赞助，这个七集系列将为您带来那些有目标地扩展AI应用的领导者们的实战洞见。

Hosted by me, Bethany Whitimore, and powered by KPMG, this seven part series delivers real world insights from leaders who are scaling AI with purpose.

从文化领导力对齐到建立信任、数据准备和部署AI智能体。

From aligning culture and leadership to building trust, data readiness, and deploying AI agents.

无论您是高管、战略家还是创新者，这档播客都将带您前排领略企业级AI的未来。

Whether you're a C suite executive, strategist, or innovator, this podcast is your front row seat to the future of enterprise AI.

立即访问www.kpmg.us/aipodcasts，或在Spotify、Apple播客等平台搜索《AI赋能》。

So go check it out at www.kpmg.us/aipodcasts, or search You Can With AI on Spotify, Apple Podcasts, or wherever you get your podcasts.

精简敏捷的团队总能胜过臃肿的咨询公司。

Small, nimble teams beat bloated consulting every time.

Robots and Pencils与各组织合作开发基于AI的智能云原生系统。

Robots and Pencils partners with organizations on intelligent, cloud native systems powered by AI.

他们洞悉人性需求，设计AI解决方案，破除复杂迷局，无需层层官僚就能实现实质影响。

They cover human needs, design AI solutions, and cut through complexity to deliver meaningful impact without the layers of bureaucracy.

作为AWS认证合作伙伴，Robots and Pencils兼具大公司的资源与可信伙伴的专注。

As an AWS Certified Partner, Robots and Pencils combines the reach of a large firm with the focus of a trusted partner.

团队遍布美加欧拉，让客户既享本土智慧，又获全球规模。

With teams across The US, Canada, Europe, and Latin America, clients gain local expertise and global scale.

在AI演进浪潮中，他们助您从容应对变革。

As AI evolves, they ensure you keep peace with change.

这意味着更快见效、可衡量的成果，以及持久稳固的伙伴关系。

And that means faster results, measurable outcomes, and a partnership built to last.

合适的合作伙伴让进步成为必然。

The right partner makes progress inevitable.

与机器人和铅笔携手合作，请访问robotsandpencils.com/aidailybrief。

Partner with robots and pencils at robotsandpencils.com/aidailybrief.

本节目由Blitzy赞助播出，这是一个拥有无限代码上下文的企业级自主软件开发平台。

This episode is brought to you by Blitzy, the enterprise autonomous software development platform with infinite code context.

Blitzy运用数千个专业AI代理，它们花费数小时来理解数百万行代码的企业级代码库。

Blitzy uses thousands of specialized AI agents that think hours to understand enterprise scale codebases with millions of lines of code.

企业工程领导者们每个开发冲刺周期都从Blitzy平台开始，输入他们的开发需求。

Enterprise engineering leaders start every development sprint with the Blitzy platform, bringing in their development requirements.

Blitzy平台提供计划，然后为每个任务生成并预编译代码。

The Blitzy platform provides a plan, then generates and pre compiles code for each task.

Blitzy自主完成80%以上的开发工作，同时为剩余20%需要人工完成的开发工作提供指导。

Blitzy delivers 80% plus of the development work autonomously, while providing a guide for the final 20% of human development work required to complete the sprint.

上市公司将Blitzy作为预IDE开发工具后，工程效率提升了5倍，结合他们选择的编程助手，将AI原生SDLC引入组织。

Public companies are achieving a 5x engineering velocity increase when incorporating Blitzy as their pre IDE development tool, pairing it with their coding pilot of choice to bring an AI native SDLC into their org.

访问blitzy.com并点击'获取演示'，了解Blitzy如何将您的SDLC从AI辅助转变为AI原生。

Visit blitzy.com and press Get a Demo to learn how Blitzy transforms your SDLC from AI assisted to AI native.

认识Robo，您的人工智能队友。

Meet Robo, your AI powered teammate.

Robo通过AI驱动的搜索、聊天和代理释放团队潜力，您还可以用Studio构建自己的代理。

Robo unleashes the potential of your team with AI powered search, chat, and agents, or build your own agent with Studio.

Robo由您组织的知识驱动，运行在Atlassian可信赖的安全平台上，始终围绕您的工作场景运作。

Robo is powered by your organization's knowledge and lives on Atlassian's trusted and secure platform, so it's always working in the context of your work.

将Robo连接到您最爱的SaaS应用，确保知识永不遗漏。

Connect Robo to your favorite SaaS app so no knowledge gets left behind.

Robo运行在Teamwork Graph上——这是Atlassian的智能层，能统一整合所有应用数据，从第一天起就提供个性化AI洞察。

Robo runs on the Teamwork Graph, Atlassian's intelligence layer that unifies data across all of your apps and delivers personalized AI insights from day one.

Robo已内置于Jira、Confluence以及Jira服务管理标准版、高级版和企业版订阅中。

Robo is already built into Jira, Confluence, and Jira Service Management Standard, Premium, and Enterprise subscriptions.

可曾体会过AI从工具转变为队友的感觉？

Know the feeling when AI turns from tool to teammate?

若你使用Robo，自会明了。

If you Robo, you know.

探索Robo——您由Atlassian驱动的新AI队友。

Discover Robo, your new AI teammate powered by Atlassian.

请访问rov,asinvictory,0.com开始体验。

Get started at rov,asinvictory,0.com.

欢迎回到AI每日简报。

Welcome back to the AI Daily Brief.

一周内第二次，OpenAI因其过于随意的沟通方式，未能意识到自己已非昔日可轻率发言的古怪初创公司，而掀起了一场完全没必要且极具病毒式传播的激烈评论与批判风暴。

For the second time in a week, OpenAI has stirred up a completely unnecessary and incredibly viral and virulent hornet's nest of commentary and critique by being too loose with their communication and failing to understand that they are no longer some quirky startup that can just talk flippantly.

若你觉得我对此感到恼火，没错——因为这关乎整个行业未来一年与政治、市场及文化的互动方式，影响重大。

And if you think I sound annoyed about this, you are right that I am because it has big implications for how this entire industry will interface with politics, markets and culture in the year to come.

我指的是OpenAI首席财务官Sarah Fryer周三在加州《华尔街日报》TechLive活动上的发言。

I'm referring in this case to comments from OpenAI CFO Sarah Fryer at The Wall Street Journal's TechLive event in California on Wednesday.

这场对话包含多个不同层面的内容。

There were a number of different elements of the conversation.

Fryer驳斥了AI泡沫的说法。

Fryer rejected the idea of an AI bubble.

她评论道：'考虑到AI的实际应用潜力及其为个人带来的变革，我认为当前对AI的热情还远远不够。'

She commented, I don't think there's enough exuberance about AI when I think about the actual practical implications and what it can do for individuals.

我们应当继续全力推进。

We should keep running at it.

针对资金交易的循环性质疑，她表示：'我完全反对这种前提假设。'

She discussed the circularity critique of all these funding deals, saying, I kind of reject the premise completely.

如今我们都在建设完整的基础设施，让更多计算能力进入世界。

We're all just building out full infrastructure today that allows more compute to come into the world.

我完全不认为这是循环的。

I don't view it as circular at all.

过去一年的大量工作是为了实现供应链多元化。

A huge body of work in the last year has been to diversify that supply chain.

弗莱尔还推迟了关于IPO的讨论，否认OpenAI目前正在做准备，并表示IPO'目前不在考虑范围内'。

Fryer also deferred chatter about an IPO, denying that OpenAI is currently making preparations and saying that an IPO is quote not on the cards right now.

但这些评论都没有被采纳。

But none of that was the commentary that got picked up.

相反，被关注的是这部分内容——《华尔街日报》精炼地总结为'OpenAI希望联邦政府为新投资提供担保'。

Instead, it was this section, which the WSJ so crisply summed up as OpenAI wants federal backstop for new investments.

为求全面，让我们听听弗莱尔谈论OpenAI如何受限于计算能力：始终试图构建最先进技术，始终试图使用最先进芯片，因此必须解决所有资金问题。

For the sake of completeness, let's listen to the clip where Fryer is talking about how OpenAI is compute constrained, always trying to build at the state of the art, always trying to use state of the art chips, and thus has to figure out how to finance all of it.

对话始于讨论芯片实际价值和使用期限对资金动态的影响差异。

Where the conversation starts is with a discussion of the difference in the dynamics of funding based on how long we find chips are actually valuable and useful for.

那么问题在于：一颗芯片能保持前沿地位多久？

So the question is how long does a chip remain on the frontier?

是三年、四年、五年，还是更久？

Is it three years, four years, five years, or even longer?

在当前计算资源匮乏或受限的世界里，我们确实在使用相当于100颗的旧芯片——这些芯片可能已经存在六七年了。

Now in a world where we have no compute or compute constrained, we are absolutely using chips that have are the the like a 100 equivalents that have been around like maybe six, seven years at this point in time.

如果是这种情况，芯片融资就会容易得多。

If that's the case, financing chips gets a lot easier.

如果芯片的生命周期很短，那就更难了。

If the timeline on the chip stays short, that gets harder.

因此我们正在寻求银行、私募股权甚至政府等各方共同构建的生态系统——探索政府可能的介入方式。

And so this is where we're looking for an ecosystem of banks, private equity, maybe even governmental, the ways governments can come to bear.

是指像联邦补贴之类的吗？

Meaning like a federal subsidy or something?

首先是指那个保障机制，即确保融资能够进行的担保。

Meaning like just first of all, the backstop, the guarantee that allows the financing to happen.

这确实能大幅降低融资成本，同时提高贷款价值比。

That can really drop the cost of the financing, but also increase the loan to value.

所以你们可以在股权融资基础上承担的债务额度，为芯片投资提供某种最终保障。

So the amount of debt that you can take on top of an equity portion for So some final backstop for chip investing.

正是如此。

Exactly.

我认为我们正在见证这一点。

And I think we're seeing that.

我认为美国政府尤其表现出惊人的前瞻性，他们真正理解人工智能几乎已成为国家战略资产，在思考与例如中国的竞争时，我们必须保持审慎态度。

I think the US government in particular has been incredibly forward leaning, has really understood that AI has is almost a national strategic asset, and that we really need to be thoughtful when we think about competitive competition with, for example, China.

我们是否正在采取一切正确措施以尽可能快速发展我们的AI生态系统？

Are we doing all the right things to grow our AI ecosystem as fast as possible?

以上就是那些评论。

So those are the comments.

这里重要的是，是弗莱尔本人重新引入了‘后盾’这个词。

And what's important here is that it was Fryer herself who introduced the word backstop.

换句话说，这并不是一个过分热心的标题撰写者将话语强加于她，这种情况我们已屡见不鲜。

In other words, this wasn't an overzealous headline writer putting words in her mouth, which as we've seen plenty of times can happen.

现在，一些人试图善意理解弗莱尔，并解释她想表达的意思。

Now, some folks tried to give Friar the benefit of the doubt and explain what she was trying to say.

公关领袖兼评论员Lulu Chang Mezervi写道：不幸的沟通失误，使用了这个充满包袱的词汇‘后盾’。

Lulu Chang Mezervi, who is a PR leader and commentator, wrote unfortunate comms fumble to use the baggage laden word backstop.

在视频中，弗莱尔显然在寻找合适的词汇来描述政府支持。

In the video, Friar is clearly reaching for the right word to describe government support.

本可以采用公私合作或跨金融、工业和政府的协作方式，就像我们过去对大型基础设施投资所做的那样。

Could have gone with public private partnership or collaboration across finance, industry, and government as we've done for large infrastructure investments in the past.

相反，她有点仓促地使用了‘后备方案’这个词，随后被《华尔街日报》采访者重复引用，最终成了新闻标题。

Instead, she kind of stumbles into using Backstop, which was then repeated by The Wall Street Journal interviewer and then became the headline.

后来弗莱尔本人和OpenAI新闻室都试图澄清，在领英上写道：OpenAI并不寻求政府为我们的基础设施承诺提供后备支持。

Now, Fryer herself and the OpenAI newsroom later sought to clarify, writing on LinkedIn, OpenAI is not seeking a government backstop for our infrastructure commitments.

我用了‘后备方案’这个词，结果模糊了重点。

I used the word backstop and it muddied the point.

正如我完整回答所显示的，我的观点是美国的技术优势将来自建设真正的工业产能，这需要私营部门和政府各司其职。

As the full clip of my answer shows, I was making the point that American strength in technology will come from building real industrial capacity, which requires the private sector and government playing their part.

不幸的是，这次失言很难轻易挽回。

Unfortunately, this was not one that could be easily walked back from.

宏观研究公司创始人朱利安·布里金写道：我闻到了老鼠的味道。

Macro research firm founder Julian Bridgen writes, I smell a rat.

如果山姆·奥特曼能赚取数千亿美元，为什么还需要纳税人来担保他们的债务？

Why does Sam Altman need the taxpayer to guarantee their debt if they're going to make hundreds and hundreds of billions of dollars?

他是不是刚做了现金流分析，发现自己资金短缺？

Has he just done a cash analysis and realized he's cash flow short?

企业家兼投资者山姆·莱森写道：这是‘救助前的救助’。

Entrepreneur and investor Sam Lesson writes a pre bailout bailout.

我欣赏这种厚脸皮的请求，这真是时代的标志。

I appreciate the balls to ask, and what a sign of the times.

他接着补充道：笑死，政府担保当然会降低你的资金成本。

He followed up, LOL no crap that a government backstop will drop your cost of capital.

顺便说一句，如果现在派发这种福利，我也想要一个。

By the way, I would love one of these too if we were handing them out.

就在大约一周前，金融评论员卢·格鲁曼写道：当那些‘大到不能倒’的银行在全球金融危机前做这类事时——这里指的是用表外债务为元数据中心融资——格鲁曼继续写道：他们心里清楚当危机爆发时自己会被救助。

Just about a week ago, finance commentator Lou Grumman wrote: When the too big to fail banks were doing stuff like this prior to the GFC, in this case referring to off balance sheet debt to finance a metadata center, Grumman continues: They knew at some level they would get bailed out when the S hit the fan.

请为这个打上时间戳。

Please timestamp this.

如果事情变得一团糟，超大规模企业将像那些‘大到不能倒’的银行一样获得联邦救助。

If and when this goes pear shaped, the hyperscalers will get federal bailouts just like the too big to fail banks did.

卢克随后引用了《华尔街日报》的标题说：哇，从我推测这事最终会发生到现在才过了五天。

Luke then followed that up with a headline from The Wall Street Journal saying, Wow, only five days elapsed from the time I speculated this would happen eventually.

前白宫AI顾问迪恩·鲍尔写道：弗莱尔描述的是一种比我们在任何美国立法中见过的监管俘获更糟糕的形式。

Former White House AI advisor Dean Ball writes, Friar is describing a worse form of regulatory capture than anything we have seen proposed in any U.

S.

S.

无论是州级还是联邦立法，据我所知。

Legislation, state or federal, I am aware of.

一家游说这种结果的公司，实际上就是在为监管俘获进行游说，而非表面功夫。

A firm lobbying for this outcome is literally, rather than impressionistically, lobbying for regulatory capture.

投资者杰夫·帕克写道：OpenAI作为一个非营利组织，现在既想要联邦为所有新资本支出投资提供担保，又想在明年以万亿美元估值为其专属股东进行IPO。

Investor Jeff Park writes, OpenAI is a nonprofit that now wants a federal backstop guarantee for all new capex investments but also wants to IPO at a trillion dollars next year for its exclusive shareholders.

然后你会奇怪为什么曼达米能以压倒性优势当选。

And you'll wonder why Mamdami was elected in a landslide.

确实，无法回避的是——显然弗莱尔和OpenAI完全没有考虑到——这些言论将会引发的舆论背景。

And indeed, what is inescapable from this, and what was clearly completely ill considered by Fryer and OpenAI, was the context into which these comments were going to come.

芬恩·墨菲写道：对于那些抱怨曼达米的科技人士，我想指出：联邦为无限制投入数据中心的风险资本提供担保（最终让OpenAI股东受益），这实际上比免费公交车更糟糕的社会主义形式。

Finn Murphy writes, For all the tech people complaining about Mamdami, I would like to point out that a federal backstop for unfettered risk capital deployment into data centers for the benefit of OpenAI shareholders is actually a much worse form of socialism than free buses.

X上的尼莫写道：不确定现在是不是软性宣布未来救助计划的最佳时机。

Nemo on X writes, Not sure this was the best time to soft launch a future bailout.

Spectrum Markets总裁布伦特·唐纳利用一条简短的推文捕捉了大众情绪：滚蛋。

And Spectrum Markets President Brent Donnelly captured the mood with his simple tweet, F off.

现在需要重点注意的是：这不仅仅是弗莱尔个人的行为。

Now, one thing that's important to note is that this was not just Fryer out on her own.

阿尔特曼在谈到泰勒·考恩时也表达了类似观点。

Altman on Tyler Cowen said something similar.

在那篇同样发布于昨日的访谈中，他表示：某种程度上，当某件事物发展到足够庞大时，无论是否在纸面上有所体现，联邦政府终将成为最后的担保人——正如我们在多次金融危机和保险公司搞砸时看到的那样。

In that interview, also published yesterday, he said, At some level, when something gets sufficiently huge, whether or not they are on paper, the federal government is kind of the insurer of last resort, as we've seen in various financial crises and insurance companies screwing things up.

考虑到我对AI经济影响的预期规模，我认为政府最终确实会成为最后的担保方。

I guess given the magnitude of what I expect AI economic impact to look like, I do think the government ends up as the insurer of last resort.

但我的理解角度与你不同，我并不期待他们会像对待核能那样实际制定政策。

But I think I mean that in a different way than you mean that, and I don't expect them to actually be writing the policies in the way they maybe do for nuclear.

MECE麦克写道：这个混蛋现在几乎是在明目张胆地嘲讽我们——他正在变得'大而不能倒'，而我们就是那个兜底的人。

MECE Mike writes, This MF is now pretty directly taunting us that he's becoming too big to fail and we're the backstop.

从弗莱尔的评论中可以明显看出，这里存在对全球地缘政治的评估。

Now as you heard from Fryer's comments, there is clearly an assessment of global geopolitics here.

OpenAI非常清楚AI是地缘政治议题，而美国政府正是这样对待它的。

For OpenAI, they see very clearly that AI is a geopolitical issue and that the US government is treating it as such.

因此抛开'兜底'这个说法，探究弗莱尔真正想表达的，美国政府与这个具有重大地缘意义的行业之间必将形成某种关系。

And so removing the word backstop and trying to get at what Fryer was actually trying to say, there is inevitably going to be some sort of relationship between the US government and the industry with this much geopolitical relevance.

英伟达黄仁勋当天发表的言论强化了这一点——他用最直白的语言表示他认为中国将在AI竞赛中战胜美国。

This was reinforced by the fact that these comments came on the same day of comments from Nvidia's Jensen Huang who put in the starkest terms yet that he thinks that China will win the AI race with The United States.

在周三英国《金融时报》AI未来峰会的采访中，他直言中国将赢得AI竞赛。

In an interview at the Financial Times Future of AI Summit on Wednesday, he plainly said China is going to win the AI race.

他表示西方正被犬儒主义所束缚，并补充道：'我们需要更多乐观精神'。

He said that the West is being held back by cynicism, adding, We need more optimism.

黄仁勋认为，比起消极态度，更严重的是美国正受到...

Perhaps even more than a negative attitude, Huang believes The U.

监管负担

S.

的限制。

Is being limited by regulatory burden.

他指出，基于州的规则制定意味着AI公司将需要应对50项新法规。

He noted that state based rulemaking means AI companies will need to deal with 50 new regulations.

相比之下，黄仁勋提到中国为运行国产芯片提供的新能源补贴，称电力是免费的。

In contrast, Huang pointed to new Chinese energy subsidies for running domestically made chips commenting power is free.

目前英伟达相比中国产芯片的一大优势是能效。

Now one of NVIDIA's big advantages over Chinese made chips is energy efficiency.

但本周早些时候，北京宣布对使用国产芯片的数据中心提供50%的电费补贴，使得这一优势荡然无存。

But earlier this week, Beijing introduced a 50% electricity subsidy for data centers using Chinese made chips, making that advantage irrelevant.

当然，这些言论发表前不久，特朗普总统刚刚彻底禁止英伟达向中国出口Blackwell芯片——即使是降级版本，因此这些言论可能被解读为充满挫败感和/或利己主义。

Now these comments, of course, come shortly after President Trump shut the door on sending NVIDIA's Blackwell chips to China even in a downgraded form, meaning that it's possible to read these comments as frustrated and or self interested.

与此同时，黄仁勋全年都在将这种情况描述为芯片禁令的必然结果。

At the same time, Huang has consistently presented this as the inevitable outcome of chip bans all year.

事件发酵后，英伟达和黄仁勋发布澄清声明，其中黄仁勋表示：'正如我一直所说，中国在AI领域仅落后美国纳秒级'。

After the story started to get traction, Nvidia and Huang released a clarifying statement where Jensen commented, As I have long said, China is nanoseconds behind America in AI.

美国必须通过加速发展并赢得全球开发者来取得胜利，这至关重要。

It's vital that America wins by racing ahead and winning developers worldwide.

很多人都看出这两组言论是相互关联的。

Many people got that these two sets of comments were interrelated.

DC投资者写道：'OpenAI现在希望联邦政府为其数据中心建设提供资金保障，而黄仁勋更公开地谈论如果我们不快速扩大能源产能就会输给中国'。

DC Investor writes, OpenAI now wants a federal funding backstop for their data center build out, and Jensen talking much more openly about how we will lose to China if we don't rapidly build out our energy capacity.

越来越明显的是，这些公司之间看似资本主义循环投资的表象，实际上是在执行国家优先政策。

It's becoming very obvious that what appears like broken capitalism circular investment between these firms is actually implementation of a national policy priority.

他们试图让这一切看起来是通过自由市场完成的，但这很可能是以尚未被广泛讨论的方式进行协调的。

They are trying to make it look like it's being done through the free market, but this is very likely orchestrated in ways that are not being broadly discussed yet.

简而言之，AI很可能不是泡沫，而是一个自由市场版的曼哈顿计划。

TLDR AI is likely not a bubble, it's a free market Manhattan project.

几周前gmoney.e写道：'如果AI竞赛被视为国家安全问题和赢家通吃的局面，那么只需看看军费开支就知道这场竞赛不会很快放缓'。

A couple of weeks ago gmoney.e wrote: If the AI race is considered an issue of national security and winner take all, then one needs to only look at military spending to see that it won't slow down anytime soon.

首次有意义的下降直到冷战结束才出现。

The first meaningful downtick wasn't until the Cold War ended.

投入AI的每一分钱都物有所值，因为胜利的回报实在太高。

Every dollar spent on AI can be justified because the spoils of winning are so high.

GMoney昨天刚跟进写道：如果你将这场竞赛视为美国对中国的较量，那么为保障AGIASI投入的资金将毫无上限可言。

GMoney followed that up just yesterday, writing, If you view this as a race of USA versus China, there is literally no ceiling on the money that will be spent to secure AGIASI.

中美两国的掌权者——无论是美国两党还是中国领导层——都愿意通过通货膨胀牺牲大多数国民的生活，以赢得这场竞赛。

And everyone in power, both sides of the aisle in The US and leadership in China, will be willing to sacrifice the lives of most of their citizens through inflation to win this race.

所以这绝不能像2007年始于Countrywide破产，两年多后以雷曼兄弟倒闭告终的次贷危机那样发展。

So this can't be like the housing crisis which started in 2007, with the collapse of Countrywide and culminated more than two years later with the collapse of Lehman.

如果AI投资连续两年崩盘，我们就完蛋了。

If we have an AI spending collapse for two years, we will be toast.

因此这次纾困将在几周内——甚至几天内——实施。

So this bailout will happen in a matter of weeks, if not days.

耐人寻味的是，AI高管们现在把潜台词都挑明了。

The interesting thing here is AI execs are now saying the quiet part out loud.

他们正在亮出底牌。

They're tipping their hands.

他们直言我们需要天文数字的资金才有一线希望。

They're telling us we need ungodly amounts of money in order to have a shot.

通常政府要印钞时，总会用些稍显隐晦的说辞。

Usually when the government will print our Es, they call it something slightly less obvious.

过去二十年我们见识过TARP（问题资产救助计划）、QE（量化宽松）、Operation Twist（扭转操作）、SIVB（硅谷银行）救助等各种字母汤式的纾困方案。

We got that with TARP, QE, Operation Twist, SIVB bailout, or whatever other alphabet soup bailouts we've gotten over the last twenty years.

在这个背景下，昨晚Mamdani的胜利就显得合情合理。

Using this as a backdrop, the Mamdani victory really makes sense last night.

如果金融高管能在2008年获得救助，AI高管甚至在需要之前就呼吁纾困，那普通公民凭什么不能得到救助？

If financial execs got bailouts in 2008, and AI execs are already calling for bailouts before they even need it, then why shouldn't the average citizen get the bailout?

现在至少有一位OpenAI人士表示，他们认为政府兜底不应成为政策。

Now at least one OpenAI voice said that they didn't think that government backstopping should be the policy.

鲁恩在周三晚间写道：我认为美国政府不应为数据中心贷款提供担保，也不该将资金输送给英伟达这种毛利率90%的企业。

Rune wrote on Wednesday night: I don't think the USG should backstop data center loans or funnel money to Nvidia's 90% gross margin business.

相反，他们应该通过补贴和完善规则来大幅降低能源生产门槛。

Instead, they should make it really easy to produce energy with subsidies and better rules.

建设对全民有益、能让我们与中国持平的基础设施。

Infrastructure that's beneficial for all and puts us at parity with China.

金融分析机构'Finance a Lot'指出普通美国人不太可能被'赢得AI竞赛'的理念打动。

Finance a Lot pointed out why the average American is unlikely to be compelled by the idea of winning the AI race.

他们写道：'他们定义过什么是赢得AI竞赛吗？'

They wrote, Have they defined what winning the AI race is?

是第一个实现25%失业率的国家吗？

Is it the first country to reach 25% unemployment?

评论员康纳·桑德指出：'这个周期过后即将到来的政治反扑，对40岁以上人群来说简直显而易见。'

Commentator Connor Sand points out, The epic political backlash coming on the other side of this cycle is so obvious for anyone over the age of 40.

我们花了十五年时间把银行塑造成反派角色。

We turned banks into the bad guys for fifteen years.

祝AI从业者们好运吧。

Good luck to the AI folks.

我们正在补贴那些即将夺走你们工作岗位的公司，而它们尝试这样做时还会推高你们的电费。

We're subsidizing the companies who are going to take your job, and you'll pay higher electricity prices as they try to do so.

这引出了我对整件事的质疑，也是为何尽管身处AI行业，我却要加倍强调这个批评。

And this brings me to my issue with all of this, and why despite being in the AI industry I am doubling down on this particular critique.

我们当前所处的世界存在两种完全割裂的经济体系。

We are operating in a world right now where there are two entirely separate economies.

一边是蓬勃发展的AI经济，另一边——说得客气些——则是毫无起色的其他经济领域。

There is the AI economy that is booming and there is the rest of the economy which is putting it charitably not.

在我们的小天地之外，最激烈的政治辩论之一就是政府是否要继续为食品券项目拨款。

Outside of our little corner of the world, some of the biggest political debates going on are whether the government is going to continue to pay for food stamps.

与此同时，美国首次购房者的平均年龄

Meanwhile, the average age of first time U.

在

S.

过去三年间已从历史平均的30岁左右飙升至40岁。

Homebuyers has surged in the last three years to 40 from a historical average closer to 30.

而纽约市刚刚选出了一位社会主义市长。

And New York City just elected a socialist mayor.

现在你可能会问，OpenAI在发表声明时真的需要考虑所有这些背景因素吗？

Now, you might say, is it really OpenAI's job to have to think about all of this context when they make statements?

毕竟他们不过是个初出茅庐的初创公司。

They're just a scrappy startup after all.

不，他们早已不是了。

No, no they are not.

OpenAI，如果你们在听，请记住：

And OpenAI, if you are listening, please hear this.

你们已不再是那个可以口无遮拦的硅谷神奇初创公司了。

You are no longer some wonderkin startup from Silicon Valley that can afford to be flippant.

美国所有媒体对待OpenAI领导人的每句话，都像对待白宫或参议院领袖等重要人物的言论一样。

Every media outlet in The United States covers every comment out of any OpenAI leader's mouth like they are comments from the White House or a Senate leader or some other wildly significant actor.

你们不能一方面努力将自己塑造成未来最重要的公司，另一方面却忽视这个角色带来的传播影响。

You don't get to, on the one hand, work incredibly hard to position yourself as the most essential company of the future and then not understand the communications implications of that role.

那种可以信口开河、措辞轻率的时代已经结束了。

The time where you guys get to just show up and speak loosely in ill considered terms is over.

除非你们想让AI行业在未来数年都遭受严厉的政治报复。

Unless you want to just doom the AI industry to intense political retribution for years.

我很抱歉，但现实是现在你不仅对自己的公司负有责任，还对整个行业的其他人负有责任。

I'm sorry, but the reality is now you have a responsibility not just to your own company, but to everyone else in this industry.

而你做不到这一点。

And you can't do this.

你不能出现在布拉德·格斯特纳的播客上表现得刻薄傲慢，即便我理解这种挫败感从何而来。

You can't show up on Brad Gerstner's podcast and get snide and condescending, even if I understand where the frustration comes from.

你不能出现在《华尔街日报》上开始随意抛出诸如政府兜底这样的术语和概念。

You can't show up at The Wall Street Journal and start banding about terms and concepts like government backstops.

当你这样做时，你不仅害了自己，也害了我们所有人。

When you do that, you're screwing yourselves and you're screwing the rest of us.

所以拜托，我恳求你，开始意识到你所处的角色。

So please, I am imploring you, start to appreciate the role that you are in.

你不再是一家初创公司了，至少在沟通方面不是，你不能再表现得像初创公司一样。

You are no longer a startup, not when it comes to communications, and you don't get to act like it anymore.

今天的AI每日简报就到这里。

That's going to do it for today's AI Daily Brief.

下次见，祝安好。

Until next time, peace.