OpenAI的音乐竞赛，蓬勃发展的二级市场内幕，AI助力心理健康护理 | 2025年10月27日

欢迎各位收看资讯频道TI TV。我是科里·温伯格，今天代替阿卡什·帕特里查主持节目。今天是10月27日星期一。TI TV位于加利福尼亚州——金州，未来几天将如同世界大赛般精彩。此刻我正在洛杉矶家中为您播报，距离道奇体育场仅三英里。

Welcome everyone to the information's TI TV. My name is Corey Weinberg. I'm in for Akash Patricha, and it is Monday, October 27. TI TV is in California, the golden state, the next few days, just like the World Series. I'm coming to you from my home in Los Angeles, just three miles from Dodger Stadium.

明天阿卡什将在洛杉矶Adobe MAX大会现场主持节目。周三则由娜塔莎·马斯卡雷尼奥斯在纳帕谷举行的'科技、媒体与金融领域女性峰会'上接棒。展望本周，我们将迎来科技新闻三重奏：今天起英伟达GTC大会在华盛顿特区开幕；

Tomorrow, Akash will host from the Adobe MAX conference in LA. And on Wednesday, Natasha Mascarenhos will take the mic from Napa Valley at the Information's Women in Tech, Media and Finance conference. Looking ahead, we have a trifecta, a big tech news event this week. It's going to be busy. We have NVIDIA's GTC conference in Washington DC starting today.

明天开始'科技七巨头'中的多数将陆续发布财报；此外还有一家科技企业即将IPO——商旅初创公司Nivan预计本周将在纽约证券交易所敲钟。这些事件的深度分析将在后续节目中呈现。今天我们先聚焦我最钟爱的主题——私募市场。

Most of the rest of the magnificent seven reports earnings this week starting tomorrow. And we will have a tech IPO. Corporate travel startup, Nivan, is likely to ring the bell at the New York Stock Exchange later this week. We'll have plenty of analysis on those stories later in the week. Today, we're sticking with the private markets, my favorite subject.

我们将探讨风投领域正在兴起的基金集中度争议，与记者们聊聊OpenAI进军AI音乐生成领域的动向（这应该会很有趣），并专访心理健康独角兽Spring Health的CEO。不过首先让我们进入首个环节——关于不断壮大的初创企业二级市场。

We'll discuss a debate on fund concentration that is cropping up at venture capital. We'll talk with our reporters about OpenAI's push into AI generated music. That'll be interesting. And I'll interview the CEO of Mental Health Unicorn Spring Health. But let's first get to our first guests and our first segment on the growing startup secondary market.

现场有两位嘉宾：密切追踪二级市场的数据服务商Caplight的CEO哈维尔·阿瓦洛斯，以及本月初刚同意以最高9.65亿美元价格将公司出售给高盛的Industry Ventures创始人兼CEO汉斯·威尔登斯。哈维尔、汉斯，欢迎来到TI TV。

We have Javier Avalos, the CEO of Caplight, one of the data providers that tracks the secondary market most closely. And we have Hans Wildens, CEO and founder of Industry Ventures, which just agreed earlier this month to sell itself to Goldman Sachs in a deal that could be worth up to $965,000,000 Javier and Hans, welcome to TI TV.

谢谢科里，很高兴见到你。

Thanks, Corey. Good to see you.

汉斯，你是节目常客了。据我所知贵公司成立约25年，高盛CEO大卫·所罗门特别提到你们是二级风投领域的先驱——这正是收购原因。能否为我们勾勒下背景？当前哪些因素使得二级投资对风投生态变得比以往更重要？

Hans, you're a returning guest to the program. Your firm is, I think, about twenty five years old, and Goldman's CEO David Solomon pointed to the fact that industry was a pioneer in secondary venture investing as a reason to want to wanna pick you guys up. Set the backdrop for us here. What what's happening now that makes secondary investing important like, more important to the venture capital ecosystem?

再次感谢邀请我参加。非常感谢。是的，我们在二级市场已经深耕二十五年，最初是在互联网泡沫破裂时起步，帮助构建了风险投资和科技增长的二级市场。高盛之所以有兴趣与我们深化合作并收购我们公司，其中一个原因就是我们二十多年来在市场发展过程中始终与他们并肩同行。我认为我们现在已经到达了一个转折点。

Thanks again for having me on. Appreciate it. So yeah, we've we've been pioneering the secondary market for twenty five years, started originally when the dot com collapse happened and help build a secondary market for for venture and tech growth. And, you know, one of the reasons why Goldman was interested in partnering more deeply with us and acquiring our firm is that we work with them along this whole journey since over twenty years as the markets developed. And I think we've reached an inflection point here.

他们也认同这个市场规模庞大、持续增长且不会消失。这是市场中的结构性组成部分——风险基金、有限合伙人、企业投资者、对冲基金、共同基金都通过它获得流动性，就像高盛多年前开创IPO市场帮助股东融资和获取流动性一样。我们公司正是开创二级市场流动性的先驱之一。

And they would agree that the market, it is large, growing, and not going away. It's a structural component in the market of how venture funds, limited partners, corporate investors, hedge funds, mutual funds get liquidity, similar to how Goldman pioneered the IPO market in terms of helping raise capital and getting liquidity for shareholders a long time ago. Our firm was one of the one of the pioneers pioneering the secondary market to generate liquidity.

这个市场显然在增长。哈维尔，你的公司Caplight上周报告显示，9月份二级市场交易量增长了29%，今年迄今交易量增长了约50%。推动增长的因素是什么？

And that market is obviously growing. Javier, Caplight, your firm, reported last week that it saw, I think, a 29% increase transactions in the secondary market in September, and and trading volume was up by about half so far this year. What's driving the growth?

没错科里。今年对VC二级市场来说确实是个大年。正如你提到的，交易量同比去年增长了50%左右。关于增长驱动力，我们认为主要有三个因素。

Yeah. That that's right, Corey. It's been a a huge year so far for the VC secondaries market. And as you mentioned, about 50% up on a volume basis from where we are this year versus where we were last year same time. As far as, like, what's driving the growth, we really see three things.

首先是准入机会。AI这个新事物迅速改变了VC格局——全球只有少数基金能获得纯粹的AI投资机会。二级市场因此成为新渠道，即使不是全球前十的风投机构，也能通过这些平台投资热门AI公司。随之而来的是融资轮次规模越来越大，为解决风投机构可能没有数十亿美元闲置资金但仍想参与大型AI融资的问题，我们看到SPV交易激增，大量资金通过特殊目的载体涌入这些超额认购的融资轮。

I mean, one is access. This little thing called AI has has very quickly, like, shifted the VC paradigm, where there are only a few funds in the world that can get, like, pure a pure AI exposure. And so the secondary market has popped up as, like, one place where you don't have to be one of like the top 10 venture funds in the world. You can still go and find an avenue to get exposure to some of these really hot AI companies. On the back of that, rounds are getting larger and larger As a solution for venture funds potentially not having billions of dollars of dry powder just sitting around, but wanting to participate in some of these massive AI rounds, we're seeing an increase in SPV transactions with a lot of money flowing through special purpose vehicles to get into some of these severely oversubscribed funding rounds.

其次是估值高涨。经历2021至2022年风投调整后，我们已达到新峰值。早期投资者将二级市场视为获取DPI回报的机会，同时维持VC生态系统的资金流动。

And then valuations are high. We've hit a new peak following the 2021 to 2022 venture correction. With that peak comes folks who were invested early, who are seeing the venture secondary market as an opportunity to get some DPI return capital to their investors and keep the VC ecosystem flowing.

那么是否可以认为相对沉寂的IPO市场对你们两家业务都有利？这种情况如何助推你们看到的二级市场增长？

So is it fair to say that the relatively dormant IPO market has been pretty good for both of your businesses? Like like, how how is that sort of feeding into the growth that you're seeing in the secondary market?

好的，我很乐意开始这个话题。我认为存在一个普遍的误解，即当科技IPO市场火热时，这对风险投资二级市场是负面影响。实际上恰恰相反。当出现活跃的IPO发行窗口时，往往正是我们看到风险投资二级市场买方资金增加的时期。

Yeah. Happy to start this one off. I think there's a common misconception that when the IPO the tech IPO market is hot, that's a negative for the venture secondaries market. It's actually the opposite. When there is an active IPO offering, IPO window, that tends to be when we see additional dollars coming in on the buy side of the venture secondaries market.

我们在2021年周期中就看到了这种现象，今年再次见证。我认为只要不出现所有独角兽企业同时退出的极端情况（这种情况本身就不合理），就能看到积极的IPO活动与强劲的二级市场环境呈正相关。

We saw it in the 2021 cycle. We're seeing it again this year. I think as long as there's not a situation where every unicorn hits the exit at the exact same time, which it doesn't make sense for that to happen, you can see positive IPO activity correlated with like strong secondary market environment.

我想深入探讨下这个市场的实际运作机制。汉斯，你如何看待二级市场的成熟过程？它目前处于什么阶段？还需要在哪些方面发展？众所周知，风险投资在构建二级市场基础设施方面一直落后于私募股权行业，过去几年你观察到哪些标志这个成熟过程正在发生的信号？

And I wanna get a little bit into into into how this market actually functions. Hans, how have you seen sort of the secondary market mature? Where does it, how do they still need to go from here? You know, venture has generally been behind, I think the private equity industry and sort of building up the infrastructure around secondaries, sort of what are you seeing in the last couple of years that signals maybe that maturation process is happening?

确实，这个观点很到位。纵观风险投资和成长型投资领域，在流动性获取的成熟度方面确实滞后于私募股权。坦白说，我们一直在追赶私募股权市场的发展步伐，这包括房地产基础设施、信贷等另类投资领域——虽然主体还是收购业务。就我们市场而言，正在快速崛起的新领域是延续基金。

Yeah, I think that's a really good point. When you look at venture and growth, it's been delayed from private equity development and sophistication in terms of getting liquidity. And so we're paying, we've always played catch up, honestly, to the private equity markets. That includes, you know, real estate infrastructure, credit, kind of alternative alternatives in, you know, the bulk of it's, you know, buyout. But in terms of what we're seeing in our market, the new kind of area of the market that's growing rapidly is continuation funds.

这个市场领域大约十年前在收购市场开始快速增长，而我们直到最近两三年才感受到这种增长势头。你会看到越来越多原本被收购市场接受的流动性结构正在应用于风险投资市场。延续基金本质上是对原有风险基金或投资组合进行资本重组或结构调整，以延长对投资组合公司的管理期限和资金支持。去年我们市场就完成了近十笔这类交易，今年预计会达到二十多笔，明年可能三十笔。从市场规模角度看，最值得注意的是这些交易体量都非常庞大，因为涉及的是传统风险基金持有的整个投资组合——你知道的，这些基金通常在多家公司持有大量股权。

This is an area of the market that grew rapidly starting about ten years ago in in the buyout market, and we just started feeling that growth in the last couple years. And and so what you'll see is more and more liquidity structures coming from the buyout market that were already kind of accepted being applied into the venture market. And so continuation fund, if you don't know what it is, it's a fund that it's basically an old venture fund or a portfolio of investments that is, you know, recapped or restructured to continue the management and funding of the portfolio companies and holding them longer. And so, you know, we've we saw almost a dozen of these last year, in our market, we'll see probably two dozen this year, maybe three dozen next year. And the thing that's really interesting about them from a growth perspective, in terms of our market sizing, is that these transactions have been are very large, because it's a whole portfolio of traditional venture fund ownership, which you know, as you as you know, has large ownership stakes across a number of companies.

因此交易规模变得越来越大，过去几年我们目睹了这个趋势，并认为会持续下去。这将推动整体潜在市场规模的进一步增长。

And so the transaction sizes are larger. So we're seeing transaction sizes become larger and larger over the last few years, and we think that'll continue. So that'll drive more market growth in terms of total TAM.

有限合伙人喜欢延续基金这种载体吗？他们会投票支持或签署参与这类投资吗？

Do LPs like continuation vehicles? Like, are they voting for these, or are they they, you know, signing up for these?

是的。我认为收购市场曾经历过LP们不太清楚的阶段，对吧？这对收购市场是好事还是坏事？最终他们达成共识这是好事，因为它让LP们可以选择是否变现，也让GP们能选择继续持有投资以最大化价值。

Yeah. I think the the buyout market had that period of time where LPs didn't really know, right? Is this a good thing or bad thing for the buyout market? And what they all settled on was it was a good thing because it allow it gives LPs an option to take liquidity or not take liquidity. It gives the GPs an option to continue holding the investments to maximize the value.

而且作为风投基金或投资组合中的LP，你并非被迫出售给续存基金，你可以勾选选项表示愿意转入。因此你不会受影响，能继续保持所有权。你只是从原基金转入续存基金，或原基金直接变成续存基金。续存基金的法律架构有几种不同形式，但你可以持续运作。所以LP们最终认可这种模式，因为他们能自主选择是否变现。

And, you know, as an LP in a venture fund or in a portfolio, you're not forced to sell into the continuation fund, you can check a box and say, I'll roll into it. And therefore, you're not affected, and you can continue your ownership. You just switch from the fund that you're in to the continuation fund or the fund you're in becomes the continuation fund. So there's a couple different structures of how you structure the actual continuation fund legal process, but you can just keep going. And so LPs have settled on the fact that they like it because they can take liquidity or not take liquidity.

他们可以选择部分变现、全部变现或完全不变现。说实话，这正是维持市场良好运转必须提供的选择。我想快速谈谈这个市场中的一些矛盾点，比如...

They can take partial liquidity, full liquidity, or not take any. And that's honestly what you need to offer folks to have a good fair functioning market. I quickly just wanna hit on some of the tensions in this market. You know, some of

最热门的AI公司通常不希望大量场外交易绕过他们进行。他们会拒绝提交过来的潜在交易，或行使优先认购权。今天有个关于Anderol的案例——他们在二级市场有多火爆？其COO被我引用说：'任何声称能提供该公司年度二手股票的人，很可能在歪曲事实，甚至涉嫌欺诈买家，更别提收取层层离谱费用了'。他说得有道理吗？哈维尔，这些公司与市场之间存在哪些矛盾点？买家又该如何应对？

the hottest AI companies often don't want that many secondary trades happening without their involvement. They'll they'll reject potential trades brought to them or or exercise their right of first refusal. I had a story today about how Anderol is how hot they were in the sec how hot they are in the secondary market, and their COO, I quoted as saying, anyone claiming they can offer annual secondary shares is probably misrepresenting the facts and maybe defrauding buyers to say nothing of charging layers of egregious fees. Does he have a case? Like, what's the what's the what are the tensions that that lie here with some of these companies, Javier, and and how are buyers navigating it?

是的，我认为他说得没错。我觉得你的报道写得很好，建议大家阅读。这里各利益方都有自己要解决的问题。对公司而言，他们要确保资本表上的投资者符合质量要求，同时避免公司运营受干扰，让这些流程能相对有序地进行。

Yeah. I think I think he does have a case, and I I thought your story was really well written. I encourage folks to to read it. The the stakeholders here all have their own thing that they're solving for. For the companies, that's making sure that investors on their cap table are of a quality that they care about, and it's also preventing distraction from operating the company and and therefore sort of running these processes in so somewhat orderly fashion.

正因如此，任何可能导致公司失控的临时交易都会成为极其敏感的痛点，这很合理。归根结底，这个市场存在的意义是为早期风险投资者提供流动性，让他们能回笼资金维持整个生态系统运转。二级市场需要像减压阀那样，在流动性成为刚需时发挥作用，但必须以相对有序的方式进行。因此长期来看，会有更多公司建立系统化的流动性方案，比如以更可控方式开放特殊目的载体进行联合投资，让公司能审核每个SPV的参与者。

I think because of that, any ad hoc trading of this stuff or or any trading that could cause the company to lose control is automatically gonna be, like, a really sensitive point for the company, and I think that that makes sense. At the end of the day, this market exists to get liquidity for some of the earliest investors, the risk takers that are kind of backing these companies as a solution for them to sort of return cash to their investors to keep the whole ecosystem going. The secondary market needs to exist as like a release valve for when pressure builds up, where liquidity is a must have, but it needs to be done in somewhat orderly fashion. And so I think over time, you see more companies leaning into like more systematic programs for offering liquidity. You see companies that are open to doing special purpose vehicles for syndication, but in a much more controlled manner where the company is getting the ability to kinda okay participants in each SPV.

如果未来要持续进行交易，会有公司选择加入这种模式。不过要说明的是，目前我们还没有这样的机制——私营公司还没有类似纳斯达克或纽交所的平台，但我认为这将来会出现。

And if there is gonna be ongoing continuous trading, you have companies that are sort of opting in for that. And to be clear, we don't we don't have that yet. Like, there is no there is no Nasdaq or or New York Stock Exchange for private companies just yet, but I do see that happening in the in the in the future.

嗯，这里还有很多值得深入探讨的内容。哈维尔和汉斯，感谢你们的参与。这是我最喜欢的话题之一，所以非常感谢这次对话。好的，在桑德霍尔路和旧金山南公园的玻璃墙会议室里，风险投资家及其有限合伙人支持者正在讨论这个基本问题。

Well, there'll be a lot more to dig into there. Javier and Hans, thanks for joining me. This is one of my favorite topics, so so thanks for the conversation. Alright. Well, in glass walled boardrooms across Sandhole Road and in San Francisco's South Park, venture capitalists and their limited partner backers are debating the basic question.

我们是什么样的公司？有些公司追求规模，积累大量资金并将其分散投资于众多初创企业。另一些则坚守风险投资的小众本源，进行更为集中的押注。思考并撰写这一辩论的风险投资人芬恩·巴恩斯，在First Round Capital长期任职后，与人共同创立了总部位于旧金山的早期风投公司The General Partnership。他上周写了一篇题为《Alpha VC与Beta VC：风险投资的两种对立逻辑》的博文。

What kind of firm are we? Some firms are in search of scale, amassing a huge amount of capital that they spread out over a wide set of startups. Others are sticking to the more boutique roots of venture capital and making more concentrated bets. One VC thinking and writing about this debate is Finn Barnes, who co founded the San Francisco based early stage VC firm, The General Partnership, after a long tenure at First Round Capital. He wrote a blog post last week called Alpha VC versus Beta VC, the two competing logics of venture capital.

芬恩，欢迎来到TI TV。

Finn, welcome to TI TV.

嘿，科里。很高兴见到你。谢谢邀请。

Hey, Corey. Good to see you. Thanks for having me.

当然，当然。嗯，谢谢...谢谢你能来。你是...你是从你在旧金山的办公室连线吗？

So, yeah. Of course. Yeah. Thanks thanks for thanks for coming in. Are you you you dialing in from your your SF office?

是啊，旧金山办公室的最后一堵墙？好的，很好。

Yeah. The last wall of the SF office? Okay. Good. Good.

最后一堵墙。

Last wall.

那么，什么是阿尔法风投，什么是贝塔风投？为什么现在会出现这场辩论？

So so what is an alpha VC and what's a beta VC? And and why is this debate happening now?

是的。我最近和几位创始人聊天，他们正在考虑下一轮融资中要建立怎样的合作关系。他们试图比较手头的不同投资方案，思考要形成何种伙伴关系。当谈到这些投资方的动机和机构运作方式时，我明显感觉到随着市场不断成熟——就像我们刚才在前一个环节讨论的——私募市场正在涌现各种新型金融产品。似乎出现了分化：有些机构追求市场覆盖率，而另一些则专注于寻找市场中具有特殊特质的创始人，相信这样能获得超额回报。

Yes. I was actually talking with a couple founders who were trying to decide how to think about the partnership they would form in their next round of financing. And they were trying to figure out the difference between the offers that they had, the type of partnership they would form, And they started talking about the incentives of those partners, the way the firms operated. And it just became clear to me that as the market continues to mature, and as you're just talking about with the previous segment, you know, we're sort of seeing all these new financial products get created in the private markets. It feels like there's a bifurcation and that certain firms are sort of chasing market coverage and other firms are chasing very specific sort of idiosyncratic founders within a market believing they can outperform.

因此在我看来，这种差异可以简单分为两类：一类人追求阿尔法收益，另一类则寻求市场敞口，主要提供贝塔收益——也就是让投资者广泛接触科技私募市场的机会。

And so I think the the difference to me became sort of a a pretty simple split between people seeking alpha and people looking for market exposure and generally providing access to beta, meaning access to the broader technology private markets to their to their investors.

明白了。所以你是从金融术语角度说的阿尔法和贝塔？好的。

Right. Okay. So you mean alpha and beta from like a a financial term perspective? Yeah. Okay.

那么什么是...不

And so what are No

完全没有评判的意思。稍等一分钟。好的。

judgment whatsoever. Just just a minute. Okay.

我...我可以补充一下。你应该是站在阿尔法这边的。我绝对没有任何评判。不过我想稍后详细聊聊你们公司。

I yeah. I can add someone. You're on the side of the alpha, I think. So I no no judgment for sure. Well, I wanna go a little bit into your firm in a second.

不过首先，你能举些更广泛的例子吗？就是风投们在市场上普遍讨论的那些。比如这两种投资方式的案例。可能公司们不会明确承认偏向某一种，但你知道，有哪些我们能辨认出来的典型例子？

But but first, what are some examples you see more broadly, you know, sort of around across the market that VCs are talking about? Like, examples of these two approaches. Maybe firms would would would not admit to taking necessarily just one versus the other, but, you know, sort of what are what are examples we might recognize here?

是的。我认为这关乎运营风格。小型合伙机构往往呈现头重脚轻的结构，资深合伙人仍非常活跃，相对于基金规模进行较大额投资，但标的公司数量较少，与这些企业开展更深度的合作并建立长期伙伴关系，这可以说是阿尔法策略的核心。不论处于哪个阶段，这种策略都体现在时间和资本相对于基金规模的集中配置上——毕竟每人每天都只有24小时。而在贝塔策略这边，你会看到更多金字塔结构，随着发展，这类机构会越来越像华尔街的大型资产管理公司。

Yeah. So I think it's I think it's operating style. I think you look to smaller partnerships tending to be top heavy, you know, senior partners who are still very active, making larger investments relative to fund size into a smaller number of companies, working more deeply with those companies and forming long term partnerships with them, I think, is is sort of the alpha approach. And regardless of stage, I think it really plays out in terms of concentration of of time and and capital relative to fund size and the the twenty four hours that everyone has in a day. And then I think on the on the beta side, you tend to see more pyramid structures and and firms that as they continue to evolve, look increasingly similar to large asset managers from from Wall Street.

正如我们见证私募股权的演变过程，这种模式也在那里重现。你会看到大型团队、市场版图分析，以及覆盖式银行策略的忙碌运作——不仅要接触你真正看好的那家公司，还要接触另外10家符合市场版图中同一细分领域的公司加以筛选。同时你还会看到快速跟风现象。当某家头部贝塔风投通过投资认可某个领域的合法性后，其他贝塔机构就会迅速跟进同一领域或同类竞品公司。因为作为贝塔投资者，你承受不起错过市场驱动因素的风险，否则就会落后于基准。

And as we saw the the evolution of private equity, you sort of saw this play out there as well. And I think you you see large teams, you see market maps, and coverage coverage banking approaches hustling not just to meet the one company that you really believe in, but to meet the 10 others that you believe fit that same box in market map and then choosing amongst them. And you also see fast following. So if a leading beta VC firm designates a category as legitimate by investing in it, You'll see fast follows from many of the other beta firms into that same category or companies that are competing for that that same market because none of them can afford to if you're a beta investor, you can't afford to miss the drivers of the market, at least you fall behind the benchmark.

好的。那你在和创始人交流时会直接建议说'别拿大型多阶段机构的钱'吗？

Alright. Are you going as far as arguing to founders when you're talking to them? Like, hey, don't take money from a big multistage firm?

不，不是的。我认为创始人只需要清楚自己所做的选择。事实上你既能与贝塔风投建立出色合作，也能与阿尔法风投缔结优质伙伴关系。这完全取决于你希望如何建设公司，以及你认为自身在资源和策略方面最需要什么。

No. No. I I don't think I think founders should just know the choice they're making. And I think you can form fantastic partnerships with beta VCs, and you can form fantastic partnerships with alpha VCs. And it really depends how you wanna build your company and and where you think your your needs lie in terms of resources and and and approach.

观众可能不太了解你们机构。能否介绍一下GP的战略定位？

Viewers may may not totally know about your firm. You know, can you can you walk us through what the the strategy is at at the GP?

当然。我们自认为是以人才为核心的投资机构，这种基因源自创业生态中的人才板块。我的合伙人Dan Portillo曾是硅谷最优秀的人才合伙人之一，后来在Greylock离职后加入Sweat Equity Ventures开始投资。之后我们联手创立了GP，始终保持高度集中的投资组合策略，为创始人投入大量时间，并在人才方面提供全方位服务。

Sure. So we're we're at we'd say we're a talent centric firm, and that really comes from the DNA of the firm comes out of the talent side of the of the startup ecosystem. So my partner, Dan Portillo, one of the best talent partners in the Valley until he started investing at a firm called Sweat Equity Ventures after he came out of Greylock. And then we we joined up and and formed the GP. And and we maintain this approach of a very concentrated portfolio, tremendous amounts of time with founders and providing service across talent.

因此，我们通过招聘组建团队、开拓市场、协助创收，以及产品与工程部门共同打造支撑业务的技术基础。我们采用独特方式运作：不同于平台团队作为收费的共享资源池，我们将资源投入到建设者团队中，这些团队被派驻到企业中进行高情境深度协作，以专属模式开展工作。作为回报，我们以顾问形式获得普通股权授予。

So recruiting to build your team, go to market, helping you find revenue, and then product and engineering, helping build build the technology that underlies the business. And we do that in a unique way where rather than a pooled resource of a platform team that is paid for with fee, We invest resource into a team of builders that are deployed into these companies for high context, deep engagements where they're working in a dedicated basis. And in exchange for that, we earn common equity grants in the form of advisory.

好的。在我看来，要成为所谓的'阿尔法风投'并采取这种更集中的策略，你必须判断准确，否则会比采取市场覆盖或'贝塔策略'消耗更多资金。目前有哪些最佳证据能证明你们这种策略是有效的？有实证数据吗？有第一手证据吗？

Okay. So it strikes me that to be a so called alpha VC and to take this more concentrated approach, you better be right or you're gonna blow up a lot more capital than if you're taking kind of a market exposure or beta approach. What's the best evidence you have so far that your approach or an approach like yours is working? Is there empirical evidence? Is there firsthand evidence?

某种程度上我

Sort of what I

我觉得...我认为你是对的。当观察那些能将资本集中投入少数公司的机构时，我认为它们往往能带来最佳的倍数回报，可能IRR（内部收益率）也是如此。长期观察风投组合会发现，通常少数公司贡献了大部分回报，这是历史规律。因此，如果能选中这些公司并集中投资，相比广泛的市场表现指标，我认为可以取得超额收益。

feel like the I think I think you're correct. And and, you know, when you look at the the firms that are able to concentrate capital into a small number of companies, if you look, I think those will drive the best returns in terms of multiples and and probably IRR as well. When you look at VC portfolios over time, I think what you tend to see is there are few companies within any given portfolio that drive the majority of returns. That's been the historical precedent. And so if you can choose those companies and and pour capital into them versus a broad based market performance metric, I think you can outperform.

当然也可能错判并以惊人方式跑输市场。但如今投资者作为资本配置者选择风投时，'阿尔法风投'在其组合中的角色与'贝塔风投'截然不同。作为有限合伙人进行资本配置时，这个合作伙伴在你组合中的角色定位至关重要。

You also can miss and and underperform the market in dramatic ways. But I think when you think about the decision of an investor now, a capital allocator, to choose VC, the role of VC within their portfolio is quite different if that VC is is seeking alpha versus versus the beta VC. And so when you're an LP and you're choosing where to place your capital, the role of that partner in your portfolio really matters quite a bit.

最后我想探讨的是当前AI初创企业动态如何影响这种思维模式或差异化策略。因为最近从风投界听到的另一个观点是：判断AI初创企业收入的可持续性比以往任何时候都更难。当你重仓押注某个领域或初创企业时，这种不确定性会显得尤为突出。你们是如何思考这个问题的——比如部分AI收入几年后是否真的还能存在？

The last thing I wanted to hit here is how sort of the current dynamic around AI startups affects this mindset or this this different strategy. Because like another thing I hear from VCs these days, along with this debate, is how it's harder than ever to figure out if revenue in an AI startup is sticky or durable. And and I feel like when you go big and concentrated into a category or into a startup, that kind of uncertainty I would think would loom large. How does that you know, sort of how are you thinking through, you know, that question of of is is some of this, like, AI revenue actually gonna be there in a few years?

我认为...无论是阿尔法风投还是贝塔风投，在投资时都面临同样问题。当前市场大量资金集中在AI公司，但这些企业的客户群体正处于技术应用极早期阶段。收入缺乏粘性或质量不确定的原因在于：从客户视角看，目前阶段用户被鼓励无限试用各种工具且预算不受限。因此很难判断收入来源是持续稳定的需求，还是实验性使用。虽然市场泡沫比过去少，但类似情况在2010-2011年广告技术领域也出现过——当时每个大品牌都想尝试新型广告单元或DSP/ISP平台。

I I mean, think I think the both both Alpha VC and Beta VC face that that same question as you're as you're making your investments. Much of the market today is in these AI companies. And I think the customers for these companies were at a stage we're so unimaginably early in this technology and the adoption of this technology that I think the reason we see a lack of stickiness or uncertainty around the quality of revenue is because from a customer perspective, from a buyer perspective, we're still at a place where every user is told use as many tools as you want and you have unlimited budget. And so I think it's very hard to tell when looking at revenue if you have something that's sticky and durable or if you have something that's being used as part of an experiment. There are many fewer zeros involved in the market, but AdTech went through this, I think, in the 02/2011 time when every large brand wanted to find a new ad unit or try a new DSP or ISP.

我认为无论是AI编程、AI营销还是AI驱动的CRM系统等，我们都看到了相同的现象。每个人都可以尝试各种可能，但我认为真正的赢家尚未显现。

And I think we're seeing the same thing when it comes to whether it's AI coding, AI marketing, AI enabled CRM, etcetera. Everybody is allowed to try everything, and I think the the winners have really yet to emerge.

好的，芬恩，我们会持续关注。让我们每隔几个月或至少每隔几年回顾一次，看看这场关于alpha与beta VC的辩论如何发展。芬恩，非常感谢你的参与。GP的芬恩·巴恩斯，也很高兴见到你。

Well, Finn, we'll we'll check back. Let's check back every every few months or at least every few years to see how this debate is playing out on alpha versus beta b c VC. Finn, thanks so much for for joining us. Finn Barnes from GP. Good to see you too.

OpenAI的产品线已扩展至网页浏览器、购物推荐，当然还有视频生成应用Sora。现在他们正在研发音乐生成AI，让用户能从零开始创作歌曲。我们的记者之一艾琳·吴是周五那篇报道的幕后功臣，她现在和我一起。嗨，艾琳，欢迎来到节目。

Well, OpenAI's chorus of products has swelled to a web browser, shopping recommendations, of course, the video generation app, Sora. Now it's working on music generating AI that would allow users to create songs from scratch. One of our reporters, Erin Wu, was one of the journalism maestros behind that story on Friday, and she joins me now. Hey, Erin. Welcome to the show.

嗨，科里，非常感谢邀请我。

Hey, Corey. Thanks so much for having me.

当然，当然。那么OpenAI在音乐方面正在研发什么？你的报道具体讲了哪些内容？

Of course. Of course. So what is OpenAI working on for music? What what what are you is is your story talking about here?

是的。OpenAI正开始研发能生成音乐的AI。目前还很难判断他们的进展程度，但他们已开始收集所需的训练数据。他们正与茱莉亚音乐学院的学生合作，让学生们标注乐谱来生成训练数据。同时他们也开始探讨通过文本或音频提示生成音乐的可能性。

Yeah. So OpenAI is starting to work on AI that could generate music there. It's hard to tell, like, how far along they are towards this, but they've started working to collect the kind of training data that they would need to receive. So they're working with Juilliard students to have them annotate scores to produce training data. They're also starting to discuss the idea of being able to generate music from text or audio prompts.

这意味着比如你有一段音轨，然后想添加吉他或钢琴等乐器部分。

So that can mean something like having a track and then wanting to add guitar or add piano or something like that.

好的。我一直在用ChatGPT的语音功能，还特意测试了一下，比如让它给我播放一首歌会怎样？结果它确实做不到。所以这是OpenAI目前还无法实现的功能，对吗？

Okay. And, yeah, I've been using the ChatGPT voice, and I did test it out to see, like, would it actually play me a song if I asked it to? And and no, it did not. So this is this is not something OpenAI currently can do. Is that right?

没错。他们过去确实在音乐领域做过一些尝试，2019年发布过NewsNet模型，2020年又推出了Jukebox，但这两个模型目前都没有向用户开放。

Yeah. That's correct. So they've tried they've done some work on music in the past. They've released two models years ago prior to ChatGPT NewsNet in 2019 and Jukebox in 2020. Neither of those are currently available for users.

此外他们还有些语音转文本和文本转语音的模型，这些在音频生成方面有部分功能重叠，但现阶段还没有能根据提示词直接生成完整歌曲的产品。

And then they also have some speech to text and text to speech models that would have some overlapping capabilities because it's about generating audio, but they don't currently have something where you can give it a prompt and it'll spit out a full song.

听起来很有挑战性，不过视频生成似乎难度更大？为什么音乐领域对他们或其他AI模型来说会是个棘手的类别呢？

This sounds challenging, but like video almost sounds like more difficult in a way. Or like why is music potentially, you know, sort of a challenging category for them or any other, you know, sort of AI model?

确实。音乐领域其实有很多初创公司在专注研发，比如谷歌也有自己的音乐模型。目前最知名的音乐AI初创公司要数Suno和Udio。对谷歌这类公司来说，这个方向很有吸引力，因为他们一直在强调这项技术对营销人员和广告应用的潜力。

Yeah. So, I mean, music is something that has been a focus of a lot of startups. Like, Google also has a music model. So Suno and Udio are probably the best known startups for this. It's something that's exciting for exciting for Google, for example, because they have been really touting this ability for marketers and to be able to use it in advertising.

OpenAI最近也开始讨论类似方向，认为这个领域具有巨大的商业潜力。

And OpenAI has kind of started talking about a similar thing, the idea that it's something with a lot of commercial potential.

明白了。那你觉得这个功能未来会更偏向商业用户和营销人员，还是作为ChatGPT订阅用户的附加功能向个人开放？目前还不好说对吧？

Okay. So do you think this would put I guess we don't know yet whether this would be geared towards more commercial users or marketers or or individuals to kind of tack on to their ChatGPT subscription?

是的。我是说，虽然我们不清楚具体细节，但看看OpenAI迄今为止发布的产品类型，我认为可以稳妥地说，这适用于上述所有情况。

Yeah. Mean, looking at I think we don't know specifically, but looking at the kinds of products OpenAI has released so far, like, I think it's a safe bet to say that this would be for, like, all of the above.

好的。现在这肯定会引发大量争议，就像OpenAI许多产品发布时那样，随着他们逐步推进。如果能解决这个问题，初期发布时显然会有很多人——比如我居住的好莱坞或其它地方——对潜在的版权问题提出诸多质疑，比如他们的音乐被窃取或用于OpenAI音乐产品的训练中。他们会说，嘿。

Okay. Now this is sure to generate a ton of controversy, like a lot of OpenAI product launches as like, yeah, they move closer. Infant launches, if they're able to figure this out. Obviously, there's a lot of people around, yeah, you know, around where I live in Hollywood or or elsewhere that is going to take a lot of issues with the potential copyright with, you know, with their music being stolen or used in OpenAI's training for a music product. They're gonna be like, hey.

这听起来有点像我想出的旋律。你对音乐生成相关的版权问题有什么理解吗？

That sounds a little bit of like a melody I came up with. You know, what do you what do you have sort of an understanding of is is the kind of copyright issues around around sort of music generation?

没错。音乐产业非常强大，同时也以好讼著称。他们已经起诉了这个领域的两家领先初创公司Suno和Udio。本质上，这与其它类型的训练数据面临的情况类似。这些初创公司声称他们的行为属于合理使用。

Right. So, I mean, the music industry is very powerful and also famously very laudicious. And so they've already sued the two leading startups in the space, Suno and Udio. Essentially, and I mean, this is similar to what's going on with other kinds of training data. The startups are claiming that what they're doing is fair use.

唱片公司则认为这等同于盗窃。不过话说回来，唱片公司并非只采取纯诉讼策略。有报道称他们正在与一些从事该领域的初创公司进行谈判，比如Suno、Udio、谷歌等。根据我的报道，OpenAI已经开始向音乐产业示好。

The labels are contending that what it amounts to is theft. Like, that said, though, it's not that the labels are pursuing, like, a purely litigation based strategy. There's also been reporting that they're in talks with some of the startups working on this. So, like, Sunu, Udeo, Google. And my for my reporting, like, OpenAI has already started making advances to the music industry.

例如我们看到与Spotify的整合，用户可以通过ChatGPT点播歌曲。根据我的消息源，他们至少已与一家大型唱片公司展开谈判。显然OpenAI正处于交易达成模式。而从唱片公司的角度来看，他们也认为这种合作可能带来巨大价值。

So we saw the Spotify integration, for example, where users are able to use ChatGPT to cue songs. And from my reporting, they've also been in talks with at least one of the major music labels. And so OpenAI is obviously in a deal making mode. And I think the labels also, from their perspective, see a lot of value to the idea that, okay. Like, maybe there could be some kind of partnership.

也许AI可以帮助人们发现音乐，在这方面能发挥积极作用。所以各方确实都持一定开放态度。

Maybe AI is something that could be good for helping people discover music, for example. So there is definitely some openness on all sides.

谷歌的产品在这方面遇到过什么特别的问题吗？我本以为他们会更加规避风险或更注重法律层面，

Has Google run into any particular issues here with their product? I would think that they would be a little bit more risk averse or legally,

你知道的，某种程度上

you know, sort

比像OpenAI这样的公司更谨慎。

of cautious than someone like an OpenAI.

是的。目前没有公开信息显示谷歌在AI数据授权讨论中的具体进展，但谷歌通过YouTube与音乐行业已有合作关系，这对该产品非常重要。所以，他们已经建立了部分这样的关系，能够达成这些交易。

Yeah. So there's nothing that's publicly known about where Google is in its AI data licensing discussion specifically, but Google does have an existing relationship with the music industry through YouTube, which is very important to that product. And so, like, they have they have some of that relationship already built to be able to strike these deals.

好的。亚伦，非常感谢你加入我们。这是个非常有趣的独家新闻，感谢你来和我们讨论。谢谢。

All right. Well, Aaron, thanks so much for joining us. This was a fascinating scoop, thanks for coming on and talking us Thanks

谢谢邀请。

for having me.

我们正处在一个AI显然不仅重塑音乐，还改变我们生活方方面面的时代，包括健康管理。Spring Health就是其中一家公司，这家独角兽企业致力于通过数据和AI创造更个性化的体验来改变心理健康护理。它通过分析数千个信号，更快地为个人匹配合适的支持。现在和我连线的是Spring Health的联合创始人兼CEO April Ko。

Well, we're in an era where AI is obviously reshaping not just music, but every other part of our lives, including how we take care of our health. Spring Health is one company. It's a unicorn working to transform mental health care by using data and AI to create a more personalized experience. It's analyzing thousands of signals to match individuals with the right kind of support more quickly. And joining me now is the co founder and CEO of Spring Health, April Ko.

欢迎来到TI电视台。

Welcome to TI TV.

你好。非常感谢邀请我。

Hi. Thanks so much for having me.

感谢你的到来。你们公司自2016年成立至今，我认为你们的发展历程非常有趣，尤其是在当今世界背景下。能否为我们概述一下，大语言模型如何从广义上影响心理健康服务提供者？

Thanks for coming on. You guys have been around since about 2016. I feel like your your journey has been a really interesting one, especially now in the world we live in. So so set the scene for us. How are LLMs affecting mental health providers kind of kind of broadly?

这对像你们这样的公司产生了怎样的影响？

How is this affecting a company like yours?

好的。首先我想简单介绍一下我们公司及其业务。我们是一个全球综合性心理健康平台，致力于让用户轻松获得适合的心理健康服务。九年前在耶鲁大学，我与计算精神病学领域的全球顶尖专家共同创立了这家公司。他是人工智能与心理健康交叉领域的专家，事实上是首个通过机器学习技术实现从初始阶段就将患者与最适合其的治疗方案匹配的先行者。

Yeah. Well, first off, I just wanna quickly introduce, you know, the company and what we do. So we're a global comprehensive mental health platform that makes it really easy to get mental health care that works for you. And nine years ago at Yale, I started the company with the world's leading expert on computational psychiatry. So he was an expert in AI and mental health, he was actually the first to prove that we could match people to the right care for them from the start using machine learning.

时间快进到现在，我们已成为心理健康领域增长最快的企业之一。我们为全球2000多万用户提供服务，合作方包括微软、塔吉特等世界顶级雇主和医疗计划机构。我们为其员工及家属精准匹配最适合的心理健康服务方案，无论是心理辅导、治疗还是药物干预，都通过数据分析实现个性化推荐，并将他们与最匹配的专业服务提供者对接。

And so fast forward to now, you know, we're one of the fastest growing, if not fastest growing, mental health companies out there. We serve, you know, 20 plus million, lives with some of the world's largest employers and health plans like Microsoft and Target. And, you know, for for their employees and their family members, we match them to the right mental health care for them. So it might be coaching, might be therapy, it might be medication. We really use our data to to figure out what would be right, and then we match them to the right provider for them.

我们提供当日或次日达的线上线下服务。大语言模型彻底改变心理健康服务的潜力令我们非常振奋。几个世纪以来，心理健康服务虽有演进，但某些核心模式始终未变——比如传统的45-60分钟面谈咨询，两周后再次复诊的模式，无论是线下办公室还是远程诊疗，这种形式在过去几百年间基本保持不变。

And then we deliver same day, next day care virtually or in person. We're really excited about the potential for LLMs to change mental health care completely. You know, mental health care has evolved over the centuries, but really, in some ways, stayed largely the same. So, you know, you generally see a therapist for, let's say, forty five to sixty minutes, and you, you know, have a session with them, and then two weeks later, you're back in their office, or taking taking the session virtually. And, really, that format has stayed the same over the past few centuries.

现在有了大型语言模型，护理可以从局限于这些疗程内转变为持续性的。因此我们相信，除了在这些疗程中与治疗师交谈外，你还可以在疗程间隙与AI伴侣互动，使护理持续进行。这是我们今年早些时候推出的功能。我们对LLMs彻底改变心理健康护理格局的潜力感到非常非常兴奋。

Now with LLMs, the care can go from, you know, contained within these sessions to continuous. And so we believe that, in addition to talking to your therapist in these sessions, you can engage with an AI companion in between the sessions to make your care continuous. And that's something that we've launched earlier this year as well. So we're really, really excited for the potential for LLMs to change the landscape of mental health care completely.

回到你们如何运用AI来确定最佳干预措施或最适合人们使用的服务这个话题。你们关注哪些信号？哪些信号是你们用来判断的最强依据？

And going back to kind of how you you've been using AI to, you know, sort of determine what the best intervention is or or kind of what the best service is for for people to use. What signals are you looking at? What signals are are kind of the strongest that that you use to kind of determine that?

是的。创始研究确实表明，自填问卷数据本身就包含大量信号，能比精神科医生单独工作时更精准地为人们匹配适合的治疗方案。如今，Spring真正的差异化优势在于我们拥有一个统一平台——13,000名全球服务提供者、每位用户和客户都在这个平台上。我们从服务提供者、用户、疗程记录、治疗师笔记以及客户机构中收集数据，将所有信息整合在一个平台上。

Yeah. So, you know, the the founding research was really showing that self reported questionnaire data in and of itself had a lot of signal to to match people better to match people to the right treatment for them better than a psychiatrist working alone could. And now, you know, the the big differentiator behind Spring and what makes us truly different and unique is that we have a single platform where every provider so we have 13,000 global providers, every user, and every customer is all on the single platform. And we're gathering data from the providers and the users, from their sessions, from the provider's notes, from the customer's organization. We're bringing this all together on one platform.

每个数据点都包含信号，指导我们如何为人们匹配恰当的护理方案，如何在护理过程中实施治疗，以及如何推动他们实现最快的康复。这项技术令人惊叹，它使康复速度达到我们最强竞争对手的两倍，堪称当前最有效的心理健康护理方案。

And every single data point contains signal for how we match people to the right care for them, how we treat them through their course of care, and how we drive the fastest recovery possible for them. So this is incredible technology. It drives two times faster recovery for than our nearest competitor. And so really represents the the most effective mental health care out there.

你们最近发布了Vera MH——一个用于评估心理健康聊天机器人的新开源测试工具。它是如何运作的？AI在整体评估心理健康时起到什么作用？

And you recently announced the release of Vera MH, a new open source test for mental health chatbots. How does it work? How is AI used overall when evaluating someone's mental health?

没错。我们自始至终都是AI公司，正如我提到的。但当生成式AI出现时，我们全力拥抱这项技术，因为它的本质就是对话式的。而心理治疗的主要形式正是对话。

Yeah. So look. So we were an AI company from the start, as I mentioned. But when generative AI came on the scene, you know, we fully embraced it because it's it's conversational in nature. Therapy, the main modality is conversational.

因此LLMs彻底颠覆心理健康护理形式是必然的。但当我们开始用生成式AI创新时，我们秉持Spring Health一贯的科学优先理念，以临床验证为导向，将成员安全置于首位。结果发现，在开始生成式AI创新时，心理健康领域的AI几乎没有任何防护措施、监督或标准。这显然是个严重问题——要知道半数美国成年人表示会使用语言模型或聊天机器人来支持情绪或心理健康。

So it really made sense that LLMs would completely disrupt mental health care and the format of mental health care. But when we started to innovate with generative AI, we we approached, you know, the innovation like we approach everything at Spring Health with a science first mentality, with a mentality towards clinical validation, and an approach that puts member safety first. And what we, what we found was that there were virtually no guardrails or oversight or standards, around AI and mental health when we started innovating with generative AI. And so we said, you know, this is a huge problem. You know, half of US adults, say that they use LMs or chatbots to support their emotional or mental health.

目前对AI在心理健康领域的应用几乎没有任何监管。所以我们提出，我们需要设立防护栏。我们不能因为缺乏这些防护措施就放慢创新步伐。于是我们与专家委员会合作创建了这些防护机制，这个委员会由技术专家、临床医生和该领域的顶尖学者组成。

And there's virtually no oversight over over the use of these AIs for for mental health. And so we said, look. We need guardrails. We're not gonna slow down on innovation just because these guardrails don't exist. So we we we created these guardrails, in partnership with a, an expert AI council that we put together of technologists, clinicians, the the foremost academics on the topic.

我们召集了这些委员会，并提议共同制定一个基准。通过中立第三方进行管理，确保为AI在心理健康领域设立真正的标准。这就是我们上周刚发布的VeriMHC项目——心理健康领域伦理与责任AI验证体系。

We pulled these councils together, and we said, look. We, let's create a benchmark together. Let's govern it through a neutral third party, and let's let's make sure that we're setting real standards for AI and mental health. And so that's what we released just this past week. VeriMHC, it stands for validation of ethical and responsive responsible AI in mental health.

它的运作方式非常巧妙。作为开源项目，它旨在与整个行业协作共建。本质上这是个自动化系统，不需要进行冗长的学术分析来判断某个LLM或AI是否安全，而是直接测试目标LLM，用一个LLM模拟用户，再用另一个LLM评估对话过程。

And the way that it works is really neat. It's it's open source, so really meant to be built in collaboration with the rest of the industry. And and, essentially, it's automated. So instead of, you know, having to run a really long academic analysis on whether an LLM or AI is safe, it, you know, takes the takes the LLM in question, and it uses an LLM to mimic a user. And then it also uses an LLM to judge the conversation that's happening.

它能评估LLM在提供心理健康支持时的伦理和责任表现。这是我们与专家委员会合作开发的非常棒的系统，希望整个行业都能将其采纳为该领域的领先基准。

And it assesses the LLM's ability to ethically and responsibly, you know, deliver emotional and mental health support. So it's it's a really cool system that we've we've created in collaboration with this expert council, and we really hope that the entire industry embraces it as as the leading bench for the space.

显然你们在开发过程中投入了大量心血。最让你担忧的是什么？当人们转向LLMs来补充或直接提供心理健康服务时...我是说，那些主要模型提供商虽然发布了很多声明，表示正在调整系统以适应这类工作，但他们显然要处理太多事务，最终目标都是开发出最好的模型。

This is you've obviously put a lot of care and effort into developing this. What worry what worries you the most about how people are turning to LLMs to supplement or to actually provide mental health care? I mean, you you these are companies that are particularly the the main model providers that have put out a lot a lot of statements about how they're they're trying to tailor their systems to do a good job with this kind of this kind of work. But they're obviously they're trying to do a lot. They're all trying to develop the best model, period.

他们试图吸引客户，成为广泛用途的辅助服务。心理健康显然是种非常棘手的干预领域。作为关注这个问题的人，你如何看待现状？这些都是好的吗？请结合实际案例详细谈谈。

They're trying to attract customers and be sort of a helpful service for a broad range of uses. Mental health is a really, obviously difficult know sort of intervention to use. I mean how are you you know how are you thinking about this from a sort of someone who cares about this issue? Is this all good, mostly good? Walk me through some of that on a on a real real world basis.

我认为LLMs确实有潜力以积极方式彻底改变心理健康服务的可及性。在Swimming Health我们全力拥抱这项技术，它确实非常强大。但最让我夜不能寐的是：人们没意识到目前对这些工具在心理健康领域的使用完全没有监管措施。

Look. I think that, you know, LLMs have the potential to completely transform in a really positive way, mental health access, mental health care. And so we're fully embracing it at Swimming Health. And, you know, it's it's it's definitely very powerful. But I think that the thing that keeps me up at night is people don't understand that there are no regulations or oversight over the use of these tools for mental health specifically.

遗憾的是，许多年轻人、青少年以及美国成年人正专门为心理健康问题使用这些聊天机器人，他们认为只需向AI或LLM提问就能替代专业临床医师。但现实是，所有人都应该明白这些LLM更像是朋友——可能是知识渊博、非常聪明的朋友，但它们既没有专业资质，也未接受过同等伦理标准和临床标准的培训或问责。这正是让我夜不能寐的问题：人们根本意识不到这其中缺乏监管与规范。

And, unfortunately, a lot of youth, a lot of teens, a lot of US adults are engaging with these chatbots specifically for their mental health, and they think that it's enough to just prompt the AI or the LLNs to be like a licensed clinician. But the reality is, you know, everyone should know that these LLNs are kind of like you should think about them as, like, friends. They're probably like a friend that's very knowledgeable, very smart, but also not licensed, right, and not not trained or held accountable to the same ethics standards and clinical standards of mental health clinicians. And so that is that is for me the thing that keeps me up at night. People just do not understand that there is no oversight or regulation.

所有这些发展与创新都在闭门造车。因此我们通过RMH向行业发出强烈行动呼吁：让我们公开共同遵守的基准测试标准，协作制定这些规范，齐心协力确保我们的作为对社会和人类负责。

All this development and innovation is happening behind closed doors. And so that's why we have a huge call to action to the industry with RMH where we say, you know, let's open up the benchmarks that we're all holding ourselves accountable to. Let's co develop these standards together. And let's all work together to make sure that we're doing right by by society and humanity.

那么具体来说，你们正在推动哪些新型监管措施或干预方案？或者说，你们协助制定了哪些基本规则？

What's the one, you know, sort of sort of either new regulatory in a sort of inter you know, approach or intervention or, you know, sort of what is the, you know, sort of rules of the road that you help kind of get gets put in place here?

我希望所有人都能参与这个开放式基准测试。我们上周发布了这个开源框架设计，目前正在积极征集反馈。我们不仅需要语言模型开发方，更需要学术界领袖、临床专家共同提供真实意见。最终目标是让行业——包括心理健康服务和AI工具的采购方——能通过该基准测试，判断他们提供给用户、员工或会员的这些语言模型或聊天机器人是否安全。

I hope that everyone will contribute to this open benchmark. So this is open sourced by design. It is so we released the the the framework last week, and we're o we're actively soliciting feedback. So we want, you know, not only the LMs, but also academic leaders, clinical leaders, everyone to join us in giving us real feedback around this. And then ultimately, I want the industry, so, you know, procurers, buyers of mental health care or AI tools to use this benchmark to understand whether these LMs or chatbots that they're putting in front of their users or their employees or their members are safe.

这才是我们希望能为行业做出的贡献。我们希望这个基准测试能超越Spring Health的范畴——虽然我们共同创立了这个测试标准，共同组建了监督委员会，

So that is ultimately what we would like to contribute to the industry. And, you know, we want this benchmark to be bigger than Spring Health. You know? You know, we cofounded the this benchmark. We cofounded the council that oversees the benchmark.

但我们更希望竞争对手加入，需要整个行业参与其中，因为这已远不止是Spring需要面对的单一问题。

But we want competitors to join in. Know, You we want the entire industry engaged because this is this is a problem that's much bigger than than a problem specifically for spring.

好的April，非常感谢你的参与，这次对话让我受益匪浅。

Well, April, thanks so much for joining me. Really appreciate the conversation.

好的，谢谢。

Yeah. Thank you.

在节目结束前，我想提醒大家接下来一周内即将举行的精彩活动。TITV将亲临洛杉矶的Adobe MAX现场。10月28日周二，阿卡什·帕特里查将与Adobe高层及合作伙伴就人工智能与创意展开对话。次日，10月29日周三，在纳帕谷举行的WTF峰会（科技、媒体与金融领域女性峰会）绝对不容错过，《信息报》的娜塔莎·马斯卡雷尼亚斯将主持一场特别版的TI TV节目。

Before we end the show, I wanna remind you about some exciting events happening within the next week. TITV will be on the ground at Adobe MAX in LA. Akash Patricha will be talking to some key Adobe executives and partners about AI and creativity on Tuesday, October 28. And the day after that, you don't want to miss the information's WTF Summit, which is women in tech media and finance. The information's Natasha Mascarenhas will host a very special edition of TI TV from WTF in Napa Valley on Wednesday, October 29.

我们对这场节目和峰会都充满期待，那将是一场盛会。今天的节目就到这里。提醒大家我们每周一至周五太平洋时间上午10点（东部时间下午1点）在此频道直播。感谢亚马逊网络服务对本节目的冠名赞助，也感谢各位的收看。

We're all very excited for that show, and that conference is is a great one. And that does it for today's show. A reminder that we are live on this stream Monday to Friday at 10AM Pacific, 1PM eastern. And I wanna thank Amazon Web Services who is our presenting sponsor for this production. And I wanna thank you for tuning in.

非常感谢大家的观看，我已经开始期待下次再来担任客座主持了。希望他们还会邀请我回来。明天见，硅谷。

We appreciate your viewership, and I'm already very excited for the next time I get to guest host again. Hopefully, they'll let me back, and we will see you tomorrow, Silicon Valley.