全球百强生成式AI消费应用

文化变革和文化采纳的速度会比技术实际可行性的变化更慢。

The cultural change and the cultural adoption will be slower than the technology change in what's actually possible.

如果你看一下正在Claude和ChatGPT上涌现的应用商店，它们都有200多个应用，但重叠率只有11%。

If you actually look at the app stores that are emerging on Claude and ChatGPT, they both have 200 plus apps, but there's only 11% overlap.

本质上，你可以用你的ChatGPT账户登录，并携带你的记忆和令牌，这样其他产品就能借用这些资源，为你提供更强大、更有帮助的服务。

Essentially, you'd be able to log in with your ChatGPT account and take, like, your memory and your tokens with you, and then that other product would be able to kind of borrow those things to be even more powerful and helpful for you.

我认为，目前还没有一个完全由AI内容构成的社会化产品取得成功。

I don't think we've seen a social product yet succeed that's, like, entirely AI content.

在某些方面，情感上的投入感似乎更低。

The emotional stakes just feel lower in some ways.

ChatGPT是全球最大的AI产品。

ChatGPT is the biggest AI product in the world.

但它每周仅覆盖了全球人口的10%。

It's also only reaching 10% of the global population on a weekly basis.

我们仍然处于早期阶段。

We're still early.

在a16z人工智能行动榜单的第六期中，各平台正在分化。

Six editions into the a16z top 100 AI action port, the platforms are diverging.

Claude正在全力投入面向专业用户的工具。

Claude is doubling down prosumer tools.

Gemini的用户增长几乎完全跟随创意模型的发布节奏。

Gemini's traction tracks almost perfectly to creative model releases.

而ChatGPT正在打造一个更像谷歌的全能应用，通过广告、交易和订阅实现盈利。

And ChatGPT is building towards something that looks more like Google, an everything app that monetizes through ads, transactions, and subscriptions.

全球数据令人惊讶。

The global data is surprising.

新加坡、香港和阿联酋在人均AI使用率上位居前列。

Singapore, Hong Kong, and The UAE lead per capita AI usage.

美国位列第20名。

The US sits at number 20.

对人工智能的信任度差异巨大，美国为32%，而中国高达80%。

Trust in AI varies wildly, from 32% in The US to 80% in China.

a16z的普通合伙人阿尼什·阿查里亚与a16z的合伙人奥利维亚·摩尔对话。

Anish Acharya, general partner at a16z, speaks with Olivia Moore, partner at a16z.

奥利维亚，欢迎你。

Olivia, welcome.

谢谢你们邀请我。

Thanks for having me.

这一年最令人兴奋的时刻到了，因为今天就要发布顶级100报告了，你觉得呢？

It's the most exciting time of the year, which is the top 100 report is coming out today, think.

对吗？

Is that right?

是的。

Yep.

这已经是三年来的第六期了。

It's been six editions over three years.

跟我们聊聊吧，有哪些没变，哪些变了，你的兴奋点在哪里？这份报告到底有什么新动向？

Talk to us about what's the same, what's changed, what's your excitement level, what's up with the report?

是的

Yeah.

在很多方面，自2023年我们首次发布这份榜单以来，发生了巨大的变化，取得了令人难以置信的增长。

In many ways, like so much has changed and there's been just an incredible amount of growth since the first time we put out this list in 2023.

但从宏观层面来看，我们仍然处于非常早期的阶段。

On the other hand, from like a macro level, we're still so early.

比如，ChatGPT 是迄今为止全球最大的AI产品，但全球只有10%的人口每周活跃使用它。

Like ChatGPT is by far the biggest global AI product and still only 10% of the global population is using it on a weekly active basis.

所以，还有更多潜力等待释放。

So there's like a lot more to come.

我认为过去六个月可能是我最喜爱、最激动人心的时期，因为我们看到了诸多转变。

I do think this past six months has been maybe my favorite time and the most exciting time because of the shifts that we've seen.

其中之一就是，面向消费者的竞争正在急剧升温。

One of them has been that the race for the consumer is really heating up.

当然有ChatGPT，但Gemini和Claude也在加大对消费者和准专业用户群体的投入。

So ChatGPT of course, but also Gemini and Claude are kind of doubling down on their own ICP within consumer and prosumer.

我认为我们开始看到这些平台随着时间推移可能产生复利优势。

And I think we're starting to see how these platforms might have compounding advantages over time.

因此，谁在获取最多用户这一点上显得尤为关键和有趣。

And so that makes it especially kind of existential or interesting of who is acquiring the most users.

相关的是，这是我们的第一期报告，纳入了并非原生AI产品但现已主要由AI功能驱动的产品。

And then on a related note, this was actually the first issue that we included products that were non AI native but are now majority AI enabled.

比如Canva、Notion、Freepik。

So things like Canva, Notion, Freepik.

Notion实际上宣布，现在他们认为一半的新年度经常性收入是由AI优先功能驱动的，这非常棒。

Notion actually announced that now they think half of their new ARR is driven by AI first features, which is very cool.

最后，我认为我们看到AI的应用已经远远超出了网站或应用的对话框范围。

And then lastly, I think we've seen a big expansion of AI outside of just like the website or app prompt box.

我们看到了许多新推出的浏览器，比如Dia、Comet、Atlas。

So we have all of the browsers that have come out like Dia, Comet, Atlas.

我们有Claude集成到Excel中，PowerPoint在Chrome中的应用，还有桌面应用如Cursor、Whisper Flow、Granola，因此人们使用AI的方式正经历一场令人兴奋的爆发。

We have Claude in Excel, PowerPoint in Chrome, and then we have desktop apps like Cursor, Whisper Flow, Granola, and so there's been just a really kind of exciting explosion in the ways that people are using AI.

太令人兴奋了。

So exciting.

这里有很多内容要谈。

There's a ton to cover here.

那我们先从大型基础模型开始吧。

So let's start with the big foundation models.

你能谈谈你认为 Gemini、Claude 和 ChatGPT 各自的专长领域是什么吗？

Can you talk a bit about what you think are the respective areas of specialization for Gemini, Claude, and then of course ChatGPT?

因为感觉这更像是一个共同提升的故事，而不是这些模型彼此竞争。

Because it feels like it's been a rising tide story more than these models trading off with each other.

是的。

Yes.

我同意。

I agree.

尽管过去一周可能有些戏剧性，比如凯蒂·佩里在 Twitter 上就 LLO 表态，这真是我从未料到的。

Despite the drama maybe of the past week where we have Katy Perry taking sides on Twitter in the LLO which is something that I didn't ever see coming.

我认为，从基础层面来看，如果你观察AI的使用情况，ChatGPT仍然是一个非常明显的赢家。

I think at a base level still if you look at AI usage, ChatGPT is a very, very clear winner.

在网页端，它的用户量是Gemini的2.7倍，在移动端则是Gemini的2.5倍。

So on web they're 2.7 times bigger than Gemini, on mobile they're 2.5 times bigger than Gemini.

尽管像科技推特上的讨论那样，Claude在网页端的用户量几乎只有它的1/30，在移动端更是只有它的1/80。

And then despite again like the kind of tech Twitter discourse, Claude, they're almost 30 times bigger than Claude on web and almost 80 times bigger than Claude on mobile.

我们之前看到过萨姆·阿尔特曼在超级碗广告大战时期发的那条推文。

So we had seen that Sam Altman tweet back in the Super Bowl ad wars era.

德克萨斯推文。

The Texas tweet.

是的。

Yes.

他说，在德克萨斯州使用ChatGPT免费版的人数，比Claude全球所有用户加起来还多，这确实是事实。

He was like, we have more people using ChatGPT free version in Texas than Claude has all users globally, which is true.

话虽如此，我认为我们正在看到，或许用‘两极分化’这个词不太准确，但用户使用的工具种类和用途确实在扩展，这在一定程度上改变了市场份额。

That being said, I think we are seeing, I don't think bifurcation is the right word, but maybe expansion in the number of products people are using and what they're using different products for, which has kind of changed the market share a little bit.

特别是Claude，它大力聚焦于专业消费者市场，推出了Cowork、Claude Code、Claude在Excel和PowerPoint中的应用等功能。

Claude in particular is really doubled down on prosumer with things like Cowork, Claude Code, Claude in Excel and PowerPoint.

如果你仔细查看Claude和ChatGPT上新兴的应用商店，两者都拥有200多个应用，但重叠率仅有11%。

If you actually look at the app stores that are emerging on Claude and ChatGPT, they both have 200 plus apps, but there's only 11% overlap.

Claude非常专注于高端数据源、研究工具、科学工具和金融数据，而ChatGPT则主要深耕消费类市场，如旅游、营养和消费金融等领域。

Like Claude is very much doubling down on like premium data sources, research tools, science tools, financial data, and ChatGPT is really doubling down on like consumer marketplaces, travel, nutrition, consumer finance, things like that.

而Gemini也占据着自己独特的一席之地。

And then Gemini is kind of in its own little corner as well.

其用户增长主要得益于创意类工具的推动。

And the traction has largely been driven by creative tools there.

因此，如果你观察他们的活跃用户和付费用户数量，几乎完美地与VO3、Nano Banana One、Nano Banana Pro、Nano Banana Two等版本的发布相对应。

So if you look at their kind of active users and paying users, it's nearly perfectly correlated to releases of like VO3, Nano Banana one, Nano Banana Pro, Nano Banana two.

他们在专业消费者领域也做了一些拓展。

They're doing a little more on prosumer.

他们正在将AI功能集成到Gmail、Sheets和日历中。

They're adding AI to Gmail, Sheets, Calendar.

嗯哼。

Mhmm.

但这些都被他们现有的产品所涵盖，而不是全新的体验。

But that's all being captured by like their existing products versus like a net new experience.

也许我们可以深入探讨一下应用商店的动态，因为你说的是ChatGPT所谓的应用目录吗？

Maybe let's dig into the App Store dynamic a little bit because Can that's so you talk about the bull case for ChatGPT with I think what they call the apps directory?

是的，对。

Yeah, yeah.

我认为ChatGPT采取的方法，正如萨姆自己在推特上所说，是成为每个人的AI。

I think the approach we're seeing with ChatGPT, and Sam said this himself on Twitter, is we want to be the AI for everyone.

这意味着他们试图吸引每一位消费者，并以不同的方式实现盈利。

And that means that they're trying to acquire every consumer and they'll monetize them in different ways.

我认为Claude已经非常明确地表示，他们只会通过订阅来盈利，这对能够支付订阅费的个人和公司来说很好，但不会覆盖所有人。

So like I think Claude has been very clear that they're just going to monetize via subscriptions, which is great for people and companies who can pay for subscriptions but it won't be everyone.

我认为你在他们所侧重的插件上也能看到这一点，这些插件是付费的、高ACV的、面向工作数据的工具。

I think you see that with the plugins that they're leaning into which are like paid, high ACV, like work data tools.

类似Similar Web这样的工具。

Similar web, things like that.

是的，类似这样的工具。

Yeah, things like that.

比如PitchBook，如果你是投资者、科学家或数学家，你会用到这些工具。

Like PitchBook, like things that you would use if you're an investor, a scientist, a mathematician.

我认为ChatGPT正采取一种类似谷歌的路径，即构建普通用户都会想用的产品，虽然目前只有较少比例的用户会转化为订阅用户，但他们可以通过广告，也可能通过交易来实现这些用户的变现。

And ChatGPT I think is going more of the somewhat of a Google type approach in that they're building things that like the average person will want to use and maybe a smaller percentage of those convert to subscriptions right now, but they will be able to monetize those people through ads and probably also I would guess through transactions.

比如，如果他们正在构建预订旅行或完成其他各类长尾消费者购买的入口，理论上他们最终应该能从自己带来的流量中抽取一定比例的分成。

Like if they're building the gateway to book a trip or do all of these other kind of long tail consumer purchases, hypothetically they should eventually be able to take some kind of cut of that at least for the traffic that they're driving.

因此，我认为这就是ChatGPT应用商店的看涨理由，尽管目前数据中尚未体现，但未来一两年内可能会变得更加明显。

And so I think that that is the bull case for the ChatGPT app store that isn't yet showing up in the data that will probably become like even more evident in the next year or two.

是的，这确实很有趣，因为它触及了你报告中提到的复合优势以及上下文如何累积的观点。

Yeah, it's really interesting because it touches on your point in the report about compounding advantages and how context compounds.

你能谈谈这个概念吗？

Can you talk a little bit about that concept?

那么，你用什么指标来衡量它呢？

And then what's your proxy in terms of a metric for it?

是会话时长吗？

Is it session times?

是会话次数吗？

It number of sessions?

是你提供的数据量，还是有其他因素？

Is it the amount of data you've provided or is there something else?

是的，这个问题让我非常兴奋，因为我认为到目前为止，对于像ChatGPT、Claude、Gemini、Perplexity这样的通用大语言模型，我们一直生活在一个上下文和记忆相对容易迁移的世界里。

Yeah, this is a really exciting question to me because I think thus far with these horizontal LLMs like ChatGPT, Claude, Gemini, Perplexity, we've kind of lived in a world where the context and the memory is somewhat easily exportable.

Claude最近就围绕这一点做过推广，但我认为未来会越来越难以迁移，而且我认为这实际上可能对ChatGPT这类更广泛的通用工具有利，原因有几个。

Claude ran a campaign around this recently, but I think there's going to be increasing lock in and I do think that probably actually benefits the broader more horizontal tools like ChatGPT for a few reasons.

所以我认为，第一点，我们已经看到ChatGPT开始专注于或构建一些让你在平台上与其他用户互动的产品。

So I think one, we've already seen ChatGPT focus on or start to build out products where you interact with other people on them through the platform.

比如群聊，想象一下如果ChatGPT的群聊功能变得更加成功，而你的所有朋友都在上面，那么如果你想要离开ChatGPT，你还得说服他们所有人一起转到别的平台——这就是网络效应。

So the group chats, imagine if there's an even more successful version of ChatGPT group chats and all of your friends are on there, then if you wanted to churn from ChatGPT, you'd also have to convince them all to go to Classic network effect.

没错。

Exactly.

我认为第二个方面有点像苹果和谷歌的对比：随着这些应用商店的兴起，开发者很可能会根据谁拥有最多用户，或者在某些情况下谁更愿意付费，来集中精力优先为谁开发应用。

I would say the second one is kind of also like an Apple Google comparison in that as these app stores emerge, it is likely that developers might start to concentrate their time and effort in who they build for in the most sophisticated way, who they ship to first, depending on who has the most users or maybe in some cases who's the most willing to pay.

但对于许多消费类工具来说，关键还是谁拥有最多的用户。

But for a lot of these consumer tools it'll be who has the most users.

所以我认为这也有利于ChatGPT。

So I think that also benefits ChatGPT.

另外一件让我今年最期待的事情，是萨姆·阿尔特曼曾暗示过的ChatGPT身份认证层。

And then the other thing probably that I'm most excited for this year that Sam Altman had kind of hinted at is this like authentication with ChatGPT layer.

是的。

Yeah.

本质上，你可以用你的ChatGPT账户登录，并携带你的记忆和积分跨平台使用。

So essentially you'd be able to log in with your ChatGPT account and take like your memory and your tokens with you.

然后其他产品就能借用这些信息，为你提供更强大、更有帮助的服务。

And then that other product would be able to kind of borrow those things to be even more powerful and helpful for you.

如果是这样，你就希望你的核心身份更多地存在于ChatGPT中，这样它就能将这些能力借给其他对你更有帮助的工具。

And if that's the case, then you're wanting to have more of your core identity live on ChatGPT because then it can lend it to these other tools that are even better for you.

这非常聪明，而且确实发挥了他们的优势，因为他们已经有九亿注册用户，而第三方开发者理想情况下并不想为推理付费。

It's so smart and it really plays to their advantages in that they have sign ups for 900,000,000 people, and then the third party developer ideally would not want to pay for the inference.

因此，用户可以带着自己的推理能力走，这对开发者来说是一种优势。

So the user can bring their inference capacity with them, There's an advantage for the developer.

ChatGPT获得了用户粘性。

ChatGPT gets the lock in.

用户则获得了个性化的好处，一切都很顺畅。

The user gets the benefit of personalization and it all kind of works.

是的。

Yes.

我完全同意。

I totally agree.

我对此仍有一个疑问，我认为这可能会对消费者产品的用户粘性产生正面或负面的影响，那就是你的工作内容与什么企业合同绑定。

The one question mark I still have on this that I think could play both positive and negative in terms of increasing lock in for the consumer product is what your work goes with, like what your enterprise contract is.

所以，例如，如果我的公司用ChatGPT来工作，这对我来说其实是好事，因为那样我就熟悉这个产品了。

So for example, in some ways it's good for me if my company uses ChatGPT for work because then I know how to use the product.

而作为普通消费者，他们可能只试用过一两个AI产品。

And as a normal consumer, they might have tried one or two AI products.

他们更有可能对已经使用过的东西感到舒适并继续使用。

They're more likely to be comfortable and keep using something that they've already used.

另一方面，有些人可能不希望在个人和工作场景中混用身份和记忆。

On the other hand, some people might not want to mix identity and mix memory across their personal and work use cases.

确实如此。

True.

所以我非常感兴趣，我认为OpenAI最近也暗示过这一点，我很想知道我们如何在你自己使用的不同身份之间划分记忆。

And so I'm really interested, I think OpenAI hinted at this recently, but I'm really interested in how we kind of segment memory across different personas that are within yourself that are using these products.

别让能量束交叉。

Don't cross the streams.

对，没错。

Yeah, exactly.

没错。

Exactly.

好吧，也许我们暂时换个话题，谈谈Gemini。想想谷歌早期AI产品BARD带来的那种氛围，他们将会

Well, maybe actually switching gears to Gemini for a moment, you know, think about the kind of just the vibes around Google with their early AI products, BARD, which they'll

是的，那段艰难时期永远不会被忘记。

Yes, never live some tough times there.

再到我们今天拥有的像Nano Banana这样的产品，甚至给它起名Nano Banana，都完美地缩影了谷歌走了多远。

To where we are today with products like Nano Banana, like even naming it Nano Banana is such a perfect microcosm for how far Google has come.

是的。

Yeah.

而且看起来他们在多模态方面有很多打算。

And it seems like they have a lot of intentions around multimodality.

对。

Yeah.

你对他们的方法有何评价？

What's your assessment of their approach?

我印象深刻。

I've been impressed.

我认为他们一直比较谨慎，某种程度上，这种谨慎恰恰符合我们对将AI融入核心功能时的预期，因为存在风险——可能会蚕食自己的产品，或者有太多人已经使用这些工具长达十年、二十年、三十年甚至四十年，因此转换成本确实很高。

I think they have been hesitant, maybe in some ways more hesitant in same ways as exactly what we would expect to kind of bake AI into the core features because there's a risk of either like cannibalizing their own product or like there are so many people who are have used these tools for ten, twenty, thirty, forty years and so the switching costs there is like a little bit high.

他们不希望当AI突然出现时吓到用户，这一点我能理解。

They don't want to scare users when AI is suddenly popping up and everything which I understand.

但他们真正做得非常出色的是这些全新的创意产品，这些产品基本上由DeepMind团队主导，我认为这个团队整体上非常出色。

But what they've done a really, really good job at is these new creative products that are basically very model driven from the DeepMind team who I think is generally fantastic.

我认为NotebookLM是首次展现出这种模式的产物，它在消费级AI音频领域确实是全新的东西。

I think NotebookLM was actually the first look at this and that was something truly new in like consumer AI audio.

现在我们有了图像和视频模型。

And now we have the image and video models.

因此，对于这种大型公司来说，他们必须先克服自身障碍，才能真正实现创新。

So in some ways with a big company like this, they kind of have to get over, get out of their own way in terms of being able to actually innovate.

看起来他们正在这么做，但你也曾在谷歌工作过，所以我很好奇你的看法

And it seems like they are, but you also worked at Google, so I'd be curious for your

这很有趣，很高兴你提到Notebook，因为Notebook在公司产品线中是一个全新的领域，不会有十位副总裁争抢它。

Well, it's interesting to just I'm glad you brought up Notebook, because Notebook is sort of this greenfield area within product area within the company, so you don't have 10 VPs fighting over it.

是的。

Yeah.

因此，我认为Notebook的进展非常显著。

And as a result, I think just the progress on Notebook has been tremendous.

你知道，他们刚刚推出了视频生成功能，是的。

You know, they just launched a video generation feature Yeah.

这有助于以视觉方式展示你工作空间中的所有概念，挺酷的。

That helps visually demonstrate all the concepts in your sort of workspace, which is cool.

相反，当你看现有的产品界面，比如Sheets或Docs时，过去积累的惯性和推动力太多了，再加上管理上的负担。

Conversely, when you look at the existing product surfaces like Sheets or Docs, there's just so much one sort of momentum and inertia from the past, but then management overhead around those.

他们很难做出任何超出最明显增量改进的事情。

It's harder for them to do anything other than the most obvious incremental thing.

是的，我同意。

Yes, I agree.

我们看看未来几年那里会发生什么。

We'll see what happens there in the next few years.

我觉得他们在某些产品上会奋力抵抗，因为他们不想失去用户基础，但正如你所说，他们已经与众多企业深度绑定，至少在近期内可能不需要做太多就能维持现状。

I feel like they're gonna put up a fight on some of those products because they don't want to lose that user base but to your point they're already locked in with so many enterprises that they might not have to do that much at least in the near term to kind of keep up.

在这场对话中隐含的是，我们在西方经常体验和谈论人工智能。

You know implicit in this conversation is we experience and talk a lot about AI in the West.

稍微聊聊全球人工智能的趋势。

Talk a bit about the sort of global AI trends.

我在这里看到了一些令人惊讶的事情。

There was a few surprising things I saw in there.

我们在这份报告中扩大了研究范围，最终发现非常有趣。

We kind of expanded our scope in terms of what we looked at for this report which ended up being like very fun and interesting.

有两个方面可能很明显，那就是它们与世界其他地区的差异：俄罗斯和中国。

Two things that are probably obvious in terms of how they differ from the rest of the world would be Russia and China.

俄罗斯方面，我认为中国大家都清楚，很多人工智能产品都被审查或禁止了，因此他们的使用率几乎最低，ChatGPT和Gemini的综合使用率仅为15%。

So Russia, I think China everyone knows like a ton of AI products are kind of censored or banned and so almost all of the usage, they actually have the lowest combined ChatGPT and Gemini usage of any country, it's only 15%.

所以他们主要使用的是字节跳动开发的豆包、DeepSeek、通义千问、Kimi这类模型。

So they're mostly using like Daobao which is made by ByteDance, DeepSeek, Quen, Kimi, those kinds of models.

让我有点意外的是，俄罗斯的情况也非常相似，由于受到某些制裁等因素，他们不得不建立自己的平行AI生态系统，无法使用所有美国工具。

Somewhat of a surprise to me was that Russia actually is a very similar story where they have also their own kind of parallel AI ecosystem out of necessity because they have some level of sanctions and things like that that prevent them from using all The US based tools.

因此我们看到像GigaChat和Yandex这样的产品，这些都是由俄罗斯公司、通常是与国家有关联的企业开发的，在俄罗斯有巨大的使用量，同时DeepSeek也在其中。

So we've seen products like Giga Chat and Yandex which are Russia specific built by Russian often state affiliated companies have big, big usage there and then DeepSeek.

因此，俄罗斯是DeepSeek的第二大市场，仅次于中国。

So Russia is the number two market for DeepSeek after China.

所以如果你看一下各国采用率的数据，会发现有些小波动，比如某个国家使用Claude多一点，另一个国家使用Gemini多一点，但两个巨大的异常值是俄罗斯和中国。

And so if you look at the kind of like per country adoption data, like, there's some blips where like this country uses Claude a little bit more, this country uses Gemini a little bit more, but the two huge outliers are Russia and China.

这两个都是巨大的市场。

And those are like big, big markets.

是的。

Yes.

所以我认为值得关注那里正在发生的事情。

And so I think it's worth watching what's going on there.

不过这很有趣，因为俄罗斯和中国都是例外，原因在于模型使用方面的限制以及可能的文化偏好。

It's interesting though because both Russia and China, they're outliers because of restrictions around how models can be used and and maybe cultural preferences.

还有其他一些国家有地域性的趋势吗？还是说这是一种全球性的AI行为？

Are there any other countries that have geo specific trends or is this a sort of global AI behavior side?

是的。

Yeah.

我认为在模型开发方面，尤其是能够部署专有AI产品的专有模型开发上，

I would say in terms of like model development, proprietary model development that allows you to deploy proprietary AI products.

大部分研究都来自美国和中国，可能还有少量来自俄罗斯。

Most of that research is coming out of The US and China, maybe a little bit out of Russia.

我认为我们正在看到其他地区出现一些本地化的生态系统。

I think we are seeing a few kind of native ecosystems in other places.

韩国就有几个自己的产品，比如Naver和Kakao，它们已经构建了不错的LLM界面。

Would, Korea has a couple of their own products like Naver and Kakao that have built out nice kind of LLM interfaces.

印度可能是另一个我密切关注的国家，因为那里人口众多，足以支撑专注于印度市场的大型独立公司。

India is probably the other one that I watch really closely just because there are so many people that you can have standalone big companies focused on India.

印度另一个有趣的地方是，它有这么多不同的语言，范围非常广，无论是大语言模型产品还是语音产品，都未必能很好地支持。

The other interesting thing about India is there's so many different languages like such a range that both LLM products and even voice products don't necessarily support very well.

如果你的主要语言是这些语言之一，试图使用像ChatGPT这样的产品，体验会更差。

Like it's a worse experience if you're a primary user of one of those languages and you're trying to use something like a ChatGPT.

到目前为止，我们还没有看到这方面有太大的差异，但我不会惊讶于未来看到更多来自美国的创业者瞄准印度市场做AI产品。

So, so far we haven't seen a huge amount of variance there yet, but I would not be surprised maybe to see more founders even from The US like targeting the Indian market for AI.

另外我想提一下，我们首次做了一张热力图，显示哪些国家在人均基础上采用AI最多、最少。

And then the other thing I wanted to mention we did for the first time also kind of like a heat map essentially of which countries are adopting AI the most and the least on per capita basis.

我们考察了十大主流大语言模型产品在网页和移动端的表现，来看看情况如何。

So we looked across like the 10 biggest LLM products to see on web and mobile to see what this might look like.

新加坡位居第一。

So Singapore is number one.

太疯狂了。

Crazy.

是的。

Yes.

然后是香港，接着是阿联酋，再是韩国。

Then Hong Kong, then The UAE, then South Korea.

美国排在第20位。

The US is down at number 20.

所以不算特别低，但也算不上非常高。

So not super low, not incredibly high.

俄罗斯和中国在榜单上非常靠后，低于第50名。

Russia and China are like very far down the list, sub 50.

我认为这些数据中蕴含着许多有趣的故事。

And there's a lot of interesting stories I think that live in that data.

是的。

Yeah.

第一个是，如果你看看前五名，比如新加坡、韩国、香港，它们的劳动力人口结构非常偏向科技、白领、高技能。

The first one is if you think about those top five like Singapore, South Korea, Hong Kong, it's a very like the demographics of the workforce are very like tech first, white collar, high skill.

而美国有大量的工作岗位，AI至今尚未真正触及，比如零售、交通运输和其他一些领域。

And The US has a giant chunk of jobs where AI hasn't really touched them yet, like retail and transportation and some of these other things.

我认为，围绕人工智能的文化规范差异惊人地大。

I think also the cultural norms around AI are shockingly diverse.

如果你在美国，你可能已经内化了这种持续的焦虑和对人工智能的质疑。

If you're in The US, you have probably internalized this ongoing angst and questioning around

是的，我正想问你这个。

Yeah, was gonna ask you about this.

它

Did it

按工作类型。

by job.

百分之百。

100%.

或者，人工智能对艺术家来说很糟糕，或者所有这些其他因素，都会让人选择使用或不使用人工智能。

Or, know, AI is terrible for artists or all of these other things that make people pick up or not pick up AI.

去年，全球媒体公司爱德曼实际上进行了一项大规模调查，结果显示美国对人工智能的信任度相对较低。

There was actually a big survey last year from Edelman, the global media company, and The US had a fairly low rate of trust in AI.

当时只有32%，而榜单上大多数其他国家都达到了五六十甚至百分之七十。

It was like 32%, and most of these other countries that are high on the list are like fifty, sixty, 70%.

所以我认为，尽管我们是最大产品的发源地，但这也在一定程度上拖慢了美国的步伐。

So that I think has also held The US back despite the fact that we are where the biggest products come from.

我们的人均使用率低于许多其他市场，这些市场人口可能更少，但更早地接受了AI。

Our per capita usage is lower than a lot of these other markets that have maybe smaller populations but have embraced it more.

我觉得这完全正确。

I think that's exactly right.

我读到过，在中国，人们对AI的正面看法达到了80%。

You know, was reading that in China the sort of favorability views on AI are 80%.

是的。

Yeah.

有80%的人持正面看法。

80% hold a favorable view.

我知道阿联酋和新加坡，它们的文化似乎天生就对科技持乐观态度。

And I know UAE and Singapore, I think they've sort of culturally wired to be tech optimistic.

是的。

Yes.

这是一个优势，

Which is an advantage,

是的，没错， definitely。

Yes, yes, definitely.

看看一些小国家的 per capita 采用率，挺有意思的。

It's interesting to see some of these smaller countries like the per capita adoption rate.

在美国，大约有三分之一的人是像 ChatGPT 这类产品的月度活跃用户。

In The US it's around probably a third of people are monthly active users of something like a ChatGPT.

甚至一些欧洲国家或东欧国家，虽然基数较小，但采用率在四五十到百分之六十左右，它们比我们这里更快地接受了它。

Even some of the European countries or Eastern Europe, it's like fifty, forty five, 60% on smaller bases, but they've kind of embraced it more quickly than we have here.

是的，真的很有趣。

Yeah, really interesting.

你知道，我正在关注并感兴趣的一件事是，当你从 AI 的谱系来看，从最功能性的、几乎像谷歌搜索替代品这样的应用开始。

You know, one one thing that I'm sort of watching and I'm interested in is as you look at the spectrum of AI from the most functional, almost like a Google search replacement Mhmm.

从最功能性的，几乎像谷歌搜索替代品，到最文化性、创造性、个性化的领域，我们将会看到各国之间的差异越来越大，因为显然印度制作的电影与中国或美国的电影截然不同。

To the most cultural, creative, personal, we should see more divergence country by country because obviously the culture the movies they make in India couldn't be more different than the movies they make in China or The US.

是的。

Yeah.

那么，为什么他们使用创意工具的方式不会不同呢？

So why wouldn't their use of creative tools be different?

对。

Yeah.

事实上，这正是我们在这份报告中开始关注地理细分的原因之一：在生成式AI的前两年半到三年里，绝大多数消费者可能只使用一种产品，但现在情况已经广泛多样化了，我认为我们会看到更多针对特定市场的工具。

And this is honestly part of the reason why we started looking at the geographic segmentation in this report is because for the first two and a half, three years of generative AI, the vast majority of consumers were maybe interacting with one product and now it's broadening quite a bit and I think that we will see more of these market specific tools.

如果这些工具能够抓住足够大的市场份额，比如一些俄罗斯公司或中国公司，只要当地市场足够大，它们就有可能进入全球榜单。

And if they are, if they capture enough of that market, like some of these Russian companies or Chinese companies, they can actually surface up to the global list if it's kind of the market is big enough.

谈谈创意工具的演变，以及你认为这在多大程度上反映了文化？

Talk a bit

谈谈创意工具的演变，以及你认为这在多大程度上反映了文化？

about the evolution of creative tools and how much you think that that is that reflection of culture?

这是否在推动文化？

Is that driving culture?

我们什么时候会跨越那个门槛？

When do we cross that threshold?

创意工具的趋势一直非常有趣。

The creative tools trend has been fascinating.

我的意思是，第一个大型生成式AI产品实际上是Midjourney，它在ChatGPT之前就推出了。

I mean, obviously the first big generative AI product was actually Midjourney, which came out before Chat ChatGPT.

确实如此。

True.

没错。

That's right.

在我们列表的前几期中，创意工具占据了主导地位。

And in our first few editions of the list, was very much dominated by creative tools.

我以前说过，但创意工具受益于早期模型的幻觉特性，因为它们生成的东西更具惊喜性、美感或原创性。

I've said this before, but the creative tools benefit from kind of hallucination of the early models because they produce things that are more kind of surprising or beautiful or original.

所以有一段时间，这些确实是消费级AI中唯一真正起作用的东西。

And so for a while that those are the only things working in consumer AI really.

但现在情况发生了很大变化。

Now it shifted a lot.

创意工具仍然是列表中的重要部分，但作为独立大型企业的创意工具类型已经改变了。

Creative tools are still a huge chunk of the list, but like the type of creative tool that is a standalone big business has changed.

我认为最大的变化是我们看到的独立图像生成器变少了。

I would say the biggest change is we're seeing fewer standalone image generators.

如果你只是制作一些基础的普通图像，比如表情包、基本的营销图片或信息图，现在ChatGPT和Gemini的核心模型已经相当擅长这些了。

A lot of this activity, if you're making like a basic commodity image, you know, a meme or a basic marketing image or an infographic, like the core models in ChatGPT and Gemini are quite good at those things now.

因此，那些仍然出现在列表中的产品，比如Ideogram或Midjourney，要么具有非常鲜明的美学风格，要么拥有更复杂的 workflows，这些是像ChatGPT这样的平台无法提供的。

So the products that are still surfacing on the list, an ideogram or a mid journey are either very aesthetically opinionated or they have very more sophisticated workflows that you can't get on something like a a ChatGPT.

与此形成对比的是，我认为音乐、语音、视频这些领域，大型模型公司可能投入得相对较少。

Contrasting that I would say like music, voice, video, all seem to be things that the model, the biggest model companies have maybe invested less in.

因此，我们看到了像Suno这样的音乐公司和Eleven Labs这样的语音公司彻底崛起，进入榜单前20甚至前15，并长期保持这一位置。

And so we've seen players like Suno in music and eleven Labs in voice kind of completely break out and rise to top 20, top 15 on the list and then like hold their spot there over time.

而且，由于社区和大量企业客户等因素，形成了累积性的锁定效应。

And then there's like a compounding lock in from like the community and, you know, the big base of enterprise customers and all of that.

视频是我问题最多的领域。

Video is where I have the most questions.

OpenAI 一直在通过 Sora 投资视频，谷歌也在做 Vio，但中国模型之所以这么强，是因为它们可以利用任何数据进行训练。

OpenAI has been investing in it with Sora and of course Google with Vio, but the Chinese models are so good because they can train on any data.

所以，Cdance 2 可能是最好的例子，它在某些方面远远超越了美国公司迄今为止所能达到的水平。

So, Cdance two is probably the best example of this where it's just kind of in some ways head and shoulders above what The US companies have thus far been able to do.

所以我认为我们会看到。

So I think we'll see.

我认为这实际上有利于像 Korea 这样的平台，因为你可以在一个地方使用所有模型；我妹妹 Justine 写过一篇关于这个的文章，但根据视频的发展趋势，不太可能有一个模型能一统天下。

I think this actually benefits platforms like a Korea where you can use all the models in one place because my sister Justine wrote an article about this, but the way video is shaping out, there's it's unlikely to be like one model to rule them all.

因此，你基本上需要能够在这几个模型之间切换。

And so you kind of need to be able to switch between them.

这似乎也适用于大多数模型领域，你

That seems true of most of the model spaces, you

知道吗？

know?

聊天模型、创意模型，甚至代码模型都有各自的专长领域。

Chat models, creative models, even code models have their areas of specialization.

人们经常谈论Opus的易用性。

You know, people talk about kind of ergonomics of Opus Yes.

以及编码器的准确性。

Versus the accuracy of codecs.

是的。

Yes.

这其实是一种权衡，你知道的，你得根据具体问题选择合适的工具。

And that's just that's a trade off, you know, and you have to choose what tool you wanna use for which problem.

对。

Yeah.

当然。

Absolutely.

SORA 真的让我很感兴趣，因为它不仅代表了模型的重大进步，也是一次极具雄心的社会性实验。

SORA is really interesting to me because it represented both a sort of a big step forward in the model, but also a really ambitious experiment around social.

在 SORA 的早期阶段，有一些数据，比如创建内容的人的比例，比我们之前见过的高出整整十倍。

And there was data in the early days of SORA, like the percentage of people that were creating which was dramatically 10x higher than we'd seen before.

你如何评估 SORA 在社会性尝试与模型本身之间的平衡？你认为未来会走向何方？

You know what's your kind of assessment of the SORA social effort versus the model effort and where do you see that going?

SORA 非常迷人，我认为这是一次非常有趣的早期实验，它让我们所有人对创意工具，以及更重要的是，对人工智能时代消费者社交形态有了很多新的认识。

SORA is so fascinating and I think was a very interesting early experiment that I think taught us all a lot about kind of both creative tools but also maybe more importantly what consumer social in the AI era might look like.

从数据上看，他们的发布规模非常庞大。

So by the numbers they had a massive launch.

他们在美区 App Store 连续二十天位居榜首，这非常难做到。

They were number one on the app store, The US app store for twenty consecutive days, which is very hard to do.

要想登上 App Store 榜首，如今你可能每天需要获得大约十五万次下载。

Means you're probably getting, be number one on the app store you probably have to get these days a 150,000 daily downloads.

这是一个极高的下载量。

It's like a high download volume.

而且他们达到一百万用户的速度比ChatGPT还快。

And they actually hit a million users faster than ChatGPT itself.

所以简直是大规模发布。

So like huge launch.

实际上，我认为很多人低估了它至今仍保持的显著使用量。

And actually I think what a lot of people underestimate is it still is very significant usage.

每个传感器塔有300万DAO，这已经很不错了。

So 3,000,000 DAOs per sensor tower, which is not bad at all.

关于Sora，下降的是新用户下载量。

What has dropped off about Sora is the new downloads.

所以他们可能在十一月时月活跃量峰值达到六百万。

So there may be, they peaked to like 6,000,000 a month in November.

现在看起来大约是一百五十万。

It's looking like a million and a half now.

我认为Sora真正成功的地方在于它是一个非常出色的视频模型，而且他们创新性地引入了‘客串’这一概念，即真实人物可以授权自己的形象给Sora，从而让本人和其他人能够生成包含他们的视频。

I think that what has really worked about Sora is that it's a very good video model and they kind of innovated and introduced this concept of cameos, which is where a real person can grant their likeness to Sora so that they and others can generate videos of them.

所以早期很多人都在制作朋友的梗视频。

So like a lot of people in the early days were doing like meme videos of their friends.

比如杰克·保罗走红了，因为他是第一个大力拥抱Sora的知名名人。

Like Jake Paul went viral because he was the first big celebrity to lean into Sora.

你看到的都是疯狂的杰克·保罗视频，当然，杰克·保罗。

You were seeing like insane Jake Paul videos Of course Jake Paul.

是的。

Yeah.

我的意思是，他做得不错。

I mean, good for him.

对。

Yes.

对。

Yes.

我认为Sora做得不够好的地方在于，由于内容可以导出，人们会把它发到TikTok、Instagram短视频和YouTube上，而那里竞争的是最顶尖的人工创作内容。

I think what worked less about Sora is that because the content was exportable, people would take it to TikTok, they would take it to Instagram reels, they would take it to YouTube and there it competed against the best human made content.

对。

Right.

因此，整体的动态信息流体验更好了，因为你看到的是两者中最优秀的部分，而不仅仅是Sora最优秀的内容。

And so the overall feed experience was just better because you were seeing the best of both, not just like the best of Sora.

对。

Right.

我认为，到目前为止，还没有任何社交产品能够完全依靠AI内容取得成功。

I don't think we've seen a social product yet succeed that's like entirely AI content.

在某些方面，情感上的投入感更低了。

The emotional stakes are just feel lower in some ways.

对。

Right.

因此，我预计我们会看到更多类似的情况：Source作为创意工具仍然具有非常明显且巨大的使用量和收入，但作为社交应用则不然。

And so I would imagine we'll see more examples like these where Source still has clearly very, very significant usage and revenue as a creative tool, but not so much as a social app.

对。

Right.

我不知道会不会出现一个完全由AI驱动的社交网络，但我觉得我们目前还没看到它是什么样子。

And I don't know if there'll be, there probably will be an a massive AI native social network, but we haven't seen what it looks like just yet, I would say.

这会很有趣。

It'll be interesting.

我们经常讨论这个问题，但每个社交产品都有一套地位竞争机制。

Know, we discuss this frequently, but every social product has a status game.

是的。

Yes.

比如在Instagram上，可能是最火的才算厉害，在X上，可能是最有意思的才算厉害。

You know, and on Insta, it's maybe be the hottest, and on X, it's be the most interesting.

而在Sora上，新兴的地位竞争似乎是看谁最搞笑。

And it felt like the emerging status game on Sora was be the funniest.

是的。

Yes.

我认为这正是内容难以跨平台传播的原因之一。

I think this is one of the reasons why it's hard for the content to cross over.

是的。

Yeah.

因为这是两种不同的方式来评判什么是有意思和出色的。

Because it's just two different ways of judging what is interesting and great.

我同意。

I agree.

如果我要想象他们可能会在哪个领域找到更多细分市场，他们现在已经与像Right这样的大型媒体公司达成了许多协议。

What they might do, if I had to imagine where they might find more of a niche, they have now inked a bunch of deals with big media companies like Right.

所以，如果Sora是唯一能制作受版权保护的、关于受欢迎角色和娱乐人物的粉丝视频的地方，那就会非常有趣。

And so if Sora is the only place where you can make like licensed like fan videos of like beloved kind of characters and entertainment figures, then like that's very interesting.

完全正确。

Totally.

但我觉得我们还处于这一趋势发展的早期阶段。

But we're early I think in how that rolls out.

这还太早了。

It's so early.

我知道。

I know.

我们一直这么说。

We keep saying it.

好的。

Okay.

我们换一个话题吧。

Let's switch gears.

如果我们不谈谈智能代理、OpenClaw、Manus、Genspark、Mold Book，这场对话就不完整。

We can't have this conversation without talking about agents, Open Claw, Manus, Genspark, Mold Book.

你能给我们概述一下过去六十天代理领域发生了什么吗？这份报告告诉我们什么？

Give us an overview of what has happened in the last sixty days in the world of agents and what does the report tell us?

我认为这正是为什么我说，过去六个月，甚至最近两个月的这份报告，都是我们见过的最有趣的。

I think this is mostly why I say the last, you know, six months, but actually even two months of this report have been like the most interesting that I think we've seen.

所以Opaquat实际上如你所见，并没有出现在我们的排名中，因为它在二月突然爆火了。

So Opaquat actually, as you'll see is not on our rankings because it blew up in February.

我们的数据截止到一月，但我们确实收集了二月的数据，如果它符合资格，会在我们的网页榜单上位列第30名，这是一个相当惊人的首秀。

Our data ends in January, but we did pull the data for February and if it had been eligible, it would have been number 30 on our web list, which is a pretty big debut.

我认为OpenClaw真正有趣的地方在于，它在技术社区中的使用量一直在持续加速。

I think the really interesting thing about OpenClaw is the usage has just continued to accelerate in the technical community.

所以现在它已经是GitHub史上获得星标数最多的项目了。

So now it's, I think number one GitHub stars of all time.

它超过了React，也超过了Linux。

It passed React, it passed Linux.

真的超过Linux了？

Really, really Past Linux?

是的。

Yes.

天哪。

Holy cow.

非常了不起。

Very impressive.

是的

Yeah.

但从整体新用户来看，增长已经趋于平稳。

But in terms of overall new users, it's kind of plateaued.

所以我们查看了访问入门页面或注册页面的次数，自二月初以来，这一数据周与周之间基本持平，我认为这表明，如果你是技术人员，它确实是一款了不起的产品。

So we looked at kind of visits to the get started or sign up page, and that is kind of flat week over week since early February, which I think indicates that like it is an amazing product if you're technical.

但它尚未完全突破技术圈，进入非技术人员群体，而后者显然规模更大。

It has not yet fully escaped containment to non technical people, which of course is like a bigger population.

它们被OpenAI收购了。

They were acquired by OpenAI.

所以如果我要猜测，或者我期望OpenAI做的，是把OpenClaw产品化，打造成一款面向主流消费者的产品。

So if I had to guess, or what I'd love to see OpenAI do is build like productize OpenClaw into something that is usable for a mainstream consumer.

我认为我们也已经看到，OpenClaw架构背后的理念激励了众多其他创始人。

And I think we've also just seen the ideas behind the OpenClaw architecture inspire so many other founders.

我们每天收到多少份商业计划书，其中创始人都说：‘我想做这个领域的OpenClaw’，或者‘OpenClaw让我意识到这是可能的’。

How many pitches do we take a day where the founder is like, want to be Open Claw for this or Open Claw made me realize this was possible.

所以我认为OpenClaw本身将继续取得成功，成为一个庞大的产品。

And so I think we're going to see Open Claw itself will continue to succeed and be a massive product.

我猜测我们会看到更多针对不同使用场景的、垂直化的OpenClaw版本。

And I'm guessing we'll see more kind of like verticalized focused versions of OpenCLaw for different use cases.

是的。

Yeah.

这非常有趣，因为OpenClaw运作得如此出色，原因之一是它能够跨所有模型和所有方向运行。

It's so interesting because it feels like one of the things that makes OpenCLaw work so well is it can operate across all models in all directions.

是的。

Yes.

我有点怀疑，如果OpenClaw只提供单一模型，是否会削弱其价值，从而与各大实验室形成对立。

And I sort of wonder if it dilutes the value of OpenClaw to have it be sole model provided and therefore it's sort of counter positioned against labs.

完全正确。

Totally.

是的。

Yeah.