从加油站到 AI 先驱：布雷特·泰勒的旅程

Brett Taylor 拥有我见过的最令人印象深刻的硅谷履历之一。然而，也许他最令人印象深刻的一点是，他对自己在职业上取得的成就非常谦逊。这里有几个亮点。作为 Google 的集团产品经理，Brett 在创建 Google Maps 中发挥了关键作用。他曾担任 Facebook（现为 Meta Platforms）的首席技术官，在那里他参与开发了“点赞”按钮。Brett 曾担任 Salesforce 的总裁兼首席运营官，后来成为联合 CEO，在那里他监督了 2021 年对 Slack Technologies 的收购。在 Elon Musk 收购 Twitter 期间，他是该公司的董事长。最后，Brett 目前是 AI 初创公司 Sierra AI 的 CEO 和联合创始人，同时也是你可能听说过的一家 AI 公司 OpenAI 的董事会主席。尽管拥有如此令人印象深刻的背景，并且似乎总是身处重大技术发展的核心，Brett 始终展现出一种非凡的能力，保持脚踏实地，并以非专业人士能够理解的方式解释高科技。因此，他是帮助我们理解 AI 最新发展以及这些发展如何帮助我们改革医疗保健系统，使其成为值得我们所有人的系统的完美嘉宾。我很荣幸欢迎 Brett Taylor 来到我们的“值得”播客。Brett，非常感谢你来到这里。再次见到你，并能有时间聊一聊，真是太好了。我想先请你谈谈你是如何决定进入科技领域的。科技吸引你的地方是什么？

Brett Taylor has one of the most impressive Silicon Valley resumes that I have ever seen. And yet, perhaps the most impressive thing about him is how humble he is about what he's accomplished professionally. Here's a few highlights. As a group product manager at Google, Brett was instrumental in creating Google Maps. He served as the chief technology officer of Facebook, now Meta Platforms, where amongst other things, he was instrumental in developing the like button. Brett was the president and chief operating officer and then later the co-CEO of Salesforce, where amongst other things, he oversaw the acquisition of Slack Technologies in 2021. He was the chairman of Twitter during Elon Musk's acquisition of that company. And finally, Brett is currently the CEO and co-founder of the AI startup Sierra AI, as well as the chair of a board of this AI company you might have heard of before called Open AI. Despite this impressive pedigree and always seeming to be in the room where it happened for major technology developments, Brett has always displayed a remarkable ability to stay grounded and explain high-tech in a way that is understandable to people who are not experts in the field. Therefore, he's the perfect guest to help us understand the latest developments in AI and how they might help us transform our health care system into one that is worthy of us all. It is my pleasure to welcome Brett Taylor to our worthy podcast. Brett, thank you so much for being here. It's such a pleasure to see you again and for us to take a little time to chat. I would love to start by having you talk about how it is that you decided to get into technology. What is it about technology that appeals to you so much?

对我来说，这非常个人化，而且真正关乎赋能。高中时，我在加油站工作过。我还在一家暖通空调工程公司工作过，当时还有蓝图，我在那里负责传递蓝图，你知道，就是在公司的办公楼之间。那时互联网已经出现了。1996 年，我 16 岁，碰巧那天我正穿上绣有徽章的衬衫去加油站上班，隔壁有一家宝马服务中心。经营那家店的先生抱怨说建网站太贵了。长话短说，我说：“如果我能给你做个网站，你愿意付我多少钱？”他说 400 美元。作为背景，我在加油站的时薪是 4.25 美元。所以，我去了当地一所社区大学，自学了如何建网站，然后就开始做了。显然，作为一个 16 岁的孩子，能赚到 400 美元真的很棒，我当时感觉手头很宽裕。但对我来说，最重要的是没有“守门人”。我觉得在我学会做那个网站的那一刻，我就能做任何网站了。也许这不完全是真的，但在我的脑海里是真的。我建不了办公楼，建不了桥，但这种数字技术感觉在某种程度上是无边无际、不受约束的，而且非常赋能。所以，我就开始大量摆弄它。到了上大学的时候，我真的爱上了数字技术、软件，以及那种你可以把数字世界改造成任何你想要的样子。所以我就这样进入了这个领域，经历了网络和移动革命，现在又经历了 AI，我想我们会花很多时间讨论这个。但我认为，这也是很多年轻人进入科技领域的原因：这种感觉，你可以 16 岁就去创造伟大的东西，而不必得到很多人的许可。我认为，在那个年纪，这是一种非常了不起的情感体验，尤其是如果你是一个有创造力的人，或者想从零开始创造东西。

For me it was quite personal and really about empowerment. In high school, I had worked at a gas station. I had worked at an HVAC engineering firm, kind of running blueprints back when there were blueprints, you know, between office buildings and the firm. And the internet had come out. In 1996, when I was 16, just by coincidence, I was putting on my embroidered shirt to go work at this gas station, and there was a BMW service center next door. The gentleman who ran it was complaining about how expensive it was to make a website. Long story short, I said, "How much will you pay me if I can make you a website?" and he said $400. For context, my wages at the gas station were $4.25 an hour. So I went to a local community college, taught myself how to make a website, and got into it. Obviously, that was really great as a 16-year-old to earn $400, and I was feeling flush. But the big thing for me was just a lack of gatekeepers. I felt the moment I learned how to make that website that I could make any website. Perhaps that wasn't exactly true, but in my head it was true. I can't build an office building. I can't build a bridge, but this digital technology felt to some degree unbounded and unconstrained and so empowering. So I just ended up tinkering a lot with it. By the time I got to university, I really just fell in love with digital technology, with software, and the sense that you could sort of terraform this digital world into whatever you wanted it to be. So that's how I ended up in it, and have now gone through the web and the mobile revolution, now AI, which I think we'll end up spending a lot of our time on. But I think it's why a lot of young people ended up in technology: this sense that you could be 16 and go create something great, and not have to get permission from a lot of people to do it. And I think that's a pretty remarkable emotional feeling when you're that age, especially if you're someone who's creative or want to build from scratch.

我喜欢这个故事。在加油站工作，然后得到一份 400 美元的工作来做网站，然后自学成才。这是个很棒的故事。跟我们谈谈你现在的公司 Sierra AI 吧。它是什么？它是做什么的？

I love that story. Working at a gas station and then getting a $400 job to do a website and then self-teaching yourself how to do it. That's a great story. Tell us a little bit about your current company, Sierra AI. What is it? What does it do?

在 Sierra，我们是领先的平台，帮助公司为他们的客户体验打造 AI 智能体，或者考虑到我在和谁说话，我会说患者体验。本质上，我们帮助各种公司，从 Nordstrom 到加州蓝盾。如果你给他们打电话，希望你不用等待。如果你打电话给 Nordstrom，我想他们的智能体叫 Nora，你会听到 AI 接起电话，而不是按 2 转服务。这些智能体不仅能回答你的问题，还能代表你采取行动。对于像 Credential 这样的保险公司，或者像 Sofi 这样的金融科技公司，这些 AI 智能体真的正在成为他们客户体验的前门。我开玩笑说，如果我把工作做好了，你就不用再等待了。但我认为更有说服力和有趣的是，长期以来，接电话真的非常昂贵，所以大多数行业真的负担不起提供他们可能想要的客户体验。显然，在健康保险领域，你有足够的收入，特别是作为非营利组织，你没有过多的利润可以再投资，但你至少负担不起接电话，而且在大多数情况下有义务接。我总是开玩笑说，我给 Sundar Pichai 打电话比给 Google 客服打电话还容易，因为当你运营一个大型消费品牌时，你的每客户平均收入实际上低于一个电话的成本。所以你真的负担不起提供你想要的那种个性化、高触感的体验。这不是因为公司不喜欢你，而是他们真的做不到。数学上不成立。现在有了 AI，我们已经将最后一个模拟渠道——电话——数字化了。我们让这些智能体真正实现了多模态。所以像我这样不喜欢打电话的人可以聊天。喜欢打电话的人也可以。我希望这不仅意味着我们可以提供一些高触感的体验，永远不用再等待，而且也许企业在为客户打造体验类型时会少一些限制。所以即使在低利润的业务中，我们也能提供高触感的体验，我认为这对消费者来说是非常令人兴奋的事情。我知道我们今天会大量讨论医疗保健。我认为在一些世界上最重要的行业尤其令人兴奋。

At Sierra, we're the leading platform that helps companies make AI agents for their customer experience, or I'll say patient experience given who I'm talking to. Essentially, we help firms, everyone from Nordstrom to Blue Shield of California. If you're calling them up on the phone, hopefully you don't need to wait on hold. If you call up Nordstrom, I think their agent's name is Nora, and you'll hear an AI pick up the phone rather than press two for service. These agents can actually not only answer your questions, but take action on your behalf. For insurance companies like Credential, or for FinTech like Sofi, these AI agents really are sort of becoming the front door to their customer experience. I joke if I do our job well, you'll never have to wait on hold again. But I think the more compelling and interesting thing is that for a long time, answering the phone was really, really expensive, and so most industries couldn't really afford to provide the customer experience they might want to. Obviously, in the health insurance space, you have enough revenue that, especially as a nonprofit, you don't have excessive profits to reinvest, but you at least can't afford to pick up the phone, and in most cases have an obligation to. I always joke it'd be easier for me to call Sundar Pichai on the phone than to get Google customer service on the phone, because when you're running a large-scale consumer brand, your average revenue per client is actually less than the cost of a phone call. So you literally can't afford to provide the kind of personal, high-touch experience you'd want to. It's not because companies don't like you. It's just they literally can't. The math doesn't work. And now with AI, we've digitized the last remaining analog channel, which is the telephone. And we've enabled these agents to truly be multimodal. So people like me who don't love talking on the phone can chat with it. People who prefer to talk on the phone can. And I'm hopeful that it means that not only can we provide some high-touch experience, never wait on hold again, but perhaps businesses will be a little less constrained in the types of experiences they want to build for their clients. So we can provide a high-touch experience even in low-margin businesses, which I think is a really exciting thing for consumers. And I know we'll talk a lot about healthcare today. I think it's exciting in some of the most important industries of the world in particular.

你之前谈到，身处一个无边无际的世界，可以创造任何你想创造的数字事物，并且早早拥有那种探索感，是多么令人兴奋。我现在甚至无法想象，当你看到 AI 时，那种感觉会是什么样。但展望未来，最让你兴奋的技术前景是什么，又让你担心什么？

You were talking about how exciting it was to be in this kind of boundless world where you could create whatever digital thing you wanted to and having that sense of discovery early on. I can't even imagine what that feeling is now when you look at AI. But what is it when you look forward? What excites you most about the prospects of technology and what worries you?

我是个乐观主义者。我开玩笑说，你很难找到一个悲观的创业者。我觉得这两者其实是相辅相成的。我先讲一些风险，最后再讲让我兴奋的事，但先给个结论：我对 AI 将给社会带来的积极影响极为乐观。不过我先从风险说起。AI 已经开始接近某种程度的超人智能，我们还没完全达到。但几周前，我相信 GPT-5.4 证明了一个未解的数学猜想，这个猜想已经存在了几十年，甚至超过一百年，这太惊人了。但随之而来的是，很多工作都涉及查看信息、思考并采取行动，你会想，哇，如果 AI 真的擅长这个，那我的角色是什么？律师事务所里律师助理的角色是什么？软件工程公司里程序员角色是什么？如果一个智能体能像我一样写代码，你顺着这个思路想下去，就会意识到这项技术如此复杂，以至于它将比我职业生涯起点——编程——做得更好。

I'm an optimist. I joke you'd be hard-pressed to find a pessimistic entrepreneur. I think the two sort of go hand in hand. I'll start with some of the risks and end with what I'm excited about, but I'll put the headline which is I'm extremely optimistic about the positive impact AI will have on society. But I'll start with the risks. AI has already started to approach some degree of superhuman intelligence. We're not fully there. But a couple weeks ago, GPT-5.4, I believe, proved one of the unproven math conjectures that had been there for decades if not over a hundred years, which is amazing. But with that, there's a sense that there's a lot of jobs which involve looking at information, thinking, and doing something, and you're like, wow, if an AI is really good at that, what's my role? So what is the role of a paralegal in a law firm? What is the role of a computer programmer in a software engineering firm? If an AI agent can write code as well as I can, you start pulling that thread and you realize that this technology is so sophisticated that it's going to be much better than I am at where I started my career, which is programming a computer.

而且它让人思考，这意味着什么？它是在取代那份工作，还是在创造新的工作？这意味着什么？

And it has a sense of, okay, what does that mean? Is it displacing that job? Is it creating another one? What does it mean?

这就是一个风险。在回到让我兴奋的事情之前，我再列几个其他风险。另一个我想说的是安全。我认为这些技术极其强大，但如果你看看人类过去的伟大发明，比如分裂原子，你从那个技术中得到了非常惊人的影响，但你也得到了切尔诺贝利。你得到了很多这样的东西：你拿一项伟大的技术，或者潜在伟大的技术，加上人性的所有缺陷，把这些放在一起，就可能产生意想不到的、往往是蓄意的负面后果。对于 AI，存在对齐问题：它是否与创造它的人意图一致？是否与使用它的人意图一致？还有更微妙、我认为同样重要的事情，比如心理健康。出现了一种连接危机，尤其是在年轻一代中。我们花在彼此身上的时间越来越少，而花在盯着屏幕上的时间越来越多。这项技术确实放大了所谓“焦虑一代”的一些风险。你是愿意和一个反社会的智能体待在一起，而不是和朋友们在酒吧里闲逛？对于那些选择这样做的人来说，答案会是肯定的。这是另一个风险。你有很多这样的风险：工作替代、安全、心理健康。总的来说，我非常乐观，因为我认为这项技术实际上是根本积极的，并且是民主化的，如果你退一步看。想想计算机编程，我的职业，智能体可能已经比我做得更好了，现在如果我回到 16 岁时的自己，那时我在建网站，现在作为一名软件工程师，我可以使用智能体来单独创建软件，这远远超出了我以前个人能力。就像我们不再用打孔卡来制作软件，而是发明了更高层次的东西。

And that's one thing that's a risk. I'll list a couple others before coming back to what I'm excited about. The other one I would say is really around safety. I think these technologies are extremely powerful, but if you look at previous great inventions by humanity, like splitting the atom, you got some really amazing impacts from that technology, but you also got Chernobyl. You got a lot of things where you take a great technology, or a potentially great technology, and all the flaws of humanity, and you put those things together, and there can be unintended, often intentional negative consequences from the technology as well. And with AI, there's issues of alignment: is it aligned with the intentions of the people who create it? Is it aligned with the people who use it? There's more subtle and I think equally important things like mental health. There's been a kind of a crisis of connection, particularly in younger generations. We're spending less time with each other and more time staring at screens like this. And this technology really amplifies some of the risks of the so-called anxious generation. Can you spend time with a sociopathic AI agent rather than hanging out with your friends at the pub? And the answer is going to be yes for people who choose it. And that's another risk. You have a lot of risks like that: job displacement, safety, mental health. Overall, I'm quite optimistic because I think this technology is actually fundamentally positive and democratizing if you zoom out a bit. If you think about things like computer programming, my profession, which AI agents are already probably better than me at, now if I go back to my 16-year-old self building that website, now as a software engineer, I can use AI agents to create software individually that was well beyond the capabilities of what I could do individually before. And just like we're no longer using punch cards to make software and we invented something higher order.

是的，工作正在改变，但我认为它实际上变得更具高杠杆性，更具创造性。

Yeah, the job is changing, but I would argue it's actually getting more high leverage, more creative.

我看看心理健康之类的风险，再看看积极的一面，我会说：“哇，现在世界上每个孩子都能拥有一个个性化导师了。”曾经只有世界上的超级富豪才能享有的东西，现在每个孩子都能得到。如果你是一个视觉学习者，无论你说什么语言，这都将是最无限耐心、最无限个性化的导师。再看看获取医疗建议。像你我这样的人都能获得优质的医疗保健，但世界上很多人没有。而且，它永远不会取代初级保健医生的角色。但我想，如果你想象一下，比如说你在撒哈拉以南非洲的一个国家，无法获得良好的医疗建议，你可以上传一张图片，描述一些症状，然后得到建议，这可能比你在过去一千年文明中能得到的任何建议都要好。这真的很了不起。即使在我的世界里，在我和医生交谈之前，我也会更新我的化验结果，并询问应该问什么问题。我们实际上已经以一种非常显著的方式民主化了某种程度的专业知识：财务建议、法律建议。你仔细想想，这本质上是一股非凡的民主化力量，我们以某种方式民主化了获取专业知识的途径，我认为我们还没有完全内化这一点。所以如果你退一步看，你问的是我对什么感到兴奋，有什么风险，答案是两者都有很多。我认为这就是技术进步的有趣之处。我认为从历史的角度来看，当我们有适当的距离来看待技术变革时，我们俩都不会活着看到，但我强烈的直觉是，我们会把人工智能的出现视为我们作为一个物种、一个社会所做的最显著积极的事情之一，因为我认为它将带来清洁能源生产，它将帮助我们探索科学中最有趣的问题。我还认为，它将继续提高那些今天无法获得建议和技术的人们的获取门槛。但我也认为这一转变将非常艰难。所以身处其中，我对两者都有清醒的认识。这就是我们生活的世界。

I look at things like the risks of things like mental health. Then I look at the positives and I say, "Wow, every child in the world can have access to a personalized tutor now." Something that was so exclusive to the ultra wealthy of the world is now available to every child. Then if you're a visual learner, no matter what language you speak, this will be the most infinitely patient, most infinitely personalized tutor. Look at getting medical advice. Folks like you and I have access to great health care. A lot of the world doesn't. And, well, it will never replace the role of say a primary care physician. But I think if you imagine being in, let's say you're in a country in sub-Saharan Africa without access to great medical advice, you can upload an image, you can describe something, and get advice that's probably better than anything you could have gotten for the previous 1,000 years of civilization. It's really amazing. And even in my world, before I talk to my doctor, I'll update my lab results and ask what questions should I ask. And we've essentially democratized some degree of expertise in a way that's really remarkable: financial advice, legal advice. You go through it, and fundamentally it's this remarkable democratizing force where we've democratized access to expertise in a way that I think we haven't completely internalized yet. So if you zoom out, and your question was what are you excited about and what are the risks, there's a lot of all of the above. I think that's the interesting thing about technological progress. I think through the lens of history, where neither of us will be alive to sort of, you know, when you have the appropriate level of distance from a technological change, but my strong intuition is we'll look back at the advent of artificial intelligence as one of the most remarkably positive things that we've done as a species, as a society, because I think it will lead to things like clean energy production, it will help us explore the most interesting questions in science. I think it will also lead to the continuing sort of raising the floor of access to advice, to technology, for folks who don't have it today. But I also think this transition will also be really hard. And so being in the middle of it, I'm cognizant of both. And that's kind of the world that we live in.

你作为专业人士和领导者有很多经验。我一直发现，我有时从最痛苦的错误中学到最多。我很好奇，你认为你职业生涯中最有价值的错误是什么？

You've had a lot of experiences as a professional, as a leader. I've always found that I've learned the most from sometimes my most painful mistakes. I'm curious as to what you'd consider your most valuable mistake in your career.

这是个非常好的问题。我给你讲一个轶事。我不确定那算不算错误，但那是一次失败，而失败是错误的一种形式。我当时是谷歌的一名产品经理，为一位名叫玛丽莎·梅耶的女性工作。

That's a really good question. I'll give one anecdote. I'm not sure mistake was right, but it was a failure, which is one form of mistake. I was a product manager working for a woman named Marissa Mayer at Google.

嗯，在谷歌相对早期的阶段。那是在它上市之前，但已经远远超过了它达到逃逸速度之后。我当时是搜索产品经理，积累了一定的信誉，我记得那时我大概 23 或 24 岁，被分配去从零开始打造一个新产品。这对产品经理来说是一个更高的门槛，而不是仅仅维护或增长一个产品，从零开始做事情。

Um, in Google's relatively early days. It was before it went public, but well beyond after it sort of reached escape velocity. I had been a product manager for search and sort of earned enough credibility, as I think I was 23 or 24 at the time, to be assigned a new product, creating it from scratch. Which was something that was a higher bar for a product manager rather than just sort of like maintaining or growing a product, to do something from scratch.

当时的背景是，黄页仍然是我们寻找本地企业的重要方式。我不知道你还记不记得那些送到家门口的黄页。

And the premise was, at the time the yellow pages was still a huge part of how we found local businesses. And I don't know if you remember getting those on your doorstep.

我记得。

I do.

当时，那些企业中的大多数你找不到，更不用说在谷歌上，你在互联网上根本找不到。你知道，不是每个小企业都有网页。所以有整整一个世界的信息你在谷歌上找不到。而我们的使命是组织世界的信息。所以他们说：“嘿，做一个本地搜索产品。”当时主要的竞争对手是像雅虎黄页这样的地方，每个国家都有自己的黄页列表。

And at the time, most of those businesses you couldn't find, let alone on Google, you couldn't really find on the internet. You know, not every small business had a web page. And so there was this whole world of information you couldn't find on Google. And our mission was to organize the world's information. So they said, "Hey, make a local search product." And at the time, the dominant competitors were places like Yahoo Yellow Pages, and every country sort of had their own yellow pages listing.

我和一群工程师做了一个叫谷歌本地（Google Local）的产品。显然不是我一个人做的，是一个团队。它做得不太好。它没有真正的观点。如果你问：“嘿，为什么我要用谷歌的产品而不是雅虎黄页？”没有一个清晰的答案。就像，嗯，如果你更喜欢谷歌而不是雅虎，也许就用它吧。但它给人的感觉非常像，虽然不是完全的复制品，但基本上，相对于当时的现有产品，它没有足够显著的差异化。

I made a product called Google Local with a group of engineers. I didn't make it by myself, obviously, it was a team. And it was not very good. It didn't really have a point of view. If you were saying, "Hey, why should I use Google's product versus Yahoo Yellow Pages?" There wasn't really a clear answer. It was like, well, if you like Google more than Yahoo, maybe use it. But it very much felt like not quite a carbon copy, but for all intents and purposes, it didn't have a differentiator that was significant enough relative to the incumbents at the time.

而且因为谷歌就是谷歌，它在谷歌首页上有一个链接，所以它并不是真正的失败，因为你知道，谷歌做什么，人们都会用。但很明显，这不是一个富有灵感的产品。我记得和拉里·佩奇以及玛丽莎进行了一次相当艰难的产品评审。是的，不是说我被解雇了，但我经历了那种“好吧，第二版最好比这个好”的事情。你知道，实际上比那更礼貌一些，但即使作为年轻人，我也足够精明，知道可以说我挥棒落空了，而且剩下的机会不多了。

And because Google was Google and it had a link from the Google homepage, it wasn't truly a failure just because, you know, it turns out when Google does anything, people use it. But it was clear that it wasn't an inspired product. And I remember having a pretty tough product review with Larry Page and Marissa. And yeah, it's not like I was fired, but I had one of those things which was like, "Okay, version two better be better than this." You know, it was a little more polite than that, but I was savvy enough even as a young kid to know I had sort of swung and missed, so to speak, and there weren't that many strikes left.

那一刻，我们花了很多时间思考我们能做什么新的、不同的事情来真正在这个领域增加价值。我不会详细讲所有的，但我们最终找到了两位了不起的工程师，拉斯和延斯·拉斯穆森，他们正在探索地图技术，并讨论地图和黄页的融合。快进大约 6 个月，我们推出了谷歌地图，结果证明，可能直到今天，它都是我帮助创建的最引以为傲的产品之一。

And that moment, we spent a lot of time thinking about what was new and different that we could do to really add value in this space. And I won't walk through all of it, but we ended up finding two incredible engineers named Lars and Jens Rasmussen who were exploring mapping technology and talking about the convergence of mapping and yellow pages. And you fast forward 6 months or so and we launched Google Maps, which turned out to be, probably to this day, one of the products I'm most proud of having helped create.

而且正是经历了失败，以及在失败时刻的反思，才真正把那个问题拆解开来。它教会了我很多关于第一性原理思维、产品设计，以及技术融合的东西。而且直到今天我都非常感激，拉里个人真的在推动地图。我在那个过程中得到的很多指导也让我反思，现在我的职位比那时更高了，当我对别人做的产品不满意时，我该如何处理那些时刻，帮助人们从错误中学习，而不是被涂上柏油粘上羽毛什么的。所以作为领导者和产品构建者，我都从中学到了很多。

And it took kind of failing and having the moment of reflection that you get in a moment of failure to really break that problem apart. And it taught me a lot about first principles thinking, about product design, about sort of convergent technologies. And I'm very grateful to this day that Larry individually was really pushing on mapping. A lot of the mentorship I got through that process made me reflect too, now that I'm in a position of more seniority than I was then, when I am not happy with a product someone makes, how do I approach those moments and help people learn from mistakes rather than get tarred and feathered or whatever. So I learned a lot from it both as a leader and as a builder of products.

你对其他技术领导者有什么建议，尤其是当他们进入医疗保健领域时，因为医疗保健在采用和有效使用技术方面有着相当长的免疫历史。你现在也涉足其中。那么你对那些试图这样做的人有什么建议？

What advice would you have for other technology leaders as they particularly approach healthcare, because there's been this quite a long history of healthcare being somewhat immune to adopting and using technology effectively. And you're kind of into that now. So what advice would you have for folks that are trying to do that?

是的，我本来想把这个问题抛回给你，但我觉得这违反了播客的规则。

Yeah, I was going to flip this back on you, but I think it's violating the rules of podcast.

不，如果你想的话，你可以抛回给我。

No, you can flip it back on me if you want.

我会的，你知道，我先开始，但我确实有那个问题要问你。你会给其他创业者什么建议？

I will, you know, I'll start, but I do have that question to you. What advice would you give other entrepreneurs?

嗯，正如我们谈到的，为了给那些不太了解医疗保健的听众铺垫一下，你知道，大多数估计是我们将 GDP 的 18% 左右用于医疗保健。在美国的医疗体系中，我们在一些真正重要的方面取得了很好的成果，但我们也有一些令人尴尬的部分。你知道，我们有很多非受迫性失误，正在拉低这个国家某些人口群体的预期寿命，而且我们存在劳动力短缺。所以我们缺少护士、医生、行政人员。我的意思是，你对医疗保健的了解比我多得多，但这些东西的融合，我认为人们理所当然地对目前的结果不太满意，而且我们有劳动力短缺，我们花费巨大。所以这是一个如果你是个经济学家，你会看着它说“哇，技术应该能在这里真正发挥作用”的领域，尤其是 AI，它确实是生产力的驱动力。

Well, as we've talked about, just to lay the land for people listening who aren't as plugged into healthcare, you know, most estimates are we spend on the order of 18% of our GDP on healthcare. It's something that we get great outcomes in some really meaningful ways in the US healthcare system, but we also have some sort of embarrassing parts of it too. You know, we have a lot of unforced errors that are driving down life expectancy for some demographics in this country, and we have a labor shortage. So we have a shortage of nurses, doctors, administrators. I mean, you've forgotten more about healthcare than I know, but this convergence of things, which is I don't think people are rightfully not very satisfied with the outcomes we're getting, and we have a labor shortage, and we're spending a lot. So it's one of those areas where if you're an economist, you look at that and say, "Wow, technology should be able to really help here," especially AI, where it's really a driver of productivity.

所以我认为重要的一点，尤其是对创业者来说，是要有谦逊的态度，明白在医疗保健领域你不知道自己不知道什么。但我认为如果创业者因此远离医疗保健，我们实际上不会解决社会上一些最重要的问题。所以如果你带着谦逊进入，但不回避解决重要问题，我认为这很重要。

So one of the things I think is important, especially for entrepreneurs, is to have humility and understand that you don't know what you don't know in healthcare. But I think if entrepreneurs stay away from healthcare because of that, we won't actually solve some of the most important problems in society. So if you go in with humility but you don't shy away from solving important problems, I think that's important.

我想说的是，找到合适的设计伙伴非常重要。我非常感谢你个人，以及 Ascendion 和加州蓝盾团队，与像 Sierra 这样的公司合作，在医疗保健领域我们不一定知道自己不知道什么，但你看到了一个很棒的技术伙伴，结果这变成了一个非常富有成效的合作。我认为对于更年轻的创业者来说，找到那些对技术未来感兴趣、并且有兴趣将 AI 的杰出头脑与医疗保健的杰出头脑结合起来的医疗生态系统成员，我认为这非常重要。

And what I would say is really finding the right design partners is important. I'm very grateful to you individually and Ascendion and the Blue Shield of California team for partnering with companies like Sierra, where we didn't necessarily know what we didn't know in healthcare, but you saw a great technology partner, and it's ended up being a really fruitful partnership as a consequence. And I think for a younger entrepreneur, finding those members of the healthcare ecosystem who are interested in the future of technology and are interested in bringing the great minds in AI together with the great minds in healthcare, I think that's really important.

然后第二件事是不要试图煮沸整个海洋。

And then the second thing is not trying to boil the ocean.

呃，你知道，我，我，呃，事实证明，仅仅帮助医生不用记笔记就价值巨大。它能解决医疗保健的所有问题吗？不能。但对他，对那位医生来说，这是巨大的生活质量提升。我们花在医疗保健上的 GDP 的 18%，我认为其中 500 个基点是行政成本。所以，你知道吗，当你给医疗服务提供者打电话，你只是想，呃，重新预约，或者你想弄清楚这个提供者是否在网络内，你知道，所有这些都是医疗保健中的日常琐事。

Uh, you know, I, I, uh, it turns out just helping a physician not have to take notes is actually worth a lot. Is it solving all the problems in healthcare? No. But for him, for that physician, it's a huge quality of life benefit. The 18% of GDP we spend on health care, I think 500 basis points is just administrative. So, you know what, when you call a health care provider on the phone and you just want to, you know, reschedule your appointment or you're trying to figure out is this provider in network, you know, all the stuff that's just day-to-day in healthcare.

嗯，是的，你说得对。这不是处方药依从性，也不是解决肥胖流行病，但你正在切入这个领域。而且我还想说，当你在考虑技术采用时，找到机会并迈入门槛，因为这是医疗保健专家找到技术专家的方式，而愿景是五年后实现，不是五天后，并且也从那里开始。但我很好奇你的建议。我的意思是，你处于一个非常独特的位置，既经营过技术公司，又经营过支付方，而且坦率地说，你还在加利福尼亚。你知道，你在硅谷的中心。你的建议是什么？我的意思是，身处这一切的中心。

Well, yeah, you're right. It's not prescription drug adherence. It's not solving the obesity epidemic, but you're cutting into this. And I would also say that as you're thinking about the adoption of technology, find the opportunities and get your foot in the door, because this is the way the experts in healthcare will find the experts in technology, and the vision is to get there five years from now, not five days from now, and sort of start there as well. But I'm curious about your advice. I mean, you're in a really unique position having run both a technology firm and a payer, but also being here in California, candidly. I mean, you know, you're at the epicenter of Silicon Valley. What is your advice? I mean, being at the center of all of this.

嗯，你看，我认为你和 Sierra AI 是正确做法的绝佳例子，因为当我谈到核心原则时，你首先强调的就是解决对社会重要的问题。那么，你如何让这个极其昂贵的系统更可负担，降低成本，提高生产力？你如何改善大多数人在普通日子里体验不佳的服务？你如何提高医疗保健的质量？你不必一次做所有事情，但要找到真正有价值的东西，不仅对你可能服务的公司或客户有价值，而且对我们整个社会的医疗保健系统有价值。当然，Sierra AI 做到了。你改善了服务体验，而且成本更低。所以这显然符合。我想说的另一点是，你必须以极大的谦逊态度对待医疗保健，而你显然做到了。所以你和你的团队知道你们擅长什么，带来什么专业知识，也知道你们在哪里没有专业知识，需要提问和参与。我认为有很多人说，如果我能从技术上解决这个问题，那就解决了整个问题。而我认为他们错过或误解了激励措施可能多么不一致。所以，如果做一件好事与他们的财务自身利益相悖，那不一定别人会愿意做。

Well, I look, I think you and Sierra AI are a terrific example of how to do it right, in the sense that when I get to core principles, the first thing you highlighted is solve a problem that matters to society. So how are you going to make this incredibly expensive system more affordable and take costs out, improve productivity? How are you going to improve a service experience that has been like not very good for most people on the average day? How are you going to improve the quality of health care? You don't have to do all of those things all at once, but find something that actually is valuable not just to the companies you might be serving or the customers you might be serving, but to the overall health care system for us as a society. And which of course Sierra AI does. You are improving the service experience and doing it at a much lower cost. So that clearly fits. I'd say the other is you have to approach healthcare with a lot of humility, and you clearly do that. So you and your team know what you know well and the expertise you bring, and you know where you don't have expertise and need to just ask questions and engage. And I think there's a lot of folks that say if I can solve this technologically, then it solves the entire problem. And I think they miss or misunderstand how misaligned the incentives can be. So doing a good thing is not necessarily something that others will want to do if it runs counter to their financial self-interest.

这是一个复杂的网络，很多科技公司都会在这里绊倒。

That's a complex web that a lot of technology firms stumble upon.

嗯，然后我想说的最后一点是你提到的不要试图一口吃成胖子，要专注，但是，你看，你确实需要执行。我的意思是，这很重要。你需要能够出现并说：“我正在做这件事，它正在取得成果，对任何客户来说都有明确的回报。”就像，有明确的好处。有点像你关于 Google Local 与 Google Maps 的观点。所以我的意思是有些东西真的很基础，但我想说这些就是我会给出的三大建议。

Um, and then I'd say the final thing you talked about not boiling the ocean and being focused, but, look, you do need to execute. I mean, it does matter. You need to be able to show up and say, "I'm doing this thing and it's getting results and there's a clear return for whoever the customer is." Like, there's a clear benefit. Kind of back to your point about the Google Local versus Google Maps. So I mean some of the stuff is really basic, but I'd say those are the things that I would give as the three big pieces of advice.

这是很好的建议。

That's great advice.

是的。谢谢。当然，你已经在遵循这些建议了，这就是为什么这个合作如此顺利。

Yeah. Thanks. And of course you're already following it, which is why this partnership is working so well.

我偶尔听你的。

I listen to you occasionally.

是的。嗯，你在开头提到你在职业生涯中经历了几波技术变革，你知道，智能手机的出现，互联网和智能手机，云计算，现在又是 AI。你怎么看待 AI，以及它对我们社会意味着什么？

Yeah. Well, you had mentioned in your beginning how you'd been through waves of technological change in your professional lifetime, you know, the advent of smartphones and the internet and smartphones and cloud computing, and now AI. How do you look at AI in terms of what it means to us as a society?

嗯，这很有趣。我在网上看到一个图表，描绘了某些技术在社会中的传播，你知道，比如从汽车发明到大多数家庭拥有汽车需要多长时间。

Well, it's interesting there. I saw a graph online that charted the diffusion of certain technologies through society, you know, like how long from the invention of the automobile to most people, most families having automobiles.

而真正有趣的是，如果你看这些图表，它们都变得越来越陡峭，你知道，基本上我们采用技术的速度比以往任何时候都快。

And what was really interesting, if you looked at the graphs, they all got progressively steeper, you know, and essentially we've been adopting technologies faster than ever before.

呃，这有道理，对吧？你知道，我们现在有集装箱船，而 1900 年没有。嗯，同样，AI 的有趣之处在于它乘着之前几波浪潮的东风。所以，我们现在口袋里都有一台超级计算机。它通过互联网连接。我们有云计算。所以，我们有这些数据中心可以容纳所有这些用于 AI 的 GPU。因此，对大多数消费者和企业来说，AI 是你可以直接开启的东西。显然比那要复杂一点，但你不需要建造数据中心。你不需要像互联网早期那样铺设光纤电缆。

Uh, it makes sense, right? You know, we have container ships now and we didn't, you know, in 1900. Um, similarly, what's interesting about the AI though is it's riding on the coattails of those previous waves. So, we now all have a supercomputer in our pocket. It's connected by the internet. We have cloud computing. So, we have these data centers that can house all these GPUs for AI. And so, for most consumers and businesses, AI is something you can sort of turn on. It's obviously a bit more complicated than that, but you don't need to build a data center. You don't need to lay fiber optic cables in the ground like we did when the internet was young.

是的。

Yeah.

从根本上说，尽管在数据中心上投入巨大，但它本质上是一场软件中介的技术革命。因此，你知道，作为技术史，微软的使命曾是将一台 PC 放在每张桌面上。他们相对成功，但我认为我们实际上达到了约 20 亿台个人电脑的峰值，这远低于世界人口。但我们现在手机数量超过人口。所以那项技术真正触达了世界，因此现在你有了软件中介的技术，世界上几乎每个人都通过口袋里的超级计算机连接到互联网，云计算也最终存在。所以这项技术快速产生影响的能力是现成的。因此，像 ChatGPT 这样的服务达到 8 亿、9 亿的周活跃用户，比历史上任何产品都快。你知道，在公司内部，你可以通过决定来做来推出像 Codex 或 Claude Code 这样的技术。所以相对于之前的技术，我认为互联网和 AI 在前后对比方面有很多相似之处。我的意思是，没有互联网就不会有亚马逊和谷歌。我认为这将是非常相似规模的变化。但不同的是，你知道，1998 年不是每个人都能上网。让宽带普及到世界花了很长时间；那真是花了很久。但现在你可以像拨动开关一样。所以这有很多好的和坏的影响。

It's fundamentally, notwithstanding the huge investment in data centers, it's fundamentally a software-mediated technology revolution. And as a consequence, you know, being a history of technology, Microsoft's mission at one point was to put a PC on every desktop. And they were relatively successful, but we actually I think peaked at about 2 billion personal computers, which is much less than the population of the world. But we now have more mobile phones than people. So that technology truly reached the world, and as a consequence now you have a software-mediated technology and everyone in the world roughly is connected to the internet with a supercomputer in their pocket, and cloud computing exists at an end. So the ability for this technology to have an impact quickly is just available. And so services like ChatGPT reach 800, 900 million weekly actives faster than any product in history. You know, within companies, you can roll out technologies like Codex or Claude Code, you know, essentially by deciding to do it. And so relative to the technologies that preceded it, I think there's a lot of parallels between the internet and AI in terms of just a before and after. I mean, you wouldn't have had Amazon and Google without the internet. I think it'll be a very similar scale of change. But the difference is, you know, not everyone had access to the internet in 1998. It took a long time to get broadband to the world; it just took forever. But now you can like flip a switch. So there are so many implications of that that are good and bad.

你知道，从好的方面看，这项技术的益处能传播得非常快，我觉得这真的很了不起。但坏处是，比如我姐姐是加州中部的公立学校老师，课程设置并没有考虑到每个学生口袋里都有一台超级计算机，可以用来帮助学习，也可以用来作弊，取决于你怎么看。所以结果就是技术超越了社会的防护措施，比如老师们不得不重新用蓝皮书考试。我们花了那么多时间给每个孩子配笔记本电脑，现在却要说‘请合上你们的笔记本电脑’。所以，正如我所说，我对这项技术是坚定乐观的。

You know, on the good side, where there are benefits of this technology, it can diffuse really quickly, and I think that's really amazing. But the downside is like my sister's a public school teacher in central California, and the curriculum did not contemplate every student having a supercomputer in their pocket that could be available for helping or cheating, depending on your perspective. So you end up with the technology sort of outpacing society's guardrails, whether it's teachers having to go back to blue books. We spent all this time trying to give every kid a laptop, and now we're like, 'Please close your laptops.' So I'm, as I said, decidedly optimistic about the technology.

但采用的速度令人不安。对于公司的员工个人来说确实如此。如果你 50 岁，可能计划在 5 到 10 年内退休，取决于你的计划和退休安排，突然之间，当你本该处于专业领域的巅峰时，却被 AI 颠覆了，现在你必须学习这个新工具才能保持竞争力。

But the pace of adoption is uncomfortable. It is, for individual employees of companies. If you're 50 years old and maybe you're planning to retire in five or 10 years, depending on your plans and your retirement plan, all of a sudden, when you're supposed to be at the peak of your expertise in your field, it gets upended by AI, and now you have to learn this new tool to remain relevant.

是的。

Yeah.

这确实有点糟糕。你知道，就像你本该是导师、专家，处于职业巅峰，现在突然一切都变了。

That kind of sucks. You know, it's like you were supposed to be the mentor, the expert, at the peak of your profession, and now all of a sudden it's completely different.

所以我觉得这种速度是我们以前从未真正经历过的，尽管社会在过去经历过这类革命。我总是喜欢提醒人们，当有人说‘工作要消失了’时，我并不相信。

So it's just the level of pace that I think we haven't really experienced before, even though society has experienced these types of revolutions in the past. I always like to remind people, when people are like 'the jobs are going away,' which I don't believe.

是的。

Yeah.

1776 年，这个国家大多数人都是农民，从那以后我们经历了工业革命、全球化，现在我们有了服务经济，我觉得如果你试图向开国元勋们描述这种经济，他们根本无法理解。而且我们的失业率比以往任何时候都低。所以作为一个社会和经济体，我们完全有能力吸收技术变革、旧工作消失和新工作被创造，但我们从未以这样的速度做过。所以这就是我真正不知道的事情。但我对我姐姐这样的人，那些一线工作者，深表同情。

In 1776, most of the country were farmers, and since then we've had the industrial revolution, we've had globalization, and now we have a services economy that I think most of the founding fathers wouldn't understand if you tried to describe it to them. And we have lower unemployment than we have ever had. So as a society and as an economy, we are more than capable of absorbing technology change and jobs going away and new jobs being created, but we've never done it at this pace. So that's the thing I honestly don't know. But I have a lot of empathy for people like my sister, the frontline people who are dealing with it.

他们在应对这些。

Dealing with it.

公立学校老师只是众多例子之一。我认为这可能是这次技术变革中最具挑战性的部分。

And public school teachers are just one of many. And I think that's probably the most challenging part of this particular technology shift.

嗯，我们之前讨论过技术整体的潜在益处和风险，尤其是 AI。你想到有哪些具体步骤，是我们现在应该考虑来减轻潜在风险的？

Well, we had talked earlier about the potential benefits and risks of technology in general and AI particularly. Any particular steps that come to mind that for you that we should be considering right now to mitigate the potential risks?

首先，我认为重要的是我们一直在谈论风险。其中一点，你和我都很独特，因为我们都参与了非营利组织。OpenAI 现在有两个组织，但我是非营利基金会的主席。非营利基金会和公共利益公司都有一个使命，那就是确保通用人工智能造福人类。这就是 OpenAI 成立的目的。以使命为导向的好处是，你不仅能够，而且有义务去问这些关于风险的问题。实际上，OpenAI 和 Anthropic 都很棒的一点是，虽然表面上它们当然存在竞争，但两家公司都是以使命为导向的。两家公司都真正专注于这项技术的益处，这让我对未来充满希望，因为最重要的两个实验室都如此注重益处和减轻风险。所以首先，我认为重要的是我们都在谈论它，因为我信奉负责任迭代部署的原则。我的意思是，我认为在所谓的象牙塔里，仅仅靠苦思冥想就想出所有风险并做对，是非常困难的。如果你做过任何复杂的工程，比如发射火箭或建造核电站，其中有很好的科学，但很大一部分是科学遇到现实时会发生什么。这就是工程。做法不是让一个人坐在火箭上，而是先发射一枚你知道可能无法飞出大气层的火箭，然后不断迭代，了解科学与现实之间复杂的特殊交汇点，通过工程过程获得真正安全的技术。所以当我们思考风险时，我的想法是，让 GPT-8 安全并造福人类的最好方法，就是让 GPT-7 安全并造福人类。

First, I think it's important we're always talking about the risks. One of the things, you and I are unique in the sense both of us are involved with nonprofits. And OpenAI is, well, two organizations now, but I'm the chairman of the nonprofit foundation. And both the nonprofit foundation and the public benefit corporation have a mission, and that mission is to ensure that artificial general intelligence benefits humanity. That's what OpenAI was created to do. And what's nice about being mission-driven is that you not only can but you have an obligation to ask those questions about risks. And actually, what's really nice about both OpenAI and Anthropic is that, ostensibly there's some rivalry between them of course, but both companies are mission-driven. Both companies truly are focused on the benefits of this technology, which gives me a ton of hope for the future, that the two most important labs are so oriented towards benefits and mitigating risks. So first, I think it's important we're all talking about it because I believe in the principle of responsible iterative deployment. And what I mean by that is, I think it's very hard in a proverbial ivory tower to just think really hard and come up with all the risks and get it right. If you've ever done any deep complex things in engineering, like sending a rocket to space or building a nuclear power plant, so much of it, there's good science, but so much of it is like what happens when that science meets reality. That's engineering. And the way you do it is you don't start with a person on the rocket. You start with a rocket that you send up knowing it's probably not going to make it out of the atmosphere, and you iterate and you iterate and you learn about the complex idiosyncratic intersection of science and reality, and from that you get really safe technology through the process of engineering. So as we're thinking about the risks, the way I think about it is, the best way to make GPT-8 safe and have positive benefits for humanity is if GPT-7 was safe and had positive benefits for humanity.

而且通过在这个过程中犯的错误，希望是小的错误，我们学习、迭代并不断进步。

And through the mistakes, hopefully minor, that we made during that process, we learned and iterated and kept progressing.

我认为这可能是正确的做法。实际上，如果你看看，说到我姐姐是公立学校老师，随着这项技术越来越先进，这些学校已经有三四年使用 ChatGPT 的经验了，不管现在是多少年，他们可以从中学习。所以随着技术更先进，他们不是从零开始，不是冷启动。我认为这很重要。另外我想说的是，我认为在所有这一切中，公私合作非常重要。政府要监管一项如此新颖且变化如此之快的技术是非常困难的。所以我认为，要取得好结果，就需要在技术公司、世界各国政府和公立学校系统之间建立真正的伙伴关系和对话。因为我认为，你和我之前谈到过，我开玩笑说你在医疗保健方面忘掉的都比我永远知道的多。我们之所以有良好的伙伴关系，是因为这是一种伙伴关系，你把医疗保健专业知识和 AI 专业知识结合起来，我认为大多数领域都是如此。所以我认为硅谷有义务参与进来。没有选择可以让你袖手旁观，如果你在这些艰难的社会讨论中置身事外，你就无法完成你的使命。

And I think that's probably the right way to do it. And actually, if you look at, talking about my sister being a public school teacher, as this technology gets more and more advanced, these schools have had three years of ChatGPT or four years of ChatGPT, whatever it is now, to learn from it. So as the technology is more advanced, they're not starting from scratch, they're not starting from a cold start. And I think that is important. The other thing I would say is that I think it's really important that we have public-private partnership in all of this. It's very challenging for governments to regulate a technology that's so new and changing so fast. So I think the way that you get good outcomes is when you end up with a true partnership and dialogue between the companies building the technology and governments around the world, public school systems, all of it. Because I think, you and I were speaking about, I joke that you've forgotten more about healthcare than I will ever know. The reason why we have a good partnership is it's a partnership, and you bring the healthcare expertise and the AI expertise together, and I think the same is true of most domains. So I think it's an obligation for Silicon Valley to engage. There's no option here where you sit on the sidelines and you're not going to be fulfilling your mission if you're on the sidelines of these hard societal discussions.

是的。这项技术太强大了，发展太快了，有太多潜在的社会影响，不可能不进入政治讨论。

Yeah. The technology is just too powerful, advancing too quickly, and has too many potential social implications to think that it's not going to end up in the political discourse.

我完全同意这个观点。那么这次轮到你给一些建议，也许是给企业领导者的，你和他们中的很多人合作过。你有没有观察到他们中有人做得很好，在思考如何制定 AI 战略方面？你会给公司领导者什么建议，无论他们是在医疗保健还是其他领域，关于接触人工智能的最佳方式？

So I agree completely with that notion. So your turn to give some advice again this time to maybe corporate leaders and you've worked with a number of them. Have you observed examples of them doing a really good job thinking about strategizing about how to use AI? And what advice would you give to leaders of companies about whether they're in healthcare or elsewhere like the best way to engage with artificial intelligence?

我本来想说找 Sierra，但我是开玩笑的。也许从那个开始。

I was going to say call Sierra, but I'm kidding. Maybe start with that.

没关系。你可以把它放在那里。没问题。

That's okay. You can put it up there. That's fine.

我在开玩笑。我认为错误的方式是为了 AI 而制定 AI 战略。我有一个术语，我想你听我说过，叫做“AI 旅游”，就是你表演性地做概念验证，或者你给所有员工推出 Copilot，然后说“好了，我们现在有 AI 战略了”。我确实认为你应该从第一性原理出发，思考在这个技术存在的情况下，你想要什么样的业务形态。一个简单的开始方式是关注那些已经众所周知的应用程序。所以我要说，到目前为止受 AI 影响最大的两个市场，第一是软件工程，让你的技术团队能够使用像 OpenAI 的 Codex 这样的智能体。第二是客户服务，这就是我开玩笑说找 Sierra 的原因，那可能是那个特定应用的一个好选择。原因就是我们讨论过的所有那些，它比之前的技术更好、更快、更便宜。你不需要为我们提供的东西做艰难的商业论证。

I'm teasing. I think the wrong way is to focus on an AI strategy for AI's sake. I have this term that I think you've heard me say which is AI tourism where you're performatively doing proofs of concept or you roll out co-pilot to all of your employees and say good we have an AI strategy now. I really do think you want to start with first principles thinking about what shape of business do you want given the existence of this technology. One easy way to start is with the applications that are already well known. So I would say the two markets that have been most impacted by AI so far are number one software engineering, enabling and empowering your technology teams to use agents like OpenAI's codecs. Number two is customer service, which is my joke around calling Sierra, that probably is a good option for that particular application. The reason for it is all the things that we talked about, which is this is both better, faster and cheaper than the technology that preceded it. You don't need to make a hard business case for what we provide.

但除此之外，我认为要思考你业务中的所有核心流程。在医疗保健领域，可能是寻找网络内的提供者，或者解释福利，或者当你的提供者打电话来协商索赔时，所以是索赔处理或索赔裁决，或者为提供者做术前术后，处方药依从性，简单的事情比如你转诊，初级保健医生转诊给专科医生，有多少转诊基本上没有完成。所以就像你能如何帮助这些流程？每一个这样的流程，如果你从第一性原理出发，说现代 AI 如何能帮助这个流程，你会找到 10 到 20 种不同的方法。所以真正从流程层面开始，首先是为了你的客户、你的病人，然后是你的员工。

But then beyond that I think thinking about all the core processes in your business. So in healthcare it could be finding providers in network or explanation of benefits or when your providers call you to negotiate a claim, so claims processing or claims adjudication, or pre-op post-op for a provider, prescription drug adherence, simple things like you refer a primary care physician does a referral to a specialist and how many of those go essentially unfulfilled. So like how can you help with those processes? Every single one of those if you think about it just from first principles and say how could modern AI help with this process, you'll find 10 or 20 different ways. So really starting at the process level for your hopefully your customers, your patients first, and then your employees.

然后说我要从那个开始，并实际指派一个团队来改进那个流程。也许你可以在两天内而不是两个月内让新供应商入职。也许你可以减少，我们有很多做技术支持的客户，也许你可以减少人们因为问题没有真正解决而再次打电话的次数，减少 200%。

And then say I'm going to start with that and actually assign a team to make that process better. Maybe you can onboard a new vendor in two days rather than two months. Maybe you can reduce, well we have a lot of clients who do technical support, maybe you could reduce the number of times people call back because a problem wasn't really solved by 200%.

对。

Right.

如果你选择这些业务指标并授权一个团队，你可以把 AI 的科学变成工程，你可以真正务实、实际地应用它。所以我的建议是：从已知领域开始。我认为你不需要太有创造力就能在软件工程、客户服务等领域应用。第二是提炼你业务的核心流程，并进行一些第一性原理的讨论，鉴于 AI 已知能做什么，好的样子是什么？好，我们去做吧。但实际上要从业务结果开始。不要从技术开始。我认为这是你在 AI 中找到成功的方式。

If you pick those business metrics and empower a team, you can turn the science of AI into engineering and you can really get pragmatic, get practical about how to apply it. So that's my advice: start with the known domains. I think you don't need to be too creative to apply in areas like software engineering, customer service. And number two is distill the core processes of your business and have some first principles discussion about given what AI is known to be able to do, what does good look like? Okay, let's go do it. But actually start with the business outcome. Don't start with the technology. I think that's the way you find success in AI.

对于医疗保健领导者，你还有什么要补充的吗？因为医疗保健是美国唯一一个，至少是我所知的唯一一个主要行业，有人真的说随着技术的进步，它实际上增加了成本，而且它对我们在几乎所有其他地方看到的那种生产力提升非常抵触。那么对于如何帮助医疗保健领导者改变这一点，你有什么建议或想法吗？

Anything you'd add to that for healthcare leaders because healthcare is the only industry in the United States or at least the only major industry I'm aware of where there's people who literally say with the advancement of technology it actually increases costs and it's been pretty resistant to the types of productivity enhancements that we've seen almost everywhere else. So any advice or thoughts on how to help healthcare leaders change that?

这是一个非常重要的问题，作为这个国家的公民，看到医疗保健领域缺乏生产力增长，真是令人沮丧。部分原因可能是可以接受的。你知道，我们在人们生命的最后几年花费了很多成本。

It is a really important question and it is just as a citizen of this country it's so frustrating to see the lack of productivity growth in healthcare. Part of it might be okay. You know, we spend a lot of costs on the last few years of people's lives.

作为为我所爱的人做过这件事的人，这也许是这个国家能负担得起的奢侈品。

As someone who's done that for people that I love, that's maybe a luxury that we can afford in this country.

但其中一些也是不可接受的，这有点像我们医疗保健系统中一些更系统性的问题。我想说的一件事是，我认为你之前给医疗保健系统提建议时说得很好，但我会从一个真正可衡量的问题和流程开始，并有点坚持不懈地专注于它，而不是试图解决整个医疗保健系统，因为有一个世界，让我们以提供者为例，人们出现在他们的预约中。这是一个真正的机会。这是一个真正的成本驱动因素。对任何参与者来说都不是很好的体验。如果有人不出现，显然会花费，有很多机会成本，可能还有一些硬性成本。然后对病人来说，通常原因是重新安排太繁琐了。你可能忘了。所有这些事情，AI 都可以成为其中的一个要素。如果你坚持不懈地专注于那个问题，你可能可以为所有参与者节省成本。

But some of it's just not acceptable too, which is sort of some of the more systemic issues in our healthcare system. One thing I would say is I think you articulated it well when you were giving advice earlier about the healthcare system, but I would start with a really measurable problem and process and focus on it somewhat relentlessly rather than trying to solve the healthcare system because there is a world where, let's just take for a provider people showing up to their appointments. It's a real opportunity. It's a real driver of cost. It's not a great experience for anyone involved. If someone doesn't show up, it obviously costs, there's a lot of opportunity costs and probably some hard costs as well. And then for patients often the reason is it's so tedious to reschedule. You might have forgotten. There's all these things that on all of it there's AI could be an ingredient to. And if you relentlessly focus on that problem, you probably can save costs for everyone involved.

医疗保健的问题在于，经常发生的情况是，生产力最终被系统的其他部分吸收了。

The problem with healthcare, what often ends up happening is you end up sort of that the productivity gets absorbed by other parts of the system.

对吧？

Right?

我不知道如何解决这个问题。正如我开玩笑说的，我认为你可能比我更有能力回答这个问题。然而，我认为如果你专注于那个最终问题，并具有真正可衡量的、切实的利益和切实的结果，我们就会取得进展。也许我们最终会在其他地方吸收成本，因为我们创造了这个，你知道，有点像神话中的野兽，你砍掉一个头，长出两个，对吧？这就是我们创造的医疗保健系统。

And I don't know how to solve that. And as I joked, I think you might be more equipped to answer that than I. However, I think if you're focused on that end problem with really measurable, tangible benefits, tangible outcomes, that's how we're going to make progress. And maybe we'll end up absorbing the cost elsewhere because we've created this, you know, sort of, what's the mythical beast where you cut off a head or two grow, right? That's a healthcare system that we've made.

但我认为变得悲观是不对的。我只想说，如果我有建议，那就是不要试图一口吃成胖子。尝试解决一个我们知道正在推动系统成本的重要问题，并将其降低。我认为如果我们系统地这样做，我们可以改善结果，而且我们作为一个社会可能只是选择我们想在医疗保健上花很多钱，因为正如你所知，AI 的通货紧缩效应如此之大，我们只是选择这是我们想花钱的地方。或者也许我们实际上可以降低成本。

But I don't think getting defeatist is not right. And I would just say if I have advice is don't boil the ocean. Try to solve an important problem which we know is driving costs in the system and drive it down. And I think if we do that systematically we can improve outcomes and we may as a society just choose we want to spend a lot on healthcare because as you know so much of the deflationary effect of AI we just choose this is where we want to spend our money. Or maybe we can actually lower cost.

我的意思是，如果我们有能力做到那一点，我会很乐意，但我只想说，专注于一些问题并解决它们。我认为这才是处理这些巨大问题的有效方式。

I mean, I would love that if we have the ability to do that, but I would just say focus on some problems and fix them. And I think that is just the productive way of sort of approaching, you know, these humongous problems.

那么，Brett，在我们结束之前，你还有什么想分享的、而我还没有问到的吗？

Well, Brett, what else would you like to share that I haven't asked you about yet before we wrap things up?

嗯，我只想再次表达我的感激之情，Paul。感谢你成为医疗保健系统中如此重要的领导者。正如我所说，我认为你通过 Ascendion 所做的事情，让你处于一个独特的位置，既是技术领导者，又是医疗保健领导者。但我也想对听众说，我希望医疗保健的复杂性不会劝阻优秀的企业家进入这个领域。你知道，它很难，因为它本来就难。但我确实认为，AI 应该惠及医疗保健的第一性原理是真实的。我们存在劳动力短缺，我们没有得到想要的结果。我只想说，如果你像 Paul 建议的那样，以谦逊的态度，找到好的合作伙伴，我认为现在确实有机会做出改变。

Um, well, just want to say express my gratitude again, Paul. Thank you for being such an important leader in the health care system. And as I said, I think you're in a unique position with what you've done with Ascendion to, you know, be both a technology leader and a healthcare leader. But um and I was just going to say for the people listening, I just hope uh the complexity of healthcare doesn't dissuade great entrepreneurs from going into it. Um you know, it's uh it's hard because it's hard. Uh but I do think um the first principles reasons why AI should benefit healthcare are real. We have a labor shortage. uh we're not getting the outcomes we want. Um and I would just say that, you know, if you approach it the way Paul um suggested, which is with humility and finding, you know, good partners, um I think there's a real opportunity to make a difference right now.

所以，你知道，我希望当我们俩聊天的时候——我本来想开个玩笑，我不知道网上的人知不知道这个。我在开玩笑。你现在没在用跑步机办公桌。这好像是第一次。

And so, you know, I hope when you and I are are uh talking, I was going to make a joke. I don't know people online know this. I was joking. You're not on a treadmill desk right now. It's like the first time

这是我第一次跟你说话时你只是静静地坐着。我现在其实很不自在。

It's the first time I've talked to you when you're just sitting still. I'm very uncomfortable actually right now.

嗯，但我希望几年后当我再跟你说话时，希望那时你是在跑步机办公桌上，我们能开始看到这项技术带来一些真正切实的好处。所以，我只想表达一些乐观情绪，也许是对正在收听的企业家们发出行动号召，我认为现在确实有一些真正的机会。

Um, but I'm hopeful when I'm speaking to you in a few years, hopefully while you're on a treadmill desk that time that we can start to see some real tangible benefits from this technology. So, I just want to express some optimism there and maybe a call to action to the entrepreneurs listening that I think there's some real opportunity right now.

多么美妙的收尾方式。Brad，非常感谢你参加本期播客，祝你在 Sierra AI 一切顺利。

What a wonderful way to close things out. Brad, thank you so much for being a part of the the podcast and good luck with everything at Sierra AI.

谢谢你的邀请。

Thanks for having me.