本集简介
双语字幕
仅展示文本字幕，不包含中文音频；想边听边看，请使用 Bayt 播客 App。
OpenAI将在2026年大力进军企业市场。
OpenAI is going big on enterprise in 2026.
迪士尼将魔法带入AI视频创作领域。
Disney brings the magic to AI video creation.
在甲骨文和博通股价暴跌后，AI基础设施交易剧烈震荡，而Meta的超智能团队与公司其他部门存在分歧，真的如此吗？
The AI infrastructure trade is wobbling hard after tumbles from Oracle and Broadcom, and Meta's superintelligence team is at odds with the rest of the company, or is it?
广告之后马上回来。
That's coming up right after this.
Capital One的技术团队不仅在讨论多智能体AI。
Capital One's tech team isn't just talking about multiegenthic AI.
他们已经部署了一个。
They already deployed one.
它被称为聊天礼宾服务，正在简化购车流程。
It's called chat concierge, and it's simplifying car shopping.
通过自我反思和分层推理结合实时API检查，它不仅帮助买家找到心仪的车辆。
Using self reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love.
它能帮你预约试驾、获得贷款预批，并估算旧车置换价值。
It helps schedule a test drive, get preapproved for financing, and estimate trade and value.
先进、直观且已投入使用。
Advanced, intuitive, and deployed.
这就是他们的技术实力。
That's how they stack.
这就是Capital One的技术。
That's technology at Capital One.
事实上，AI安全就是身份安全。
The truth is AI security is identity security.
AI代理不仅仅是一段代码。
An AI agent isn't just a piece of code.
它是你数字生态系统中的一等公民，需要得到相应的对待。
It's a first class citizen in your digital ecosystem, and it needs to be treated like one.
这就是为什么Okta要率先保护这些AI代理的安全。
That's why Okta is taking the lead to secure these AI agents.
解锁这层新防护的关键，在于身份安全架构。
The key to unlocking this new layer of protection, an identity security fabric.
企业需要一套统一全面的方案，用一致的政策和监督机制保护每个身份——无论是人类还是机器。
Organizations need a unified comprehensive approach that protects every identity, human or machine, with consistent policies and oversight.
别等到发生安全事件才意识到你的AI代理是个巨大的盲区。
Don't wait for a security incident to realize your AI agents are a massive blind spot.
了解Okta的身份安全架构如何帮助你保护下一代身份，包括你的AI代理。
Learn how Okta's identity security fabric can help you secure the next generation of identities, including your AI agents.
访问okta.com。
Visit okta.com.
网址是okta.com。
That's okta.com.
欢迎收听《大科技》播客周五版，我们将以一贯冷静细致的风格为您解析新闻。
Welcome to Big Technology podcast Friday edition where we break down the news in our traditional cool headed and nuanced format.
今天我们为您准备了精彩内容。
We have a great show for you today.
我们将深入探讨OpenAI的山姆·奥特曼与纽约顶级编辑及媒体CEO们的会议内容，为您揭秘OpenAI对2026年的规划。
We are going to go inside a meeting with OpenAI Sam Altman and top editors and media CEOs in New York to tell you what OpenAI is planning for 2026.
我们还将讨论OpenAI与迪士尼的重大合作，以及这对AI内容创作领域的整体影响。
We're also gonna discuss OpenAI and Disney's big deal and what it means for AI content creation in general.
同时我们注意到AI基础设施领域出现了一些波动，我们会探讨这背后的含义。
We're also seeing some tumbles in the AI infrastructure trade, so we'll talk about what that means.
当然，我们也会剖析《纽约时报》这篇引人入胜的报道——关于Meta超级智能团队与公司其他部门的分歧，我们将尽力解析报道中的说法是否属实。
And, of course, we're gonna talk about this fascinating story from the New York Times that says Meta's super intelligence team is at odds with the rest of the company, and we will do our best to dissect the wording in that story and tell you whether that's actually true or not.
每周五与我们共同探讨这些话题的是《边际》栏目的冉冉·罗伊。
So joining us as always on Fridays to do it is Ranjan Roy of Margins.
冉，很高兴见到你。
Ranjan, great to see you.
欢迎回来。
Welcome back.
很高兴见到你。
Good to see you.
当我能通过迪士尼与OpenAI的新协议，在Sora上对米老鼠做些奇怪的事情时，我就要宣布人工通用智能实现了。
When I am able to do weird things with Mickey Mouse on Sora, thanks to this new Disney and OpenAI deal, I will declare artificial general intelligence.
所以，我准备好了。
So, I'm ready.
我准备好了。
I'm ready.
看到OpenAI和迪士尼的协议时，我第一个念头就是直奔Sora平台，试图生成米老鼠吸食违禁药物的视频——这想法是不是有问题？
Is it a problem that my first thought when I saw this OpenAI, Disney deal was like, I'm gonna go straight to Sora and try to get a video of Mickey Mouse doing some form of illicit drugs.
我还没付诸行动。
I didn't do it.
我知道内容过滤器会阻止我，但这确实是我冒出来的第一念头。
I I knew that the content filter would get me, but that was my initial desire right off the bat.
但我们会实现的。
But we will do it.
多亏这次合作，我们很快就能实现。
We will do it very soon, thanks to this deal.
还是说我们不会？
Or will we?
好的，我们稍后会深入讨论这个话题。
Alright, so we're gonna get into that in a moment.
但让我们从本周的播客开始，带您走进周一中午时分的Rosemary's Midtown餐厅。
But let us begin our podcast this week by bringing you into the Rosemary's Midtown on Monday around noon time.
在纽约中央车站附近的Rosemary's Midtown高级餐厅里，山姆·奥特曼与来自《大西洋月刊》、《纽约客》和《纽约时报》等媒体的编辑及CEO们共进午餐。
And there at Rosemary's Midtown, a fine establishment near Grand Central Station in New York City, Sam Altman had lunch with editors and CEOs from places like the Atlantic, The New Yorker, and The New York Times.
根据我的报道，与会者包括《纽约时报》的安德鲁·罗斯·索尔金、《纽约客》主编大卫·雷姆尼克、《大西洋月刊》CEO尼古拉斯·汤普森、《财富》杂志主编艾莉森·钱特尔等众多人士。
This is a group that included, according to my reporting, Andrew Ross Sorkin, New Yorker editor David Remnick, Atlantic CEO Nicholas Thompson, Fortune editor in chief Alison Chantel, and many others.
这群人围坐进行了一场广泛而散漫的谈话，山姆·奥特曼则一如既往地展现出他迷人的亲和力。
And this group sits around for a unwieldy and wide ranging conversation and Sam Altman is his typical disarming and charming self.
除了雷姆尼克询问奥特曼是否觉得自己像奥本海默这个可能最有趣的点（我们稍后会谈到），他们还问了OpenAI的首要任务是什么。
And the most interesting thing that happened at this meeting outside maybe of Remnick asking, Sam Altman if he felt like Oppenheimer, which we can get to, is they asked what is the priority gonna be for OpenAI?
根据几位知情人士向我透露的信息，奥特曼表示，即便不是前八大重点，企业级服务也将会是OpenAI明年的重要优先方向。
And what Altman said, according to a couple of people who I've heard, who I've spoken with, who know what was said in that room, Altman said that that if not the top eight, a very big priority next year for OpenAI is going to be enterprise.
为企业构建人工智能，这很有趣，因为如果你想想OpenAI，它一直是70%消费者、30%企业的比例分配。
Building AI for businesses, which is interesting because if you think about OpenAI, it's been about a 7030% split, 70% consumer, 30% enterprise.
Anthropic确实是企业领域的领军公司，但这个类别是有史以来增长最快的软件类别。
Anthropic has really been the company that's led in enterprise, but the category is the fastest growing software category in history.
根据Gartner预测，明年该领域收入预计将达到375亿美元。
It's expected according to Gartner to bring in 37,500,000,000 in revenue next year.
企业级人工智能，这个数字在2022年还是零或者说基本为零。
Enterprise AI, that's up from zero in 2022 or effectively zero in 2022.
Ranjan，现在交给你了。
Ranjan, turn it over to you.
你怎么看这个消息？
What do you make of this news?
这是OpenAI的正确决定吗？它让你感到意外吗？
Is it the right decision from OpenAI and did it surprise you at all?
嗯，对此我有非常强烈的看法，长期听众会知道我在writer.com工作，我们专注于企业AI并与多家财富500强公司合作。
Well, have very strong thoughts on this given long time listeners will know I work for writer.com and our focus is enterprise AI and work with a number of fortune 500 companies.
因此我对这个领域的现状有着非常清晰的认知。
So I have a very clear view into what this world looks like.
当我听到这个消息时，虽然确实制造了轰动性标题，但对我来说只是再次印证了他们缺乏专注度。
When I heard this news, it's definitely, it makes for a splashy headline, but all it did for me is kind of just affirm the lack of focus.
我们上周就多次讨论过OpenAI缺乏专注度的问题——他们是要开发AI云服务吗？
The lack of focus we talked a lot about for OpenAI last week, are they gonna develop an AI cloud?
还是要转型成为消费级设备公司？
Is it gonna be a consumer devices company?
他们刚聘请了前Slack CEO，显然正在布局这个领域。但对我来说，这个新闻最有趣的部分是Sam Altman的发言，而不是OpenAI官方的其他声明。
Any of the above, they did just hire the former Slack CEO, so that's obviously they're making moves into this, but to me the most interesting part of this headline is that Sam Altman, that's what he said rather than any of the other things that OpenAI is saying.
我当然可以深入分析我对明年企业级AI发展的看法，但就Sam而言，这感觉更像是他当时随口说出的一句话，反映了他那一刻的想法。
I can definitely get into, you know, what I view is going to happen in the enterprise world with AI next year, but but specific to Sam, it just felt like another off the cuff comment at that moment with for whatever was on his mind at that, you know, at that time.
好吧。
Alright.
我不得不强烈反对你的观点，不仅因为那个轰动性标题就是我写的。
I'm gonna disagree with you strongly here, not only because I'm the one that wrote that splashy headline.
这个我知道。
This is I know.
重大科技独家新闻，我们会在节目说明中附上链接。
Big technology scoop, which we'll we'll link in the show notes.
但还有另外两个原因。
But for two other reasons.
一是OpenAI联合创始人Greg Brockman看了我分享的推文故事，他几乎三重确认并写道'企业AI将成为2026年的重大主题'。
One is Greg Brockman, the OpenAI co founder, looked at the tweet that I shared of my story and he sort of triple confirmed it and wrote Enterprise AI is going to be a huge theme of 2026.
我不认为这是随口一说。
I don't think that's an off the cuff thing.
另外本周——我们稍后会详细讨论——OpenAI发布了一个新模型。
Then And also this week, and we'll get into it in a moment, but OpenAI had a, had a new model ship.
它叫GPT-5.2。
It's called GPT 5.2.
而GPT-5.2的核心就是帮助你在商业应用场景中更好地使用AI。
And GPT 5.2 is all about helping you do a better job, of working with AI for business use cases.
这包括，比如你可以通过提示生成劳动力规划表，更高效地处理待办事项，开发更深入的企业级应用。
And that includes, you know, examples where you could prompt a workforce planning planning sheet, do a better job with to dos, do like better, deeper enterprise applications.
所以这很有趣，我注意到你有些分心，对吧？
So it is interesting, I I hear your lack of focus, right?
他们是家面向消费者的公司吗？
Are they a consumer company?
还是家数据公司？
Are they a data company?
或是机器人公司？
A robotics company?
他们正在和乔尼·艾维合作这个项目吗？
Are they doing this with Johnny Ive?
他们主攻企业市场吗？
Are they enterprise?
但我认为他们确实非常重视企业市场，我就说这一点，把话题交还给你。
But I I think they're really serious about enterprise and I'll just say one thing and turn it back to you.
这就是为什么从逻辑角度来说，这对我很有意义。
This is why it makes sense to me from a logical standpoint.
这是一项非常大的业务，而且ChatGPT在赚钱方面做得很好。
This is a very big business and, you know, ChatGPT has done a good job making the money.
他们今年可能会赚200亿美元，或者年底时达到200亿美元的年度运行率，但他们明年需要展示出实质性的收入增长。
They're gonna make maybe 20,000,000,000, this year or have a 20,000,000,000 run rate at the end of the year, but they're gonna need to show substantial revenue growth next year.
而实现这一目标的方法之一就是进军企业领域，那里才是真正的金矿。
And one way to do that is by attacking this enterprise, area which is, which is where the money is.
不。
No.
我同意这将是2026年最大的机遇，而且再次说明，这也是在为自己说话，但我真心相信的是，就像Cursor和AI编程今年彻底爆发一样，我认为这股浪潮将传递到商业终端用户，企业最终将看到过去几年一直存在但未能实现的AI潜力。
I I agree it's gonna be the biggest opportunity of 2026 and I again, that's talking my own book as well, but but what I genuinely believe is gonna happen is the way Cursor and AI coding just completely exploded this year, I think is going to pass on to business end users and enterprises are going to actually see the promise of AI that has been there for the last few years and has not been realized.
所以我确实认为这是一个巨大的机会。
So so I do believe it's a massive opportunity.
但我在读你的报道时想的是，谷歌是如何应对这个问题的？
But what I was thinking about when I read your story is, you know, how did Google approach this?
微软企业版最初是如何起步的，后来又是如何逐步发展出一些面向消费者的应用？
How did Microsoft Enterprise first from the beginning and then kind of working their way towards some kind of consumer applications?
谷歌则走了相反的路，先建立了庞大的消费者业务，然后将谷歌云打造成庞然大物，而OpenAI却试图同时推进所有业务。
Google going in the opposite way, having just a massive consumer business and then building Google Cloud into this behemoth, but OpenAI is trying to do everything at once.
我不认为你能放弃消费者业务，同时在企业级这个完全不同的领域并行发展。
And I don't believe you can't give up the consumer business and to try to do this in parallel when enterprise is a completely different beast.
再说，Slack的CEO虽然在场，但最终决策权会落在谁手里？
And again, you have the Slack CEO, but who's gonna get the final word over any decision?
当Sam Altman想用AI情色内容来提升用户参与度时，这会吓坏所有企业的CIO，那么谁对这类事有最终决定权？
When Sam Altman wants to do AI erotica to actually juice engagement numbers, and that's gonna freak out every enterprise CIO, who's gonna get the last word on that?
嗯，这是个好问题。
Well, that's a great question.
顺便说下，我们确实得到了一些关于AI情色内容明年何时上线的明确消息。
By the way, we did get some, clarity about when AI erotica is coming sometime next year.
最重要的。
Most important.
期待这个。
Look forward to this.
米老鼠抽大麻和AI色情内容。
Mickey Mouse smoking weed and AI erotica.
这就是
That's the
这就是你想要的新旅程。
that's is the new trip you want.
好吗？
Okay?
还有企业AI。
And enterprise AI.
我是说，如果你能把它们都打包在一起，好吧，看起来是同一个引擎在运作。
I mean, if you can bundle it all together, but okay, it's look, it's the same engine that that does it all.
所以，也许这是件令人兴奋的事，但你说得确实有道理。
So, maybe that is, an exciting thing, but it is it is a good point.
面向企业销售时，必须构建企业级销售基础设施。
When you sell to enterprises, need to build out an enterprise sales infrastructure.
这是谷歌花了差不多十年时间才在谷歌云上实现的。
This is something that took Google, like, what, a decade to do with Google Cloud.
说起来容易做起来难，OpenAI必须解决这个问题。
It's much easier, said than done, and OpenAI is gonna have to figure this out.
我今天甚至接到一个从事企业AI销售的人的电话，他问我：你觉得他们会如何组建团队？
And I got a call today even from someone who works in enterprise AI sales and was like, what do you think is how do you think they're gonna build out a team there?
我当时就想，这正是他们必须解决的核心问题。
And I'm like, that's like the core the core problem that they're gonna have to deal with.
所以我们拭目以待吧。
So we'll see what happens there.
不过我们还是来看看他们做出这种转变的可能原因。
But let's go let's actually look at, like, why they might have made this shift.
可能是为了钱，也可能是其他原因——毕竟这个时间点恰逢谷歌Gemini在功能上实现匹敌，甚至在某些方面有所超越。
So maybe it's about money, maybe it's about something else because it does come on the heels of Google's Gemini effectively equaling and some might say surpassing and surpassing in some ways.
可以说，谷歌的Gemini目前基本上与GPT模型、也就是ChatGPT旗鼓相当。
Let's just say it's Google's Gemini is basically on par with, with chat with GPT right now, the GPT models.
这是我的观点。
Here's my thesis.
OpenAI一直在努力推进它认为可能直达AGI（人工通用智能）的路径，并相信自己能引领这一方向。
OpenAI had been working on, like, effectively what it believed could be a straight shot, towards AGI, which it believed it could lead.
但今年下半年发生了一件有趣的事。
Then an interesting thing has happened in the second half of this year.
首先，尽管模型在改进，但你开始看到模型的局限性。
First of all, even though models are improving, you're starting to see the limits of the model.
不再是全面进步，而是在特定领域的改进。
It's not broad improvement anymore, it's improvement in specific areas.
第二件事是模型已经商品化了。
And the second thing that's happened is the models have commoditized.
因此，既没有直达AGI的捷径，模型又已商品化，这意味着真正的竞争将围绕应用层面展开。
So you've had both no straight shot to AGI and the models are commoditized, and that means that the real competition is gonna be based off of the applications here.
午餐时萨姆·奥特曼说了一句很有意思的话：这不是训练问题。
And actually interestingly in the lunch, what Sam Altman said was it is not a training problem.
而是应用问题。
It is an application problem.
重点不在于模型的智能程度，而在于如何构建应用来最大限度发挥其智能。
It's not about the model's intelligence, it's about building the applications to get the most intelligence out of them.
因此我认为，我们当前看到的这些发展表明，OpenAI乃至整个AI领域都意识到：单纯开发更好的模型可能走不远，因为模型性能提升已开始趋缓。
And so that is, I think, what we're seeing right now with these developments is OpenAI and I think everybody else in the AI world is realizing that, you are maybe not gonna get that far developing a better model because they are starting to tighten up.
关键在于优先级排序——这就是为什么他们意识到：AGI或更聪明的模型并不能解决所有问题。
It is all about prioritization and that's why they've hit this point, they see that AGI isn't gonna solve or or even smarter models won't solve all the problems.
某些局限客观存在，所以他们现在●●●转向应用层面●●●（注：此处'front'译为'层面'更符合中文科技语境）
There's just some limits there, and now they're going to the application front.
这就是为什么你现在看到企业用例正从该公司涌现。
That's why you see the enterprise use case as the one bubbling up out of that company right now.
抱歉，Alex你能重复一下吗？
I'm sorry, can you repeat that Alex?
因为我在这里微笑，因为常听节目的人都知道这个问题——产品与应用相对于模型。
Because I'm smiling here because anyone who listens regularly knows this question product and application versus the model.
长期以来，我一直站在产品团队这边，很高兴Sam终于也转过弯来了。
For a long time, I've been on team team product, and it's nice that Sam is finally coming around.
Alex，你转变立场了吗？
Alex, are you coming around?
你现在是产品团队的人了吗？
Are you are you team product now?
嗯，这很有趣，因为我这周上了CNBC节目，有人问我，既然谷歌可能有了更好的模型，这是否意味着谷歌已经或正在赢得AI竞赛？
Well, it's so interesting because, I was on CNBC this week and I was asked, you know, now that Google has what might be a better model, is does that mean Google's won or is winning the AI race?
我强忍住了自尊心。
And I bottled up my pride.
我咽了口唾沫。
I took a gulp.
然后我在全国电视上说了这句话。
And on national television, I said this.
问题难道不在于谁拥有更好的模型吗？
Isn't the issue who has the better model?
而目前市场反馈可能是Gemini更胜一筹。
And right now, the marketplace is saying it's probably Gemini.
所以我认为现阶段绝对不取决于谁拥有更好的模型。
So I would say it absolutely isn't who has the better model right now.
我认为当前关键是如何利用这个模型并实现产品化。
I think right now what matters is what you do with that model and how you distribute it.
我可能得合上笔记本宣布工作完成了。
I I might have to shut shut off my laptop and just just my work here is done.
我结束了。
I'm done.
具体到OpenAI这个案例，我还是觉得这像是Sam在表态。
I mean, so specific to OpenAI in this case though, I I still feel this is kinda Sam saying things.
因为这根本不可能是他们的核心赌注。
Because like he that cannot be the bet.
你不可能筹集这么多资金，长期发表这些言论，然后突然来个180度大转弯就直接走人。
You can't raise this much money and, like, say all of these things for so long and then just kind of about face and just go.
我说这话的同时也承认OpenAI在产品层面确实做得非常出色。
And and I say this recognizing OpenAI has been great at the product layer.
我一直认为这实际上是他们的一大优势，比如他们在聊天界面上的创新，真正开创了这一模式。
Like, I think I've always believed that's actually been one of their stronger suits in terms of, like, they innovated on the they created the chat UI and made it a thing.
所以他们确实强化了这方面，但我的意思是，如果真是这样，他们根本不值那个估值。
So so they have strengthened that, but, I mean, if that's the case, they're not worth what they're supposed to be worth, just simply.
作为一家应用公司，他们根本算不上是OpenAI。
Like, as an application company, they're not they're not OpenAI.
他们不值那个价。
They're not worth that.
所以我觉得，这对我来说依然很意外。
So I think, like, it's still surprising to me again.
这就是为什么我没有完全深思熟虑这些声明，当时只是本能地做出了反应，这是我的真实感受。
That's why I'm, like, not thinking fully through these statements and just kind of like reacting at that moment is how I feel and read this.
好的。
Okay.
所以我想解释一下为什么我暂时加入了产品团队，以及我认为OpenAI正在发生什么——情况并不像你想象的那么糟糕。
So I wanna both explain why I've come along to team product, for the moment and explain what's happening, what I think is happening with OpenAI and it's not as dreadful as you think.
明白。
Okay.
我现在转向产品是因为模型的重要性已经下降——之前我在模型团队时，我们看到模型改进带来的实际收益：能力增强、幻觉减少。GPT5.2的设计也体现了这点，它能实现更多企业级功能。
I've come along to this the product now matters more than the model because previously when I was on team model, we were seeing real gains when the models got better and when the models got better they could just do more stuff and they hallucinated less and you're even seeing that with a little bit with the way that, this GPT 5.2 model has been designed, that it can do more more things for enterprise that it previously couldn't.
但坦率地说，模型的核心改进正在趋于平缓。
But it wouldn't be a, a real stretch for me to say that top line improvement of the models is leveling out.
实际上我们已看不到那些巨大飞跃。
That effectively we're not seeing those giant leaps.
当这种飞跃消失时，'我们终将实现AGI并由它解决一切'的承诺就开始褪色。
And when you don't see those giant leaps, you basically have started to lose this promise of like, well, we're just gonna get to AGI and then AGI is gonna solve everything.
现在的情况是：前沿模型已具备相当智能，重点是如何将这种智能引导到不同的商业应用中。
At this point, you have a model that's that's, you know, the frontier models are very intelligent and at this point what you need to do is take that intelligence and channel it into different applications for business.
我认为OpenAI确实配得上这个估值，即便——说来好笑，我现在完全是在复述你的观点。
And I think OpenAI could be worth the valuation, even if it and I'm really this is funny because I'm really channeling you here.
即便它无法实现AGI，也完全值得这个估值，因为如果能让其模型在众多专业领域发挥价值，这才是它获得经济提升的关键。
Even if it doesn't get to AGI, it could really, merit the valuation because if it makes its models valuable and applicable in many different professions, that's where it's gonna see the economic boost.
我们周三刚和《商业内幕》的玛利亚·拉塞尔讨论过这个话题，她专门做了一期关于AI如何改变法律领域的节目。
We just spoke about this on Wednesday with Malia Russell from Business Insider doing an entire episode about how AI can change the legal field.
这只是一个专业服务领域的例子。
That's just one, you know, professional services area.
随着这些应用的完善，它能应用于咨询行业吗？
And as these applications get better, can it do it for consulting?
能应用于会计领域吗？
Can it do it for accounting?
能在研究领域实现类似的应用吗？
Can it have similar applications in research?
能在医疗领域实现类似的应用吗？
Can it have similar applications in medicine?
所有这些可能性现在都已摆在桌面上，如果它能在现有智能模型基础上增加应用和协调层面的价值，或许这正是它成为预期中强大价值引擎的契机。
All this is on the table now, and if it just adds that level of application and orchestration on top of the very smart models that it has today, maybe that's where it becomes this valuable powerhouse that it's been expecting to be.
你的两个问题。
Your Two questions.
第一，为什么山姆·奥特曼在与布拉德·加斯特纳交谈时会有那种情绪爆发，还说什么'别质疑我们的收入'。
One, why did Sam Altman, when he was talking to Brad Garstner and had that kind of outburst and talk about like, don't question our revenue.
他说'我们会达到1000亿零28美元'。
We're gonna get to a 100,000,000,028.
但为什么他没提企业市场？
Like, why didn't he say enterprise?
他提到了AI云、消费级设备。
He mentioned AI cloud, consumer devices.
但当时就是没提企业市场。
He did not say enterprise at the time.
这个我可没法回答你。
That I can't tell you.
我们可以达成共识，山姆在那次采访中的表现可能是因为他那天心情不好，毕竟他看到谷歌表现不错，这也是我们推测的。
We can we can we both agree that whatever Sam did on that interview was maybe he was having a bad day because he saw that Google was getting good, which we've surmised here.
但我们别把这当作官方表态。
But let's not take that as the official Okay.
好的。
Okay.
你知道，这是第一点。
That you know, that's number one.
不要把发脾气当作OpenAI的官方产品策略。
Temper tantrum as the official product policy of OpenAI.
现在第二点，山姆说2028年营收要达到1000亿美元。
Now now number two, Sam says a 100,000,000,000 in 2028 in revenue.
你认为这个目标如何分解实现？
How do you see that breaking down?
因为我之前问过这个问题，听众反响很好，就是关于你说的1000亿目标——我查过报道但没找到任何关于他们业务实际预测和分解的具体信息，特别是考虑到这些新想法，比如制药行业这种莎拉·弗莱尔提到的药物开发，显然他们正在考虑。
Because it I I had asked this and we got a, like, a very, like, good reaction from listeners and around, like, this question of when you say 100,000,000,000, and I we've look I've looked for reporting, haven't seen anything around what are their actual projections about what their business looks like and breaks down, especially as all these new ideas, the pharma idea, drug development that Sarah Fryer had mentioned, like, it's clearly on their mind.
你们总得有个计划吧。
You gotta have some plan.
现在是2025年12月。
It's December 2025.
如果你说的是未来两三年，那应该已经制定了某种计划。
If you're talking about two to three years now, there's some plan in place.
你认为达到1000亿规模时业务会是什么样子？
What do you think that business is gonna look like at a 100,000,000,000?
其实我跟你说过，我们问过他们这些预测数据是从哪来的？
Well, I've told you that, like, we asked, like, where did they get these projections from?
我觉得他们就是在Excel表格上撒了点仙尘，指望这样就能实现目标。
And I think that's used they've been sprinkling some fairy dust on the, on an Excel spreadsheet and hoping that gets them there.
我不知道。
I don't know.
我觉得他们自己也不清楚具体要怎么实现。
I don't think they know exactly what's gonna go there.
我认为他们是在快速押注，就像之前那样，你知道的，扩大这些模型规模，增加更多数据，增强计算能力，就能快速提升使用这些模型做事的能力，并最终带动经济活动。
I think they're betting that rapidly that they've bet it previously that, you know, you scale up these models, you add more data, you add more compute, and you will get rapidly rapid improvement in your ability in this your ability to to do things with these models and they will lead to economic activity.
所以我不明白他们如何能这么快达到那个数字，因为我们俩都清楚这点——你在这个领域工作，我也做过相关报道。
So, I don't see how they get to that number that quickly because once we both know this because you're working in this, I've done reporting on this.
一旦进入企业市场，会发生什么？
Once you get into enterprises, what happens?
即使你有个表现优异的产品，也得经历业务转型，更糟的是还要走审批流程。
Even if you have a product that works really well, you have to engage in business transformation and even worse approvals.
而这些都不可能快速推进。
And that does not move fast.
这正是我想说的，如果这真是通往2027年的路径，他们需要搭建的基础设施规模——虽然我确信他们已经开始行动，比如聘请了Denise Dresser——但这仍是个巨大的赌注。
That's what I'm saying that if that truly is the path to let's say 2027, the amount like that they're gonna have to build the infrastructure, which I'm sure they are starting to with the hiring of Denise Dresser, but still, like, that's a big bet.
如果制药收入来自药物开发和合作，你就必须招人。
If pharma revenue from drug development and partnerships, you you gotta hire people.
你得把整个业务体系搭建起来。
You gotta, like, build out that entire business.
如果他们真的想把这些发展成真正的业务线，那就不仅仅是产品那么简单了。
So if they're really getting to this point where these are true businesses that they wanna build in lines of revenue, it's not just product at all.
甚至也不仅仅像你说的只是应用层面。
It's not even just application, as you said.
是的。
Yeah.
这是一整套商业计划。
It's it's a whole business plan.
需要人员配置。
It's people.
需要销售团队。
It's a Salesforce.
需要所有这些配套。
It's all of this stuff.
所以在听到或看到更具体的进展之前，这部分真的很难揣测他们实际的想法。
So so that's the part that until I hear or see more concrete things, it's it's it's just tough to read in terms of, like, what they're actually thinking on this.
再次强调，很明显这是基于招聘做出的赌注，但至于这个项目在OpenAI内部优先级排序中究竟处于什么位置——是米老鼠抽大麻的Sora项目还是其他什么——对我来说仍然不明确。
Again, it's clear this is a bet based on the hires, but how they're actually where this sits in the Mickey Mouse smoking weed on Sora or whatever else, in the totem pole of priorities within, OpenAI, it's still unclear to me.
所以我实际上问过Fiji Simo，我们和她开了个新闻发布会讨论这个新模型。
So I actually asked, Fiji Simo, we had a press call with her about this new model.
我问她企业市场对公司有多重要。
I asked her how important enterprise is to the company.
她说企业市场对我们绝对至关重要。
She says enterprise is absolutely critical for us.
我们将突破性技术转化为对人类有用的智能，而专业领域是创造价值并将智能转化为全球实用性的重要驱动力。
We take our breakthroughs and turn them into intelligence that's useful for people, and the professional space is a huge driver of creating value and turning that intelligence into usefulness for the world.
所以他们确实重视这个领域，但是
So they are into this, but
但这个优先级和广告业务相比如何？
But where where does that sit next to ads?
比如
Like
嗯，广告业务目前暂停了。
Well, ads are paused.
但这个...好吧。
But this Okay.
抱歉。
Sorry.
我觉得你触及到问题的核心了。
I I think you're hitting on the real issue here.
这个问题我也向Fiji提过，因为就像处理这类事情时的惯例，我问了她一个包含两部分的问题。
And and this is something I also brought up to Fiji because I asked her a two parter as you do on these things.
常规操作。
As you do.
人之常情。
As one does.
企业业务有多重要？
How important is enterprise?
现在风险在于，当你为企业设计时，如果企业级应用真是你想要的，你如何确保这款广受欢迎的聊天机器人依然是普通用户愿意在消费场景中使用的产品。
And, and now here's the risk, when you design for enterprise, if enterprise is really what you want, how do you then make sure that the chatbot that's so popular is still something that regular people want to use for consumer use cases.
以下是Fiji的回答。
Here's Fiji's answer.
她说，你关于企业级和消费级产品存在差异的观点是正确的。
She said, you're correct that some things are different between enterprise and consumer.
我们认为整体智能水平的提升对所有人都有利，能让整体水平水涨船高。
We think that increasing intelligence overall benefits everyone and raises all boats.
如果我们正在打造个人助手，那么重点会是解决购物、旅行等各类生活场景的需求。
If we're building a personal assistant, it's going to be more important to figure out how to do shopping, how to do travel, how to do all of these things.
而在企业级应用上，我们更关注复杂、长期且具有模糊性的任务。
Whereas on the enterprise we are much more focused on complex long horizons and ambiguous tasks.
我记得很多人认为GPT-5在人格化表现上相比ChatGPT-4是一种降级。
I I remember GPT-five was viewed by many as a personality downgrade over chat GPT-four o.
也许是因为它不再那么谄媚，或者当你问它外面天气如何时，它会反问你是否需要帮你查找附近五家卖雨伞的地方？
Maybe that's because it's not as sycophantic or maybe it's because, you know, when you ask it what the weather is out, it it asks you, do you want me to, like, find five places where you can find an umbrella nearby?
显然，在打造对企业既经济实用又人性化的产品之间需要取得平衡。
Clearly, there's a balance between making something economically valuable and usable for enterprise and also making it personable.
这就是为什么我认为他们必须将这两种模式分开，但风险在于可能出现这种自我蚕食效应：你转向企业市场，但每周使用ChatGPT作为个人应用的8亿或9亿用户会说，我不想用这个了。
This is why I believe they're really gonna have to separate those two models out, but that's the risk is you might have this this cannibalization effect where you go for enterprise, but then your 800 or 900,000,000 weekly users who are using ChatGPT as a personal whatever consumer application are like, well, I don't wanna use this.
要么用，要么走人。
Go or go somewhere else.
而Gemini一直在谷歌研发更亲切的助手。
And Gemini had been working on making a warmer assistant, over at Google.
这确实是个很好的观点。
That's actually a really good point.
比如，就拿阿谀奉承的性格和那些烦人的后续互动策略来说。
Like, let's take those two things, the sycophantic nature and the following follow on annoying engagement tactics.
这两者在企业市场都行不通。
Neither of those will work in the enterprise.
再说一次，尽管我相信每个商业用户都想要个谄媚的助手，但到某个节点，你必须得到正确答案，或者被引导至正确方向，而不是一味迎合你已有的倾向。
Again, like, even though I'm sure every business user wants a sycophantic assistant, still, at a certain point, you have to get the right answer or you have to, like, be led in the right direction as opposed to simply the one that you already wanted to go to.
如果你不这样做，将会产生负面后果。
And if you don't, there will be negative outcomes.
在用户参与度方面，比如，你想要五件额外的东西吗？
In terms of, like, user engagement, would you like five additional things?
这类东西根本行不通，因为当有人想做事时，他们只希望你完成这项工作，然后继续前进。
Like that kind of stuff just won't fly because when someone wants to do something, it's just please do this work for me, get it done, and let's move on.
而不是我想坐在这里被拉进这种参与的兔子洞。
Not I wanna just sit here and kind of be pulled into this rabbit hole of engagement.
所以试图同时在两个方向上开发产品，这很困难。
So so trying to develop the product simultaneously in those two directions, which they kind of have to, it's tough.
这很困难。
It's tough.
她甚至说他们想要两者兼得。
And she I mean, she even said they want it both.
他们想要鱼与熊掌兼得。
They want it both ways.
很明显他们正朝这个方向努力。
So it's clear they're shooting for that.
我认为最终他们不得不拆分成不同的模型。
I think eventually they're gonna have to split out into separate models.
我是说，难道要让同一个人——人们确实会这么做。
I mean, are you gonna have the same I mean, people are gonna do this.
他们会建立关系，与聊天机器人发生虚拟性行为，然后又要让它帮忙整理文件。
They're gonna have relationships and they're gonna have cybersex with their chat GPT and then they're gonna ask it to, like, organize their files.
我觉得这类功能可能需要两个独立的界面。
Like, I think you might need two separate windows for that stuff.
这个话题我就点到为止。
I'm just gonna leave that one there.
我只是说说而已。
I'm just saying.
我并不是说我希望发生这种事。
I'm not I'm not saying I want this to happen.
现实一点吧。
Let's be realistic.
这事迟早会发生。
This is gonna happen.
这是《连线》杂志的消息。
This is from Wired.
电话会议上的Fiji表示公司计划在2026年推出成人模式。
Fiji on the call said the company now plans to roll out its adult mode in the 2026.
奥特曼此前暗示，年满18岁的用户将被允许与ChatGPT进行色情对话。
Altman had previously indicated users, over 18 would be allowed to have erotic conversations with ChatchipPT.
所以期待这个功能的人，只要等到第一季度。
So for anyone who's looking forward to that, just wait till q one.
我们快实现了。
We're almost there.
再等几天就好。
Just a few more days.
把倒计时时钟打开。
Put the countdown clock on.
是啊。
Yeah.
在你能签署你的团队之前，用倒计时日历等着吧。
Advent calendar until you can sign sign your team.
好吧。
Alright.
好吧。
Alright.
我们要进入一段休息时间了。
We're gonna go to a break.
我们得在这里整理一下思绪。
We have to gather ourselves here.
让我们休息一下吧。
Let's go to a break.
当我们从休息回来时，实际上要讨论拼图的另一块，那就是：如果ChatGPT、如果OpenAI要为企业开发产品，这对它的消费者产品意味着什么？
When we come back from the break, we're actually gonna talk about another piece of the puzzle, which is, again, if ChatGPT, if OpenAI is going to be developing for enterprise, what does that mean for its consumer products?
迪士尼现在又涉足了什么领域？
And what is Disney now getting involved with?
因为迪士尼和OpenAI达成了一项重大协议。
Because Disney and OpenAI came to a very big agreement.
本周我们将报道相关内容。
This week, we'll cover that.
我们将报道AI建设进展。
We'll cover the AI build out.
我们将在休息后接着报道Meta的消息。
We'll cover Meta, right after this.
Capital One的技术团队不只是讨论多代理AI。
Capital One's tech team isn't just talking about multi agentic AI.
他们已经部署了一个。
They already deployed one.
它被称为聊天管家，正在简化购车流程。
It's called chat concierge, and it's simplifying car shopping.
通过自我反思和分层推理结合实时API检查，它不仅帮助买家找到心仪的车辆。
Using self reflection and layered reasoning with live API checks, It doesn't just help buyers find a car they love.
还能协助安排试驾、获得贷款预审批以及估算旧车置换价值。
It helps schedule a test drive, get preapproved for financing, and estimate trade in value.
先进、直观且已投入实际应用。
Advanced, intuitive, and deployed.
这就是他们的技术堆栈。
That's how they stack.
这就是Capital One的技术实力。
That's technology at Capital One.
如今，每一分钱似乎都应该更高效地工作，但决定现金投向何处可能令人困惑。
These days, it feels like every dollar should be working a little harder, but figuring out where to put your cash can be confusing.
这正是Wealthfront的用武之地。
That's where Wealthfront comes in.
Wealthfront是一个以科技驱动的金融平台，旨在帮助您将储蓄增长为长期财富。
Wealthfront is a tech driven financial platform built to help you grow your savings into long term wealth.
截至2025年11月7日，他们通过合作银行提供的高收益现金账户可为未投资资金提供3.5%的年化收益率。
Their high yield cash account through program banks offers a 3.5% APY on your uninvested cash as of 11/07/2025.
且无月费、无最低或最高余额要求即可享受该利率，您甚至可以在7×24小时内向符合条件的账户进行免费即时提款，几分钟内到账，让资金触手可及。
And there are no monthly fees, no minimum or maximum balance to earn that rate, and you can even make free instant withdrawals to eligible accounts in just minutes, twenty four seven, so your money can always be within reach.
目前Wealthfront为新客户提供为期三个月的特别优惠：在15万美元余额内，可在基础利率上额外获得0.65%的年化收益率。
Right now, Wealthfront is offering new clients an extra point 65% APY over the base rate for three months on up to a $150,000 balance.
当您开设首个现金账户时，总计可获得4.15%的浮动年化收益率。
That's a total of 4.15% variable APY when you open your first cash account.
立即访问wealthfront.com/bigtech进行注册。
Go to wealthfront.com/bigtech to sign up today.
这是Wealthfront的付费推荐广告。
This is a paid testimonial for Wealthfront.
内容不代表其他用户的体验，也不保证未来的表现或成功。
It may not reflect the experience of others, and there's no guarantee of future performance or success.
Wealthfront经纪业务并非银行机构。
Wealthfront brokerage is not a bank.
利率可能随时变动。
Rate is subject to change.
促销条款与条件适用。
Promo terms and conditions apply.
更多信息请查看节目描述。
For more information, see the episode description.
欢迎回到周五版的大科技播客。
And we're back here on Big Technology Podcast Friday edition.
我们将讨论迪士尼与OpenAI的这笔交易。
We're gonna talk about this Disney and OpenAI deal.
在开始前先快速预告：本周三我将与CNBC的吉姆·克莱默对话，探讨AI泡沫现状、AI投资趋势、市场行情以及他的个人投资见解。
Before we do, quick promo on, on Wednesday, this upcoming Wednesday, I'm gonna have a conversation with none other than Jim Kramer from CNBC about the state of the AI bubble and whether it's a bubble, the AI trade, the market, his takes on personal investing.
这期节目会非常精彩，周三务必锁定收听。
It's gonna be a really fun show, so definitely stay tuned for Wednesday.
还有一件事，现在是感恩季，如果你想回馈《大科技》播客，我们非常欢迎你在Spotify或苹果播客上给我们五星好评。
And then one more thing, is the giving season, and, if you feel, like giving back to big technology podcast, we're always happy to get five star ratings on Spotify or Apple Podcasts.
这是在这个节日季里你能帮我们的一个大忙。
That's a solid you could help help us with on, in this holiday season.
这将帮助我们邀请到更重量级的嘉宾。
It will help us book better guests.
2025年我差点就请到了'神奇七侠'（科技巨头）的几位CEO。
I got real close to landing some, magnificent seven CEOs, in 2025.
我认为只要我们证明拥有庞大听众群就能实现这个目标，当然五星好评是最佳证明方式，因为这是这些公司衡量播客规模的唯一公开指标。
I think we'll get there if we show that we have a a big audience and of course five star ratings, and reviews are the best way, to show it because they're the only publicly available metric that, these companies can look at in terms of podcast size.
好的。
Okay.
言归正传，我们来聊聊迪士尼与OpenAI的这笔交易。
With that being said, let's talk about this Disney and OpenAI deal.
消息来自《华尔街日报》。
It's from the Wall Street Journal.
迪士尼周四宣布与OpenAI达成一项重磅协议，将投资10亿美元换取这家聊天芯片茶制造商的股权。
Disney on Thursday announced a blockbuster deal with OpenAI to invest $1,000,000,000 for an equity stake in the chat chipi tea maker.
作为协议的一部分，OpenAI将获得迪士尼200多个角色的授权，用户可在Sora中创建AI生成的视频。
As part of the deal, OpenAI will license more than 200 characters from Disney so users can create AI generated videos in Sora.
通过这项为期三年的授权协议，粉丝们将能生成自己与史迪奇在夏威夷海岸冲浪，或是在R2-D2面前挥舞光剑的视频。
Through the three year licensing arrangement, fans will be able to generate videos of themselves surfing with Stitch off the shores of Hawaii or wielding a lightsaber in front of R 2 D 2.
Ranjan，你对这个消息感到兴奋吗？
Ranjan, are you excited about this?
你对此有何感想？
How do you feel about this?
你认为其中的利害关系是什么？
And what do you think the stakes are?
好吧。
Alright.
让我们...让我们先抛开我个人观点——虽然我对此确实挺兴奋的。
Let's, let's separate my personal views on this whether which I'm kind of excited.
展开剩余字幕（还有 325 条）
从创意角度来看，这确实非常有趣，但对企业层面的OpenAI意味着什么则是另一回事。
I actually from like a creative standpoint, this is very interesting versus what this means for OpenAI the business.
首先从个人角度，我觉得这很有意思。
I think first on the personal side, I find this interesting.
很难说我完全赞同，但在某种意义上我确实喜欢这个创意——粉丝文化结合生成视频的IP授权。当我开始接触这类技术时，我就确信这必将成为趋势。
It's hard to say I like it, but I kinda like it in the sense that, like, fandom and actually licensing IP for generative video, The moment I started playing with any of this, I was convinced had to become a thing.
以这种结构化方式推进时，我看到不少人讨论这有点像苹果音乐的模式。
So doing it in a kind of structured way, I saw a bunch of kind of like people talking about how this is kind of the Apple Music.
所以对于在座的年长者来说，这就像是Napster之后的iTunes。
So this is the iTunes post Napster for any of the older folks in the room.
我认为人们尝试解决这个问题是件好事。
So, like, I think it's good that people are trying to figure this out.
而且我认为这对粉丝 and 迪士尼的粉丝来说都是有益的。
And I think it's good for I do think it's good for fandom, like, fans of Disney.
我想这会很有趣的。
I think it's it's it'll be fun.
没错。
Right.
那么，这里存在一个核心问题：内容创作者是会抵制这种现象，还是会有效地顺应它？
Well, there's this whole question of, like, are the content creators gonna fight this or are they going to lean into it effectively?
我对迪士尼如此迅速表态感到惊讶，你知道吗？
I'm surprised at how quickly Disney has said, you know what?
因为迪士尼是一家注重控制的公司。
Are because Disney is a company that likes control.
对吧？
Right?
是的。
Yeah.
它希望自己的内容出现在家庭友好的环境中。
It wants its stuff in, like, family friendly environments.
说真的，一旦米老鼠开始吸毒或为你的色情聊天机器人配音，事情就会变得很糟糕。
Like, as like, legitimately, the second Mickey Mouse is, like, doing cocaine or voicing your, chattypuyty erotic pal, things are gonna be bad.
迪士尼会气疯的。
Disney's gonna be mad.
我猜他们相信OpenAI有能力将其打造成所谓的品牌安全环境。
I guess they believe in OpenAI's ability to turn this into like a quote unquote brand safe environment.
我不知道。
I don't know.
这是一个——我认为这是一笔重大交易，将促成与其他工作室的更多内容合作。
It's a it's a I think this is a major major deal and it's just gonna lead to many other content deals being made, you know, with other studios.
我是说，对我们用户来说，能做这种事会很有趣。
Mean, it's gonna be fun for us, the users, to be able to do this type of thing.
是啊。
Yeah.
我也很震惊，在所有公司里居然是迪士尼率先行动。
I think going like, I was shocked too that Disney of all companies is the one that did it.
我看到一条推文说，他们知道人们无论如何都会这么做，因为OpenAI采取放任政策，所以他们不如试着引入一些规范，这在我看来更像是勒索。
I saw one tweet where it was like, they know that people are going to do it anyways because OpenAI has a hands off policy, so they might as well kinda try to bring some structure to it, which sounded more like extortion than, anything else to me.
但我觉得从这方面来看确实令人意外，同时也表明他们可能正试图在这方面保持前瞻性。
But I felt like it was surprising on that side, but it also also shows that maybe they are trying to be a bit forward looking on it.
我认为另一个有趣的细节是他们向谷歌发出了停止侵权信函，其中一封是关于版权侵权的。
Another element of this I thought was interesting was that they'd sent the cease and desist letter to Google, one of them, a a around copyright infringement.
很明显他们在这场竞赛中选边站队了，而且是大张旗鼓地选择了立场。
So it's clear they're choosing a horse in the race, and they chose in a big way.
所以或许他们真的意识到压力——这件事注定会发生。
So so maybe it really the pressure they realize this is going to happen.
我们已经...我们无法控制局面。
We have we aren't able to control it.
因此我们至少要尝试以某种建设性的方式参与其中，这是我唯一能想到的解决之道。
So we have to try to at least take part in some kind of productive way is the only way I can kind of see this.
是啊。
Yeah.
他们给谷歌发了封措辞严厉的信，指控谷歌大规模侵犯了迪士尼的版权。
The they sent a nasty nasty letter to Google saying infringed that Google infringed on Disney's, copyrights on a massive scale.
谷歌给我的回应相当可悲且令人失望。
Google's response to me was fairly pathetic and disappointment disappointing.
我们与迪士尼有着长期互利的关系，并将继续保持合作。
We have a long standing and mutually beneficial relationship with Disney and will continue to engage with them.
更广泛地说，我们使用开放网络上的公共数据来构建AI，并建立了额外的创新版权控制措施。
More generally, we use public data from the open web to build our AI and have built additional innovative, copyright controls.
这简直就是在展示他们的漠视态度
It just, like, shows the disregard bless
那些不得不编造此类声明的公关专业人士。
the PR professionals out there who have to come up with statements like that.
我是说，这份声明难道不正暴露了这些公司对网络版权作品的普遍漠视吗？
I mean, doesn't this statement just betray the disregard that so many of these companies have for copyrighted work, online?
简直让我抓狂。
Mean Drives me nuts.
是啊。
Yeah.
我是说
I mean
他们并不反对起诉他们。
They don't disagree for suing them.
是啊。
Yeah.
不。
No.
所以这部分我觉得又挺有意思的。
That's why that that part of it I found interesting again.
我同意。
I agreed.
就像，我们不是要起诉，但同时也要直接对你采取行动。
Like, it's like we're going to not we're gonna sue, but we're also gonna take action here directly at you as well.
所以我觉得在这方面是合理的。
So so I think on that side, it makes sense.
不过我开始从迪士尼的角度思考这个问题，唯一的方式就是培养角色粉丝是业务核心，他们通过无数方式将其变现。
The other the only way I was starting to think about this from Disney standpoint though is building fandom for your characters is the core of the business, and they monetize it in a million ways.
就像迪士尼这个企业，他们明白电影不仅仅是电影。
Like, Disney the business is the one that understood that a movie is not just a movie.
那个角色以各种方式与观众互动，这里面大有文章可做。
Like, that character engaging with them in all different types of ways is there's a lot of money to be made there.
任何去过迪士尼世界的人，近年的票价如何，比如迪士尼游轮。
Anyone who's gone to Disney World and in in recent years for how much those tickets cost, like, have Disney Cruise.
他们拥有整个迪士尼生态系统。
They have the the the entire ecosystem of Disney.
所以，任何实际上能增强粉丝与这些角色联系的事物，都可能为他们带来丰厚收入。
So so anything to actually boost that connection fans have with these characters should probably make them a lot of money.
从这个角度看，我想我有点理解了。
So from that standpoint, I guess I can kind of see it.
是的。
Yep.
这会在哪里出现？
And where is this gonna appear?
它将会同时出现在Sora中，但迪士尼CEO鲍勃·艾格周四在CNBC上表示，计划是精选一些Sora视频，然后放到Disney+上，并允许在其流媒体服务上创建由Sora驱动的视频。
So it'll both, appear in in Sora, but, Iger, Bob Iger, Disney CEO Bob Iger on, Thursday on CNBC said that the plan is to curate some of these Sora videos and then put them on Disney plus and allow them to create Sora powered videos on its streaming service.
所以这不仅像是'好的，继续吧'，而是他在说这也是我们的内容。
So not only is this like the okay go ahead, it's him saying this is also content for us.
好的。
Okay.
我开始理解这一点，也能看到一些真正有创意的内容从中产生。
I both start to see that and can see some, like, genuinely creative things coming out of it.
但另一方面，用户生成内容的世界试图将其引入Disney+，我觉得这实际上抓住了迪士尼严密守护的花园特质，而不是让它变成一个更加粗放的地方。
But on the other hand, the the world of user generated content and trying to, like, bring that into Disney plus, I feel that actually kinda captures the tightly guarded garden of Disney versus making it more of a rough and tumble place.
那不是迪士尼的风格，从来都不是。
Like, that's not Disney and never has been.
所以那些东西，我认为还是不会发生。
So that stuff still, I don't think it's gonna happen.
真的吗？
Really?
好吧。
Alright.
我想这么说吧，无论他们最终采取什么行动，这对迪士尼来说都是巨大的净收益。
I I think let me put it this way, I think whatever they end up doing here, this is a massive net positive for Disney.
这就是我对这个新闻的看法。
That's just the way I look at this story.
这是个大胆的举动，也是个冒险的举动——就像我们一直在调侃的可卡因米老鼠那样，我确信他们考虑过这种风险。但这会让人们更投入迪士尼角色，比如这种让你置身于场景中的创意，比如和R2-D2或者高飞同框。
It's a bold move, it's a risky move, like we've been joking about Cocaine Mickey Mouse through this story, and that is a risk that I'm sure that they thought about, but this is gonna get people more invested in the Disney characters like this idea of like putting yourself in a scene, with you know r two d two or whoever it is, goofy.
这对人们来说就是件很酷的事。
It's just, it's a cool thing for people.
我认为这会给他们带来很好的效果。
I think this is really gonna work out well for them.
等等。
Wait.
不过我有个问题。
One question I do have though.
为什么迪士尼既要投资OpenAI并承诺购买企业许可证，又要将他们的角色授权给OpenAI使用？
Why does Disney have to invest in OpenAI and commit to buying OpenAI enterprise licenses if they're giving their characters to OpenAI?
我不确定这是否属实，但看到投资金额时，我的直觉是迪士尼那10亿美元投资OpenAI的资金，某种程度上可能是循环融资。
I don't know this to be true, but my hunch in seeing the funding was that, like, the 1,000,000,000 investment that Disney's making in OpenAI is has to be some form of circular funding.
比如，OpenAI可能会用其股票向迪士尼支付将角色植入Sora的授权费，这可能就是一笔十亿美元的交易。
Like, Disney like, OpenAI is gonna maybe pay Disney for the right to put its characters in Sora with OpenAI stock and that might just be a billion dollar trade.
我不确定这是事实，这只是我的猜测。
I don't know this to be true, it's just my speculation.
我知道。
I know.
我也试图查证这一点，从我看到的信息来看，这似乎是一笔附带认股权证的直接股权投资，但在我脑海里，这绝对是最初的走向。
I I was trying to look into that as well and it seemed to be a straight equity investment with warrants like from what I saw, but to in my mind too, that that's definitely where it went to start.
但或许他们因此获得了免费的ChatGPT企业版许可，这实际上是OpenAI大规模进军企业市场的首次尝试
But maybe they're getting free ChatGPT Enterprise licenses as part of this and this is actually the first push and foray into enterprise in a big way for
看吧。
See.
OpenAI。
OpenAI.
好吧。
Okay.
转回来了。
Coming around.
也许这就是一切如何串联起来的。
Maybe this is how it all comes together.
也许吧。
Maybe.
好的。
Alright.
我还想告诉你一件事，这周我再次打开Sora时想到的。
I wanna tell you one more thing that I sort of, came came to, this week, when I opened up Sora again.
不知道你有没有注意到这些Sora视频在互联网上流传，但它们确实已经传开了。
I don't know if you've been seeing these Sora videos make their way around the internet, but they certainly have.
关于Sora有件事特别让我恼火，我经常在TikTok上看到它们被冒充成真实视频，而且Sora有一种特定的语法模式——当人们在Sora视频里说话时，你能听出来是这种效果，听起来像是'天啊他要翘车头了 他真的翘了'。
And one thing that really annoys me about Sora, I I see it on TikTok all the time passed off as the real video, and there is a particular Sora syntax that's like you can hear when when people are, speaking, when when people speak in SORA videos it sounds like this it sounds like oh my god he's gonna do a wheelie He did a wheelie.
不可能。
No way.
对吧？
Right?
它有种生硬的、可能过度兴奋的节奏感。
It has that sort of stilted, like, kind of maybe overly excited rhythmic.
我是这么想的，想跟你探讨下。
I here's what I think sort of I wanna run this by you.
我认为Sora在训练时吸收了太多游戏音频，所以这些视频里传递的对话带有明显的游戏对白特征，这就是为什么如此独特。
I think Sora trained on so its audio on so many video games that it is a distinctly video game type dialogue that you're hearing come through in these videos, and that's why it's so distinct.
我...我喜欢这个理论。
I I like that theory.
我其实更感到惊讶。
I'm actually more amazed.
我有一阵子没上Sora了。
I've not been on Sora in a while.
实际上就在我们说话这会儿我刚打开它，看到两只老虎戴着法国厨师帽在做饭。
I actually just opened it right now as we're speaking, and I see two tigers and French chef hats cooking.
但我觉得这其实值得肯定，我的推荐流和算法里很少刷到这类内容。
But I wonder it's actually to the credit, I don't come across it very much in my feeds and algorithms.
不过想到如果存在某种标志性语音——我记得TikTok刚崛起时那种单调的默认配音就成为了文化符号。
But the fact that if there's sort of voice because I remember when TikTok really started gaining prominence that there was that kind of monotone TikTok voice that was the default for any video created that actually became kind of a cultural artifact.
所以我觉得如果Sora真能做到这点，对这个平台来说确实很了不起。
So I think to Sora's credit, if they actually are getting there, that's actually impressive for the platform.
但...我也不确定。
But but I don't know.
比如，他们完全没有公布这方面的用户数据。
Like, there's certainly not advertising any user numbers around it.
而且，再次回到关注点的问题上，他们一开始声势浩大。
And, again, going back to the focus thing, they came out big.
后来热度有点消退。
It kind of faded away a bit.
现在，我猜它变成了一个为其他平台服务的内容创作工具。
Now, I guess it's a content creation tool to use for other platforms.
所以我不太确定他们的发展方向。
So so I'm not sure where they're going with it.
不过本周确实从Apptopia获得了一些数据。
Well, did get some numbers from Apptopia this week.
他们把这些数据发过来了。
They sent them along.
Sora刚推出时，日均下载量大约是24万次。
So when Sora came out, it was doing around 240,000 downloads, looks like a day.
后来增长到了每天40万次左右。
Bumped up to around 400,000 or so per day.
实际上它已经大幅下降，这对你来说应该不会感到意外。
It has actually dropped significantly and this won't come as a surprise to you.
目前看起来每天的下载量大约在15万次左右。
Looks like it's like around maybe a 150,000 downloads a day right now.
所以，这可能就是Apptopia仍然存在的原因。
So, this might just be what Apptopia still around.
这可能接近Apptopia测量的数据，但它确实显示出明显的下滑趋势。
It might be around what Apptopia measured, but it just shows like a real decline there.
再次强调，我们之前讨论过这个问题，就是那种同质化问题——每个Sohr视频都有非常相似的风格。
And it's again, we talked about this before, it's this like sort of sameness problem that every video, every Sohr video has a very similar feel.
所以我们会继续关注它的发展。
So we'll keep monitoring it.
我们拭目以待。
We'll see.
我们拭目以待。
We'll see.
明年对Sohr来说可能是米老鼠的大年。
Next year, big year for Sohr maybe in Mickey Mouse.
我正式声明这是毒品米老鼠。
I said weed Mickey Mouse for the for for the record.
哦，抱歉。
Oh, sorry.
我是不是把毒品米老鼠升级成可卡因米老鼠了？
Did I I upgraded weed Mickey Mouse to cocaine Mickey?
我志向远大。
I I dream big.
你知道有多嗨。
Know how hyper.
是啊。
Yeah.
不。
No.
我们了解你们俩。
We know you both.
我们知道他为什么兴奋。
We know why excited he is.
是啊。
Yeah.
所以很明显，米奇的形象不该是那种瘾君子米奇。
So, obviously, what what Mickey is, it's not like that that personality is not stoner Mickey.
饶了我吧。
Give me a break.
我们知道他的底细
We know what he's been
但现在还不是，不过很快就是了。
doing But behind there but it will be.
等我拿到那个Sora就会是了。
It will be once I get my hands on that Sora.
但你认为魔法对迪士尼意味着什么？
But what do you think magic means, for Disney?
天啊。
Oh my god.
我要被起诉了。
I'm gonna get sued.
但你不会被起诉的。
But you but you won't get sued.
迪士尼会投资你的
Disney will invest in you for
那还不错。
That'll be good.
为了那个创意授权。
For that creative licensing.
我得告诉你，我正在采取萨姆·奥尔特曼对鲍勃·艾格的方式，而不是埃隆·马斯克的方式。
I have to tell you, my I'm I am taking the, the Sam Altman approach to Bob Iger and and not the Elon Musk.
是啊。
Yeah.
不客气。
You are welcome.
欢迎你在这里与我合作。
You are welcome to partner with me here.
你觉得我们会
Do think you we'll
收到律师函吗？
get a lawyer letter?
希望不会。
I hope not.
我只是在开玩笑。
I'm just joking about all this stuff.
这就是我们的工作。
This is our job.
我们必须探讨各种可能性。
We have to talk about the possibilities.
我们讨论基础设施。
We talk about infrastructure.
我们来谈谈基础设施。
Let's talk about infrastructure.
甲骨文和博通这周不好过，对吧？
Rough week for Oracle and and Broadcom, right?
这些正是AI基础设施交易的核心支柱。
These are the the sort of cornerstones of the AI infrastructure trade.
今天有报告称甲骨文将把部分数据中心交付OpenAI的时间从2027年推迟到2028年。
Oracle, there's a report out today that Oracle is gonna delay some of its data centers to OpenAI, but from '27 to 2028.
另外市场开始意识到，他们为AI建设的设施利润率低于其他数据中心，这倒不足为奇。
Also, the market is starting to see that their build out for AI is lower margin than the rest of their data centers, surprise surprise.
博通也受到波及，虽然接到了AI订单，但其AI硬件规模远未达到市场许多人的预期，所以两家公司都在经历估值下调。
Broadcom also sort of getting hit on this, it's gonna have, it has orders that are coming in for AI, but it just hasn't been as big as AI hardware but hasn't been as big as a lot of people in the market, believed and so they're they're getting a haircut, both of them are.
你认为这里的大新闻是什么？
What do you think the the big story is here?
是不是因为这些交易存在不确定性，每次出现波动市场就会逃离，因为他们知道如果真的恶化会有多糟糕？
Is it just that like this is a and these trades are based on, have a level of uncertainty and every time they wobble, the market just runs away because they know how bad it could get if it really goes south?
是的。
Yeah.
我想我们每次都能看到真实数据，比如甲骨文的合作消息最初是什么时候出来的？是九月吗？
I think it's just we see real numbers each one of these times, like, or when was it September the Oracle partnership news first came out?
我记得那天它涨了37%，好像是。
There was that day it was up 37%, I think it was.
就像现在，我们开始看到真实数据了，连甲骨文也开始明白，在独立于实际计算需求的情况下建设这些数据中心的具体物流问题，以及实际建设工作所需付出的努力——即使需求存在，这些真的可行吗？
Like, now we're starting to see real numbers even with Oracle starting to understand, like, the actual logistics of building out these data centers independent of the actual demand for the compute, but at the actual work it's going to take to build these out, that even if the demand is there, is that gonna be possible?
所以我认为整体而言，这给整个行业带来了一定程度的现实感，提醒人们事情不会那么简单直接。
So I think, like, overall, it's just bringing a degree of reality to this entire the trade, like the entire infrastructure trade that just is, you know, reminding people that it's not gonna be that straightforward.
没错。
Right.
甲骨文本周公布了一些财报，并表示其数据中心资本支出在未来几个季度将超过收入。
And Oracle brought delivered some earnings this week and said that its capital expenditures from the journal, on data centers will outrun revenue for several quarters to come.
这就像在问，为什么市场对此感到惊讶？
And it's like, why why is the market surprised here?
也许市场对AI的信仰过于坚定，以至于对这项技术能如何运作描绘了过于乐观的图景。
Maybe the market believed so much in AI that it took this overly rosy picture of how this stuff can work.
我不知道。
I don't know.
告诉我一点你为什么认为市场如此乐观。
Tell me a little bit about why you think the market was so positive.
不。
No.
我认为市场之所以乐观，是因为它确实觉得甲骨文与OpenAI的公告标志着基础设施投资真正进入主流时刻的开始。
I think the market was positive because I think it really felt like the Oracle OpenAI announcement was the begin kinda like the real mainstream moment for the infrastructure trade.
而且这种趋势不会仅限于OpenAI。
And this idea that it's not just gonna be OpenAI.
实际上，所有建设者都将推动数据中心等设施的建设，随着AI使用量的增加，这些设施将在未来多年持续提供计算资源。
It's gonna be all of the builders who are actually gonna, like, drive the data center build out and the like, that's gonna feed in all of that compute over the years as everyone is using more AI.
这是个合理的说法。
It was a sound story.
它某种程度上捕捉到了那种狂热时刻，但很快就回归现实了。
It kinda captured that moment of mania, but it's pretty quickly coming back down to earth.
我认为这是好事——关于这一切如何发展的讨论变得更加理性，事情开始变得更真实，人们也开始关注具体数据，因为这些数据至少开始部分公布了。
And and I think in a good way, like, the sense that having that more reasonable discussion around how this all plays out It's just becoming a little more real, and people are actually looking at some numbers because they're starting to come out at least a little bit.
尽管这些都是两到三年的长期故事，但一个季度的数据终究是一个季度的数据。
Even though these are all two to three year stories, still one quarter of data is one quarter of data.
没错。
Right.
我的意思是，OpenAI与甲骨文签订了价值3亿美元、为期五年的计算资源采购协议。
I mean, OpenAI is has a $300,000,000 deal to buy computing from Oracle over five years.
据我了解，OpenAI占甲骨文剩余履约义务的55%。
OpenAI is 55%, from what I understand, of the remaining performance obligation for Oracle.
所以如果OpenAI做不到，对甲骨文来说是个相当糟糕的消息
So if OpenAI can't do it, it's pretty bad news for Oracle
到十亿
to billion.
看
See.
3300美元
$3,300
我说的是百万吗？
I said million?
十亿
Billion.
是的
Yeah.
这就是那种不健康的债务组合，以及对OpenAI的依赖
It's just this sort of unholy combination of debt, of dependency on OpenAI.
本周又提到了你永远不想听到的三个字母。
Well, the three letters you never want to hear were mentioned again this week.
CDS。
CDS.
信用违约互换。
Credit default swaps.
信用违约互换。
Credit default swaps.
所以你能简单解释一下那里具体是什么问题吗？
So so can you just for a second explain exactly what the issue is there?
信用违约互换我们之前讨论过，就像为任何企业信用违约投保的成本。
Well, credit default swaps, again, we've talked about in the past, like the cost of insuring against the default of any kind of corporate credit.
这个数字越高，保险费用就越贵，表明市场认为违约可能性越大。
And it's just the the higher that number goes, the more expensive that insurance is, the more it's indicating that the market believes a default is more likely.
而且这些数字在某个节点会开始变得很小。
And again, these numbers start to be small at a certain point.
本周，甲骨文的信用违约互换(CDS)达到了自2009年4月以来的最高日内水平。
Now this week, the Oracle CDS hit the highest intraday level since April 2009.
所以人们开始意识到，这可能是需要投保的风险，而最初并未充分定价这种风险要素。
So so it's people are starting to recognize that this is something that maybe you want to be insured against, and it's there is a element of risk to this that that was not priced in initially.
但正如我们之前讨论过的，你永远不想谈论CDS。
But but as we as we discussed before, you never wanna talk about CDS.
每当它进入主流讨论时，对任何人都不是好事。
Like, anytime it's entering the mainstream conversation, it's not a good thing for anyone.
而我们越频繁听到甲骨文CDS的消息，就意味着大家应该更加警惕了。
And the more we start to hear about Oracle CDS, I mean, that's where everyone should get a little bit more worried.
2009年中后期发生了什么不好的事吗？
Did anything bad happen in mid to late two thousand nine?
抱歉。
Sorry.
我一时想不起来了。
I'm forgetting.
2008年。
2008.
2008年。
2008.
哦，是2008年。
Oh, it's 2008.
那实际上是我们上次都在讨论2008年时提到那三个字母的时候。
That was the last time we were all talking actually from 2008 talking those three letters.
老实说，除了信用违约互换交易员，我认为在2008年之前没人会想到信用违约互换这种东西。
I I honestly outside of like CDS traders, don't think anyone anywhere had ever thought about a credit default swap until 2008.
所以希望我们不会听到太多相关消息。
So, hopefully, we don't hear that much more.
是啊。
Yeah.
关于这件事最后再说一点。
One last thing about this.
这似乎包含在OpenAI的轴线上。
This seems to be contained in the OpenAI axis.
对吧？
Right?
比如，博通不算。
Like, Broadcom no.
不过博通是谷歌系的。
Broadcom is Google, though.
他们负责谷歌和Anthropic的合作。
They do Google Anthropic.
但说真的，今天Orweave跌了，英伟达也跌了。
So but, like, seriously, like, Orweave is down today, Nvidia is down.
你觉得这些是系统性问题，还是目前比较局部？或者现在下结论还为时过早？
Do you think that these these are, like is it a systemic problem or is it sort of more contained right now or too early to tell?
我认为这是行业性问题。
I think it's an industry problem.
所以我觉得，我见过这类情况，比如你可以交易OpenAI生态系统和谷歌生态系统的份额。
So I like, I've seen kind of these, like, you can trade buckets of open the OpenAI ecosystem versus the Google ecosystem.
我仍然认为，无论什么影响到甲骨文和OpenAI，至少从人们定价基础设施的角度来看，也会影响到博通和谷歌。
I I still think whatever hits Oracle and OpenAI will hit a Broadcom and a Google at least in the in terms of how people are pricing in the infrastructure side of things.
但我仍然不认为这是系统性的问题。
So but I I I still don't think it's systemic.
我认为尽管股市的涨幅很大程度上依赖于这个因素，但我仍然不认为会导致某种大规模的经济灾难。
I think even though the stock market is gains have been heavily predicated on this, I still don't think it leads to some kind of, like, large scale, like, economic calamity.
我觉得这是市场小幅下跌、401k计划受影响、人们有点不高兴的情况之一，但我不认为是系统性的。
I think it's one of those market goes down a bit, four zero one k's get hit, people are a little unhappy, but I don't think it's systemic.
让我这么说吧。
Let me let me put it this way.
这是AI交易开始瓦解的开端，还是你认为AI交易依然强劲？
Is this the beginning of the unraveling of the AI trade or do think the AI trade remains strong?
我认为这是AI交易性质的一种转变。
I think it is a shift in what the AI trade is.
我认为对我来说，最关键的是如果相信Sam Altman之前的说法，即应用层将是一场大战，那么就必须开始设想构建能以更高计算效率运行的应用程序将成为竞争优势，尤其是在需要成本效益的企业领域。
I think that to me, that's like the biggest thing that I think everyone piled in the actually, if we are to believe Sam Altman earlier and, like, the idea that the application layer is going to be this big battle, you have to start to imagine, like I mean, building applications that run-in a more compute efficient way start to become a competitive advantage, especially if you're going to the enterprise where you need to be cost efficient.
所有这些都与实际的数据中心基础设施计算需求背道而驰。
Like, all of that side of the story runs counter to the actual compute data center infrastructure side of the story.
二者虽然可以共存，但这两者之间存在许多相互对立的压力，而市场正试图同时讲述这两个故事。
Like, there's some coexistence of them, but there's also like a lot of, you know, like opposing pressures within those two stories and they're trying to tell them both.
这基本上就是我想下周问Kramer的第一个问题：AI泡沫是这样破裂的。
That's kind of what I think I'm gonna go to, Kramer with first question when we speak next week is basically like, here's how the AI bubble bursts.
你可以更高效地进行计算。
You can compute much more efficiently.
我们仍未获得明确的迹象，表明这些东西能达到人们想象中的经济价值规模。
We still don't have, real clear indications that this stuff is gonna be economically valuable on the scale that people imagine.
因此我们可能会看到从高基础设施成本、大规模建设、宏大梦想转向更实用、更高效，然后突然间它变成了具有经济价值的事物，但不再是那种追逐全能技术的大规模资金投入。
So what we could see is a conversion from a high infrastructure cost, big build up, big dreams to, like, getting more practical, getting more efficient, and then all of a sudden, it becomes an economically useful thing, but not this, you know, sort of god pile of money chasing, an all knowing technology.
是的。
Yeah.
不。
No.
不。
No.
我是说，这就是我的看法，尤其是在企业环境中，当人们变得更加注重成本和竞争时，创新就体现在真正让这些技术更高效地运行这些流程、计算和工作流上。
That that's I mean, how I view it that, like, especially in the enterprise when people are gonna be a lot more cost conscious and, like, competitive around it, that the innovation happens in terms of actually making these technologies more cost efficient to actually run this run these processes and compute and and workflows.
所以我认为，是的，你可以让AI生成。
So so I think, yeah, you can you can have AI generate.
黄仁勋说过，它将从100万亿的GDP总量中产生50万亿的GDP增长。
Jensen said it's gonna generate a $500,000,000,000,000 of GDP, like, from a 100,000,000,000,000 aggregate GDP.
也许这会发生，但这并不意味着基础设施投资真的会有回报。
Maybe that happens, but that doesn't mean that the infrastructure trade actually pans out.
我知道这听起来很荒谬，但我确实认为这是一种可能性。
I know that sounds ridiculous, but I do think that's a possibility.
现实的可能性。
Realistic possibility.
是的。
Yeah.
好的。
Alright.
我们就此结束这个话题吧。
Let's close on this.
Meta的新AI明星团队与公司其他部门产生了摩擦。
Meta's, new AI superstars are chafing against the rest of the company.
这是《纽约时报》的报道。
This is from the New York Times.
今年马克·扎克伯格重组Meta人工智能业务时，他招募了一位新领导者——28岁的企业家亚历山大·王，从OpenAI和谷歌等竞争对手那里组建了一支顶尖研究团队。
When Mark Zuckerberg revamped Meta's artificial intelligence operations this year, he recruited a new leader, Alexander Wang, a 28 year old entrepreneur, to build a team of top researchers from rivals like OpenAI and Google.
这支团队被安置在Meta总部中心扎克伯格先生办公室隔壁的一个独立空间里，据称是为了让他们远离公司的官僚体系。
The team was placed in a siloed space next to mister Zuckerberg's office at the center of Meta's headquarters, probably just, this what the story says to remove them from the bureaucracy of the company.
五个月后，这种隔离已不仅仅是物理上的。
Five months later, the divide has become more than physical.
王先生私下向人透露，他不同意扎克伯格的一些长期副手，包括首席产品官克里斯·考克斯和首席技术官安德鲁·博斯沃思的观点。
Mister Wang has privately told people he disagreed with some of Zuckerberg's longtime lieutenants, including Chris Cox, the chief product officer, and Andrew Bosworth, the chief technology officer.
在一次事件中，
In one case, Mr.
考克斯先生和
Cox and Mr.
博斯沃思先生希望
Bosworth wanted Mr.
王先生的团队专注于利用Instagram和Facebook的数据来训练Meta的新基础AI模型，以改进公司的社交媒体动态和广告业务。
Wang's team to concentrate on using Instagram and Facebook data to help train Meta's new foundational AI model to improve the company's social media feeds and advertising business.
但开发该模型的王先生予以回绝。
But mister Wang, is developing the model, pushed back.
他主张目标应是在专注产品前先追赶OpenAI和谷歌的竞争性AI模型，知情人士表示。
He argued that the goal should be to catch up to rival AI models from OpenAI and Google before focusing on products, the people said.
我能快速分享一下我对这段文字的看法吗？因为它对我来说不太说得通。
Can I just share quickly what my perspective on this Please, paragraph is, because it doesn't make too much sense to me?
我是说，如果你在Meta内部运营一个超级智能项目，你肯定会利用Meta旗下的Instagram和Facebook数据进行训练。
I mean, if you're running an a super intelligence operation within Meta, you're gonna use Meta's Instagram and Facebook data to train.
对吧？
Right?
你拥有那些数据，那些是可以用于训练的专有数据。
You'll have that data that says that's proprietary data you can use to train.
那你为什么不想用这些数据来训练你的模型呢？
So why wouldn't you want to use that that data to train your models?
而且使用这些数据并不一定只会改善社交媒体推送和广告业务。
And using that data doesn't necessarily improve, you know, only improve the social media feeds and the advertising business.
这是个训练场景，不是产品场景。
It's a it's a training situation not a product situation.
不得不说，这故事在我看来非常奇怪。
Very weird story to me I have to say.
也许这事正在发生，但就是感觉很奇怪。
Maybe this is happening but weird.
是啊。
Yeah.
我也觉得这很奇怪，毕竟Meta旗下的平台可能拥有全球最丰富的数据集之一。
I I definitely found it weird as well in the sense that I mean, Meta properties probably have one of the richest datasets in the world.
我能想到的唯一可比的是YouTube，就整体数据的丰富性而言，但Meta现在也有大量视频数据可供利用。
I mean, the only comparison I can imagine is YouTube, which, like, in terms of just richness of overall data, but Meta's got plenty of video to work with as well now.
所以，对于没有在观点纯粹性或珍贵性上表现出某种差异，比如'我们不想被广告业务玷污'这种说法，
So, like, to not to have some kind of difference in opinion and purity or preciousness around, like, we don't wanna be sullied by the the advertising business.
我觉得这确实让我感到意外。
I think, like, that was surprising to me.
但我认为更重要的故事是，我们一开始就讨论过，从个性角度来看，这事可能会彻底失败。
But I think the bigger story is we talked about this from the beginning like this could be this could just fall on its face from a personality standpoint.
在这种事情上投入如此高调的个性元素，以及他们砸进去的巨额资金，
To bring in this level of personality around this splashiness, the amount of money they're throwing at this.
我已经在推特上看到几个人宣布将在短短几个月内离开TBD实验室了。
Already, I saw on Twitter a few people, like, announcing their departure within just a matter of months from TBD Labs.
因此我认为更令人担忧的是，如果政治因素真的阻碍了技术发展，这对项目来说可不是个好兆头。
So I think that that's the more concerning part to me that if the politics of this are actually getting in the way of the technology, that's not a good start for this.
没错。
Right.
《时代》的报道确实提到离职潮已趋于稳定，可能在首次股权兑现期过后，100人中只有2人离开。
The time story does say that the departures have stabilized and maybe two out of a 100 have left after that first vesting period hit.
有意思的是——我们且看后续发展——我完全同意利益冲突必然存在，正如报道所述：实验室研究人员认为多数Meta高管只关注提升社交媒体业务，而实验室的抱负却是要创造神级AI超级智能。
I mean the interesting thing here is I think that I mean we'll see what happens because I I don't disagree there's definitely going to be clashes of interests, they did bring get brought I mean here here this is from the story the lab's researchers have come to view many meta executives as interested only in improving the social social media business while the lab's ambition to create, is to create a godlike AI superintelligence.
我觉得这很正常。
I think that's natural.
虽然这可能演变成公司内部的重大裂痕，但似乎本就是预料中事：加入Meta就意味着既要助力AI发展，也要推动核心业务。不过...
I mean, it's not very surprising to me, although you could see that really turning into a big fissure within the company, but it seems like that was always part of the equation was like if you go to Meta, yeah, like, you're gonna be there also to improve the core business along with helping cyber But
我在想谷歌处理得更好——回到几周前的讨论——谷歌巧妙地将AI整合到公司架构中，并建立了那个中心。
I wonder I feel Google has handled I mean, going back to the conversation from I think a couple of weeks ago around how savvily Google handled the overall integration of AI within the company and kind of created that center.
这种天然张力永远都会存在。
Like, it there's always gonna be that natural tension.
我认为这种张力可能是健康的，但必须要有某种总体上的明确性，比如在这些组织内部要清楚我们到底想实现什么目标？
It could be a healthy tension, I think, but there has to be some kind of overarching, like, clarity within the these organizations around, like, you know, what are we actually trying to do here?
而Meta最终拥有十亿用户。
And Meta, in the end, they have a billion.
我是说，抱歉。
I mean, I'm sorry.
我在说什么来着？
What am I talking about?
现在是有40亿用户吗？
Like, 4,000,000,000 users now?
还是30亿用户？
3,000,000,000 users?
管它具体是多少。
Whatever it is.
几十亿的差距算什么？
What's a couple billion?
几十亿算什么？
What's a couple billion?
脸书好友。
Facebook friends.
不。
No.
你看，持续改进产品本身就能带来巨大规模效应，无论你做什么，只要把精力放在这上面，实际潜在商业影响甚至社会影响都是巨大的，试图将其边缘化、保持纯研究实验室的定位是没多大意义的。
Like, improving improving the the product product just brings such a massive scale into anything that you do that focusing on that, the actual potential business impact of that or even societal impact is so massive that trying to just be kept to the side and like be a pure research lab does not make much sense.
但我觉得，马克·扎克伯格应该对此给出明确方向。
But I mean, they they Mark Zuckerberg should bring some clarity to to that.
知道我们怎么判断这事真的结束了吗？
You know how we're gonna know this is really over?
就是当Meta超级智能实验室宣布转向企业服务时。
Is when Meta Superintelligence Lab says it's focused on enterprise.
那就意味着终结。
That's that's the end.
这就是结局。
That is the end.
我想那将是他们转型为AI云服务/消费者可穿戴设备业务的时候。
I think that's that's when or they become an AI cloud slash consumer wearables business.
嗯，你知道那就是
Well, you know that's how
他们已经是了。
They are.
他们
They
是。
are.
他们确实是。
They are.
我在说什么呢？
What am I talking about?
雷朋Metis眼镜。
Ray Banned Metis.
是啊。
Yeah.
没错。
Exactly.
好吧。
Alright.
那么，我们就到此为止吧。
Well, let's leave it there.
周三的吉姆·克莱默节目。
Jim Kramer on Wednesday.
兰詹，一如既往，周五和你聊得很愉快。
Ranjan, as always, great speaking with you here on Friday.
感谢你的参与。
Thanks for coming on.
好的。
Alright.
下周见。
See you next week.
好的，各位。
Alright, everybody.
感谢您的收听和观看。
Thank you for listening and watching.
我们下次在《大科技播客》中再见。
We will see you next time on Big Technology Podcast.
关于 Bayt 播客
Bayt 提供中文+原文双语音频和字幕，帮助你打破语言障碍，轻松听懂全球优质播客。