2025年回顾，2026年预测——与里德·阿尔伯戈蒂对话

让我们回顾一下2025年，并展望2026年将要发生什么。

Let's look back at 2025 and anticipate what's coming in 2026.

我们将在接下来的内容中，由Semaphore技术编辑里德·阿尔伯戈蒂来带大家了解。

We'll do it with semaphore technology editor, Reed Albergotti, right after this.

Capital One的技术团队不仅仅在谈论多模态AI。

Capital One's tech team isn't just talking about multiegetic AI.

他们已经部署了一款。

They already deployed one.

它叫做聊天礼宾服务，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，并结合实时API检查，它不仅能帮助买家找到心仪的汽车。

Using self reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love.

还能帮助安排试驾、申请贷款预批，并估算旧车折价。

It helps schedule a test drive, get preapproved for financing, and estimate trade in value.

先进、直观，且已上线。

Advanced, intuitive, and deployed.

这就是它们的运作方式。

That's how they stack.

这就是摩根大通的技术。

That's technology at Capital One.

这里是迈克尔·刘易斯。

Michael Lewis here.

我的畅销书《大空头》讲述了2008年美国房地产市场泡沫形成与崩塌的故事。

My best selling book, The Big Short, tells the story of the buildup and birth of The US housing market back in 2008.

十年前，《大空头》被拍成了获得奥斯卡奖的电影，现在我首次将其制作成由尤瑞斯·特鲁利朗读的有声书。

A decade ago, The Big Short was made into an Academy winning movie, and now I'm bringing it to you for the first time as an audiobook narrated by Eurus Truly.

《大空头》的故事——押注市场崩盘的意义，以及谁真正为失控的金融体系买单——如今比以往任何时候都更相关。

The Big Short story, what it means to bet against the market, and who really pays for an unchecked financial system is as relevant today as it's ever been.

现在就前往 pushkin.fm/audiobooks 或任何有声书平台获取《大空头》。

Get The Big Short now at pushkin.fm/audiobooks or wherever audiobooks are sold.

欢迎收听《大科技》播客，一档致力于对科技世界及其更广泛影响进行冷静而深入对话的节目。

Welcome to Big Technology Podcast, a show for cool headed and nuanced conversation of the tech world and beyond.

现在是假日季节，这意味着一件事。

It's the holiday season, which means one thing.

我们回顾过去一年发生了什么，并思考接下来会怎样？

We take a look back at what happened over the past year and say, what's next?

如果我们能对9万条预测中的三四条猜对，我相信明年我们一定会为此庆祝。

And if we can be right, you know, maybe three or four times out of, like, the 90,000 predictions we'll make, I'm sure we'll celebrate it next year.

因此，和往常一样，或者更准确地说，在历史的关键时刻加入我们的，是七四科技编辑里德·阿尔瓦加蒂。

So, joining us as always is, or actually, joining us in key moments in history, shall we say, is, seven four technology editor, Reed Alvarghatti.

里德，很高兴见到你。

Reed, great to see you.

能在这里总是很好。

It's always good to be here.

我给你发了邮件，说我们来做一下预测吧，但我真的很想听听你对2025年的回顾。

I kind of emailed you and said, let's do predictions, but I I'd love to hear just your recap of of 2025.

我的意思是，这一年太疯狂了，资金投入、承诺，来回反复。

I mean, is there is it's been a crazy year, so much funding, commitment, back and forth.

你觉得过去一年最显著的特征是什么？

What would you say is like sort of the defining characteristic of the past year?

是的，我认为这是其中之一。

Yeah, I think that is one of them.

我认为基础设施方面的问题以及其庞大的规模，可能是最热门的话题。

I think the infrastructure sort of questions and just the sheer scale of it has been probably the big topic of conversation.

但我也觉得我们最近一直在谈论中国。

But I also think we've been talking a lot about China recently.

我认为地缘政治发挥了重要作用，比去年大得多。

I think geopolitics has played a big role, a much bigger role than it did last year.

我回顾了去年的预测，想看看哪些猜对了，哪些猜错了。

I was looking back at my predictions from last year and trying to see what did I get right and what did I get wrong.

感觉一年前的世界完全不一样了，对吧？

It just seems like it was such a different world a year ago, right?

在政府更迭之前，我们看到了海湾地区发生的各种疯狂事情。

Before the changeover in the administration, we've seen all kinds of crazy stuff happening in The Gulf.

所以我认为这是另一个重要的方面。

So I think that's another big one.

对我来说，我认为这一年是规模的一年，我会试着回顾一下自ChatGPT发布以来的这些年份，因为显然，GenderValvei一直是科技界的核心话题。

I mean, for me, I view this as sort of a year of scale where me I'm just gonna try to go through, like, the various years since ChatGPT has been released because, obviously, GenderValvei has been the animating conversation in the tech world.

我的意思是，2022年的时候，真是令人惊叹。

I mean, 2022 was like, oh, wow.

电脑能跟你对话了，而且听起来不再像个傻瓜。

The computer can talk to you and doesn't sound like an idiot.

2023年基本上感觉像是原地踏步。

2023 was, basically, it it felt like treading water.

当时有很多关于这东西最终会走向何方的讨论，因为能力还不够先进。

There was a lot of, like, where is this gonna go because capabilities weren't advanced enough.

萨姆·阿尔特曼被解雇，然后又复职。

Sam Altman gets fired, reinstated.

我认为2024年和2025年才是真正开始看到应用落地的年份。

And then I think 2024 and '25 have really been the years where, like, the applications are starting to material.

比如，2024年可能是应用开始显现的年份，你真的感觉到这些东西可能行得通，而2025年则像是所有人都把资金全部投入这个项目，希望它能继续像以往那样持续增长。

Like, maybe '24 was when the application was starting to materialize, like, you really had a sense that this stuff might work, and now 2025 has been like, everybody's taking all of their money and putting all of it into this project in the hopes that, you know, it really will continue growing the way that it has.

这是否很好地反映了我们当前所处的位置？

Is a good sort of representation of where we are?

我喜欢这个说法。

I like that.

我喜欢这个说法。

I like that.

是的，我认为这是对的。

Yeah, I do think that's right.

去年我在预测中写过，关于基础设施，以及这其实是个显而易见的预测：我们今年的算力规模将比去年高出数倍。

And I did write in the predictions last year about, you know, infrastructure and the fact that it was probably an obvious prediction, but that, like, we were just going to see huge multiples more compute this year than we did last year.

我还提到，这些数据中心将变得如此庞大，以至于大型前沿模型的训练将跨越多个地点——我们在上个季度已经开始看到这种趋势。

And also talked about how these data centers were going to get so big that these large frontier models are going to be trained across multiple locations, we started to see in the last quarter.

所以，是的，我完全同意这一点。

So yeah, I mean, I totally agree with that.

我认为明年这种情况还会继续。

And I think we'll see that continue next year.

那么如果这些是2025年的标签，你认为2026年的标签会是什么？

So if those are the labels for 2025, what do you think the label is going to be for '26?

比如，当我们明年此时回望时，你觉得我们会处于什么阶段？

Like if we are going to look back this time next year, where do you think we are?

嗯。

Yeah.

我的意思是，我认为人工智能正达到这样一个阶段：它仍然是人们讨论的主要话题。

I mean, I think actually we're reaching this point in AI where it is still going to be the major topic of conversation.

但我觉得它的新奇感现在正在消退，我认为它会趋于平稳。

But I think some of the novelty of it is wearing off now and I think we'll see it plateau.

然后，我认为产品会变得重要得多。

Then I think the products will become much more important.

你提到过2024年的产品，但我觉得明年真正重要的将是产品。

You mentioned the products back in 2024, but I think next year it's really going to be about the products.

我不认为人们会对任何新模型的发布及其功能感到兴奋，因为对大多数人来说，现有的模型已经足够好了，真正重要的是推出新产品，而不是底层模型能力的提升。

I don't think people are going to be getting excited about whatever new model's coming out and what are the capabilities because I think for most people they're basically good enough and it's really about just adding new products rather than the underlying model capability.

所以我认为，至少对于消费者而言，明年将围绕这一点展开。

So I think that's what next year, at least for consumers, will be about.

在企业领域，情况可能略有不同。

In the enterprise, it's maybe a little different.

是的，我认为会有一些公司继续突破界限，并带来一些令人印象深刻的消息。

Yeah, I mean, I think there's going to be some companies that will continue to push the envelope and that will be some some pretty impressive news.

但我同时也认为，洗牌即将来临。

But I also think the shakeout is coming in a way.

在我看来，2026年将真正成为充满混乱的一年。

Like, to me, I think 2026 is really gonna be a year of chaos.

我们已经开始看到一些迹象，比如甲骨文和其他一些基础设施建设者，他们的赌注变得如此之大，而赌博的一点就在于，并不是每个人都能赢。

I mean, we're already starting to see some of this happen with Oracle and some of the other infrastructure builders and the bets have become so big and, you know, that's one of the things about gambling is not everybody wins their bets.

所以，没错。

So Yeah.

嗯。

Yeah.

没错。

That's true.

所以，我打算这样：我们即将有一期预测节目，嘉宾是兰詹·罗伊。

So so what I'm gonna do, think that, like, we have a prediction episode coming up with Ranjan Roy.

所以，与其只是随意提出一些疯狂的预测，我更想先为2026年做个铺垫，谈谈一些公司可能经历的事情，这种混乱会是什么样子，然后听听你的看法，里德。

So instead of just going, like, for our wild predictions, I just kinda wanted to set the table for 2026 and sort of throw some ideas about what the companies, you know, might be going through, what this chaos might look like, and then sort of hear your perspective about it, Reid.

这样好吗？

Does that sound good?

当然。

Sure.

嗯。

Yeah.

我也很想听听你是否不同意，因为我觉得争论这些话题很有趣。

And I'd love to hear if you disagree because I think this is it's fun to argue about this stuff.

当然。

Definitely.

好的。

Alright.

我们开始吧。

Let's do it.

抱歉打断一下这种方式。

Pardon the interruption style.

对于Meta，问题是那个超级智能实验室会怎样？

Meta, the question for Meta is like, what is gonna happen to that super intelligence lab?

我的意思是，这是一个数十亿美元的项目。

I mean, that is it's a multi billion dollar effort.

它产生过任何成果吗？

It hasn't has it produced anything?

明年，这个部门真的将面临要么交付成果，要么陷入混乱或危机的局面。

And it is really gonna be sort of deliver or or chaos or crisis mode for that division next year.

是的，我认为这将继续是混乱和危机状态。

Yeah, I think it's going to continue to be chaos and crisis mode.

我的意思是，你已经看到一些新闻出来了。

I mean, you're already seeing stories coming out.

我想《纽约时报》曾报道过产品团队和超级智能团队之间的一些摩擦。

I think it was in the Times about some sort of friction between the product teams and the super intelligence team.

我认为这种情况会持续下去。

I think that will continue.

我认为非常引人注目的是，Meta在人工智能领域起步很早，曾试图在谷歌之前收购DeepMind，但后来却逐渐失去了焦点。

I think it's just so remarkable that Meta, they were so early on AI, they tried to acquire DeepMind before Google did and then sort of just lost focus on it.

而现在，它似乎到处都乱糟糟的。

And now it just seems to be all over the place.

我认为对他们来说，未来将完全聚焦于产品，对吧？

And it's going to be all about, I think for them, again, it's products, right?

比如这些眼镜，把AI融入眼镜以及其所有产品中，但用户体验一直很笨拙，对吧？

It's like these glasses, the glasses incorporating AI into those and across its products, which has been sort of a clunky experience, right?

我认为他们推出的Facebook和Instagram AI产品并没有特别受欢迎。

I don't think the Facebook and Instagram AI products that they put out have been necessarily that popular.

所以我认为，在他们试图理清头绪的过程中，情况只会更加混乱。

So I think it'll just be more chaos as they try to figure it out.

我的意思是，我们来详细展开一下。

I mean, let's flesh that out a little bit.

所以很明显，第一个问题是亚历山大·王是否会留下。

So obviously, there's one question about whether Alexander Wang is gonna stay.

另一个问题是，他们是否会推出一个领先的模型，或者至少是一个让人尊重的模型？

The other is, are they gonna ship a leading model or at least a model that people respect?

因为自从Llama 3之后，它曾与一些顶尖的闭源模型不相上下，但此后他们却急剧下滑，尽管他们拥有所有专业知识和所需的全部GPU。

Because after, let's say, Lama three, which was neck and neck with some of the state of the art closed models, they've really fallen off a cliff and they have all the expertise and all the GPUs they could want.

当然，这类事情需要很长时间，但那里的核心优势在哪里呢？

And obviously this stuff takes a long time, but yet where's the cream filling?

我认为他们在推出开源模型并试图将其设为标准的策略上实际上已经失败了。

I think they've actually failed at that strategy of sort of putting out open source models and then trying to make that the standard.

他们在开源软件方面非常擅长。

They're really good at open source software.

他们多次通过PyTorch做到了这一点，但我认为AI不同，因为它的成本实在太高了。

They've done that many times with PyTorch, but I think AI is different because it's just so expensive.

我写过很多关于这个的话题，这些模型最终会不会贵到根本没必要发布一个领先的前沿模型并免费提供？

I wrote about this a bunch, like eventually do these models get so expensive that it just makes no sense to put out a leading frontier model and make it free?

现在你面临着来自中国和法国Mistral的激烈竞争，因此我认为，如今的新模型要么会作为他们产品的一部分，几乎成为后台技术，要么就会像OpenAI一样成为闭源模型。

You have so much competition now from China, from France with Mistral that it's just I think at this point the new models are either going to just be part of their products, like almost backend technology, or they'll be closed models just like OpenAI.

这是我的猜测，我觉得这是一个有根据的猜测。

That's my guess, I I think, like an educated guess, would say.

我不知道你是否同意这一点。

I don't know if you'd agree with that.

我也这么认为。

I think so.

我认为他们的开源战略已经彻底结束了。

Think they're not gonna their open source strategy is definitely done.

真正的问题是，他们能否构建出能与当今领先者相媲美的封闭模型？

The question really is can they build a even closed model that rivals today's leaders?

老实说，我认为他们没必要这么做。

Well, I don't think they have to, to be honest.

用这种方式思考，就像很多人对人工智能的固有看法一样。

Think that thinking about it that way is like the way a lot of people still think about AI.

就好像存在一场通往超级智能的竞赛。

It's like there's this race to super intelligence.

我不认为情况真是这样。

I don't think that's the case.

我认为现在关键是产品，如果人们喜欢Meta推出的那些基于基础模型的产品，老实说，这些产品其实可以用许多现有的不同模型来运行。

I think at this point it's about products and if people like the meta products that they put out there that are run on these foundation models, to be honest, they could probably be run on a lot of different models that are out there.

我认为这部分是出于自尊，部分是炫耀资本，也许还为了向华尔街展示自己站在技术前沿，但我真的认为他们没必要这么做。

I think it's partly ego, it's partly bragging rights and maybe a nod to maybe it's really for Wall Street to look like you're on the cutting edge, but I really don't think they have to.

我认为现在的模型已经相当不错了，彼此之间的差距也很小，没必要非得有什么突破性进展。

I think the models are pretty they're good enough and they're close enough that you don't have to have some breakthrough.

事实上，我认为OpenAI也是如此。

In fact, I think the same is true for probably OpenAI.

这可能是一个有争议的观点，但你同意吗？

Again, that's maybe a controversial opinion, but do you agree?

我本来想说，让我提出另一个观点，这基本上是对你的问题的一个回答。

Well, I was gonna say so let me let me throw this this other thing out there, which which is, which is basically is one answer to your question.

当你谈到Meta时，显然Meta完全可以不必构建超级智能。

As you're talking about Meta, it it does seem clear that Meta can effectively fail to build super intelligence.

即使他们构建出最差的模型，依然可以让他们的产品变得更好。

They could fail to build, you know, the worst model and still make their products better.

但他们的产品会是什么样子呢？

But then what do their products look like?

与以往相比，他们的产品中与AI的互动多了很多。

There's a lot more interaction with AI in them, than there were previously.

你从与人类朋友互动，转变为与你的AI机器人互动，而我们已经看到这种应用具有很强的吸引力。

They you go from, you know, human you friend a human to you friend your AI bot, and we've already seen that there's compelling uses for that.

所以也许明年真的是一个关键时刻，Facebook——我的意思是，我们已经看到一些迹象，比如招聘方面——但Facebook和OpenAI之间的竞争将彻底公开化，扎克伯格和萨姆可能产生分歧，Meta甚至可能开始看到用户使用量下降，因为人们会想：我不想再看别人在Instagram上展示的精彩生活。

So maybe next year really is, the moment where Facebook and I mean, we've already seen some of it with the recruiting, but Facebook and OpenAI's battle just explodes out into the open and maybe Zuck and Sam are at odds with each other and Meta might even see start to eventually see some use usage declines as people are like, I don't wanna look at other people's great lives on Instagram.

我想和我的AI治疗机器人聊天，它会告诉我我有多棒。

I wanna speak with my AI therapy bot who tells me how great I am.

嗯，这很有趣，因为所有这些科技公司总是在各个领域相互竞争。

Well, you know, that's interesting to think about, like, so so All these tech companies are always competing with each other on every front.

它们都在每个品类推出产品，而OpenAI现在已经推出了一款社交媒体应用。

They all make a product in every single category and OpenAI has now made a social media app.

我觉得这看起来像是失败了。

I think that's looking like it's failed.

它们把意大利面往墙上扔，看哪些能粘住。

They threw the spaghetti against the wall.

我觉得实际上，Facebook或Meta和Instagram更多是在与YouTube和TikTok竞争，这已经不再是纯粹的社交媒体了。

I think actually it looks like Facebook is or Meta and Facebook and Instagram are sort of competing more with YouTube and TikTok and it's not so much social media anymore.

它看起来更像是一种媒体和电视。

It looks a lot like just media and television.

我认为YouTube的策略是，如果你去和YouTube的CEO迪尔·莫汉聊聊，他们会说，他们并不在意内容是AI生成的还是人类生成的。

I think that's and the YouTube strategy is like if you talk to Deal Mohan, he's the CEO of YouTube, it's like they don't care whether it's AI generated content or human generated content.

他们只是想把你想要看、想要点击的内容推到你面前。

They just want to put content in front of you that you want to see and that you want to watch.

我认为这最终才是Meta的核心所在。

And I think that's ultimately what Met is about.

我的意思是，这本质上就是争夺眼球，对吧？

I mean, it's eyeball farming, right?

所以我不知道在AI时代，社交媒体和内容的未来会是什么样子。

And so I don't know what the future of social media and content is in the AI world.

我们可以对此做出一些猜测。

We can make guesses about that.

但Meta会观察实际情况，而且它们非常擅长模仿。

But Meta will see what it is and they're very good at copying that.

因此，无论眼球流向哪里，它们都会推出相应的产品，而且它们通常在这方面做得不错，我觉得。

So they will just mimic whatever, wherever the eyeballs are going, they'll do those products and they'll probably do, they generally do a good job of that, I think.

再多讲讲为什么你不认为OpenAI需要最顶尖的模型。

Talk a little bit more about why you don't think OpenAI needs the best models.

嗯，据《今日信息》报道，一个有趣的故事指出，OpenAI已经意识到，模型方面的进展并不一定转化为更好的消费者产品，随着模型能力的提升，这在公司内部引发了一些紧张。

Well, I mean, there was an interesting story in the Information Today that said, AI so OpenAI's realized that there's less progress on the model front consumers don't it doesn't necessarily translate into better consumer products as they increase the capabilities of the models and that's creating some tension within the company.

当然，公司回应称，这基本上不属实。

Of course the company responded and said that's more or less not true.

但我一年多来一直强调这一点：自从ChatGPT推出后，OpenAI就彻底从一家研究实验室转变为一家消费者产品公司。

But I've been calling this now for over a year, which is that real OpenAI, as soon as ChatGPT came out, completely changed from a research lab to a consumer products company.

是的，他们有一个正在增长的企业业务，我认为这可能是一个不错的收入来源，但他们真正拥有、而别人没有的是ChatGPT和大量的用户关注。

Yes, they have an enterprise business that's growing and I think that could be a good revenue source, but really what they have that nobody else has is ChatGPT and a lot of eyeballs.

如果只看数据和增长速度，它可能是历史上增长最快的消费者产品。

Think it was or is the fastest growing consumer product in history if you just look at the numbers and the speed of growth.

所以，他们要做的就是保持这些用户的参与度。

So they're just what they need to do is keep those people engaged.

这才是他们的产品，他们必须降低成本。

That's their product and they have to reduce costs.

我认为，当你谈到OpenAI自研芯片和数据中心时，这一切都是因为他们预见了未来：会有大量用户使用这些服务，运行数据中心的成本会非常高，因此他们必须制定一种策略来控制成本并实现垂直整合。

I think when you talk about OpenAI building its own chips and data centers and all this stuff, that's all about them foreseeing the future where there's so many people using these things, it's expensive to run-in the data center, they're going to have to have a strategy to control the costs and vertically integrate there.

这并不是关于我们能否构建超级智能。

It's not about can we build super intelligence?

我认为他们必须保持前沿地位，但这更多是因为他们仍有许多投资者，这些投资者相信他们投资这家公司是因为它将发明通用人工智能，并告诉他们如何赚钱，对吧？

I think they have to stay on the cutting edge, but that's more because they still have a lot of investors who believe that they've invested in this company because they're going to invent AGI and it's going to tell them how to make money, right?

他们必须持续讲述这个故事。

There's a story that they have to keep telling.

我认为这对招聘和其他方面都很重要，但最终我认为，OpenAI内部将仅保留一个研究实验室，就像谷歌多年来一直拥有的AI研究实验室一样。

And I think it's important for recruiting and all of that, but ultimately I think that will just become there will be a research lab within OpenAI just like Google has had for many years, like an AI research lab.

而现在，这正越来越多地融入产品之中。

And now that's becoming much more a part of product.

所以，这就是我如何看待这个问题的，不知是否说得通。

So that's kind of how I see it, if that makes sense.

对，说得通。

Yeah, makes sense.

我的意思是，我在这一点上不同意你的看法。

I mean, I think I will go against you on this one.

我认为他们确实需要最好的模型。

I do think they need the best models.

我认为他们的故事之所以重要，是建立在他们始终领先于所有人之上的。

I think their story and their story is important, is predicated on them being out ahead of everybody else.

有趣的是，在谷歌发布Gemini 3之前、Code Red事件发生前，公司内部一个被忽视的评论是萨姆说会有一些——我想他是不是说了‘负面情绪’，或者一些经济上的不利氛围，这恰恰说明‘最佳模型’的故事仍在继续，然后你才能逐步构建交付产品所需的基础设施。

And it was interesting how when Google released Gemini three, before the Code Red happened, the underappreciated comment within that company was Sam saying there are gonna be some, you know, I guess, like, did he say bad vibes or some economic bad, like, tough economic vibes, which which I think just indicates that best model story continues and then you can sort of build the infrastructure you need to be able to deliver the products.

这些不好吗？

Are those bad?

你请说。

Go ahead.

哦，好的。

Oh, good.

抱歉，刚才打断你了。

Sorry, cut you off then.

是的，不是的。

Yeah, no.

我只是想说，如果你在模型开发上落后太多，一些神奇的东西就会消失。

I was just going to say if you fall too behind on model development, then some of that magic evaporates.

我同意这一点。

I agree with that.

但我希望听听你的看法：当出现负面情绪、经济上的负面情绪时，是客户对产品不满意，还是更多关乎投资者、融资或借贷这类事情？

But I think the thing that I'd love to get your thought on is when the bad vibes, the bad economic vibes, is that their customers with the products or is that more about investors and raising money and, you know, or borrowing money, that kind of thing?

我认为这更多是关于投资者。

I think it's more about about investors.

这是我猜的。

That would be my guess.

那就是对的，对吧？

That's what that's what right?

是的。

Yeah.

但这笔钱现在对这家公司来说至关重要。

But that money right now is so important to the company.

完全正确。

Totally.

完全正确。

Totally.

这正是我的意思。

And that's what I mean.

我认为他们必须继续这样做，但这更多是关于表象。

I think they have to keep doing it because but it's more about appearances.

并不是说，如果他们不保持在模型前沿，用户就会去使用谷歌之类的平台，对吧？

It's not like, you know, if they don't stay on the cutting edge of models, the users will go to Google or something, right?

我认为在用户层面确实存在竞争，但更重要的是实际产品的体验如何，消费者能用它做什么，我认为这与底层模型能力的关系越来越小，而更多地取决于他们在这些模型周围构建的所有东西。

I think there is a competition for users there, but it's much more about what does the actual product feel, what are consumers able to do with this, Which has, I think, less and less to do with the underlying model capabilities and more to do with all the stuff they build around those models.

所以，这就是我的观点。

So that's sort of my view.

好的。

Okay.

这需要时间。

That'll take.

这需要时间。

That'll take.

我明白，我明白。

I I get I get that.

这是重要的细微差别，我们在这里很喜欢做这种分析。

That's important nuance, and we love doing that here.

好吧。

Alright.

哦，我们还有很多公司要谈，我想我们才真正完成了其中一个半。

Oh, we have so many companies, left to do, I think we've only really done one and a half.

好吧。

Alright.

谷歌？

Google?

速度。

Speed.

快速问答。

Speed round.

关于谷歌的事情是，我们将在明年看看他们是否能维持这种势头。

The Google thing is, like, we're gonna see next year whether or not they can maintain this momentum.

如果他们做到了，考虑到他们的起点和迄今为止的进展，他们实际上将成为人工智能领域的无可争议的领导者，这对他们来说是一次令人印象深刻的逆转。

And if they do, they effectively become the undisputed leader in AI given where they started and how they've been tracking, and that is that is a an impressive turnaround for them.

但但我的，好吧。

But but my alright.

我先说说我的看法。

I will go with my perspective on this first.

我不认为他们能做到。

I don't think they're gonna do it.

我觉得关于Gemini有多么出色的讨论太多了。

I think there's been all this buzz about Gemini being so great.

你使用这个产品，它确实不错，能做些很酷的事情，拥有许多有趣的功能，比如可以和视频对话，这一点我很喜欢。

You use the product, it's good, it does cool things, it has really interesting features, you can talk to videos which I like.

但它给人的感觉没有其他产品那么成熟，比如Claude和ChatGPT。

But it it doesn't doesn't feel as developed as as the others do, as Claude and as, ChatGPT.

如果你说：快去用个AI完成点事，我根本不会选Gemini。

Like, if you said run, go use an AI, and you need to get something done right away, I don't go to Gemini.

它就是，就是，抱歉。

It just it it sorry.

我本来想说点刻薄的话。

Well, I was gonna say something mean.

我只是觉得这跟你没关系。

I just think not about you.

是关于我，那我就直说了。

It's about I'll say it.

我说了。

I'll say it.

我说了。

I'll say it.

对我来说，Gemini 感觉像一篇 Axios 文章。

Gemini to me feels like an Axios article.

对吧？

Right?

就是给你列要点。

Like giving you the bullet points.

哇。

Wow.

我本来以为你要说 Google 的坏话，但你其实还行。

I thought you were gonna say something mean about Google, but you're okay.

去说 Axios 吧。

Go for Axios.

好吧。

Alright.

我喜欢Axios，但你知道，它就是给人感觉，哦，在这儿。

I like Axios, but, you know, it just feels like, you know, oh, here.

我们就直接给你列个要点，然后你就走吧。

Let's just give you, you know, the bullet points then you go away.

嗯。

Yeah.

不。

No.

我明白你的意思。

I see what you mean.

其实，我也喜欢Gemini。

I actually, I like Gemini too.

我经常用它。

I use that.

我认为它非常适合你所说的那种视觉类内容。

I think it's good for, like you kind of said, anything visual.

我用它来尝试重新装修我的房子，那时候家里一团糟，它真的帮了大忙。

I used it to try to redecorate my house which was just in mayhem and so it was really helpful for that.

我试过ChatGPT。

Tried ChatGPT.

我认为在多模态方面，它们更胜一筹。

I think on the multimodal front they're better.

不过，我的看法并不是谷歌赶上了什么的。

The way I look at it though is not that Google caught up or something like that.

我认为谷歌在人工智能方面本来就已经领先了，只是之前还没考虑如何将其产品化。

I think Google was already ahead on AI but they just hadn't been thinking about productizing it yet.

他们当时的想法是，这项技术还没成熟。

They were sort of like, This technology's not here yet.

还没准备好，所以我们打算等等。

It's not ready yet so we're going to wait.

你可以说那是个错误，也可能不是，但实际情况是，他们被拖入了这个新世界，不得不追赶的不是AI技术本身，而是如何将AI研究转化为产品，而他们基本上已经赶上了。

You could say that was a mistake or not, but whatever happened is they got dragged into this new world and they had to catch up not on AI, but just on how to turn this AI research into products and they pretty much have caught up.

但我认为他们仍面临一个大问题，那就是他们80%的收入我认为来自搜索广告生态系统。

But I think that they still have a big problem which is that their underlying they get 80% of their revenue I think from the search advertising ecosystem.

而整个行业正在发生变化。

And that's totally changing.

整个行业正在发生变化。

That whole industry is changing.

传统的搜索将会完全消失。

Think traditional search is going to be just gone.

就连谷歌搜索对我来说也完全不一样了。

Even Google search is to me totally different now.

这对收入会有什么影响？

And what does that do to revenue?

我认为这只是一个平衡问题，他们必须走一条钢丝，不能太快地破坏自己，但又必须主动进行自我颠覆。

And I think it's just a question of that's a very tricky balance that they have to it's a tight rope they have to walk where they can't disrupt themselves too fast, but they kind of have to disrupt themselves.

这中间有很多地方可能出错。

And there's a lot of ways that that can go wrong.

所以某种程度上我同意。

So I kind of agree.

这并不是说谷歌现在领先了，就可以继续在AI上变得更好。

It's not like Google's now in the lead and they can just continue getting better at AI.

这其实并不是关于如何在AI上做得更好。

This isn't really about being better at AI.

但另一方面，我认为把Gemini看得太狭隘了，因为他们的AI在很多地方都有体现，比如自动驾驶汽车，对吧？

But then again, I think that's also looking at Gemini as very narrow because in a lot of ways their AI is showing up in places like self driving cars, right?

Waymo的应用无处不在。

The Waymo thing is all over the place.

如果我们愿意，可以聊聊自动驾驶的预测，但还有量子计算，对吧？那里有潜力实现突破，进而推动Isomorphic Labs——他们的生物科技子公司，这可能会成为全新的收入来源。

We could talk about self driving car predictions if you want to, But quantum computing, right, where you've got a potential there for breakthroughs in quantum which then would feed into isomorphic labs, their biotech spin off, right, Which could be whole new sources of revenue.

所以我认为，在很多方面，他们是一个庞大的帝国，拥有许多我认为非常出色的东西。

So I think in a lot of ways, they're this giant empire that has a lot of, I think really good things going.

所以我会把范围扩大到Gemini之外，如果这样说得通的话。

So I'd sort of broaden it out beyond Gemini, if that makes sense.

是的。

Yeah.

这确实有道理。

That that does make sense.

我的意思是，Waymo几乎每几集就会被提到一次。

I mean, this is like a place where I mean, Waymo is brought up every, you know, couple episodes.

对于Waymo，你唯一能说的就是：他们真的在做。

And there's the only thing you can do with Waymo is just be like, they're really doing it.

谷歌在Waymo、AI、量子计算方面，还有收购DeepMind的事实，都说明他们是一家聪明的公司。

It's Google like between Waymo, AI, Quantum, I mean the fact that they acquired DeepMind, they're a smart company.

是的，我认为这是一家非常聪明的公司，我认为桑达尔虽然经常被批评，但他确实能快速动员和变革。

Yeah, I think it's a very smart company and I think Sundar, you know, he gets a lot of crap I think but like to be able to quickly mobilize and change.

如果你去问谷歌的员工，现在的氛围和以前完全不同了。

If you talk to people at Google, it's a totally different vibe now than it used to be.

现在的感觉更像是‘咱们赶紧干起来’，节奏很快，而我以前觉得它有点沉闷。

It's much more of a, hey, let's get going, fast paced vibe versus I thought it was it's kind of sleepy before.

这就是我的感受。

That was my vibe, so.

是啊，某种程度上就像个养老院。

Yeah, was a retirement home in a way.

我知道这么说不对，但这笑话里有几分真道理。

I mean, know that's wrong, but there's truth in the joke.

是啊，不是对每个人都这样，对吧？

Yeah, not for everyone, right?

公司规模很大，但确实有些人一直过得挺轻松，我想他们正以某种方式被请走。

It's huge a company, but there were some people who for sure were taking it easy and I think they're being ushered out in one way or another.

但要扭转这艘大船很难，所以我完全同意他们面临巨大挑战，但这只是更大的一部分——我甚至还没提到他们的TPU，对吧？

But it's hard, it's tough to turn around that ship, so I totally agree that they have huge challenges, but it's such a bigger I mean, haven't even talked about their TPUs, right?

我觉得这么多AI相关的事情都关乎成本和垂直整合，而他们在这方面是领先的。

I think so much of this AI stuff is about cost and vertical integration and they're leading there.

我认为所有这些数据中心的建设，从我的角度看，训练部分固然有趣，但真正关键的是推理。

I think all this data center build out, the way I think about it is it's like the training part is interesting but it's really about inference.

真正重要的是以最高效的方式向消费者和企业交付产品。

It's really about serving the products to consumers and enterprise in the most efficient way possible.

亚马逊在这方面做得也很不错，微软也是如此，但谷歌似乎在这方面略胜一筹。

Amazon's doing a good job of that too, so is Microsoft, but Google seems to be just ahead there.

我不知道你怎么看这一点。

I don't know how you feel about that.

不，我也同意这一点。

No, I agree with that as well.

我认为，Gemini 的炒作会逐渐消退，但谷歌公司整体状况非常好。

Perspective on this is the Gemini hype is going to fade a little bit, but Google the company is in really good shape.

好的。

Alright.

我需要去休息一下。

I need to, take a quick break.

我们回来继续。

Let's come back.

我们来谈亚马逊。

We'll do Amazon.

我们来谈苹果，再看看还能聊些什么。

We'll do Apple, and we'll see what else we can get to.

奈飞？

Netflix?

我不确定。

I don't know.

我们就在接下来聊这个。

Let's let's do that right after this.

Capital One的技术团队不只是在谈论多智能体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，它不仅能帮助买家找到心仪的汽车。

Using self reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love.

还能帮助预约试驾、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 in value.

先进、直观且已投入应用。

Advanced, intuitive, and deployed.

这就是它们的优势所在。

That's how they stack.

这就是Capital One的技术。

That's technology at Capital One.

我们回到《大型科技播客》，嘉宾是丽塔·阿尔布拉加蒂。

And we're back here on big technology podcast with Rita Albragati.

他是Symaphore的技术编辑。

He's the technology editor at Symaphore.

来自未来的关于亚马逊的头条新闻。

Headline from the future about Amazon.

我这里有一条。

I have it here.

亚马逊与OpenAI和Anthropic合作，将购物功能引入聊天机器人。

Amazon partners with OpenAI and Anthropic to bring shopping into chatbots.

我们会看到吗？

Are we gonna see it?

会的。

Yeah.

我的意思是，这对我来说完全合理。

I mean, that would make total sense to me.

这完全说得通，因为现在人们正是想要这样的服务。

I would make total sense because right now that's what people want.

现在每个人都在试图通过亚马逊实现这一点，他们非常希望绕过亚马逊。

Everyone is trying to basically do that with Amazon now and they would love to disintermediate Amazon.

所有这些聊天机器人公司，包括Perplexity，都希望你能直接在聊天中完成亚马逊购物，而无需访问亚马逊网站。

All these chatbot companies, Perplexity too, they'd love to just have you be able to shop on Amazon and never have to go to Amazon.

当然，这对亚马逊来说是个糟糕的交易。

Of course, that's a bad deal for Amazon.

所以我认为自然的解决方案是他们必须合作，无论通过MCP服务器还是其他方式，共同打造一个顺畅的购物体验。

So I think the natural solution is they've got to work together and they've to figure out how with whatever it is, MCP servers or something else, to just kind of make shopping experience work.

是的，我觉得这正在发生。

Yeah, no, I think it's happening.

而且亚马逊显然在关注这100亿美元对OpenAI的投资，这很有趣，因为他们已经在Anthropic投入了数十亿美元。

And of course, Amazon's eyeing this $10,000,000,000 investment into OpenAI, which is fascinating because they already have many billions in Anthropic.

这真的很有趣。

That's really interesting.

我听说一个传言，虽然还没被证实，就是OpenAI可能会使用亚马逊的Tranium芯片，这也很有意思。

And I think the rumor I don't think it's been confirmed that OpenAI is going to actually use the Tranium, the Amazon Tranium chips is another really interesting one.

如果真发生了，Anthropic和OpenAI都使用Tranium芯片，那对亚马逊在定制芯片领域将是一个巨大的胜利。

That to me, if that happens, you've got Anthropic and OpenAI using Tranium, that's a huge win for them on the custom chip front.

这将帮助他们开发下一代芯片，我认为这将在垂直整合和成本方面成为对谷歌TPU的真正竞争者，尽管可能不是直接竞争。

It will help them develop their next generation of chips and I think will be a real competitor on the, not directly maybe to the TPUs at Google, but just in terms of vertical integration and cost.

对。

Right.

这是来自未来的苹果头条新闻。

Here's Apple headline from the future.

苹果。

Apple.

我和MGC不久前讨论过这件事，虽然有争议，但我仍然坚持这个观点。

And MGC and I talked about this a little while ago and it's been disputed, but I'm still going with it.

蒂姆·库克在第一季度电话会议上宣布，他将在年底辞去公司职务。

Tim Cook announces on first quarter conference call that he's leaving the company at the end of the year.

他功成身退，将送给世界最后一份礼物——折叠屏iPhone，这将成为公司的旗舰手机。

He's going out on top, that he's going to give the world his last little gift, is the folding iPhone, which will be the flagship phone for the company.

嗯，严格来说不是旗舰，但将是苹果有史以来利润最高的iPhone，而蒂姆会像迈克尔·乔丹那样耸耸肩说：‘我还能做什么呢？’然后功成身退。

Well, not the flagship, but the the most profitable iPhone Apple has ever made, and Tim will shrug his shoulders like Michael Jordan and say, what more can I do and be done?

我不知道这是否会成真，但正如你所说，我读过一些相互矛盾的报道，不过这对他来说可能是个不错的决定。

I don't know if that will happen, but I've read, as you said, conflicting reports on that, but probably a good move for him.

只要你看看这家公司的市值，他所做到的简直不可思议，而且以他的年纪，就个人声誉而言，已经没什么太多可追求的了。

Just if you look at the market cap of that company, what he's done is incredible, and I think at his age, there's just very little upside in terms of just his legacy.

我不知道他还有多少，但你面对着这么多与中国相关的问题，所有这些事情。

I don't know how much he And you've got just so many issues with China, all this stuff.

苹果的未来真是个巨大的头疼问题。

It's just like the future of Apple is such a massive headache.

如果是我，我肯定会做出这样的决定。

I think if it were me, I would totally make that move.

但话又说回来，这正是我不是苹果亿万富翁CEO的原因。

But then again, that's why I'm not a billionaire CEO of Apple.

他们不一样。

They different.

但没错。

But yeah.

我喜欢，我喜欢这样。

I like I like that.

我喜欢这样。

I like that.

特斯拉。

Tesla.

特斯拉最重要的事情是自动驾驶。

So Tesla is the the big thing about Tesla is self driving.

我想知道2026年会不会成为特斯拉自动驾驶的一大关键年份。

And I wonder if 2026 could be a big year for Tesla's self driving effort.

我刚在达拉斯，坐了一辆由人类司机驾驶的Uber特斯拉，我就说：嘿。

I was just in, in Dallas, and, I got picked up in a Uber Tesla and with a human driver, but I said, hey.

你能把那个功能打开吗？

Can you can you flick that thing on?

他照做了。

And he did.