人工智能初创企业何在？——与里克·海茨曼对话

AI初创企业都在哪里？它们真的会出现吗，还是ChatGPT会吞噬一切？稍后我们将与FirstMark Capital的里克·海茨曼讨论这个问题。Capital One的技术团队不仅在多模态AI上高谈阔论，他们已经实际部署了一个。

Where are the AI startups? Are they actually coming, or will ChatGPT gobble it all? We'll talk about it with Rick Heitzman of FirstMark Capital right after this. Capital One's tech team isn't just talking about multiegetic AI. They already deployed one.

它被称为聊天礼宾服务，正在简化购车流程。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，还能协助预约试驾、获得贷款预批以及评估旧车置换价值。先进、直观且已投入使用——这就是他们的技术实力。

It's called chat concierge, and it's simplifying car shopping. Using self reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love. It helps schedule a test drive, get preapproved for financing, and estimate trade in value. Advanced, intuitive, and deployed. That's how they stack.

这就是Capital One的技术。欢迎收听《大科技播客》，这是一档冷静而细致探讨科技界及更广阔领域的节目。我们一直在思考：那些独立的AI初创公司都去哪儿了？当然，我们知道ChatGPT、Claude这些大型聊天机器人，但为什么没有涌现出一批基于生成式AI的独立初创企业呢？今天我们有幸请到最合适的嘉宾——里克·海茨曼。

That's technology at Capital One. Welcome to Big Technology Podcast, a show for cool headed and nuanced conversation of the tech world and beyond. Well, something we've been wondering on the show is where are all the individual AI startups? We know, of course, about ChatGPT and Claude and the big chatbots, but why hasn't there been a wave of individual startups building on top of generative AI that has emerged alongside this wave? And we have the perfect person to speak with us about this today because Rick Heitzman is here.

他是First Mart Capital的创始管理合伙人，今天亲临演播室与我们深入探讨这个话题。里克，欢迎来到节目。

He is the managing partner, and founder of First Mart Capital, and he is here with us in studio today to talk all about it. Rick, welcome to the show.

谢谢，非常感谢邀请。作为长期听众却是首次做客，总是令人兴奋的。

Thank you. Thank you for having me. A longtime listener, first time guest, so it's always exciting.

很高兴你能来。我总喜欢在我们要上CNBC前碰到你，通常我们一个刚结束另一个就接上。今天终于有时间好好聊聊了。

It's great to have you here. I love running into you before we're about to go on CNBC. Usually, one of us is right before, right after. So today we actually have some time to speak with each other

我算是你的开场嘉宾了。

I'm one on your opening act.

或者反过来。那么让我们回到最初提出的核心问题。如果你相信生成式AI是一项变革性技术，或者至少有能力在科技界掀起波澜——我认为这基本上是业内的共识——那么所有的AI初创公司都去哪儿了？当然，有一些针对特定领域的解决方案，比如Harvey，对律师非常有用。但如果你将生成式AI内置的功能释放给所有初创开发者，而不只是ChatGPT

Or the other way around. So let's go to the big question, right, which we started with here. If you believe that generative AI is a transformative technology or at least has the ability to make some waves in the tech world, which I think is basically consensus in this world, where are all the AI startups? Of course, have some point solutions like Harvey, which is really good for lawyers. But if you took the functionality that's baked within generative AI and you sort of unleashed it to all these startup developers without ChatGPT

嗯。

Mhmm.

我猜我们会看到一大批AI初创公司涌现。有的可能用于健身，有的能帮你找到最佳冲浪点，

My guess is we would see a swarm of AI startups. Something that you could use for fitness, something that you could use to find the best surf break,

这个我知道是

which I know is

你曾经试用过的应用，但我们还没看到那波浪潮。所以到底发生了什么？

a application that you've played with, but we haven't seen that wave. So what's happening?

其实我们已经开始看到一些迹象。这通常与数据的专业性和规模有关。我认为有几个因素造就了当前市场的这种态势。首先，OpenAI和ChatGPT做出了卓越的产品，兼具广度与深度。作为风投，我们总希望行业龙头会变得懒惰、自满、迟缓，从而容易被颠覆。但OpenAI恰恰避免了这些陷阱，他们招募优秀人才，持续快速迭代产品。

So we're starting to see some things. It generally has to do with how specific and how big your data is, And I think there's a couple things which create this dynamic that we're seeing in the market. First of all, I think OpenAI and ChatGPT have done a great job of making a very good product that has both breadth and depth. So, you know, the leader not being complacent is something you know, we hope as venture capitalists that there's leaders, and then they get lazy, they get complacent, they get slow, and they get easy to disrupt. I think in this case, OpenAI has done an excellent job of not being any of those things, hiring great people, continue to develop product very quickly.

另一个关键因素是数据。AI的优劣取决于底层数据和训练数据。目前通用消费级AI的训练数据主要来自全网抓取。现在围绕数据训练的诉讼层出不穷：到底用了哪些数据？是世界上的所有书籍？谷歌爬虫抓取的内容？还是整个开放网络的抓取数据？

The other thing is a lot of the the data. So your AI is only good as your underlying data and your training data. So a lot of the training data in general consumer is general broad based web. And, you know, obviously, you're seeing litigation around who's training what on what data, and is it all the books in the world? Is it all the crawlers on Google or all the crawlers on the open web?

目前还没有基于数据实现差异化区分的情况，这与你们提到的Harvey和一些企业AI公司略有不同。我们在保险领域有家公司EvolutionIQ，法律领域有Harvey，商业地产领域有Henry，它们都拥有非常独特且有时是私有的数据集，这使得它们能构建更好的模型，提供更优质的终端用户应用。但对于像你我这样想找冲浪点、度假去处或纽约最好吃的法式蘸酱三明治（答案是4 Charles）的人来说...

And there's there's not been a differentiating differentiation based on data, which is slightly different than you know, you alluded to Harvey and some of the enterprise AI companies. We had a company EvolutionIQ and insurance. There's Harvey in legal. There's Henry in commercial real estate, and they all have a very discreet and sometimes private dataset that enables them to build a better model, enables them to deliver a better end end user application. But for, you know, you and I who are trying to find surf breaks or where to go on vacation or the best place to have a French dip in New York, answer is 4 Charles.

但如果你想找这些东西而我刚好没空做播客，通常你会用通用型的ChatGPT或Perplexity。说实话，我们对目前缺乏能让我们沿此方向投资的初创公司感到有些沮丧。

But they say if you're trying to find those things and I'm not available on a podcast, you know, those things have generally been broad based chat GPT or perplexity. And we frankly have been a bit frustrated by the lack of startups we've seen in their ability to invest along those lines.

没错。顺便说一句，Harvey、Henry...我嗅到了命名趋势。

Right. By the way, Harvey Henry, I'm sensing a trend here.

确实有这个趋势。

There's a trend there.

是啊。这会不会像当年的.ly域名热潮？随便抓个男性名字就行？

Yeah. Is it the new dot l y? You just take a random guy's first name?

我...我觉得是。好记又顺口。不过想想90年代有Blue Nile、Amazon之类的命名风潮，可能这次...

I I think so. It's easy to say, easy to pronounce. We'll see. I mean, there was Blue Nile, and there was Amazon, and there was a bunch of things in the 90s. Maybe this is You can the

命名潮流总是轮回的。言归正传，这点很重要——十年前我在BuzzFeed报道消费科技时，用过一款营养应用。你可以上传饮食记录和想法，真正的营养师会评估并给出建议，帮你追踪目标进度。

go in waves there. All right. But let's drill down on this because I think this is a really important point. To give an example, there was about ten years ago when I was at BuzzFeed writing about consumer tech, there was this nutrition app that I would use. And you would upload your meals and your thoughts and stuff like that and a real nutritionist would take a look give you a rating and give you advice about how were tracking on your goals.

当时人们嘲笑我。我说，你可以直接用自然语言编程那个应用，但你真的可以。而且我知道不止我一个人这样。很多人把ChatGPT当作饮食教练，你给它设定一个目标。顺便说一句，不只是说‘当我的饮食教练’，而是实际给它一个目标。

Now people laughed at me. I was like, you can just program that app with natural language, but you really can. And it's something I know it's not just me. Many people have been using Chan Chi PT as a diet coach where you give it a goal. By the way, you don't just say be my diet coach, you actually give it a goal.

你可以说‘我想每天控制在2000卡路里以内’或‘我想吃全食物’。然后你可以上传照片，它能查看照片，结合文字上传，讨论早晨称重数据，给它这些信息，它就能很好地跟踪记录。

You say I wanna keep under 2,000 calories a day or I wanna eat whole foods. And then you can upload photos, can see the photos, upload with text, talk about morning weigh ins, give it the data, and it does a great job of keeping track

是啊。

Yeah.

这些事情。现在，再次强调，没有ChatGPT的话

Of this stuff. Now, again, without ChatGPT

ChatGPT是一个——我们回到正题——它确实是个很棒的产品。

ChatGPT is a we're going back to it's a really good product.

绝对是的。

Absolutely.

它是一款功能广泛的好产品，而且，你知道，它正在解决你的问题。

It's a really good product that has breath, And, you know, so it's it's solving your problem.

但这是我的问题，从你的角度来看。我们是否即将见证一波从未出现过的消费者初创企业浪潮？因为那确实是一家获得数百万美元融资、并成功退出的真实初创公司，我记得是被一家健康保险公司收购了。当然。而今天，我甚至很难想象这样的项目能获得融资，因为风投可能会说，我为什么不在聊天机器人里直接实现这个功能呢？

But this is my question then from your perspective. Are we about to see a wave of consumer startups that never happened? Because that was a real startup that got millions of dollars of funding, got a nice exit, I think, to a health insurance company. Sure. And today, you can't it's hard for me to even conceptualize that that would get funding because a VC might just say, why wouldn't I just do this in chat?

我们已经见证了很多案例。我们看到营养师应用，看到各种不同类型的应用涌现。明白吗？所以我会说问题可以归类为：你是否需要一个独立应用？

We've we've seen a lot. We've seen nutritionists. We've seen a bunch of different things that have come out. You know? So I would say there's buckets of, do you need a discrete application or you don't need a discrete application?

出于各种原因，包括健康科技领域的监管合规要求，某些场景确实需要独立应用。但像'我正在吃这块三文鱼，它有多少卡路里？你能帮我计算热量吗？'这类通用需求，ChatGPT就非常擅长处理。

Certain things for a bunch of different reasons, including regulatory and compliance in in areas like health tech, you need a discrete application. But certain things, including general things like, you know, I'm eating this piece of salmon. How many calories does it have? Could you count it in your calories? It's something ChatGPT is great for.

因此我们遗憾地发现，预期中可持续的初创企业浪潮并未出现。实际上只有少数基于ChatGPT的初创项目，可能在旅行规划或数学辅导方面略胜一筹，但它们的差异化程度并不显著。回顾搜索领域的发展历程，人们曾预言垂直搜索将大有可为——Indeed成为巨头企业就是明证。

So we've found, sadly, that, you know, we haven't seen this wave of startups that we believe are sustainable. So there's actually been a handful of startups that are rappers on ChatGPT that are maybe a little bit better at at travel. They might be a little bit better at being your math tutor, but they're not that step function different. And even if you go back to the other areas of search, if you remember, there was search, and then people said, oh, there could be vertical search where we get really good at something. So, obviously, Indeed is a very large company.

Indeed是专注职位的垂直搜索，Kayak则是价值数十亿美元的旅行垂直搜索平台。通过细分领域，本应能做得更专业。但ChatGPT的泛在性反而使这种垂直深耕模式变得比以往任何时候都更具挑战性。

That was vertical search for jobs. Kayak was a very big multibillion dollar outcome. That was vertical search for travel. And, you know, you're able to break down that landscape and then think about where where that goes because with a more narrow focus, you should be better at it. And I just think that the broad landscape of of ChatGPT has made that more difficult than ever.

确实。这很有趣，因为OpenAI最近发布了用户行为数据——当然数据来源是OpenAI自身。我们在节目中也讨论过，嗯。

Yeah. And it's very interesting because OpenAI recently released data, and of course, it's coming from OpenAI. Yes. But data about how people use ChatGPT, we've talked about it here on the show. Mhmm.

数据显示用户最主要的使用场景是获取实操指导。让我们做个思想实验：假设没有ChatGPT，也没有广泛可用的生成式AI技术——也就是说你无法授权使用大语言模型。

And the number one use that people go to it is for practical guidance. Yes. And let's just do a thought exercise. If there was no ChatGPT and no broadly available generative AI technology, so think about it, you can't license an LLM. Yeah.

但假设五年前有家公司找到你，他们说，我们有一款应用能通过自然语言为你的恋爱关系提供建议，告诉你是否该和男友分手，

But a company came to you, let's say five years ago, and they said, we have an app that with natural language will advise you on your relationship and tell you whether or not to, you know, break up with your boyfriend or

或是女友，比如

girlfriend, for

，或者如何改善关系。如果他们来找你说，我们有个自然语言健身教练。如果他们来说，你可以上传足球训练的照片或视频，我们会和你讨论站位与动作。这些点子在我看来个个都像是价值十亿的创意。对吧？

instance, or how to improve the relationship. If they came to you and said, have a natural language fitness coach. If they came to you and said, we have you upload photos or videos of soccer practice, and we'll talk to you about positioning and form. Each one of those ideas to me sounds like they would be like, billion dollar ideas. Right?

没错。非常容易融资。虽然未必是十亿美元级别的创意——这个有待观察——但绝对值得投资。想想人生教练、健身教练、运动教练，任何领域只要拥有海量知识，并能将其个性化输出——这其实和Harvey的本质并无不同，对吧？

Yeah. Very financeable. Very if if maybe not billion dollar ideas, we'll see where that goes, but very financeable. If you think about life coaches, fitness coaches, sports coaches, anything where you have a tremendous amount of knowledge and you could take that knowledge and make it very specific to somebody, which, you know, again, going back to Harvey's is not that different. Right?

法律就是个浩瀚的知识海洋，需要特定规则约束。过去人们只把这当作思维训练，如今我们称之为LLM（大语言模型）。它能更快速、廉价、高效地产出解决方案，提升生活或工作效率。

Law is a huge huge pool of knowledge that you put a certain rules around it. You know, historically, they just thought that was a thought exercise in rules. Today, we could call it LLM. And then that produces better, faster, cheaper results of how to how to be more efficient in your life or job.

就拿Harvey来说，我们讨论的这个法律AI，它通晓法律法规，拥有超大上下文窗口，既能提供法律咨询也能辅助律师工作。但在我看来Harvey的护城河并不牢固，因为随着上下文窗口不断扩大——Harvey的核心优势在于它能精准匹配适用法条并解答法律疑问——但我确信很快律师就能做到：下载整个州的法律压缩包，

I mean, even Harvey. We talk about Harvey, right? Which is, again, this is his legal AI that knows the laws, knows the rules, has these big context windows, so you can go to it for legal advice or a lawyer would use it to help. But even Harvey to me doesn't even seem that defensible because what we're starting to see is bigger and bigger context windows So from these what Harvey's great at is it's figured out a way to get the applicable law and then find a way to measure that against the questions you might have. As a lawyer, we are gonna get to the point, I think without a doubt, that a lawyer will be able to say, download the zip file of all the law in the state.

没错。将其导入上下文窗口，再上传案件细节，最终得到的效果可能接近Harvey的水平。

Yes. Upload it into the context window. Download the specifics of the case, upload it into the context window, and maybe get close to as good as Harvey is.

是的。我是说，在某个特定案件中你可能需要一位专业律师，这是非常具体的情况。但想想看，大约80%的法律工作——虽然不幸的是并非由律师完成——比如这位里克需要立遗嘱，或是Firstmark公司正在进行A轮融资。

Yeah. I mean, that's a very specific thing on a case where you might need a specific attorney. If you think about probably 80%, unfortunately, not a lawyer, but probably 80% of all legal work is, this is Rick. He needs a will. This is you know, here is a Firstmark company that's going through a series a financing.

能否根据这些创始人资料、争议焦点和条款清单直接生成文件？法律金字塔底层存在大量机械性工作，本应由更高效、更廉价的方式完成，而不是压榨那些过度劳累的初级律师。

Could you just reproduce documents given these are the founders, these are the issues, and here's the term sheet? So there's a lot of rote work that's done by the bottom of the legal pyramid, which should be done better, faster, cheaper than overworked, overtired associate.

没错。问题在于这些工作将在哪里完成？我想探讨的观点是——这一切最终是否都会在ChatGPT的界面里发生？最近有种争论认为所有AI应用都只是外壳，比如Perplexity不过是搜索功能的AI包装。那么投资逻辑何在？

Right. And so the question is, where does it get done? And the argument that I'm making or trying to tease out here is, does all this stuff end up just happening within the ChatGPT interface? I think it's kind of been this debate that's gone on where people say that any AI application is just a wrapper, like Perplexity is just an AI wrapper that you do search in. And so then how do you invest?

回到我们对话开始时谈到的各种独立应用场景——法律领域，还有更具普适性的教练指导、健身规划、信息检索。难道最终都会被整合进那些全能型通用机器人？想到这儿我不禁摊手追问：创业者、投资人、律师们将面临什么？

And so I'm trying to think through the beginning of our conversation here where we're talking about all these distinct and discrete different applications, legal, but even more applicable coaching, fitness, search. It's all gonna happen within these broad, multi general purpose bots. And so then I, like, throw my hands up and say, well, what's gonna happen? Like, what's gonna happen to start start up founders and investors? Lawyers?

未来会怎样？

What's gonna happen?

播客主们也是。

Podcasters.

不过说真的...从经济活动角度...我们可以先讨论就业问题，但...

But no. But but really in in terms of the the economic activity. Well, we can get to So so there's probably get to jobs, but

经济活动可能涉及两个方面。其中一部分人正进行轻度红队测试。这没问题。那么，ChatGPT能否在所有方面都比所有人强？大概不行。

the economic activity is Probably two pieces. And some one is slightly red teaming. It is alright. So can Chad GBT be better at everything than everybody? Probably not.

总会存在局限性。奥特曼会说：是的。然后存在一个渐近线，比如最新模型是否就是最佳模型？ChatGPT是否还能看到阶跃式的改进？传统观点认为很可能不会。

There's gonna be limitations. Altman, he'll say. Yes. And then there's there's an asymptote where, you know, are the latest models the best models, and are you still seeing even a step function improvement in ChatGPT? Conventional wisdom is probably not.

你会看到它已经能处理大部分事情，这很好。但对于整个生态系统来说，要攻克最后那10%意味着什么？是否需要专门针对旅行或法律领域的特定模型？第二点是：这些模型如何通过更好的数据来提升。

You're you're seeing, like, oh, it gets most of the things, and that's good. And what does that mean for the broad based ecosystem to get maybe that last 10%? Do you need a specific model to travel or to law? The second piece is, alright. Well, how these models will get better is through better data.

那么是否存在某些特定数据，人们可能不放心交给OpenAI或ChatGPT？我们投资了几家做数据安全的公司。你向哪些模型分享了什么数据？数据是否保留在你的环境中？我们如何确保数据不会泄露到模型或生态系统的其他部分？

And then is there specific data which people might not trust in OpenAI or ChatGPT? And, you know, we're investors in a couple of companies that does to do data security. What data are you sharing with what models? Are they staying inside your environment? Are we making sure that all our pieces of that data are not leaking out into a model or into another part of that ecosystem?

如果你拥有一个包含私有数据、模型和安全措施的封闭花园，这是否会更好？因为它在个人层面（如人际关系、度假计划、财务状况或遗嘱）或企业层面（如所有交易的法律文件）都更贴合你的需求。我可能不希望这些信息外泄，但需要设定某些参数；或者我的基金回报数据，必须确保其机密性。人们是否会像警惕其他大公司那样产生担忧？

So if you have a private walled garden of your data, your model, and your security, you know, will that be better because it's more specific to you even on personal basis if you're talking about your relationship or where you're going on vacation or your finances or your will or on an enterprise basis? So if, you know, here are all my legal documents on all my deals. I probably don't want that out in the world, but I wanna have some parameters around it or here are all my returns for my funds. I wanna make sure that that's confidential. So, you know, are people gonna get scared no differently than they they than they become suspicious of other large companies.

他们是否会对OpenAI、ChatGPT这类大型模型过度警惕？数据隐私是否会成为下一代公司发展的关键制约因素？

Are they gonna become overly suspicious of OpenAI, ChatGPT, the larger models? And is that data privacy gonna be a key limiter to how the next generation of companies evolve?

我认为安全领域在这里是个极具投资价值的切入点。

I mean, I would imagine security is like a highly investable place here.

我们在数据安全的各个层面、模型安全以及整个企业环境的所有环节上都投入了大量时间。我认为我们可能甚至还没进入第一局比赛。

We're spending a lot of time around that, on every level of the data security, model security, you know, every around the enterprise environment, all of those pieces. I think we're maybe not even in the first inning.

是的。我们刚和Wiz的联合创始人伊诺·安科斯蒂卡做了期播客。对，这家公司刚以320亿美元卖给了谷歌。

Yeah. We just did a podcast with Ino Ankostika, the cofounder of Wiz. Yes. And it was just sold to Google for 32,000,000,000.

有史以来最大的风投退出案例。确实是史上最大。

Biggest venture outcome ever. Ever. Yes.

目前是。目前是。我们收到评论说你们应该多聊聊这个话题。没错。当时我们只是概述了基本情况，但显然人们对此确实存在担忧。

For now. For now. And we we we had comments coming in, being like, you need to speak about this more often. Yes. And it was just like, here's a general lay of the land, but clearly there's real concern there.

那么，安全领域是一方面。其他领域还有些特定的企业用例。在消费者端，或者传统科技投资领域——我不知道该不该这么称呼——有没有哪个方向会让你考虑投资一家生成式AI初创公司？我是说，以生成式AI为核心的初创公司。

So, okay, security is one place. There's certain specific enterprise use cases elsewhere. Is there anywhere, like, on a consumer or I don't even know if should I call it traditional technology investment place where you you would see a generative AI startup, like a startup let me put it this way. A startup using generative AI at the heart of it. Yeah.

你会愿意投资的那种。

That you would invest in.

你指的是应用层吧？对。在应用层方面，我们确实...我们看好企业级市场。我们已经投资了几家做企业AI的公司。

On the application layer, I assume. Yes. Yeah. So on the application layer, we do I think we we like the enterprise space. We're investors in a couple things in the enterprise AI.

它们通常具备两个特征。首先会定义一组客户群体，从而拥有特定的数据集；其次制定关于共享数据与专有数据的规则，这些数据本身（而非模型）才是竞争优势所在——模型只是将正确结果输出到相应应用场景的载体。有时企业会在自己的封闭生态中使用这些数据。比如某公司可能要求：'我需要所有历史租赁数据，以便全面了解旗下所有星巴克特许经营店的租赁情况'。

They tend to have two things. They tend to have defined a set of customers, which have, therefore, a defined set of data. And they have some rules around what is shared data and which which own data, and that data is the competitive advantage, not necessarily the model, that outputs to the right application and the right answers. And sometimes they use it within their own walled garden. So if you were a company that says, want to have all my leases historically, and therefore I want to understand all my leases across all of my, you know, Starbucks franchises.

没错。获取具体租赁数据与从OpenAI模型获得'纽约市中心概况'这类通用答案截然不同。当企业拥有专属数据、制定内部规则、并在特定行业建立封闭生态时，模型就能针对该行业进行优化。甚至可以通过协作或共建数据库的方式，使这种模式在中长期更具可持续性。如果说数据是这些应用和模型的氧气，那么掌握数据所有权就至关重要。

Alright. Well, getting very specific lease data is going to be very much different than getting, you know, generic answers from, you know, what Downtown New York looks like in the OpenAI models. So though that having the specific data, having specific rules around your company, and having kind of a walled garden within a particular industry, that that model can be tuned to that particular industry, and then there's some benefits of maybe even collaboration or a co op database that makes that more sustainable in medium or long term. So if data is kind of the oxygen for a lot of these applications and models, having some kind of ownership on that.

我认为当人们讨论科技初创企业时，判断其优劣的标准——我相信你也有自己的见解。

So I think when people talk about tech startups, what makes a good tech startup, you I'm sure you have a philosophy. Yes.

我认为

I think

我常听到的一个核心观点是：它必须能解决问题。

one of the the consistent philosophies I've heard is that it solves a problem.

正是。

Yes.

这个理念很妙。就像健身或饮食领域的例子——企业会聚集健身专家或营养专家，共同研究问题本质并专注解决它。如今大语言模型的表现已接近（虽未完全达到）专家水平。

And I think that's kinda nice. Like, one of the nice parts, like, take the fitness example that I Yes. Or the diet example, is that you you get a company with that gets together with fitness experts or diet experts and says, let's try to see what the problem is and pay a lot of attention to it and then try to solve it for people. Yeah. And now you have large language models that are like doing just as good or or not not just as good, almost as good.

这样一来，那个类别就会变得不那么值得投资。如果人们不再有机会从专家那里获得建议，而是转向那些我们过去二十年里看到的应用程序——它们有时表现不错，有时则不尽如人意——我们是否会失去什么？但如果这些应用程序和科技公司不再涌现，所有这一切基本上都被交给只能做到75%效果的聊天机器人处理，虽然不需要初创企业和大量资金投入，但我们是否会因此失去某些东西？

And and so that would make that category less investable for Do we lose something if people instead of getting a chance to get this advice from the specialists are instead of going to these apps that we've seen for the better part of twenty years come up and serve use cases and sometimes do a good job and sometimes not. But do we lose something if instead of seeing these apps come up and these technology companies come up, all this basically gets handed over to chatbots that do, like, 75% as good of a job, but just don't take the start up and capital to to get there.

嗯，你希望其中能有一些创造性，对吧？如果你说它们能做到75%，我本来想猜80%，总之选个中间值。只有专家会站在顶端说：‘嘿，我是你的营养师。我会像你一样使用ChatGPT的后端，但我会给你更多建议，因为我知道你今晚要去牛排馆，而你在注意胆固醇水平。’但ChatGPT能做到吗？

Well, you hope that there is a bit of creative Right? So if you say they're doing 75, I was gonna guess 80, you know, you pick a number in between, and only the expert is gonna sit on top of it and say, hey. I'm gonna be your dietitian. I'm gonna use the back end of ChatGPT like you would, but I'm gonna give you some more advice because I know you're going to this steakhouse tonight, and you're trying to watch your cholesterol, whatever that may be. So does ChatGPT, though.

所以它可能已经帮你订好了位置。

So It probably made that reservation for you.

它可能订好了，还知道菜单是什么，了解你的目标以及如何实现。但也许上面会有一个界面，甚至可能是人工的。嗯。那么你如何知道并理解自己独特的附加值？作为人类，你的独特附加值不在于能谷歌餐厅菜单并挑出鱼类。

It probably made it, and it knows what the menu is and knows what your goals are and how to do it. But there might maybe there's an interface on top of it, which might even be human. Mhmm. So how do you know and understand your discrete value add? So your discrete value add as a human is not being able to Google the restaurant menu and pick out fish.

那真的很棒。比如，那是他

That's really good. Like, is that He

为此花了很多钱。

took a lot of money for that.

他们目前确实如此。但可能是，‘我更了解你。我知道三文鱼对你来说可能是正确答案，但你就是不喜欢三文鱼。或者你前两天刚吃过三文鱼，诸如此类。所以我会找到特别适合你、我知道你会喜欢的东西。’

They they they currently do. But it might be, I know you better. I know that salmon might be the right answer for you, but you just but you just don't like salmon. Or you, you know, you ate salmon the last two nights or whenever it is. So I I'm gonna find something specific to you that I know you'd like.

或者我今天和你聊过，你说，你知道吗？我今天不想吃鱼，或者我只是想...你知道吗？今晚我只想坚持原计划。我要走那条路。所以他们的能力，也许随着时间的推移会变得个性化，你的聊天机器人知道你很累，因为它接入了你的群体数据，或者它知道过去两晚你吃了SAM，因为它还追踪了你过去两周的饮食和餐厅预订。

Or I talked to you today and you said, you know what? I'm not in the mood for fish or I I just wanna you know what? I just wanna stick tonight. I'm gonna go down that path. So their ability, and maybe this becomes personalized over time, which your chatbot knows that you're tired because it's plugging in your group data, or it knows that you had SAM in the last two nights because it also tracked your food and your restaurant reservations over the last two weeks.

你知道，它可以获得更高层次的个性化服务，但就像历史上每一次技术进步一样，人类的工作就是保持领先于技术，并理解他们如何在技术之上创造独特且独立的价值。

You know, it could get an additional level of personalization, but, like, every time through history, the human's job is to stay just ahead of that technology and understand where they could create unique and discrete value on top of technology.

是啊。我觉得这可能会很难。也许我...

Yeah. I I think that's gonna be tough. Maybe I'm

我对人类有信心。

I have confidence in the humans.

好吧。我也是。是的。能进行这样的讨论很有趣，因为显然当今的生成式AI技术还存在很多漏洞。是的。

Okay. I I do too. Yes. And it's interesting to be even having this discussion because there's clearly so many holes Yes. In the generative AI technology today.

比如，在所有这些任务上，它目前还不如人类。是的。但它正在接近

Like, at all of these tasks, it's not as good as a human today. Yes. But it's getting close

足以让这个问题变得相关。更近了。如果你看看五年前它的水平和现在的进步，它确实越来越接近了。我是说，我们正在研究AI伴侣，无论是用于约会、老年人陪伴、儿童陪伴还是辅导。即使我们在一两年前看这个，也觉得它并不怎么样。

enough to make the question relevant. Closer. And if you look at where it was five years ago and the progress it's made, it's getting closer. I mean, we're looking at AI companionship, and whether that's dating or whether that's for elderly people or whether that's for kids or whether that's for tutoring. And as we looked at it even last year or two years ago, like, this this isn't very good.

比如，我不太确定，你知道，我年迈的祖母或我的孩子真的会和一个这样表现的聊天机器人互动。现在它确实很好。现在它

Like, I'm like, I'm not sure, you know, my elderly grandmother or my kid is really gonna engage with a chatbot that acts like this. Now it's really good. Now it's

确实很好。

It's really good.

非常明显。现在，你知道，人们正在参与其中，我相信你经常读到，越来越多有意义的关系，或者每个人都能分辨出什么是AI生成的广告、AI生成的视频，甚至是AI生成的女演员。现在有一位著名的基于AI的女演员，各大人才经纪公司竞相签约，你几乎分辨不出来，那个人几乎和真人一样好。我认为这种情况会持续发生，但对于那些能调整心态或思考创造价值以保持领先的人来说，这将是非常具有颠覆性的。

Pretty clear. And now, you know, people are engaging in I'm sure you read about it all the time, more and more meaningful relationships, or everyone could tell what was AI generated advertising or AI generated video or even AI generated actress. And there's this now famous AI based actress who in a bidding war to be represented by the major talent agencies that you can't tell, and that person is almost as good as a human. And I think this is gonna continue to happen, but it's gonna be very disruptive for people who can adjust their mindset or think about creating value to stay ahead of the curve.

是的，有一些令人着迷的应用。当然，这里也有一些担忧。人们过度依赖这些机器人，机器人谄媚，鼓励他们进行自我毁灭的行为。但另一方面，也有一些惊人的应用。我们在节目中提到过，在韩国，有一种内置了LLM的毛绒玩具。

Yeah, there are some fascinating applications. I mean, of course, there's concerns here as well. People becoming overly dependent on these bots, the bots being sycophantic, encouraging them to do self destructive behavior. But on the other side, there's some amazing applications. We've talked about on the show here, there's in Korea, there is a like a stuffed animal with an LLM baked in.

是的。

Yes.

它就像是陪伴孤独的老年人。是的，陪伴他们。然后当他们发现问题时，嗯，检查他们是否按时服药，那个和这个LLM毛绒玩具成为朋友的人说，我已经吃完药了。然后他们会发送一条消息给

That's like hanging out with elderly people who are lonely Yeah. Keeping them company. And then when they sense issues or they Mhmm. Check whether they're taking their medication and the person who's become friends with this LLM stuffed animal says, I'm done taking my meds. Then they send a message to the

护士。是的。或者我不知道，或者就像是‘我摔倒了，爬不起来’的终极版，所有这些事情。它一开始非常简单。我每天早上8点发送一条短信，确保这位老年人服用了他们所有的五种药物，可能还得问问他们。

nurse. Yes. Or I don't know or it's like the nth degree of I fall and I can't get up, that all these things. And it started off very simplistically. I'm going to send a text at 08:00 every morning, making sure that this elderly person took all five of their meds and maybe had to ask them.

现在它变得更加对话化，更具吸引力。可以通过聊天实现，也可以通过音频和语音实现，这比让人逐行查看文本线程要好得多。因此它变得更加平易近人。但我不确定我们是否准备好面对那些阴暗面——你提到的自我伤害问题，这些机器人各自拥有的不同人格，以及思考这些问题的同时，还要考虑如此快速达到如此高水平可能带来的意外后果。

Now it's become much more conversational, much more engaging. It could be via chat. It could be via audio and voice, which is better than, you know, having, you know, someone have to go into each line of of a text thread. So that's becoming much more approachable. But I'm not sure if we're ready to there's always the dark side, which you touched on the self harm, the different personalities, actually, that each of these bots have, and thinking about that and what is the and what's the unintended consequence of of something getting that good that quickly.

作为投资者，这是你想涉足的领域吗？还是说...

And as an investor, is that something that you wanna touch or you're

不。我们正在观察，投入大量时间研究。我认为AI陪伴是件了不起的事，它是全方位的陪伴。对老年人来说可以是医疗助手。

No. We're looking. We're spending a lot of time. I think, you know, AI companionship is an incredible thing, and it's a broad based companionship. It could be your medical buddy if you're an elderly person.

对学生可以是数学辅导伙伴，对计划度假的人可以是冲浪向导。这些AI伙伴中有些会是类似ChatGPT的存在。然后你必须思考：这种陪伴有多少是自主引导的？它能在多大程度上理解你的个性和输入内容？

It could be your math buddy if you're a student. It could be your friend who it could be it could be your surf buddy if you're trying to figure out where to go on vacation. So all these buddies, some of which are gonna be chat GPT, you know, are gonna be out there. And then I think you have to think about, you know, how much is that self directed? So how much is it understanding your personality and what you're inputting?

它们会阿谀奉承吗？你会遇到像军训教官那样的营养师AI吗？这究竟是你想要的还是需要的？未来你将能自行调节这些特性。

And are they sycophantic? Are you know, do you have a drill sergeant type nutritionist? And is that what you want or what you need? You'll be able to tune it yourself.

或者它会自我调整，因为你给出的数据会让它意识到'我之前太严厉了，他们都不理我了，现在我要更谄媚些'。

Or it will adapt because you're gonna give it numbers, and it will be like, oh, I was too hard on them. They stopped talking to me. Now I'm a little more sycophantic.

没错。'你的体检报告没问题，吃块牛排怎么了？薯条又不是世界末日，你值得有个放纵日'。

Yeah. Losing my blood work. You coulda had the steak. French fries aren't the end of the world, and you earned a cheat day.

好的，我会报名参加。那么，在我们最初用几分钟讨论了AI将如何吞噬一切，甚至可能是所有事物之后，我现在要问你，科技行业或投资者是否在AI上投入了过多资金。这听起来有些矛盾，但我认为两者可能都是真的，因为

Okay. I will sign up for that. So now after spending our first bunch of minutes together talking about how AI is going to gobble things up, maybe everything, I'm going to now ask you whether the tech industry or investors are putting too much money into AI. It sounds inconsistent, but I think it could both be true because

我认为两者都可能成立。很难说多少资金算过多。我的意思是，很明显的是，所有超大规模企业都在尽其所能地投资。甚至可能与历史上之前的时期不同，我见过被引用最多的两个例子是铁路和互联网基础设施。我对后者比较熟悉。

I think they both could be true. It's hard to say what too much money is. I mean, what's what's been very clear is all the hyperscalers are investing as much as they possibly can. And maybe even differently than probably prior times in history, and the two ones I've seen cited the most are are the railroads, and then the infrastructure of the Internet. And I'm familiar with the last one.

令人惊讶的是，上世纪九十年代末的那段时期，风投扮演了重要角色。那些市场很大程度上依赖于外部资本。对吧？如果你在建设CLEC（竞争性本地交换运营商）或互联网基础设施、暗光纤，你依赖的是资本市场的股权或债务融资。因此，当资金来源中断、成本上升或市场不再买账时，就能控制这种‘氧气’和建设进度。

Amazingly, that was VC during that last time in the late nineties. The they were those markets were largely reliant on external capital. Right? You were if you were building out a CLEC or if you were building out, you know, Internet infrastructure, dark fiber, you were relying on equity or debt from the capital markets. And therefore, when that shut off or that became more expensive or the markets didn't buy in, it was able to control that oxygen and that build out.

这次的不同之处（或其中之一）在于，超大规模企业实际上是通过自身盈利来支付这些投资的。显然，市场会通过股价投票，但他们不需要对外宣布‘我要投资1亿美元用于数据中心能源’，他们可以直接用当季一半的EBITDA（税息折旧及摊销前利润）来建设，因为他们认为这是现金流的重要用途。也许市场会对马克·扎克伯格的选择皱眉，但他不必像伸手要资本的人那样受制于人。所以我认为这不会停止。而且这些企业的野心既宏大又广泛，他们已经‘全押’了。

The different thing this time or one of the different things this time is that the hyperscalers are actually paying for this through their own earnings. So effectively, obviously, the market gets to vote through your stock price, but they don't have to go out and say, I'm investing, you know, a $100,000,000 in energy for my data centers, and I'm just I'm just gonna take, you know, this quarter half of this quarter's EBITDA and build that out because I believe that's an important part and important use of my cash flow. And maybe the market will frown on Mark Zuckerberg if he chooses to do it, but he he he's not gonna be beholden to anybody as you are when you go hat in hand asking for capital. So I think this is not gonna stop. And I also think the hyperscalers, all of their ambitions are so big and so broad, and they're also pot committed.

我认为没人会停下。除非出现极其重大的变化——不是外部有人喊停‘我不会再为你们可能永远用不上的暗光纤开支票’或‘我不会再修建通往虚无的铁路’，而是内部决策者说‘我要退出AI竞赛的这一部分’。考虑到涉及的市值、资金和自尊心，我认为这太难了

I don't think anyone's gonna stop. So, you know, it's gonna take something incredibly material where there's not an outside person who's saying, hey. I'm stopping writing the checks for, you know, for you to buy dark fiber that you're not gonna light up, or I'm not gonna, you know, build a railroad to nowhere because that doesn't make sense anymore despite where the hype was in the market. You know, that has to be internal, and that has to be, hey. I'm I'm bowing out of this part of the AI race, which I think, given the ego's market caps and dollars involved, I think that'd be too hard

那么给我们一些背景信息。你能否立刻说出FirstMark对一家公司开出的最大支票金额？或者

to do. So just give us some context here. You Do know off the top of the head the largest check that FirstMark has put into a company? Or can

你能告诉我们

you us

大概范围？

a ballpark?

2亿美元。

$200,000,000.

好的。Jensen刚刚承诺或最近承诺了一笔1000亿美元

Okay. Jensen just committed or recently committed a $100,000,000,000

是的。给OpenAI。一天，一张支票。我是说，这比，你知道，比某些年份整个风险投资行业几年的总和还要多。

Yes. To OpenAI. One day, one check. I mean, it's more than, you know, more than what's been a couple years certain certain in certain years of venture of all venture capital.

那么，显然当你投入这些资金时，你必须考虑我能得到什么回报。

So what so you obviously, when you're putting in these checks, you have to think about what am I gonna get in return.

是的。

Yes.

如果你向一家公司投资1000亿美元，你需要得到什么？至少需要拿回一万亿美元吗？

What do what do you have to to get if you invest a 100,000,000,000 in a company? Do you need to do you need to get a trillion dollars at least back

作为回报？这取决于你是谁，某种程度上这又回到了循环利用的问题上

in return? Depends on who you are, and that kinda goes to the recycling

好的。

Okay.

或者说这些事物中某些的循环性。显然，OpenAI与微软或OpenAI的工作可以追溯到OpenAI与甲骨文的交易。嗯。而且，你知道，我会给你资金，让你投资于我们的基础设施，或者这种资本循环是如何运作的，这往往出现在这些循环的末期，对吧，当你自身无法产生足够的资金时。你可能已经耗尽了外部的资本池。

Or the circularness of some of these things. Obviously, the OpenAI Microsoft or OpenAI work goes back to the OpenAI Oracle deal. Mhmm. And, you know, I'm gonna I'm gonna give you money that you're gonna invest in our infrastructure, or how how does this cycle of capital work, which tends to be towards the end of these cycles, right, where you can't generate enough money yourself. You might have exhausted the capital pools externally.

所以现在我们都要互相提供收入和现金流，以保持这个链条继续运转。这实际上有点像煤矿中的金丝雀，预示着这种运作方式的问题。但同时，你也知道，NVIDIA的价值如此之高，以至于，可以说1000亿美元对我们来说并不是什么大问题，如果我们相信这是一家能影响一代人的公司，并且有一定的立足点。

So now we're gonna all give each other revenue and and cash flow to keep them keep the train going. So that is actually a little bit of a canary in the coal mine of how this is working. But and then also, you know, NVIDIA's worth so much money that, you know, gents can almost say a $100,000,000,000 is is not that big of a deal if we believe this is a generational company and have somewhat of a leg to stand on.

对。你知道吗？

Right. You know?

而且，就在不久前，1000亿的市值还只有极少数公司能够超越。所以，这些数字的比例悬殊到令人难以置信，很难真正将它们置于具体情境中去理解。

And, you know, not that long ago, a 100,000,000,000 was greater than the market cap of all but a few companies. So, you know, it's it's the numbers are so mind blowingly disproportionate, it's hard to really contextualize them.

没错。我们得再次说明，对于许多OpenAI的投资来说，至少在初期，这有点像有趣的数学游戏。最初是100亿，并计划分阶段再追加900亿。

Right. And and we should say again, with with many OpenAI investments, it's kind of funny math, at least to the beginning. It's 10,000,000,000 to start with plans to contribute another 90,000,000,000 in increments.

而且，你知道，OpenAI作为私营公司的一个好处，或者说优势之一，就是他们能达成许多这类交易

And, you know, the best part of of one of the good parts of, you know, OpenAI being private is they could do a lot of these deals

没错。

Right.

这些交易不必——显然也肯定不需要公开披露，毕竟英伟达是上市公司。他们必须披露与甲骨文或其他任何合作方的交易，就像九十年代末美国在线那样展示漂亮的营收数据。但这本质上是一种实物贡献，附带某些里程碑条款和其他条件，这非常像市场泡沫期的特征——我们讨论的并非真实财务指标、GAAP收入或实际到账现金。

Where they don't, you know, definitely don't have to be disclosed, obviously, because NVIDIA's public. They're gonna have to disclose that or Oracle or whatever it is that they're able to put up, you know, great top line numbers no different than maybe AOL did in the late nineties of, hey. Here's the top line. But it was really a contribution in kind, and there's really some milestones to it, and there's really some other things, which also is very much a symbol of, like, a very frothy market of, hey. We're not talking about actual financial metrics or actual GAAP revenue or actual cash on the barrel head.

我们讨论的是基于理论里程碑的广义现金等价物价值，这些可能并非真实货币。

We're talking about a theoretical milestone based broader in cash, in kind dollar amount, which might not be real dollars.

对。我记得黄仁勋将其称为'合作优先，投资次之'。有意思的是，这很可能成为史上最大规模的投资。

Right. I think Jensen has referred to it as a partnership first and investment second. Yes. And that's interesting because it would be by far the biggest investment in history.

是的，没错。但严格来说又不是投资。

Yes. Yes. And So But but not an investment.

正是如此。你曾与许多寻求融资的创始人面对面交流，想必遇见过脚踏实地的创业者，也见过只会画大饼的。我很好奇你是否遇到过...

Right. Exactly. You've you've spoken, sat across the table with lots of founders that are trying to pitch you on on fundraising. I'm sure there's a spectrum of really grounded founders to founders who will try to sell you a dream and Yes. I'm curious if you've ever heard

其中很多项目兼具双重属性，我们也投入了大量资源在这些双轨项目上。

And a lot of them were both, and we've invested a lot that are both.

所以我很好奇你是否听过类似的说法。我们在节目里讨论过这个——这是山姆·奥特曼谈及英伟达投资时说的：'这个超级大脑将孕育出的成果会非凡到以我们目前认知都难以理解的程度。'没错。

So I'm curious if you ever heard anything like this. We've talked about this on the show. This is from Sam Altman when he was talking about the NVIDIA investment. He says, the stuff that will come out of this super brain will be remarkable in a way I think we don't really know how to think about yet. Yes.

如果有人来告诉你，即将发生的事情——你正在投资的东西会震撼到让你完全无法理解。你...你会作何反应？

Is that if someone came and told that to you, that what what what is coming? What you're investing in will be so amazing, we you don't even know how to wrap your head around it. What what's your reaction?

是的。我正想问这个

Yeah. I'm asking this

顺便说句真心话，因为

by the way earnestly because

我...我几年前确实遇到过一位创始人，当时我问他问题，他说：'我可以解释给你听，但这可能是在浪费我的时间，因为你大概率理解不了。'

I I did have a I did have a founder a couple years ago, several years ago, who basically said, I asked him a question. He said, I can explain it to you, but it's probably not worth my time because you probably couldn't understand it.

好吧。

Okay.

我说，试试我，他们说，不。我不认为我不认为你能做到。令人惊讶的是，我们投资了，我想我们

And I said, try me, and they said, no. I don't think I don't think you could get it. Amazingly, we we invested, and I think we

之后我们确实做到了。

made we made after that.

我们赚了钱。

We made money.

那么这是个好策略。

So this is a good strategy then.

是的，也许吧。也许吧。不，我，我认为那是，嘿。

It's yeah. Maybe it is. Maybe it is. No. I I think the that is hey.

我的抱负如此广阔，而我设定的期望——我设定的期望如此之高。言语无法充分表达这些期望，这也是另一个警示信号，可能实际上是我的性格使然。我喜欢具体的事物。所以，就像，嘿，我们要做这个。

My ambitions are so broad, and my expectations I'm setting I'm setting expectations so high. Words cannot do the expectations justice, which also is another little canary in the coal mine of I actually maybe it's my personality. I like concrete things. So, like, hey. We're going to do this.

我们将成为一家大公司，之所以能成为大公司，是因为我们认为根据这个世界的运作方式，我们可以卖出价值十亿美元的这款产品。你可能会说，哦，那是个宏大、艰巨、大胆的目标，但我可以追踪它，因为这对我来说是有意义的。但当人们说，我们将成为有史以来最大的公司，因为我们要做你的大脑甚至无法追踪的事情，你知道，那感觉有点难以追踪。但是，考虑到Sam的成就，如果有人有资格说这样的话，也许是他，Elon，少数能发表这种言论的人。

We're gonna be a big company, and we're gonna be a big company because we think we could sell a billion dollars of this product given how this world works out. And you'd be like, oh, that's big, hairy, audacious goal, but I could track that because that makes sense to me. But when people say, we're gonna be the biggest company ever because we're gonna do things that your brain can't even track, that's you know, that it feels a little bit harder to track. But, you know, given what Sam has done, if anybody has earned the right to say things like that, you know, maybe him, Elon, rare air of folks who who could get away with that type of comment.

确实如此。我的意思是，这里存在一种平衡，比如你可以欣赏——我当然也赞赏萨姆在OpenAI掌舵期间以及即便失去大量人才后仍在持续取得的成就。是的，公司仍在不断推出产品。但当你提出这个更广泛的问题：事情是否有些过热了？

Definitely. I mean, there's there's a balance here between, like, you can appreciate, and I certainly do what Sam has done at the helm of OpenAI and continues to do even though they've lost a lot of talent. Yes. The company continues to ship. But then when you're asking this broader question of is are things a little frothy?

是的。

Yes.

当你

When you

在关于这笔1000亿美元投资的报道中看到这样的引述时，那就是我开始质疑的地方

see a quote like that in a story about this $100,000,000,000 investment, that is that's where I started

要问的不是投资本身，而是子公司在实质层面做了什么。这个千亿级合作的关键支柱之一，是他们将开展那些我无法向你解释的事情——因为你不会理解。就像，你知道，这可能会成为某本书的封面故事，《我在市场巅峰所见》，由某位五年后不愿具名的作者撰写。

to ask non investment in the in the the subs made substantive. One of the key pillars of what's the the the $100,000,000 partnership is they're gonna do things that I couldn't explain to you because you wouldn't understand. You're like, you know, that might be that might be on the cover of a book of what I saw what I saw at the top of the market by, you know, writer to be unnamed in five years.

是啊。我最好赶紧推销这个点子。没错。但但但接下来我们该讨论这对市场意味着什么。对吧？

Yeah. I'm I better get pitching that. Yes. But but but then we should talk about then then what it means for the market. Right?

因为你当然关注着私募市场和公开市场。想想公开市场对萨姆表现优异、兑现承诺寄予了多大期望。

Because you follow, of course, the the private markets, the public markets. And if you think about how much the public market is relying on Sam to do well, Sam to deliver on that promise that he's made.

嗯，萨姆必须这么做，因为依赖关系确实如此

Well, Sam has to do that because the relying Because of

甲骨文没错。CoreWeave、英伟达，它们现在都依赖OpenAI来交付成果。我甚至不清楚具体要交付什么。

Oracle Yeah. CoreWeave, NVIDIA, they are now all relying on OpenAI to deliver. And I don't even know what more I mean, to deliver what exactly.

然后你再想想所有这些生态系统。

And then you think about all those ecosystems.

确实如此。

So Yep.

要知道，能源公司正依赖CoreWeave建设基础设施。想想微软正在做的那些事，都依赖于OpenAI的部分成果。单就人们基于现有合同、远期合约、承诺、合作伙伴关系或握手协议所预估的纯粹估值而言，这种预期和承诺正在不断升级，说实话，有时会让人感到些许不安。

You know, the energy companies are relying on CoreWeave to build out the infrastructure. You think about, you know, all the things that Microsoft's doing that are reliant on some of the things that OpenAI is doing. You know, if just nothing else, the pure valuation that people are baking in given all the contracts or forward contracts or promises or partnerships or handshakes are done, that it's just escalating the expectation and commitment, which again, you know, starts to make you feel a bit uncomfortable at times.

没错。我认为OpenAI的解决方案必须是：为了满足这些巨大期望——虽然我刚搭建好框架，也清楚他们的目标方向——他们需要实现大量白领工作的自动化。是的。

Right. And I think the answer for OpenAI has to be that in order to meet these enormous expectations, I just set it up, I know what they're building towards. That wasn't quite right. What they need to do is to automate a tremendous amount of white collar labor. So Yes.

我想探讨这个问题，以及Z世代的现状——他们似乎首当其冲，目前难以找到工作。广告之后我们将深入讨论。通过Agency（a g n t c y）塑造企业AI的未来，这个Linux基金会的开源项目正在为物联网代理建立可信身份与访问管理标准，打造确保AI代理能安全发现、连接并跨框架协作的交互层。

I wanna talk about that and what's happening with Gen z who's at the seems like the spear's edge of this Yes. And not not able to find jobs right now. I wanna talk about that right after this. Shape the future of enterprise AI with Agency, a g n t c y. Now an open source Linux foundation project, Agency is leading the way in establishing trusted identity and access management for the Internet of agents, a collaboration layer that ensures AI agents can securely discover, connect, and work across any framework.

通过Agency，您的组织将获得开放、标准化的工具和无缝集成，包括强大的身份管理功能，以便在任何平台上进行识别、认证和交互。让您能够自信地部署多智能体系统，加入思科、戴尔科技、谷歌云、甲骨文、红帽等行业领袖及75家以上支持企业的行列，共同制定安全、可扩展的AI基础设施标准。您的企业准备好迎接Vagintic AI的未来了吗？立即访问agency.org探索应用场景。网址是a g n t c y . o r g。

With Agency, your organization gains open, standardized tools, and seamless integration, including robust identity management to be able to identify, authenticate, and interact across any platform. Empowering you to deploy multi agent systems with confidence, join industry leaders like Cisco, Dell Technologies, Google Cloud, Oracle, Red Hat, and 75 plus supporting companies to set the standard for secure, scalable AI infrastructure. Is your enterprise ready for the future of Vagintic AI? Visit agency.org to explore use cases now. That's a g n t c y dot o r g.

Capital One的技术团队不仅在谈论多智能体AI，他们早已部署了一套系统——名为聊天礼宾服务，它正在简化购车流程。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，还能协助安排试驾、获取融资预批以及评估置换价值。

Capital One's tech team isn't just talking about multi agentic AI. They already deployed one. It's called chat concierge, and it's simplifying car shopping. Using self reflection and layered reasoning with live API checks, it doesn't just help buyers find the car they love. It helps schedule a test drive, get preapproved for financing, and estimate trade and value.

先进、直观且已投入应用——这就是他们的技术实力。这就是Capital One的科技之道。寻找合适的科技人才不仅困难，更是关乎企业存亡的关键任务。

Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One. Finding the right tech talent isn't just hard. It's mission critical.

然而许多企业雇主仍依赖过时的方法或无效的平台。当今市场中，招聘科技人才不仅是填补职位空缺，更是超越竞争对手的关键。但面对专业技能、混合工作偏好和高薪资期望，从未有比现在更困难的时刻来突破噪音，连接对的人选。这正是Indeed的用武之地——该平台每月发布超过50万个科技职位，雇主使用其先进定向功能后，通过科技网络分发的申请启动量可提升2.1倍。

And yet many enterprise employers still rely on outdated methods or platforms that don't deliver. In today's market, hiring tech professionals isn't just about filling roles, it's about outpacing competitors. But with niche skills, hybrid preferences, and high salary expectations, it's never been more challenging to cut through the noise and connect with the right people. That's where Indeed comes in. Indeed consistently posts over 500,000 tech roles per month, and employers using its platform benefit from advanced targeting and a 2.1 x lift in started applications when using tech network distribution.

若我需要招聘顶尖科技人才，我会选择Indeed。发布首个职位可享75美元优惠，请访问indeed.com/techtalent。立即前往indeed.com/techtalent领取优惠。Indeed——为当下及未来的科技招聘而打造。现在让我们欢迎FirstMark Capital董事总经理Rick Heitzman回到节目。

If I needed to hire top tier tech talent, I'd go with Indeed. Post your first job and get $75 off at indeed.com/techtalent. That's indeed.com/techtalent that claim this offer. Indeed, built for what's now and what's next in tech hiring. And we're back here with Rick Heitzman, managing director of FirstMark Capital.

Rick，我们在休息前讨论了OpenAI需要自动化大量工作岗位以匹配其估值的问题。现在进入下半场，让我们从宏观角度开始：我们是否正迈向由OpenAI、Anthropic等技术公司试图全面自动化所有白领工作的未来？如果它们成功了，将会发生什么？

Rick, we talked before the break about how OpenAI is gonna need to automate a lot of jobs in order to justify this valuation. So let's just start broad as we begin the second half here. Are we marching towards technology companies like OpenAI, like Anthropic, basically trying to automate all work, the all white collar work, and if they're successful, what happens?

呃，我认为是自动化所有工作。对吧？因为考虑到当前机器人技术和工厂自动化的发展，这既涉及蓝领也包含白领岗位。与历史上农业机械化（比如推土机和蒸汽机替代蓝领劳动）不同，本次自动化浪潮有着本质区别。

Well, I I would say automating all work. Right? Because if you think about some of the robotics things that are happening now, some factory automation things that are happening now, it's both blue collar and white collar. I think maybe differently than any kind of automation going back to, you know, farming where, you know, there's bulldozers and there's steam engines that are automating blue collar work. You know, this has been very different.

因此，我确实认为，在与那些通常雇佣大量初级员工或入门级咨询公司、业务流程外包公司的银行家和律师交谈时，他们正在暂停或采取更缓慢、更审慎的方式来填补金字塔底层的职位，这促使他们重新思考自己的业务模式。我相信他们会重新规划业务，虽然可能会流失部分人员，但这些人员将被重新调配用途。回顾二十世纪初的情况...

And so I I do think, and in talking to the bankers and the lawyers who usually hire a whole lot of folks or, you know, entry level consulting firms, BPO firms, they are pausing or taking a slower approach or a more thoughtful and cautious approach to how they fill in the bottom of the pyramid, and that makes them, you know, rethink their their business. You know, I do believe that they're gonna rethink their business. I think you're gonna lose some people, but those people are gonna be repurposed. Right? So if you go back to, the beginning of the twentieth century.

二十世纪初，约93%的美国人口从事农业经济，基本上是农民。而到了二十世纪末，美国劳动力中农民的比例降至约3%。仅看这两个数据，你可能会惊呼：天哪，一定发生了什么可怕的事情，这些人肯定都失业了。

So beginning of the twentieth century, about 93 people 93% of Americans were in the agrarian economy, farmers, basically. At the end of the twentieth century, it was about 3% of the American workforce as farmers. And if you looked at just those two stats, you're like, oh my god. Something horrible must have happened. All these people must have lost their jobs.

实际发生了什么？那简直糟透了？不，那是人类文明史上任何一个经济体所经历的最伟大的世纪。

What happened? It was terrible. What happened? Oh, it was the greatest century of an economy of any civilization's economy in the history of civilization. Yeah.

美国的二十世纪及其间发生的一切变革。我认为资本主义极其擅长这种创造性破坏。在工业革命中你看到过这种职业转型，在那个世纪的工厂自动化与技术进步的各个阶段，这种转型多次上演。我预见我们将见证对某些领域（尤其是白领工作）的重新思考。

The American twentieth century and everything that happened. So there is a a creative destruction that I think capitalism is really good at. And I think that you're you're see you saw that repurposement in the industrial revolution. You saw a repurposement several times in the automation, the ability to have factories and all the technological advances during that century. I think you're going to see a rethinking about some of those things, especially around white collar work.

类似的例子比比皆是。记得文字处理器刚问世时——具体时间记不清了，大约是35年前——人们惊呼：天哪，这将是法律行业的末日。

And, you know, there's a bunch of different things that are analogous to it. You know, word processors came out and they said, I forget what it is. We call it thirty five years ago, so I'll be imprecise. And they said, oh my god. This is gonna be the end of the legal industry.

结果呢？自从文字处理器出现后，美国的律师数量反而增长了四倍。电子表格的发展史同样如此。

You know? Now we're not gonna need you know, now people are gonna be able to, you know, print these documents, use word processors. Since the advent of word processors, there's four times more lawyers in America than there was at the time. You know? The same thing if you think about spreadsheets.

当Lotus 1-2-3电子表格软件面世时，人们说：我们不再需要这么多银行职员了，计算尺也该淘汰了，有电子表格就够了。

And, you know, Lotus one two three comes out. There's gonna be spreadsheets. And they said, well, we're not gonna need all these bankers. We're not gonna need slide rules. We're gonna have spreadsheets.

我们要把整个金融流程自动化。现在的银行从业者数量比电子表格出现前多了不知多少倍。那么问题来了：如今我们能自动化大量白领工作，现在就能实现基础业务流程的自动化，中期来看效果可能会更好。

We're gonna automate this whole finance thing. There's more bankers than there were before spreadsheets by, you know, a huge huge factor. So, you know, what is it gonna be? Now we're able to automate a lot of white collar tasks. You're able to automate basic business processes today, probably better in the medium term.

你认为Z世代是否具备足够的创造力和创业精神来实现自我重塑？我相信可以。我对此有信心。我确实为那些刚毕业的大学生朋友感到难过——去年是金融危机以来应届毕业生处境最糟糕的一年。

And do you believe that Gen Z is gonna be creative enough and entrepreneurial enough to reinvent themselves? I think so. I have confidence in that. I do feel for I have friends who are going through recent college graduates, and it's the worst. It was the worst year last year since the financial crisis for recent college graduates.

我有个读大三的儿子，他正焦虑AI是否会取代他的工作。我理解这种担忧，但也鼓励他思考：你能做哪些AI做不到的事？或者你有什么与众不同的想法？因为最优秀的人往往是那些能提前半步理解趋势的人。

I have a son who's a junior in college. He feels anxiety of, you know, what does this mean? Is AI gonna take my job? So I I I have empathy for that, but I also push him to say, well, what what what could you do that AI can't do, or what what are you thinking about that's thinking about something differently? Because the best people are gonna be the people who understand it a little ahead of time.

我们已经看到有人创办律所，其全部助理工作都由AI完成。合伙人只需专注提供判断力、客户对接等高价值工作，后端由AI支撑，不再受传统律所金字塔模式的束缚。这将催生创业浪潮和创造性破坏，总体而言对大多数人都是有益的。

And, you know, we're beginning to see people spin out law firms, that their, you know, entire associate infrastructure is AI. So, you know, they're able to be the partner who's able to add that level of judgment, client interface, all those things, and their back end is AI, and they're not beholden to the pyramid model of law firms to be able to make their business work. So you're gonna see entrepreneurship. You're gonna see creative destruction. I think that on the whole, we should all we're all almost all of us will be better off for it.

说实话我对此很矛盾。一方面AI确实越来越强大，而商业决策终究要回归本质——资本正在押注这些岗位将被自动化取代，英伟达投资OpenAI就是明证。

You know, I really go back and forth on it because on one hand, it does seem like AI is becoming more and more capable. And again, just start so many times in business, it's worth just starting at the money, right? Yes. The money is betting that all these jobs will be automated. Yes.

关键在于后续发展。如果英伟达在其投资回报周期内实现目标，必将引发巨大震荡。但另一方面，我们也要观察企业日常运作的真实情况。

That's what that money from NVIDIA into OpenAI is trying to do. And then the question is what happens afterwards? And you could have it it seems like if they get there, right, in the time that NVIDIA wants that investment to pay back, there's gonna be massive disruption. Yeah. But then you also look at what happens in the day to day of of many companies.

有个发人深省的Substack文章《企业职位的消亡》，作者追踪记录了朋友同事的工作状态。文中说：我不断遇到用非常规词汇描述自己工作的人，这些表述在他们日常对话中从未出现过。

There's a great and a thought provoking Yeah. Sub stack post in this, sub stack called Still Wandering, and it was called the death of the corporate job. And the author was trying to track what the what their friends and counterparts did in their work. Here's what the author said. I keep meeting people who describe their jobs using words they'd never use in normal conversations.

他们参加关于会议的会议。制作没人阅读的PPT，通过没人打开的邮件分享，产生不需要完成的任务。这篇帖子在Substack上被浏览了近11000到12000次，是的。在Substack上，这是非常非常多的人。这意味着

They attend meetings about meetings. They create PowerPoints that nobody reads, which get shared in emails that no one opens, which generate tasks that don't need doing. This post was like 11,000, close to 12,000 times on Substack, which Yeah. On Substack, that's a lot a lot of people. Which means like

这也很可笑。没错。但它引起了人们的共鸣。可笑是因为它是事实。

It's also funny. Yeah. But it resonates with people. Funny because it's true.

对。就像，这件事能如此引起知识经济领域这么多人的共鸣，本身就说明问题。也许人工智能会消除这些现象，也许当前招聘紧缩或停滞的局面，正是企业意识到

Right. Like, the fact that that resonated that way with so many people who are in the knowledge economy, it's just so telling. And maybe AI eliminates that stuff, and maybe this moment where we've had hiring consolidate or or stop in many ways is a realization by companies

正在发生的事。《纽约客》有幅很棒的政治漫画，讲的是有人写封邮件说：AI，把这变成100页的PPT演示稿。AI生成后他们发给同事，同事又说：AI，把这100页PPT浓缩成简短邮件。完全如此。每个人都在用AI自动化不同环节。我觉得，让那些写没人看的邮件、做只称重不阅读的演示稿的人停止这些工作，对大家都更好。

like going on. There's a great New Yorker, political cartoon, which is the same thing of, you know, someone types out an email and they say, you know, AI, turn this into a 100 page PowerPoint presentation, and they turn it into a 100 page PowerPoint and they email it to their colleague, and then says, AI, take this 100 page PowerPoint presentation and turn it into a short email. Exactly. So they're everybody is using AI to automate different pieces. You know, I I I just think that, you know, to have the have those people who are doing writing emails that no one reads or creating decks that people only weigh but not read, I think having them not do that is better for everyone.

没错。但问题是之后商业形态会怎样。有两种可能：一是这些人最终转向更高价值的工作。

Right. But the question is, like, what business looks like afterwards. So there's, like, two possibilities. One is that all those people end up on higher value tasks.

但愿如此

You hope

二是企业发现：天啊，我们只需要三分之一的人手。

Two is companies go, oh my goodness. We need one third of the people.

是的，你已经看到一些公司开始这么做了。比如Shopify在增加收入的同时裁减了大量员工。工程师岗位并未消失，但公司保持工程团队规模不变，却因各种编程工具获得了更高生产力。我认为许多业务流程外包、呼叫中心和客服岗位正因技术而缩减。所以我确实认为某些领域会出现实质性岗位流失。

Yeah, and you're seeing some companies already do that. You know, some companies, Shopify increased revenue and took out a fair amount of their employee base. You're not seeing engineers go away, but you're seeing companies keep engineering flat, but getting a lot more productivity due to all the coding tools. I think you're seeing a lot of kind of business process outsourcing or call centers and customer service things that are getting shrunk due to technology. So I do think there's going to be a substantive job loss in certain fields.

我们希望人们能从事更有意义的工作，而不是被迫去做那些——我们俩都很幸运，我们的工作不是写没人看的邮件或制作没人听的内容，但愿如此。

You hope that people will do some more meaningful work than having to go, and I think we're both very fortunate that our job is not writing not writing emails that people don't read or producing content that people don't listen to, I hope.

你出现在这里是有充分理由的。我们拥有观众。我希望屏幕前的每个人都——是的。

You're here for a good reason. We have an audience. I hope everyone out there. Yeah.

没错。所以我认为...我认为你会看到更多人从事不同工作。我的意思是，这部分源于创作者经济的崛起。

Yes. So I think that I I think you're gonna see more people doing different stuff. I mean, you you part of that is the rise of the creator economy.

对。

Right.

你会看到更多人在创作者经济中展现创业精神。即便我们与这个领域的从业者交流时发现，这通常只是副业——有时候他们的主业既缺乏意义又收入微薄，甚至显得荒诞可笑。而他们通过这种副业找到了意义、创造力和收入。有时这份副业会转变成主业，成为更有价值的机会。我认为这种情况会越来越普遍。

And you're seeing more people be entrepreneurial in the creator economy. And even as we've talked to folks out there in the creator economy, it's often a side gig, and sometimes either their their day job is not very meaningful and not very lucrative or seems like it's a, you know, it's a cartoonish type thing, and and they're they're finding meaning, creativity, and dollars in doing this side hustle. And sometimes that side hustle turns into their main job, and that becomes a more meaningful opportunity. I think it's going happen more and more.

确实。我自己就是这样。我最初从事营销销售工作，同时兼职自由撰稿，后来把写作转为全职职业，再转型到现在不仅写作还包括视频、音频和电视节目的工作。

Yeah. I mean, that happened to me. I was starting my career in marketing and sales and writing freelance journalism on the side, and then flipped that to a full time career, and then flipped that into something that's now not just writing, but is video, audio, some TV.

哪个那是

Which That's

一直很不错。但实际上，这让我回想起我的第一份工作，那时我会制定媒体计划，这些计划会经过那些没人会看的邮件链和演示文稿，最终有人会批准或不批准。当然，如今程序化广告已经颠覆了这一流程，一切都自动化了。所以那可能算是一份新工作。但我只是在想，像那份让我进入职场的初级工作，你可以用聊天机器人搞定它，五分钟就完事。

has been nice. But I actually, it brings me back to my first job, which was I would put together media plans that would go through those email chains and the decks that no one would read and eventually someone would approve it or not And approve of course that process has been disrupted by programmatic advertising where you just automate it all today. So maybe that would be a new job. But I'm just thinking like that job, that entry level job that I had that got me into the workforce, you could chat shit PT that and be done with My it in five

我的第一份工作是初级投资银行家，你知道，就是打印文件然后输入到电子表格里。那种工作已经存在多年了。我是说，如果你看看基本的Capital IQ之类的工具，我职业生涯最初几年的工作现在完全被技术自动化取代了。

first job was an entry level investment banker that I was, you know, printing off documents and then keying that into a spreadsheet. You know, that's that's been done for years. I mean, if you if you look at Mhmm. You know, some of the basic Capital IQ things. And probably the first, you know, couple years of my career are now completely automated due to technology.

但即使是现在进入投资银行业的人也在做不同的事情，这些工作变得更有意义，而且投资银行家比以往任何时候都多。所以我认为虽然行业在变化，但这并不意味着它正在消亡。

But even people who are now entering investment banking are doing different stuff, and that's becoming more meaningful, and there's more investment bankers than ever. So I think although it's changing, it doesn't mean it's ending.

我们谈到，好像外面某个地方经济很繁荣。是的。然后你想想正在发生的事，好吧，两类人。Z世代。对。正如你提到的，他们真的很难找到工作，还有那些失业或离职的人，找到新工作的时间也比往常要长。

And we talk about, like, there's a thriving economy out there somewhere. Yes. And then you think about what's happening with, alright, two groups of people. Gen z Yep. Who's really struggling like you mentioned to find jobs, and also people that lose their jobs or leave their jobs are taking longer than usual to find new work.

到底发生了什么？不可能全是AI的错。杰罗姆·鲍威尔最近表示AI可能对此有影响，但我们正处于一个缓慢升温或缓慢解雇的经济中。那么，这个让很多人感觉良好的经济背后的驱动力是什么，如果你看总体数据的话，但如果你是一个试图规划职业道路的个人，感觉一切都停滞不前。

What is happening? It can't all be AI. Jerome Powell recently came out and said AI might contribute to it, but we're in a slow to high or slow to fire economy. And so what is the driving force behind this economy that feels to a lot of people to be doing well if you look at the top line numbers, but if you're an individual trying to navigate your career path, feels like everything is just stuck.

我几乎同意你刚才说的一切。我认为经济非常强劲，基本面也很稳固，我们在企业和消费者公司中都看到了这一点。所以我们对经济实际上感觉良好。第二点是，我确实认为公司现在招聘速度很慢。我认为，从COVID时期走出来后，那是一个由低利率、低资本成本、远程工作推动的极度低效时期，基本上是创造低效的完美方式。

I I agree with almost everything you just said. I think that the economy is very strong, and the fundamentals are strong, and we see it in both our enterprise and consumer companies. So we actually feel good about the economy. The second piece of that is I do believe that, you know, companies are slow to hire. And I think the know, coming off of which was a massively inefficient COVID time spurred by low interest rates, low cost of capital, work from home, it was basically the perfect way to create inefficiency.

嗯。

Mhmm.

既不对资金负责，也不对绩效负责。因此，从那以后，即使是五年后的今天，公司们都在说，好吧，我不会再犯昨天的错误。我的意思是，公司就像个人一样，总是对上一阶段或上一次错误做出反应。

That no accountability for dollars and no accountability for performance. So coming off of that, companies now, even five years later, are saying, okay. I'm not gonna I'm not gonna do the sin I had yesterday. I mean, that's companies like individuals are always reactionary to the last phase. So, you know, companies like individuals are always reactionary to the last phase or their last mistake.

所以我认为公司现在正在思考，好吧，我们如何更高效？如何确保我们明智地投入时间和金钱，而不是雇人写没人看的邮件？我觉得这种转变比较缓慢。但与此同时，失业率仍然很低，而且有一种感觉，就像我毕业时那样，总会有缓冲时间。

So I think companies are now thinking about, alright. How are we more efficient? How do we make sure that we're spending that time and money wisely, and we're not hiring someone to write emails that no one reads? So I think that's been slower. But at the same time, unemployment remains low, and, you know, there is a sense of, you know, when I was coming out of school, that there is some time.

就像我们在学校时被告知的，如果你辞职或失业，你需要一些时间，因为找到新工作需要几个月。但在过去二十年人力资本紧张的繁荣时期，找到新工作只需要几小时。所以我认为现在正回归历史常态，可能是因为经济向好，市场表现良好，管理者更注重绩效驱动，向绩效的历史标准靠拢。

Like, they told us when we were in school, like, if you quit your job or you lose your job, you need to have a little bit of time because it takes months to find a new job. You know, in in in some of the boom times when human capital has been tight over the last twenty years, it's taken hours to find a new job. So I I think that you're you're moving more to historical norms as people are, maybe because the economy is doing well, maybe because the market's doing well, because the managers are being more performance driven and moving more towards historical norms around performance.

好的。想用剩余时间快速讨论几个你投资的项目。你投资了一些非常有趣的公司。

Okay. Want to use our remaining time to lightning round through a couple of your investments. You've invested in some fascinating companies.

谢谢。

Thank you.

有些是我经常使用的，也是我们常谈论的。

Ones that I use all the time, the ones that we talk about.

多用一些。

Use some more.

那我们就尽量过四个吧。我们还有大约八分钟。很好。Discord。

So let's just go through four of them if we can. We have about eight minutes. Great. Discord.

是的。

Yes.

你怎么看待社交媒体的大部分活力转向私域的现象？马克·扎克伯格有个转向隐私的决策。大家都以为他痴迷加密，其实不是。

What do you think about the fact that so much of the dynamism of social media has moved private. Right? Mark Zuckerberg had this pivot to privacy. Everyone's like, he's into encryption. It's like, no.

他意识到社交分享正在群聊中发生。是的。而那就是他希望活动发生的地方。所以稍微谈谈这个。

He realizes social sharing is happening in the group chat. Yes. And that's where he wants things to happen. So talk a little bit about that.

所以我认为转移到Discord服务器上。对吧？

So I think in in moving to Discord servers. Right?

你觉得这从根本上对我们有好处吗？

You have And is that good for us, basically?

嗯，某种程度上确实如此。我是说，我不认为每个人都...你知道那个老笑话，不必向全世界广播你遇到的每个人和你午餐吃了什么。对吧？那并不能推动任何人的生活或经济，也没人需要看你的金枪鱼三明治照片。所以某种程度上，我很高兴我们已过了社交媒体那个阶段。

Well, somewhat somewhat it is. I mean, I I I don't think everybody you know, the old joke, you know, you don't have to broadcast to anyone everybody you ever met we had for lunch. Right? That that's not pushing forward anybody's life or economy, and you don't need to see a picture of the tuna sandwich. So that's I think I'm somewhat glad we're out of that phase of social media.

与此同时，因此有了非常特定主题的服务器——无论你是为了下届世界棒球冠军纽约洋基队的Discord服务器，还是为了某个英雄联盟战队——你会发现它们都存在。正如我们讨论过的，负面因素是这些服务器中有很多是高度封闭的回音室。嗯。围绕着某些可能煽动情绪的特定信念。我认为Discord在内容审核上做得很好，确保这些服务器保持适当的讨论氛围，并维持正常运作。

At the same time, you know, therefore, having servers that are very specific and whether you're in a Discord server for the next world baseball champion New York Yankees or whether you're in it for, you know, a League of Legends clan, you know, all of those things you find that they're there. You know, the negative, as we've talked about, are these are very some of them are very intense echo chambers Mhmm. Around particular beliefs that can spin people up. So I think there is I think Discord does an excellent job of moderation to make sure that, you know, there's the right level of discourse in those Discord servers, and then make sure that works.

但这取决于Discord的管理员。

But that's on the administrator of the Discord.

取决于管理员。对。是服务器。服务器的管理员。

It's on the administrator Right. Yeah. This of the server. Server.

没错。

Right.

确实如此。但我认为你会看到更多社交媒体转向半私密模式，更像是群聊形式——可能围绕体育、音乐、科技或人际关系——因为人们可能有点厌倦了活在公众视野里。

So that is true. But I think you're gonna see you're gonna see more more social media move to semi private that look more like group chats, and and it can be around, you know, sports or music or technology or or relationships, just because I think that people, might might be a little bit over living in public.

是的。我们Big Technology团队超爱Discord。为付费订阅用户运营着私人Discord服务器。感兴趣的话下滑页面注册，我们会发链接给你。

Yeah. We love Discord over here, Big Technology. Have a private Discord server for our paying subscribers. Yes. If you're interested, scroll down, sign up, and we'll get you a link.

我认为这是大型科技公司多年来做得最棒的一件事。

And I think it's the best thing that big technology has done in years.

太棒了。

That's awesome.

对话质量很高，非常有趣，我很喜欢参与其中，从中获益良多。

The conversation is high quality, it's interesting, and I love being in there. Get a lot of value out of it.

是啊。在过去十年的社交媒体发展中，经过最初的爆发期后，内容筛选一直是维持社区健康繁荣的最重要因素。

Yeah. Curation has been, for the last ten years of social media, after the initial explosion, curation has been the most important thing in keeping a good thriving community.

好吧。DraftKings。是的。大谷翔平真的赌棒球了吗，还是他的翻译干的？

Alright. DraftKings. Yes. Did Shohei Ohtani actually bet on baseball, or was it his interpreter?

这个我不清楚。真不知道。我我我...就算我问了，他们也不一定会告诉我。我在想...是我个人的看法，还是DraftKings的观点？让我让我...

I do not know that. Do not know. I I I I'm not sure if I if they would tell me if I asked, I think what do I think, or what does DraftKings think? Let let me let me

换个不那么戏谑的方式问吧。显然，体育博彩已经普及化了。

ask it in a in a little bit less facetious way. Obviously, sports betting's been popularized.

是的。

Yes.

各大联赛都在推广它。是的。球员们也参与其中。是的。这有问题吗？

The leagues all promote it. Yes. The players are getting into it. Yes. Is that a problem?

嗯，你看到越来越多关于赌博使用和滥用的调查和禁赛。所以我认为，就像任何新技术一样，它一开始就爆发式增长。有点像狂野西部。我想，DraftKings作为一家大型上市公司，是行业领导者，可能比预测市场或一些灰色地带的市场有更多的防护措施。我认为，政府总是难以跟上技术发展的步伐。

The Well, you're seeing more and more investigations and suspensions around the use and misuse of of gambling. So I think, you know, like any new technology, it explodes out of the gates. It's a little bit of the Wild West. I think, you know, DraftKings being a large public company who is a leader probably has more guardrails around it than maybe prediction markets or some of these, you know, sweepstakes types gray area markets. I think, you know, the government always struggles to keep up with where technology is going.

政府往往关注的是昨天的问题，而不是明天的问题。所以我相信会有更明确的规则，尤其是针对球员、教练、裁判和管理人员，规定他们在赌博或预测市场上能做什么。好了，我们来谈谈Shopify。

It's oftentimes focused on yesterday's problem, not tomorrow's problem. So I believe there's gonna be more clearer rules around, especially players, coaches, umpires, managers, and what they're able to do on on either gambling or prediction markets. Okay. Let's talk about Shopify.

是的。你是Shopify的投资者。我是。所有的在线商务都会从应用程序和网站转向聊天机器人吗？如果是这样，会发生什么？

Yes. You're an investor in Shopify. I am. Is all online commerce gonna go from applications and websites to into chatbots? And if so, what happens?

我认为这不会是全有或全无的情况。所以我认为你会转向更多的聊天机器人。我认为你仍然需要一个生态系统，无论是无头商店还是后端基础设施。无论你是通过AI女友告诉你某件毛衣好看而购买，还是通过聊天机器人基于你的AI购买。

So I think that that's I I don't think all it's never all or nothing. So I think you're gonna move to more chatbots. I think you're gonna still need an ecosystem of whether it's, you know, headless stores or whether it's a back end infrastructure of you're still whether you're you're buying a sweater because your AI girlfriend tells you it looks good and you're buying it in a chatbot based on your AI

女友的建议。这就是我通常大部分购物的原因。

girlfriend's That's why I usually make most of my purchases.

是的。是的。你的AI女友就是你的造型师。那会是个不错的T恤设计。

Yes. Yes. Your AI girlfriend is your stylist. That's that'd be a good T shirt for

我的AI女友为我挑选了这件。

My AI girlfriend picked this out for me.

那会很棒。但你看，仍然需要有一件T恤存放在仓库，需要装进信封，需要发货给你。需要支付流程，还需要防范支付欺诈。所以我认为无论交易由谁发起，商业基础设施都不会消失。

That'd be great. So, you know, so but, you know, there still needs to be a T shirt, which needs to be in a warehouse, which needs to go in an envelope, which needs to be shipped to you. There needs to be a payment process. There needs to be fraud around that payment. So I think the commerce infrastructure is not gonna go away regardless of who initiates that transaction.

而且，不论你是通过Teespring购买，还是通过AI女友聊天机器人，或是Gap品牌，这些交易都会发生。我不认为...我认为前端会有些颠覆性变化，更多影响客户获取和店铺前端，但商业基础设施只会持续增长。在可预见的未来，我看不出电子商务有任何放缓的迹象。

And, you know, whether you're, you know, getting that t shirt on Teespring versus your AI girlfriend chatbot versus, you know, the Gap, it's all gonna it's all gonna happen. So I don't I don't I think it's it's somewhat disruptive on the front end, more to customer acquisition and and the front end of stores, but I think the commerce infrastructure is only gonna continue to grow. And I I I I don't see any way that e commerce is going to slow down in any foreseeable time.

对，所以界面可能是聊天机器人，但所有环节仍可被管理。

Right, so the interface might be a chat bot, but everything could be managed.

所有环节都可管理。我是说，你仍然需要...

Everything could managed. Mean, you're still going

系统支持。

to system.

是的，你还在继续，所有这些交易处理和欺诈预防的小流程，你知道它们如何运作，以及是否有退货？比如你说，嘿，你和你的AI女友分手了，不想要那件T恤了，能退货吗？除了前端商店，还有很多事情需要处理。我认为Shopify自从我们参与A轮投资以来，无论是独立开发还是通过其生态系统，已经构建了许多难以复制的功能。

Yes, you're still going again, just all those little pieces of flows of transaction processing and fraud prevention and, you know, how where that goes, and and is there a return? And if you say, hey. You break up with your AI girlfriend and you don't want that t shirt anymore, can you return it? There's all there's a a lot of things that have to happen besides just the front end store. I think Shopify, since we invested in the series a, has built out whether individually or through their ecosystem, all kinds of things that are very hard to replicate.

好的。最后，Airbnb。

Okay. And then finally, Airbnb.

好的。

Okay.

纽约决定禁止Airbnb，这是市政史上最大的乌龙球，还是接近于此的事情？

Is New York is New York's decision to ban Airbnb the greatest own goal in municipal history or something close to it?

你所说的最大乌龙球是什么意思？

What do you mean the greatest own goal?

我觉得这简直是搬起石头砸自己的脚。

I thought just a terrible shoot yourself. There

有一个

was a

好引用。我让我让我

great quote. I let me let me

自己进球，乌龙球。这是个乌龙球。就像，当

Own well, own goal. It's own goal. There was a like, when

你把球踢进自己的球门

you kick it into your own neck

在足球比赛中。听说

in soccer. Heard a

昨天从喷气机队球员那里听到一句话，他说，其他球队是自摆乌龙。而我们则是自取灭亡。

quote from a from a Jets player yesterday where he's like, other teams shoot themselves in the foot Yeah. And then we shoot ourselves in the head.

对。对。对。抱歉。对。

Yeah. Yeah. Yeah. Sorry. Yeah.

对。很好。那么好吧。我刚才没听清。我再问一遍问题。我会再回应一下。

Yeah. Great. So so alright. I I just didn't catch So I'll ask a question again. I'll respond to that again.

还行吧。你投资了爱彼迎。

Was okay. You invested in Airbnb.

是的。

Yes.

纽约市禁止爱彼迎的决定是否堪称市政历史上最严重的灾难之一？

Was New York City's decision to ban Airbnb one of the greatest disasters in municipal government history?

没错。我是说，这绝对是个乌龙球。如果我们想用那个框架来看——你懂的——你希望建立一个充满活力的生态系统，允许自由贸易，让人们有地方住，同时尽量减少监管套利。如果你希望纽约对商务旅客、旅游业等一切活动保持开放，就不该让酒店业通过监管手段垄断市场。所以我觉得这很愚蠢。虽然很多大城市都在试探性限制，但我认为长远来看，理智终将占上风，大家会做出正确选择。

Yeah. I mean, it was definitely an own goal. If we wanna go use that format that, you know, you wanna have a great vibrant ecosystem that allows free trade, allows people to stay in places, but and you want very little regulatory capture if you want a fervent place for you want New York to be open for traveling business people, for tourism, for everything that happens, and you don't want the regulatory capture from the hospitality and hotel industry. So I think that's it it it was silly. I think a lot of the large municipalities have played with it, but, I think in in the you hope in the long run, cooler heads prevail, and everybody winds up doing the right thing.

我理解人们担心租金过高，不希望住宅物业被改造成酒店，但总要有个平衡点。直接全面禁止的结果，就是把纽约原本能负担得起的住宿——比如招待亲友的住处——变成了现在700美元一晚的天价酒店，这简直让人抓狂。

I understand the concerns that the rents might be too high and you don't wanna have residential properties being converted into hotels, but there has to be a balance. And the fact that it just got banned, you know, effectively turned a hotel stay in New York City from something that was affordable. So if you had guests, for instance, into something that's now $700 a night, and that drives me nuts.

而且人们本应解决根本问题。你说得对，纽约存在住房危机，也存在酒店每间可售房收入问题，必须兼顾两者。本应通过激励机制来推动解决方案。

And it's not like people are they should fix the underlying problem. They're they're you're right. There's a housing issue in New York City. There's also a hotel RevPAR issue, so you need to be able to do both. You hope that by providing incentive, you could get people to do that.

无论是放宽住房建设许可的监管限制来增加保障性住房供给，还是降低各类住宅物业的开发门槛，这些才应该是政策目标，而不是搞监管垄断。

And whether it's incentive that, hey, we're limiting the regulatory boundaries to get housing permits to be able to build more housing, especially affordable housing, or you're doing things to open up to make it easier for people to build, you know, any any type of residential properties here. That should have been the goal and not trying to not trying to do regulatory capture.

好的，Rick。你主持的First Mark有个播客。能聊聊如果听众对我们今天的对话感兴趣，他们可以在哪里关注你或收听播客吗？当然，大声说出来吧。

Alright, Rick. You have First Mark has a podcast. Do you wanna talk a little bit about if people are interested in our conversation today where they can follow you or the podcast? Sure. Shout it out.

我们制作了多个不同主题的播客。你可以直接关注我，我的X平台账号很简单，就叫@Rick。

So we do a bunch of different podcasts. You know, I you can follow me. Simple x address of just at Rick.

@Rick。

At Rick.

如果你想...什么时候...

If you wanna When did

你什么时候注册的这个账号？

you get that?

这是个很长的故事，可能得留到我们对话的第二章再讲。总之我在X平台的账号是@Rick，你可以在那里找到我。实际上我的推文内容非常清晰且有针对性，主要是关于我对市场动态的观察和分析，希望能更清晰地传达这些见解。

It's a long, long story. That's probably for phase for chapter two of our conversation. So I'm at Rick on Twitter x. You can find me there. I I actually, you know, have a very a very clean and deliberate Twitter largely about, you know, what I think is going on in the markets and what, you know, what we're seeing and hopefully to be more clear about that.

另外，First Mark旗下有一系列播客节目，你可以在X和Instagram上关注@FirstMarkCap。我的合伙人Matt还主持一档大型播客《数据驱动》，探讨他所谓的'MAD领域'——机器学习、人工智能和数据，这档节目在过去十年里始终处于AI讨论的前沿，邀请了许多顶尖AI思想家。我们制作了大量关于数据、金融科技等内容，甚至包括《OK Computer》这档分析私募市场现状的节目，它属于风险逆转生态系统的一部分。你可以在这些平台找到我们，也可以在CNBC电视台Scott Wapner主持的《收盘钟声》节目中看到我。

And then, you know, First Mark, we have a a full parade of podcasts you should follow at First Mark Cap on Twitter and Instagram, but also my partner, Matt, has a very large podcast called Data Driven, which talks about what he calls the mad landscape, machine learning, artificial intelligence, and data, and that's really been on the forefront of AI with some of the best thinkers in AI on over the last actual decade. So I think we're we produce a lot of content around data, around financial technologies, around a lot of things we do, even Okay Computer. That's part of the risk reversal ecosystem about what the state of the private markets are. So find me any of those places as well as on on our friend Scott Wopner's closing bell on CNBC.

你即将前往的地方，我们会帮你准备就绪。

Which you're about to go to, so we'll get you to set.

瑞克，

Rick,

很高兴见到你。感谢你参加节目。

great to see you. Thanks for coming on the show.

这次太棒了。非常感谢。我很乐意随时再来。

This was awesome. Thank you very much. I'm happy to be back anytime.

当然。好的。我们得听听瑞克的故事。所以我们会邀请你回来聊这个以及其他更多内容。好了，各位。

Definitely. Alright. We'll have to get the story of at Rick. So we will have you back for that and much more. Alright, everybody.

非常感谢大家的观看，我们下次在《大科技播客》再见。

Thank you so much for watching, and we'll see you next time on big technology podcast.