AI 的下一个万亿美元市场不是软件

AI's Next Trillion-Dollar Market Isn't Software

马克斯·亚德伯格 Max Jaderberg · Silicon Valley Girl · 2026-11-10 · 约 44 分钟 · 原视频 ↗

打开互动全文版(中英对照 + 朗读 + 问答)→

本期速览 · Overview

Isomorphic Labs 总裁 Max Jaderberg 讲述 AlphaFold 的概念验证如何推动用 AI 设计药物、攻克所有疾病的使命。

Max Jaderberg, President of Isomorphic Labs, explains how AlphaFold's proof of concept is driving a mission to solve all diseases with AI-designed medicines.

要点 · TL;DR

核心观点 · Key points

反共识 · Contrarian takes

本期章节 · Chapters(共 32)

全文 · Full transcript(中英对照)

引言 Introduction

Host

有了 AI,我相信你需要那种宏大的雄心,以及尽可能最大的使命。

With AI, I believe you need that big ambition and that biggest mission possible.

Host

这是 Max Jaderberg,一位牛津培养的 AI 研究员,他的初创公司被谷歌收购。他帮助开发了 AlphaFold 3,现在是 Isomorphic Labs 的总裁,该公司由 Demis Hassabis 创立,旨在用 AI 设计新药。

This is Max Jaderberg, an Oxford-trained AI researcher whose startup was acquired by Google. He helped develop AlphaFold 3 and is now president of Isomorphic Labs, founded by Demis Hassabis to design new medicines with AI.

Host

所以 AI 可以生成很多假设。你怎么知道它真的值得追求?

So AI can generate a lot of hypotheses. How do you know it's actually worth pursuing?

Max

不必通过实验筛选数十万种不同的设计,我们可以制作 20 或 50 个,然后进入现实世界,实际制造这些分子并测试它们,看看是否有效。

Instead of having to screen experimentally go through hundreds of thousands of different designs, we can make 20 or 50 and then we go into the real world and actually make these molecules and test them out and see if they work.

Host

那么你认为我们什么时候会在药房看到第一批由 AI 设计的药物?

So when do you think we're going to see the first drugs designed by AI at a pharmacy?

Max

嗯,事实证明

Well, it turns out

Host

Max,欢迎来到播客。

Max, welcome to the podcast.

Max

太好了。谢谢你邀请我。

Great. Thank you for having me.

Host

很高兴你能来。你肩负着一个惊人的使命:借助 AI 解决所有疾病。你能跟我谈谈在你的工作中看到了什么,让你相信这一切都是可能的?

So happy to have you. You're on an amazing mission to solve all disease with the help of AI. Can you talk to me what have you seen inside your work that made you believe that this is all possible?

背景与AlphaFold Background and AlphaFold

Max

也许我回溯一下,说说我们从哪里开始,以及我们在哪里开始看到,实际上这个解决疾病的疯狂使命是我们真正可以追求的。我的背景是核心 AI、核心深度学习开发。我的研究始于计算机视觉和 ImageNet 之类的东西,如何训练这些大型网络让计算机看得见。我在 DeepMind 待了很长时间,研究所有通用算法、新架构设计、生成模型和语言模型的早期阶段,以及深度强化学习。通过这项工作,我参与了像 AlphaGo 和 AlphaStar 这样真正困难的挑战任务。我们一次又一次地解决了这些看似不可能的挑战。也许我们还没有达到 AGI,但很明显,我们拥有这项 AI 新技术,能够做非常不同的事情。我们开始思考,好吧,我们到底应该用它做什么?它对人类的影响会是什么?与此同时,我们开始看到 DeepMind 内部 AlphaFold 的成果开始显现。所以这是一个惊人的项目,试图预测蛋白质的结构,这是一项非常困难且基础的科学任务。我们开始看到这些结果出现,这个 AlphaFold 网络开始变得非常非常好。这导致了一个真正的信念:你实际上可以训练出在生物学和化学领域具有泛化能力的 AI 模型。之前没有人真正展示过这一点。我认为 AlphaFold 的这个证明点真的让我们敢于迈出信心的一步,说,嘿,我们正处于 AI 发展的正确阶段,现在真的可以在 Isomorphic Labs 开始做点什么,全力投入利用这个来解决人类健康问题,我们真的在考虑解决所有疾病,这是我们要努力推进的核心使命。

Maybe I'll rewind to say a bit where we began and where we started to see that actually this crazy mission of solving diseases is one we can actually go after. My background has been in core AI, core deep learning development. I started my research in computer vision and things like ImageNet, how to train these big networks to get computers to see. I spent a long time at DeepMind working on all the general algorithms, new architecture design, early days of generative models, language models, and deep reinforcement learning. And through that was working on these really hard challenge tasks like AlphaGo and AlphaStar. We were able to solve these impossible challenges again and again. And maybe we hadn't got to AGI yet, but it was very clear that we had this new technology with AI that was able to do very different stuff. We're starting to think, okay, well, what should we really use this for? What could be the impact on humanity here? And at the same time, we start seeing the AlphaFold results internally in DeepMind starting to come to fruition. So this was an amazing project of trying to predict the structure of proteins, which is a very hard and fundamental task of science. And we could start to see these results emerging of this AlphaFold network starting to get really good. And this led to a real conviction that you can actually train AI models that generalize in the space of biology and in the space of chemistry. And no one had really shown that before. And I think that proof point of AlphaFold really allowed us to take the leap of faith to say, hey, we're at the right stage of the development of AI to really actually now start something in Isomorphic Labs to go full force towards using this to really solve human health and we really think about solving all diseases, that core mission for us to go hard behind.

药物发现的挑战 Challenges in Drug Discovery

Host

那么从这个 AlphaFold 时刻到解决所有疾病,必须发生什么?

So from this AlphaFold moment to solving all diseases, what has to happen?

Max

AlphaFold 并不是制造药物、解决疾病的解决方案。它真正是证明解决方案可能存在的概念验证。你知道,要真正制造一种药物来帮助解决疾病,需要许多不同的东西结合在一起。这不仅仅是一个问题。而是十几个非常棘手的问题结合在一起。这就是为什么实际制造药物如此困难。制造一种药物实际上需要相当长的时间。目前从构思到患者可用的药物需要 10 年时间,而且失败率非常高。大约 90% 的药物在开发过程中失败。

AlphaFold isn't the solution to making medicines, to solving diseases. It's really that proof of concept that a solution could exist. You know, to actually make a medicine to help solve a disease, it requires so many different things to come together. It's not just one problem. It's like a dozen really tricky problems come together. That's why it's so hard to actually make medicines. It takes actually quite a long time to make a medicine. It takes 10 years to go from conception at the moment to medicine for patients and the failure rate is really high. About 90% of medicines fail through development.

Host

因为我们不知道它们会如何反应,比如蛋白质会如何与分子反应,据我所知,这就是 AI 发挥作用的地方,对吧?

Because we don't know how they're going to react, like how the proteins are going to react with the molecules, and this is where AI is helping as far as I understand this, right?

Max

有太多不同的维度在起作用,而且都很困难。所以,是的,你究竟如何解决那个疾病?通常你需要针对什么蛋白质靶点?你需要进入并调节体内的基本通路和细胞中的小分子机器,以使疾病状态好转,那是什么?你能否真正制造出一个好的分子,并以你预期的方式起作用?如果你把那个分子作为药物,它真的会到达身体的正确部位吗?它会安全吗?会有毒性副作用吗?你需要设计和理解的东西清单太长了。而且真的非常非常难。因此,从 AI 和模型的角度来看,解决方案实际上不仅仅是一个模型。它是一整套 AlphaFold 级别的突破。我们考虑的是可能需要六七个以上类似 AlphaFold 的突破,然后它们必须结合在一起,形成一个整体的药物设计引擎。在过去的五年里,我们一直在非常努力地系统性地攻克那六七个以上类似 AlphaFold 的突破。

There's so many different dimensions that come into play and which are hard. So yes, how do you actually go about solving that disease? What usually protein target do you need to go after? What's the fundamental pathway and little molecular machine in the body, in the cell that you want to go in and modulate to make the disease state better? Can you actually make a molecule that's good and that works in the way that you intended? If you take that molecule as a drug, will it actually get to the right part of the body? Will it be safe? Will it have toxic side effects? There's just such a long list of things that you need to design for and need to understand. And it's really really hard. And so the solution from an AI perspective and a model perspective is not actually just one model. It's like a whole collection of AlphaFold level breakthroughs. We think about maybe half a dozen plus AlphaFold-like breakthroughs that we have to happen to then come together into one holistic drug design engine. We've been working really hard over the last five years on just systematically going through those half dozen plus AlphaFold-like breakthroughs.

Host

但你们已经确定它们了吗?

But have you identified them?

Max

是的,我的意思是我们知道那些应该是什么。我们思考,好吧,我们想以新的方式制造药物。那些核心超能力是什么,那些超能力,如果你能把这些超能力赋予化学家、药物设计师或生物学家,它将从根本上提高我们制造药物的能力。所以我们真的可以确定那些是什么。这涉及到理解细胞如何工作以及疾病生物学的不同部分如何工作,如何设计分子的不同方面,理解人体毒性的不同方面。然后我们就可以系统地追求这些,创造那些超能力。所以我们已经确定了这些,并且一直在为此构建模型。

Yeah, I mean we know what those should be. We think about, okay, we want to make medicines in a new way. What are those core superpowers, those superpower capabilities that if you could just give this superpower to a chemist, to a drug designer or a biologist, it would radically improve our ability to make medicines. So we can really identify what those are. That comes into things like understanding how cells work and different parts of disease biology work, how to design different aspects of designing molecules, different aspects of understanding human toxicity. And then we can just go systematically after those to create those superpowers. So we've identified those and we've been building the models towards that.

AI在理解与设计中的作用 AI's Role in Understanding and Design

Host

AI 如何帮助你实现你想要的目标?

How is AI helping you achieve what you want?

Max

这真的在两个战线上。一个是理解战线,另一个是设计新药的能力战线。所以要制造药物,你首先需要理解疾病生物学。生物学是一个非常美丽而复杂的野兽。有太多的复杂性。我们所知道的只是冰山一角,我想是关于生物学和疾病生物学的科学,以及我们的细胞如何工作,关键的是如果你有疾病,那如何改变细胞的正常功能。所以我们使用 AI,我们创建这些模型。这些就像是生物学的基础模型,帮助我们理解这些系统如何工作,从而帮助我们理解不同的蛋白质和生命的不同构建块——蛋白质、DNA、RNA——它们如何组合在一起,如何形成通路,这些是细胞的信号机制。

It's really on two fronts. There's the understanding front and then the ability to design new medicines front. So to make a medicine, you need to understand the disease biology first and foremost. And biology is a really beautiful and complex beast. There's just so much complexity. We know such a tiny proportion of what there is to know, I think about the science of biology and disease biology and how our cells work and crucially if you have a disease, how that changes the normal function of cells. And so we use AI and we create these models. These are like foundational models of biology that help us understand how these systems work so it can help us understand how different proteins and the different building blocks of life, proteins, DNA, RNA, how they all fit together, how they form pathways which are the sort of signaling mechanisms for cells.

AI在生物学与药物设计 AI in Biology and Drug Design

Max

如果你患病,疾病如何改变这些通路的运作方式,这就能给我们线索,让我们知道如何介入、也许修复这种状况。所以这是等式的一边:理解我们如何运作、细胞生物学如何运作、疾病生物学如何运作。等式的另一边则是:好,那你如何设计一种干预手段?那就是药物,药物就像一种分子,比如可以做成药丸吞服的小分子,或者叫抗体、肽之类需要注射的东西。你究竟如何设计一个分子,让它按你的意图去调节你的生物系统,从而解决那种疾病状态?所以 AI 在这两个方面都给了我们非常非常强的帮助。

If you have a disease, how that changes how these pathways operate, and therefore that gives us a clue into how we can go in and maybe fix the situation. So that's one side of the equation: to understand how we work, how cell biology works, how disease biology works. And then the other side of the equation is, okay, how do you then design an intervention to that? So that's a medicine, and a medicine is something like a molecule, like a small molecule you could take as a pill, or something called an antibody or a peptide that you inject. How do you actually design a molecule that will go and modulate your biological system as you intended to solve that disease state? And so really AI is helping on those two fronts really, really strongly for us.

Host

我想从商业角度来理解,因为你描述的看起来像是非常长期的研究。你如何把它拆解成商业目标,又怎么知道什么时候该把你们在做的东西商业化?

I'm trying to understand from the business perspective, because what you're describing looks like very long-term research. How do you break it down into business goals, and how do you know when to commercialize what you're doing?

Max

好问题。我们创办公司时,并不想坐在黑暗的洞穴里,埋头搞核心引擎和模型搞十年,然后突然冒出来说:嘿,我们准备好商业化了。整个想法是,从第一天起就开发核心技术、AI 模型、平台,同时从第一天起就真正尝试把它用于最难的商业问题。公司很早的时候,我们就在做真正的药物设计,不只是我们自己发起的工作——这很重要——还主动出去寻找世界上最好的人、最好的制药公司,与他们签下合作,他们直接把业界最难的问题交给我们。这样你就没法自己给自己打分,你真正被置于行业的最高标准之下。那很难,但它确实迫使我们以真正能解决真实商业问题的方式来开发我们的模型和 AI 药物设计引擎。所以这让公司很早就有了商业基础。我们做药物设计大致有两个方面。我们有内部项目,自己发起、自己全额出资、拥有全部知识产权,自己开发这些分子。然后我们也与合作伙伴一起做。所以这构建了两种不同的商业模式,我认为它们配合得非常好。

Great question. When we started the company, we didn't want to go and sit in a dark cave and work away on the core engine and models for a decade and then pop up with the solution. Now, hey, we're ready to commercialize this. The whole idea was to, from day one, be developing the core technology, the AI models, the platform, and then also from day one to be able to really try and use this for the real hardest commercial problems. Very early in the company, we were doing real drug design, not just work that we instigated ourselves, which is really important, but also going out and just looking for the best people in the world, the best pharma companies, signed partnerships with them where they brought us literally like the industry's hardest problems. So you can't mark your own homework. You're really held against like the top standard of the industry. And that was tough, but it really forced us to develop our models and our AI drug design engine in a way that's going to really speak to true commercial problems. And so that started our commercial footing very early on in the company. And we think about sort of two aspects of how we do drug design. We do internal programs, making medicines that we instigate ourselves, we fully fund ourselves, we own all of the IP around, and we'll develop these molecules ourselves. And then we work with partners to also do this in partnership. So that builds two different commercial models, I think, that work really well together.

Host

在我的远程团队里,还有一件事我们没完美解决,那就是有一个地方能让人类和 AI 真正在同一个项目上协作。现在,调研资料在一个地方,旧脚本在另一个地方,制作人笔记又在另一份文档里,而 Claude 通常待在它自己的对话里。所以每次我们开始做新一期节目,总有人得来回复制旧的角度、问题、例子和上下文。这就是我们一直在测试 Miro 的原因,他们赞助了本期视频。我们会把一期节目的所有东西放到同一块 Miro 画布上:调研、过往脚本、听众反馈、问题和决策。然后我们通过 Miro 的 MCP 服务器把 Claude 连接到同一块看板。这样制作人不再在一个地方工作、Claude 在另一个地方,而是基于同一个项目上下文协作。Claude 可以读取看板上已有的内容,与之协作,并把输出写回那里。制作人可以审阅、加入自己的判断、调整角度,然后继续推进,而不必不停翻找旧的对话和文档。对我们来说,这意味着更少的复制和重复解释,更多时间真正花在节目上。如果你的团队已经在用 AI,但工作仍然散落在对话和文档里,去看看 Miro 的 MCP 服务器吧。链接在简介里。感谢 Miro 赞助本期视频。

There is one thing we still haven't solved perfectly on my remote team, and it is having one place where people and AI can actually work on the same project. Right now, research lives in one place, old scripts somewhere else, producer notes in another doc, and Claude is usually in its own chat. So every time we start a new episode, someone ends up copying old angles, questions, examples, and context back and forth. That's why we've been testing Miro, who's sponsoring this video. We'll put everything for the episode onto one Miro canvas: research, past scripts, audience feedback, questions, and decisions. Then we connect Claude to the same board through Miro's MCP server. So instead of the producer working in one place and Claude in another, they're working from the same project context. Claude can read what's already on the board, work with it, and write the output back there. The producer can review it, add her judgment, change the angle, and keep moving without constantly digging through old chats and documents. For us, it means less copying and re-explaining and more time actually working on the episode. If your team already uses AI, but the work is still scattered across chats and docs, check out Miro's MCP server. Link is in the description. Thanks to Miro for sponsoring this video.

Host

你在看这些新技术时,有没有一套商业框架?你怎么知道它们已经准备好商业化了?比如它们什么时候离开实验室、变成产品?

Do you have a business framework when you look at these new technologies? How do you know they're ready to be commercialized? Like when do they leave the lab and become a product?

Max

总的来说,我认为这因行业而异。你可以想想自动驾驶汽车,你得达到 99.99999% 的准确率,才足够安全可以部署、才真正算商业产品。在药物发现领域,在 AI 之前我们其实并不擅长。我们其实并不擅长预测一个分子会有多有效。行业的现状是一个非常经验化的过程,是实验室里试错的过程,这意味着你其实不需要拥有完美的预测能力,就能在你的药物发现项目上创造真正的价值、科学价值。所以我们真正思考的是:什么样的预测准确率水平能真正改变药物设计者的游戏规则,帮我们解锁别人无法处理的问题。所以当你达到那个准确率水平——而这个准确率水平其实不需要非常非常非常高,只要比业界其他人能获得的水平有明显提升就行——你就可以开始在内部使用它。我们有一套框架:我们做核心前沿 AI 模型的开发,有很多基准和评估作为这些模型有多好的代理指标,然后我们积极地把它们转化到真实的药物发现项目上,确保我们在内部基准、静态基准上取得进展的同时,这能转化为真正的科学突破,从而在真实的发现项目上带来经济价值。而且你常常会看到,你在机器学习基准上取得进展,却没有转化为现实,那是因为你的机器学习基准或评估这个代理指标和真实问题之间存在脱节。你必须回头重新设计你的排行榜和评估,更好地反映真实的实际情况。

In general, I think this is different industry to industry. You can think about something like self-driving cars, where you have to get to like 99.99999% accuracy before it's safe enough to deploy and it's actually like a commercial product. In the space of drug discovery, we're actually not very good before AI. We're actually not very good at predicting how good a molecule is going to work. The status quo of the industry is a very empirical process. It's sort of trial and error in the lab process, which means that actually you don't need to have perfect predictive power before you can create real value, scientific value, on your drug discovery programs. So we really think about what is that level of predictive accuracy that really changes the game for drug designers and helps us unlock problems that other people wouldn't be able to work on. And so when you hit that level of accuracy—and that level of accuracy actually doesn't need to be very, very, very high, it can be just like a significant delta above what the rest of the industry might have access to—you can then start using that internally. We have this framework where we're doing this core frontier AI model development. We have lots of benchmarks and evals that act as proxies for how good these models are, and then we just aggressively translate these onto live drug discovery programs to make sure that as we're making progress on our internal benchmarks, our static benchmarks, that translates into real scientific breakthroughs and therefore economic value on our live discovery programs. And oftentimes you can see that you're making progress on your machine learning benchmark and it's not translating into reality, and that's because there's a disconnect between this proxy of your machine learning benchmark or your eval and the real problems. You have to go back and redesign your leaderboard and your evals to better reflect the real on-the-ground situation.

Host

这很有意思。你如何管理你的雄心?你怎么看待这件事?

It's interesting. How do you manage your ambition? How do you think about that?

Max

在这个领域,这归根结底是要确保我们确实有足够大的雄心范围、足够广阔的业务模式,使我们能够为创造新模型和打造这个药物设计引擎的能力,合理化巨额投资。这其实又回到公司创立之初,真正去评估:那个巨大的市场潜力是什么?对患者可能产生的潜在影响是什么?其中有多少现在能借助 AI 以不同的方式实现?这有助于支撑我们对核心建模和核心科学的巨额投资,然后我们才能相信它最终会转化出成果。

In this space, and it comes back to making sure that we actually have the scope of ambition that's large enough, a business model that is expansive enough to mean that we can rationalize huge investment into our ability to create new models and to create this drug design engine. And this is really back to the beginnings of the company, really trying to assess what is that big market potential, what's that potential impact for patients that could be, and how much of that can be realized differently now with AI. And that helps underwrite this amazing amount of investment into the core modeling and the core science, which we then can have faith will translate at the end of the day.

使命与雄心 Mission and ambition

Host

当目标如此宏大时,你如何忠于自己的使命?你有没有经历过这样的阶段:你知道这事行不通,技术还没到位?

How do you stay true to your mission when the goal is so ambitious? Have you ever had periods when you know this is not working out, the technology is not there yet?

Max

是的,我们的使命很疯狂,你知道,解决所有疾病,这相当大胆。我们是板着脸说这句话的,我们真的感受到并认真对待这个使命。我们知道它不会一夜之间实现,所以对此很坦诚。坦率地说,它也不会只在 Isomorphic 的四堵墙内实现,但我相信你需要那种宏大的雄心,那种尽可能最大的使命,因为一旦你开始缩小你认为自己影响力能触及的范围,你就会以不同的方式搭建公司,开始设计一些与那份雄心不太匹配的目标。

Yeah, our mission is crazy. You know, solve all diseases. It's pretty audacious. We say that with a straight face. We really feel and mean that mission. We know it's not going to happen overnight, so we're honest about that. And frankly, it's not going to happen just within the four walls of Isomorphic, but I believe you need that big ambition, that biggest mission possible, because if you start reducing what you think your scope of impact can be, you start building out the company in different ways and you start designing different goals that don't quite match that ambition.

Host

但话说回来,这个使命太大,任何人都无法只装在心里。你如何把它转化到日常工作中?

But then saying that this mission is too big for any one person to just hold in mind. How do you translate that into the day-to-day?

Max

在我脑子里,我会想一棵使命树。你有那个宏大的“解决所有疾病”的使命。好,往下一层是什么?好,我们需要解决制药的方式,然后一层一层往下。所以我们会在整个公司层面思考使命,甚至细化到每个人。我希望每个人都有自己个人的使命,当你把这些全部拼在一起时,它们都指向我们顶层的“解决所有疾病”使命。然后再想,好,有哪些里程碑,哪些实际的里程碑可以开始衡量,帮我们做试金石式的检验:我们还在这个路径上吗?我们推得够狠吗?我们是否需要做些不同的事?

So in my head I think about a mission tree. You have the big solve-all-disease mission. Okay, what's one level down from that? Well, we need to be solving the way we make medicines, and then on and on and on. So we think about missions throughout the company, even down to every individual. I want every individual to have their own personal missions that, when you put these all together, are speaking to our top-level solve-all-disease mission. And then thinking about, okay, what are the milestones, the practical milestones that we can start to measure to help us litmus test, okay, are we still on this path? Are we pushing hard enough? Do we need to be doing something different?

Host

这真的很有意思。我刚和 OpenAI 联合创始人 Greg Brockman 聊过,那次对话我的关键收获是,他告诉我我们必须致力于自己的雄心,因为 AI 让那么多事情成为可能。还有你刚才说的,如果你的雄心更低,公司就会以完全不同的方式运作,你会开始去解决更小的问题。所以你必须要有那颗北极星。那么,如果创始人们正处在大使命上、但周围所有人都在怀疑、甚至研究也没证明他们是对的,那些最黑暗的日子里,你有什么建议?他们能对自己说什么,又能做什么?

This is really fascinating. I was just talking to Greg Brockman, co-founder of OpenAI, and my key takeaway from that conversation was he told me we have to be working on our ambition because AI makes so many things possible. And something that you just said, if you have lower ambition, then the company works in a completely different way and you start solving for smaller problems. So you have to have that North Star. So if you have any tips for the darkest days for founders when they're on a big mission but everyone around them is doubting and maybe even research is not proving that they're right, what can they tell themselves and what can they do?

生命信号与排行榜 Signals of life and leaderboards

Max

首先,我认为这就是为什么真正拥有一个使命很重要:知道它是什么,并花时间去把它想透,做到“是的,好,我可以完全支持它,然后拼命推进”。对我个人来说,Isomorphic 的使命是我可以为之奉献一生、无论多难都无怨无悔的。我认为你必须把它转化成一些“生命的信号”。这样说有道理吗?这样你就不会为了追求一个需要很长时间的宏大目标而过于疯狂。那些小小的目标桩是什么,哪怕它们是人为设定的?

First of all, I think that's why it's important to actually have a mission, to know what that is, and to have spent the time to rationalize that and be like, yeah, okay, I can fully get behind this and just drive hard on this. For me personally, the Isomorphic mission is one that I can just dedicate my life to and have no regrets behind, no matter how hard it is. I think you have to translate this into some signals of life. Does that make sense? So that you don't go too crazy ahead trying to pursue something really big that's going to take a long time. What are those little goal posts, even if they're artificial?

Host

是的。

Yeah.

Max

我举个例子。我们有一个特定的建模任务,从公司成立之初就存在。我们说,我们想解决所有疾病,当然,但这个团队在这个特定建模任务上每天的感受如何?于是我们设定了一个疯狂宏大的量化目标:你需要在这个特定基准上达到 80% 的准确率。哦,顺便说一句,全世界在这上面只有 10% 的水平。所以它感觉遥不可及,

I'll give you an example. We had this particular modeling task and it was one that has been there since the beginning of the company. And we said, we want to solve all disease, sure, but how does this team on this particular modeling task feel that day to day? Well, we set this crazy ambitious quantitative target. You need to hit 80% accuracy on this particular benchmark. Oh, and by the way, the world is only 10% good at it. So it feels so out of reach,

Host

是的。

Yeah.

Max

但它仍然是一个非常具体的数字。每天你都能看到它爬升 1%、2%、3%。所以你在朝它前进。它看起来仍然很远,但你能看到相对于它的进展。而实际上,当你年复一年地复利累积,你就能真正开始达到这些目标。所以我认为要始终尝试把这些非常宏大的定性想法转化成某种定量的东西。哪怕它稍微有点人为,也能帮助团队聚焦,以一种非常游戏化、竞争性的方式让头脑聚焦。我们真的把这些叫做排行榜。无论你在公司的哪个部门,你都有一个正在努力磨的量化目标。

but it's still a very tangible number. And day-to-day you can see climbing 1%, 2%, 3%. So you're moving towards that. It still seems very far, but you can see progress against that. And actually as you compound this year over year, you can actually start to reach these things. So I think always trying to translate these really big qualitative ideas into just something quantitative. Even if it's slightly artificial, it can help focus teams, focus the mind in just a very gamified competitive way. We literally call these leaderboards. And it doesn't matter what part of the company you're in, you've got your quantitative goal that you're trying to grind towards.

生物学之外的AlphaFold时刻 AlphaFold moments beyond biology

Host

我很喜欢这个。这对创始人来说是非常非常实用的建议。是的。你环顾四周,觉得除了生物学,还有哪些行业已经准备好迎接自己的 AlphaFold 时刻?

I love that. That's very, very practical advice for founders. Yeah. Do you think there are industries that are ready for an AlphaFold moment like biology when you're looking around?

Max

是的,所以在生物学之外,我认为有很多机会去找到 AI 现在就能应用、但仍然需要一个类似 AlphaFold 时刻的地方。那这看起来是什么样?我们在核心 LLM、推理智能体上都有这些惊人的进展。它们帮助我们理解数据、以我们的方式对世界进行推理,但在智能上开始远远超过我们。如果你想想生物学领域,那就像是人类生物学家会看数据、看一些实验数据,试图从中得出结论,并假设下一步可以做哪个实验。现在这些推理模型能够做到这一点,甚至可能开始在这方面超过人类。但它们真的能预测实验结果本身吗?它们能模拟真实世界吗?也许不能。也许在这个阶段还不能。也许有一天它们也能训练模型做到这一点。所以我认为会有很多应用 AI 的领域,推理模型会提出假设,或帮助你理解和分析世界,但你仍然受限于进入真实世界、去戳真实世界、看看它的反应是什么。做一个物理实验。生物学是这样,材料科学是这样,几乎任何你做的物理设计工作都是这样,很多工程领域也是如此。在那些情况下、那些领域里,我认为存在类似 AlphaFold 时刻的机会,也就是有能力创建这些 AI 模型来模拟那部分物理世界,这样你就不用去做物理实验,而是去戳你的模型,看看反应是什么。这与 LLM 和推理模型非常互补。

Yeah, so outside of biology, I think there's a lot of opportunity to find places that AI can now be applied and that still require an AlphaFold-like moment. So what does this look like? We have all of this amazing advancement on our core LLMs, our reasoning agents. These help us understand data and reason about the world in the way we do, but starting to far surpass us in terms of intelligence. If you think about the biology space, that could be just like a human biologist will look at the data, look at some experimental data and try and draw conclusions from that and hypothesize about what experiment can be done next. Now these reasoning models are able to do that and starting to maybe even outperform people in doing that. But can they actually predict what the experimental outcome will be itself? Can they simulate the real world? Maybe not. Maybe not at this stage. Maybe one day they'll be able to train the models to do that as well. So I think there's going to be a lot of applied AI domains where the reasoning models will come up with hypotheses or help you understand and analyze the world, but you still are bottlenecked by going into the real world and poking the real world and trying to see what the response of that is. Doing a physical experiment. That's the case in biology. It's the case in material science and almost any form of physical design work that you're doing, a lot of spaces of engineering. And in those sort of cases, in those sort of domains, I think there's an opportunity for AlphaFold-like moments, i.e., the ability to create these AI models that simulate part of that physical world so that instead of going and doing the physical experiment, you can poke your model instead and see what the response is. That's very complementary to the LLMs and reasoning models.

判断模型何时未就绪 Knowing when the model isn't ready

Host

那你怎么知道模型还没准备好解决那个问题?你有没有遇到过这样的情况:你有了这个想法,

And how do you know that the model is not ready to solve that yet? Have you ever had a situation where you have this idea,

Host

呃,但然后你看着模型,意识到这项技术距离解决它还有十年。

uh, but then you look at the model and you realize the technology is still a decade away from solving it.

Max

是的。我认为这在很大程度上取决于有什么数据可用。所以机器学习、AI 显然只能通过用数据训练来运作。

Yeah. I think a lot of that comes down to what data is available. So machine learning AI only works, obviously, by training on data.

数据成为瓶颈 Data as the Bottleneck

Max

我觉得你得先看清楚:需要的到底是什么样的预测能力,以及有哪些数据里可能含有信息量、能帮助预测这件事。如果你数据不多,或者数据非常非常嘈杂、维度很低,数据里其实没有多少信号,那么在现有数据条件下,可能就没有多大空间能对这个建模任务产生很大影响。顺便说,这对人们来说也可能是个机会,因为这时你就该开始想:好吧,什么样的数据可能存在,能解锁这个问题?我觉得这个领域里会有很多这样的问题——你看世界上目前能拿到的数据,它很小,没有正确的信号,太嘈杂,但也许有一种新的数据生成技术,或者一种收集数据、生成新数据的方式,能突然解锁以前没被解锁过的新领域。

I think you have to start by looking at, okay, what is this predictive capability that's needed, and what data is there that might have some information content inside of it that could speak to helping predict that thing. And if you don't have much data, or the data is very, very noisy, or it's very low dimensional, there's not actually much signal in this data, then there might not be actual scope, given the current data, to make a lot of impact on this modeling task. And that could also be an opportunity for people, by the way, because then you should actually start thinking, okay, well, what data could exist to unlock this problem? And I think there'll be many problems in this sort of space where if you look at the current data that the world has access to, it's quite small, it doesn't have the right signal, it's too noisy, but maybe there's a new piece of data-generating technology or a way we can collect data or generate new data that could then suddenly unlock new domains that haven't been unlocked before.

Host

所以你认为模型已经足够强大,能做到远超我们目前所能实现的事情,只是因为我们缺了数据这一层?

So you think the models are powerful enough that they can do way more than we're able to achieve, just because we're missing the data layer?

Max

我是个深度学习极大主义者。我认为我们已经有能解决几乎任何问题的算法。我们现在有算力,算力非常充裕。当然,每个人都受算力约束,但可用的算力实在太多了。所以归根结底是:如果我们有算法、能拿到算力、有出色的人能把这些整合起来,那我们就得确保有数据。而这些极具潜力的应用 AI 领域,很多都受限于可用于训练的数据。

I'm a deep learning maximalist. I think we have the algorithms that can solve almost any problem. We have the compute now. We have just an abundance of compute. Of course, everyone's compute constrained, but there's so much compute available. So it really comes down to: if we have the algorithms, we have access to compute, we've got amazing people that can put this together, then we need to make sure that we have the data. And a lot of these amazing applied AI potential domains, they're limited by the data that's available to train on.

Host

那是个商业点子。我们看到很多企业正是在这里冒出来——收集数据,并确保这些模型能拿到数据。当你招那些没有生物学背景、但擅长 AI 的人时,你具体在找什么?

That's a business idea. That's where we see a lot of businesses emerging, collecting data and making sure it's accessible to those models. When you hire people with no biology background who are good at AI, what exactly are you looking for?

跨学科招聘 Hiring Across Disciplines

Max

像我们这样用 AI 解决疾病的使命,是个非常复杂的问题。你需要从很多不同学科招人。不管好坏,世界上并不存在用 AI 解决疾病或制药的世界级专家,因为这个领域存在的时间还不够长。如果有人声称自己是用 AI 做药物设计的世界级专家,那他用的很可能是我们内部不用的技术,是过时的东西。所以我们相信,必须招各学科里最优秀的人:最优秀的 AI 科学家、最优秀的 AI 工程师,加上最优秀的软件工程师、最优秀的化学家、最优秀的生物学家和最优秀的临床医生,把他们聚到一起,让他们开始说彼此的语言、协同工作,因为在这些 AI 进入的新领域里,这种交叉型人才还不太存在。我认为很多行业都会是这样,这意味着我们鼓励招那些就是最优秀的 AI 科学家和工程师的人。他们没有生命科学背景,对制药一无所知,但必须愿意、好奇、开放地去学这个领域。最终成为出色 AI 科学家和工程师、或出色化学家和生物学家的人,往往是相当聪明、热爱学习的人。而且我觉得这些领域里很多都是别人可以学会的。所以人进来之后,他们学怎么制药,学 AI 怎么运作,然后我们在内部培养出这种跨学科的科学家或工程师。

Like a mission like ours to solve disease with AI, it's a really complex problem. You need to hire people from lots of different disciplines. For better, for worse, there exists no world expert at solving disease or making drugs with AI, because this field hasn't existed long enough. And if someone maybe claimed to be a world expert at drug design with AI, it's probably using a technology that we don't use internally. It's something outdated. So we believe that we have to hire the best people of their respective disciplines. So the best AI scientists, the best AI engineers, with the best software engineers, with the best chemists and the best biologists and the best clinicians, and then bring them all together and then ask them to start speaking each other's language and working together, because those intersectional people haven't quite yet existed in these new spaces where AI is coming to these. And I think this will be the truth for quite a lot of industries, which means that we encourage hiring people who are just the best AI scientists and engineers. They've got no life sciences background. They don't know anything about making drugs, but they have to be willing and curious and open to learning this space. People who end up being fantastic AI scientists and engineers or fantastic chemists and biologists, they tend to be pretty smart people who love learning. And I think a lot of these fields are quite learnable by other people. So as people come in, they learn about how to make drugs. They learn about how AI works. And then internally we create this interdisciplinary type of scientist or engineer.

Max

其实这还带来另一个好处。不只是招到最优秀的人,而是你能让人以一种全新的视角去看另一个学科、另一个科学领域或技术领域,因为他们对它有点天真无知。于是你几乎像是意外地得到了对这些领域的第一性原理式分析和思考。而这正是你需要做的。他们需要做那种第一性原理式的思考,才能在这些新空间里构建东西。

And actually, there's another benefit that comes from that. Not just hiring the best people, but you get people looking at another discipline or another scientific field or technical field from a completely new perspective because they're kind of naive to it. And so you almost, like, by accident get this first-principles type analysis and thinking for these domains. And that's exactly what you need to do. They need to do that first-principles type thinking to build in these new spaces.

Host

是啊,我觉得现在顶尖公司的招聘就是这样,对吧?因为你找不到一个有五年经验的人来做那份工作,因为那工作根本不存在。你提到你在找懂得如何学习、有好奇心的人。你会问他们什么问题来判断他们是不是那种人?

Yeah, and I think this is how hiring for top companies is happening these days, right? Because you can't find someone for that job with like five years of experience just because it did not exist. And you mentioned you're looking for people who know how to learn and who are curious. Is there a question that you ask them to figure out their exact type of person?

面向学习者的面试问题 Interview Questions for Learners

Max

我会很坦诚地问:跟我讲讲学一个新领域的经历,或者当你身处一个对你完全陌生的领域时是什么感觉?那种感觉如何?你是怎么应对的?试着用这种方式看人怎么思考。很坦诚地问别人:你的技术工作在哪里失败过,你是怎么学的、怎么找到解决方案的。然后开始试着让人去回忆、去讲他们过去在这些非常难受的学习处境里的经历——那种你觉得一切都从头顶飞过、你理解不了周围世界的感觉。那一刻你是什么感受?你做了什么?最后你走到了哪里?

I ask, honestly, like, talk to me about learning a new field, or how did it feel when you operated in a space that was completely new to you? How did that feel? How did you approach that? To try and see how people think in that way. Asking people very honestly about where their technical work failed and how did you learn, how did you figure out the solution to that. And so start to try and get people to just recount and talk about what their past experience has been like in these really uncomfortable learning situations where you feel like everything's going over your head. You don't understand the world around you. How did you feel in that moment? What did you do? And where did you get to?

赞助商广告 Sponsor Break

Host

Max 说他的团队把不同领域的专家聚在一起,寻找愿意互相学习的人。而快速学习是我不断从构建 AI 工具的人那里听到的技能之一。我回顾了与 Replika、Anthropic、LinkedIn 和 Microsoft AI 领导者的对话,发现了反复出现的七项技能,从界定正确的问题、检查 AI 的工作,到随着工具变化而学习。我把第一份清单放进了我的通讯 Futureproof。你可以通过简介里的链接阅读并订阅这份通讯。它是免费的。

Max says his team brings together specialists from different fields and looks for people willing to learn from one another. And learning fast is one of the skills I keep hearing about from people building AI tools. I went back through my conversations with leaders at Replika, Anthropic, LinkedIn, and Microsoft AI, and found seven skills that came up again and again, from framing the right problem and checking AI's work to learning as the tools change. I put the first list in Futureproof, my newsletter. You can read the newsletter and subscribe through the link in the description. It is free.

用AI筛选假设 Using AI to Filter Hypotheses

Host

好,跟我讲讲工作中怎么用 AI。AI 能生成大量假设。你怎么知道哪个真正值得去追?

Okay, talk to me about using AI at work. So AI can generate a lot of hypotheses. How do you know it's actually worth pursuing?

Max

你在做药物发现,你想做出一种药,你想设计分子的下一个迭代。AI 能提出——随便你挑个大数字——成千上万、上百万种不同的下一步设计假设。这些里哪些好、哪些坏?你并不想在实验室里把成千上万甚至上百万个都测一遍。所以我们构建了一套互补的模型,也就是这些模拟模型、世界模型、虚拟实验室模型,让我们不用进实验室,就能开始预测:如果我们真的把这些分子做出来、放进实验室,会得到什么反应。

You're doing drug discovery. You're trying to make a medicine. You're trying to design the next iteration of a molecule. AI can come up with, choose your large number, like any thousands and thousands, millions of different hypotheses for next designs. Which ones of these are good and which ones are bad? You don't want to test necessarily all thousands or millions in the lab. So instead, we build a complementary set of models, which are these simulation models, these world models, virtual lab models, that allow us to use instead of going into the lab and allow us to actually start to predict what will be the response if we actually make these molecules and put them into the lab.

AI驱动的药物发现流程 AI-driven drug discovery pipeline

Max

我们会考虑一些模型、生成模型,以及设计智能体,让它们提出新的假设、设计和想法;然后还有另一类模型,像是虚拟实验室模型,我们仍然在计算机上运行它们,去精炼这些方案,把它们筛选到只剩最好的、我们最有信心在真实世界中会奏效的那些。接着我们进入真实世界,真正把这些分子做出来、测试它们,看看是否有效。那才是真正耗时又昂贵的部分。

So we think about some models and generative models and design agents that come up with new hypotheses and designs and ideas, and then other models that are like virtual lab models that we ping off on a computer still to refine those and whittle those down to just the very best that we have the highest confidence will be good in the real world. And then we go into the real world and actually make these molecules and test them out and see if they work. That's the really time-consuming and expensive part of it.

Host

那些测试的成功率是多少?给我讲讲这些数字。

What's the percentage of those successful tests? Run me through those numbers.

Max

是的,这算是整条流水线各部分的核心内部指标之一,你希望它随着时间不断变好。现在到了这个阶段:我们不再需要靠实验去筛选、走遍几十万种不同的设计,而是只做 20 或 50 个,命中率就非常高,实验成功率非常高,于是我们基本可以抛弃那种老办法——测试成千上万、许许多多的分子和设计。这就是我们在内部追踪的东西。再说,随着我们在核心模型基准上取得进展,我们希望看到它转化为真实的实验成功率。实验成功率越高,你需要的实验时间、迭代实验时间就越少。这和那种“实验室在环”的做法很不一样——那种做法只是想更快地与真实世界对接。而我们其实是想尽可能把真实世界搬到计算机上。

Yeah, and this is like one of those core internal metrics for the different parts of the pipeline that you want to see get better over time. It's at the point in time now where instead of having to screen experimentally, go through hundreds of thousands of different designs, we can make 20 or 50 and there's such a high percentage of hits, of like experimental success there, that we can just sort of get rid of the old style of just testing many many many molecules, thousands of designs. So that's the sort of thing we track internally. Again, as we make advancements on our core model benchmarks, we want to see that translate into that real experimental success rate. And the higher the experimental success rate, the less experimental time, iterative experimental time you'll need. That's quite different from trying to do like a lab in the loop style where you just want to interface with the real world quicker. We actually want to just bring the real world onto a computer as much as possible.

研究者工作如何变化 How the researcher's job is changing

Host

哇,研究者的工作正在发生变化,这很有意思,对吧?研究者还会自己提出假设吗,还是你觉得 AI 现在太擅长这个了,以至于研究者的工作已经变了?

Wow, this is interesting how the researcher's job is changing, right? Do researchers still generate hypotheses, or do you think AI is so good at it now that the researcher's job has changed?

Max

我认为它变了,而且无论你是 AI 研究者还是化学家,都变了。两边我都能举例。如果你是 AI 研究者,过去你会提出一个改动模型的假设,然后把它实现、跑实验、看是否有效。现在我们越来越多地看到自动研究发生,一个 AI 研究者不再提出单个想法,而是提出模型改动的设计空间以及这个想法的目标——我想以这种风格提升效率、吞吐量或准确率。然后由智能体和模型来提出具体想法,并跑实验。作为化学家,情况非常类似。以前,如果想改进分子设计,化学家可能会在白板上画出分子,开始改动单个原子,也就是自己提出不同的设计假设。现在 Isomorphic 的化学家、药物设计师,转而定义设计空间是什么、我下一轮设计的目标是什么。这就为我们的设计智能体定义了一个目标,然后智能体自主工作,完成所有单个设计,给出下一轮迭代。

I think it's changed, and it's changed whether you're an AI researcher or a chemist. I can give examples on both. If you're an AI researcher, you used to propose a hypothesis of a change of a model and then implement that and run the experiment and see if it works. Now we're seeing more and more like auto research happening on the grounds where an individual AI researcher is not proposing an individual idea, they're proposing the sort of design space of model changes and the goals of that idea. I want to increase the efficiency or throughput or accuracy with this sort of flavor. And the agents and models are coming up with the ideas and then running the experiments on that. As a chemist, it's a very similar thing. Before, if you want to make improvements on a molecule design, a chemist might be drawing out the molecule on a whiteboard and starting to change individual atoms, so coming up with different design hypotheses themselves. Now a chemist at Isomorphic, a drug designer, is instead defining what's that design space? What are the goals of my next round of design? And that defines an objective for our design agents, and then the agents are working autonomously to do all the individual designs and come up with the next iteration.

个人AI工具栈 Personal AI stack

Host

这真的很有意思。除了设计之外,你个人用来提升生产力的 AI 工具栈是什么样的?

That's really fascinating. And apart from designing, what's your personal AI stack to run your productivity?

Max

我有一大堆不同的智能体在跑,基本上所有东西都接进去了。我们在内部大量使用 Gemini,也大量使用各种不同的模型。有一个例子对我来说简直不可思议,就是在数据分析和数据可视化上——你可以想象我们有很多不同的运营在进行。现在有了这种智能体栈,你可以把公司所有数据都接进去,把各个项目里的不同数据都接进去,然后开始真正把它们梳理出来、可视化出来。通常这需要你身边有一位出色的设计师来做,但现在完全可以由我这样一个经营公司的人来驱动。我们大量使用 Gemini,也几乎用了所有模型,包括开源模型。

I have a whole host of different agents that are running across my, like, everything's plugged in. We use a lot of Gemini internally. We use a lot of all the different models. One example that's just been mind-blowing for me has been actually on the analysis of data and the visualization of data, where you can imagine we have lots of different operations going on. And now with this sort of agent stack, you can just plug in all of your company data. You can plug in all of the different data across your different programs and start to actually tease this out and visualize this in a way that normally this would take you having an amazing designer next to you to do this. But instead this can just be driven purely by me as someone running the company. We use a lot of Gemini. We use a lot of pretty much all the models, including open source.

长期运行的药物设计智能体 Long-running drug design agents

Host

我读到你们有一些药物设计智能体能连续运行两到四天。在提示词方面,有什么办法让你能信任它们跑这么久吗?

I read that you have some drug design agents that run for two to four days consecutively. Is there something about prompting them to trust them to run for so long?

Max

这是大量研究的成果。我们在这个领域有一支很棒的团队,研究如何打造这些设计智能体。所以这综合体现在提示词、运行框架,以及我们如何定义智能体运行时的成功。

This is the result of a lot of research. We have an amazing research team in this space of how do we create these design agents. So this comes together in terms of the prompt, the harness, how we define success as these agents run.

Host

你怎么定义一个智能体的成功?

How do you define success for an agent?

Max

这与其说是准确率,不如说是定义智能体的目标空间。也就是这个智能体想达成什么?在药物设计这类事情上——我认为这对很多设计问题都普遍适用——你要达成的不只是一个量,很可能是一整组不同的属性。所以要让智能体明白:好,这些属性都有哪些?它们的目标空间是什么?其中很多属性彼此冲突。你可能把属性 A 推得非常高、非常好,但结果属性 B 变差了。所以这不是一个贪心优化过程,而是一个真正开放式的搜索与优化过程,你需要能够探索和利用。这其实又回到了强化学习的那套核心思想。

This is less about accuracy. It's more defining the goal space of the agents. So what is the agent trying to achieve? In something like drug design, and I think this is common for a lot of design problems, there's not just one quantity you're trying to achieve. There's probably a whole host of different properties that you're trying to achieve. So getting the agent to understand, okay, what are all those different properties? What's the goal space for these? A lot of these properties are in conflict with each other. You might be pushing up property A that's getting really really good, but as a result property B gets worse. So this isn't a greedy optimization process. This is actually this really open-ended search and optimization process where you need to be able to explore and exploit. It really goes back to these core reinforcement learning type ideas.

决定自动化什么 Deciding what to automate

Host

那你怎么决定用智能体去自动化哪个问题?有没有一套框架,比如“好,这件事我重复做了五次,现在该把它放下去、搭个东西出来了”?

And how do you decide which problem you're going to automate with an agent? Is there a framework where you're like, okay, I repeated this five times, now it's time to release it down and build something out?

Max

我认为在开发上,你必须真正以评估和排行榜为驱动。我觉得现在用智能体很容易自欺欺人,比如“哦对,你看,我可以用智能体把这个自动化,多棒啊”。但结果好吗?它帮你省时间了吗?它在下游创造了更多科学价值或经济价值吗?所以真正重要的是要有你的评估、你的排行榜:什么算好?你怎么把它编码化?如果你有幸在一个可验证的领域工作,这在计算机上可能是强可验证的,比如你能判断智能体是对是错。但在生物学和药物发现领域,这在计算机上并不真正可验证。所以在判断智能体做得好不好时,要软性得多。但努力把你的评估基准定下来——

I think you just have to be really eval driven and leaderboard driven in terms of the development here. I think it's very easy to fool yourself with agents at the moment, like, oh yeah, look, I can just automate this thing with agents and isn't this great? But is the result good? Is it saving you time? Is it creating more scientific value or economic value down the line? And so it's really important to actually have your evals, your leaderboard of what does good look like, how do you codify that? If you're lucky to work in a verifiable domain, this might be strongly verifiable on a computer. Like you can say whether the agent is right or wrong. In the space of biology and drug discovery, this isn't really verifiable on a computer. So it's a lot softer in terms of understanding if an agent is doing well or not. But trying to get down your evaluation benchmark—

Host

嗯。

Mhm.

衡量智能体进展 Measuring Agent Progress

Max

对于智能体系统的每一个不同部分,都要真正评估:你在这里取得进展了吗?情况在变好吗?这个智能体在这个场景下真的比人类做得更好吗?或者它比人类工作得更快吗?真正开始进行实证测量,然后你就可以将开发时间引导到技术栈的正确部分。

For every different part of your agent system, really evaluate: are you making progress here? Is this getting better? Is this agent actually doing better than the human would in this scenario? Or is it working faster than a human would? Just actually starting to empirically measure this, and then you can direct your development time into the right part of the stack.

播客订阅请求 Podcast Subscription Request

Host

在我们继续之前,能请你帮个小忙吗?如果你喜欢这些对话,我非常感谢你订阅。这是支持我和 Silicon Valley Girl 团队的一种免费方式,对我们意义重大。谢谢。

Before we keep going, can I ask you a small favor? If you've been enjoying these conversations, I'd really appreciate you subscribing. It's a free way to support me and the team behind Silicon Valley Girl, and it means a lot to us. Thank you.

不可自动化的工作 Un-automatable Work

Host

你的工作中哪部分你还没有自动化,并且你认为你永远不会自动化?

What part of your work haven't you automated yet and you think you will never automate?

Max

是的,这是个很好的问题。我的意思是,我花大量时间与公司周围的不同人交谈,管理我们公司以及我们使命背后的人类。我认为这是目前无法自动化的,而且目前可能也不应该被解决。

Yeah, so it's a really good question. I mean, I spend a huge amount of my time talking to the different people around the company and managing the humans in our company and behind our mission. And I think that's something that I don't think is automatable at the moment and probably shouldn't be solvable at the moment.

Host

是的,我认为有时你进行 20 分钟的对话,它会改变你的整个策略和你对该领域的思考方式,这就是我创办这个播客的原因。

Yeah, I think sometimes you have a 20-minute conversation and it changes your whole strategy and the way you think about the field, which is why I started this podcast.

Max

是的,是的,是的,我认为再次像这些模型可以提出许多惊人的假设,你必须思考,好吧,真正的瓶颈资源是什么?即使这些模型达到 AGI,也将会是人类注意力、时间、物理空间。这些是限制性资源,我认为永远无法解决。

Yeah, yeah, yeah, and I think again like these models can come up with so many amazing hypotheses and like you have to think, okay, what is the real bottleneck resource? Like even as these models get to AGI, it's going to be like human attention, time, physical space. These are like the limiting resources that never solvable, I think.

FDA成为瓶颈 FDA as Bottleneck

Host

另一个瓶颈,实际上 Sam Altman 提到过,我们有一个创作者问答,他被问及使用 AI 治愈癌症,他说我们可能明年就会看到某种类型的癌症被治愈,但瓶颈将是 FDA。你同意他的观点吗?

Another bottleneck that actually Sam Altman mentioned, we had this creator Q&A and he was asked about using AI to cure cancer and he said we might see something already next year curing some type of cancer, but the bottleneck will be the FDA. Do you agree with him?

Max

所以目前,通过监管流程需要一些时间,这是正确的,需要大量证据来确保:好吧,你认为你发明了一种神奇的药物,可以开始在临床试验中在人体上测试。你必须通过 FDA 的监管流程,这非常重要,因为必须将患者安全放在首位。所以目前,收集所有不同动物模型和所有不同数据点需要一些时间,以说服专家:是的,这个分子看起来在人体中应该是安全的,也许它对特定疾病领域有效。只有那时你才能开始临床试验,并真正开始在人体上测试。然后当你真正开始临床试验并在人体上测试时,即使有所有这些事先批准临床试验的数据,仍然有 90% 的药物在临床中失败,因为有一些意外的副作用或意外的毒性。所以当你开始这些临床试验时,你必须开始得相当缓慢。最初,你在第二阶段之前进行所谓的第一阶段试验。第一阶段试验是关于弄清楚你的分子是否有毒。患者能否耐受 decent 剂量,即治疗剂量?所以你必须从非常非常小的剂量开始,逐渐增加。这需要一些时间,但为了保护患者安全,这是非常必要的。但之所以这样,是因为我们不擅长预测这些分子会有多有效。我们不知道它们的机制。我们无法很好地预测这个领域的毒性。我们做这些动物模型来研究毒性,但这些动物模型只能预测约 30% 的人体毒性,因为老鼠不是人。我们看到很多机会来真正提高我们对这些药物如何起作用的理解。随着时间的推移,最终,我认为每个人,我们、FDA、患者、所有人,都希望尽快、尽可能安全地将神奇的新药送到需要它们的患者手中。所以当你开始证明实际上我们可以更安全、更有预测力地做到这一点时,这应该能够推动改变,缩短时间,更快地将这些药物送到需要它们的患者手中。

So at the moment, it takes a bit of time and rightly so, it takes a lot of evidence to get through the regulatory process to make sure that okay, look, you think you've come up with an amazing medicine, can actually start testing this on people in clinical trials. You have to go through the regulatory process with the FDA and that's really important because you have to have patient safety sort of first and foremost in mind. So at the moment, this takes a bit of time to collect data on all the different animal models and all the different data points required to convince the experts that yeah, no, this molecule looks like it should be safe in people and maybe it'll work for this particular disease area. And only then can you start a clinical trial and actually start testing this out on people. And then when you actually do start a clinical trial and test this out on people, well, it turns out even with all of that data beforehand that goes into approving this clinical trial, still 90% of drugs fail in the clinic because there's some unintended side effect or some unintended toxicity. And so when you start these clinical trials, you have to start quite slowly. Initially, you do what's called phase one trials before phase two. Phase one trials are all about figuring out if your molecule is toxic. Can patients tolerate a decent dose, therapeutic dose of this medicine? So you have to start really, really small and gradually work your way up. That takes a bit of time, but it's really necessary to protect patient safety. But it's that way because we're not very good at predicting how well these molecules are going to work. We don't know the mechanisms of those. We can't predict very well the toxicity of this space. We do these animal models for toxicity, but these animal models, they're only like 30% predictive of human toxicity because a rat isn't a human. We see a lot of opportunity to actually improve how good we are at understanding what these medicines work. And over time, ultimately, I think everyone, us, the FDA, patients, everyone, everybody wants to get amazing new drugs to patients who need them as quickly as possible and as safely as possible. And so as you start to prove out the evidence that actually we can do this a lot safer and with a lot more predictive power, that should be able to spur change of how long this takes and how quickly we can get these drugs to the patients who need them.

AI设计药物时间线 AI-Designed Drugs Timeline

Host

那么你认为我们什么时候会在药房看到第一批由 AI 设计的药物?

So when do you think we're going to see the first drugs designed by AI at a pharmacy?

Max

是的,在药房,我不知道。通过整个流程需要一些时间,但我认为目前有很多惊人的东西正在被设计。我们开始看到一些东西进入临床,它们以某种方式接触了机器学习。我们正在非常努力地开发我们自己的完全由 AI 设计的药物管线,并尽快将这些药物推向临床。

Yeah, at a pharmacy, I don't know. It takes a bit of time to get through the full process, but I think there's a lot of amazing things being designed at the moment. We're starting to see some things entering the clinic that have had they've touched machine learning in some way. We're working really, really hard on our own pipeline of fully AI designed drugs and getting those to the clinic as quickly as possible.

令人兴奋的项目 Exciting Programs

Host

最令你兴奋的,如果你能分享的话?

The most exciting for you if you can share?

Max

是的,我们有一些我认为在肿瘤学领域非常令人兴奋的项目,所以不同类型的癌症,还有免疫学和炎症,真正尝试研究这些疾病的一些核心驱动因素。我们能否开始改变这些核心驱动因素的工作方式,当然它们在疾病状态下会发生突变或改变。所以看到进展非常令人兴奋。

Yeah, we have some just I think like really exciting programs in the space of oncology, so different types of cancer, also immunology and inflammation, like really trying to look at some of the core drivers of these diseases. And can we start to change how these core drivers work and of course they get mutated or changed in disease states. So this is really exciting to see the progress.

首批受益者 First Beneficiaries

Host

你认为哪些患者将首先从你们正在开发的药物中受益?

Which patients do you think are going to benefit first from the drugs that you're developing?

Max

我不知道。现在说哪些会首先进入临床还为时过早。是的。我能说的是,我们正在如此努力地推动这些首批潜在药物进入临床试验,开始看到对患者的影响。我们有一些非常有希望的东西。所以但愿一切顺利,我们的所有数据都像现在这样持续进展,是的,我们对此非常兴奋。

I don't know. It's too early to say which ones will hit the clinic first. Yeah. What I can say is like we're pushing so hard on getting these first potential medicines to clinical trial to start seeing that impact on patients. We have some really really promising stuff. So touch wood all of our data keeps progressing as it does and yeah we're really excited about this.

长寿与预防医学 Longevity and Preventative Medicine

Host

好的,保持 fingers crossed。嗯,我最近听到 Ray Kurzweil 说,你必须活到 2030 年,因为之后你活的每一年都会因为技术进步而为你增加超过一年的寿命。你总体上相信这个概念吗?

Okay keeping fingers crossed. Um I recently heard Ray Kurzweil said you have to live to 2030 because after that every year you live is going to add more than a year to your life because technology advances. Do you believe in this concept in general?

Max

我认为我们将在设计新药和理解疾病方面变得好得多,我不知道那个进展的速度会是多少,但我认为这其中有很大的道理。我们快进 10 年、20 年、30 年。我认为我们理解疾病的方式、我们思考解决疾病的方式将会有根本的不同。甚至思考如何理解疾病的发作、它如何运作,以及因此在疾病显现为真正危急状态之前的预防措施,都应该非常非常不同。

I think we're going to get so much better at designing new medicines at understanding disease so much better that I don't know exactly what the rate of that progress is going to be but I think there is very much something in this. We fast forward 10 years, 20 years, 30 years. I think the way that we're going to be understanding disease, how we think about solving disease is going to be radically different. Even thinking about how to understand the onset of disease, how it works and therefore preventative measures before the disease manifests itself into like a really critical state should be very very different.

个性化医疗 Personalized Medicine

Max

这将要求我们更深入地理解生物学如何运作,并测量我们的一切。你想想,几乎所有市面上的药物,我们使用的,都是全世界所有人共用一种分子。但疾病是如此细微复杂,在不同患者身上表现的方式千差万别。可以想象,有了 AI,我们将有更多机会更好地理解每个患者的个体情况,然后为这些患者设计更加量身定制的解决方案。所以我认为我们会看到走向更个性化医疗的轨迹,更加关注个体——患者群体,甚至随着时间推移可能细化到个体本身。

That's going to require us to understand so much more about how biology works and measure everything about us. You start thinking about pretty much all of the medicines that are out there and that we use—it's like one molecule for the entire population of the world. But disease is so nuanced and it manifests itself in different patients in very, very different ways. And you can imagine with AI there's going to be so much more opportunity to understand each individual case for patients much, much better, and then on the other hand design a much more bespoke solution for these patients. So I think we'll see a track towards more personalized medicine, much more focused on the individual—so cohorts of patients, even down maybe to the individual itself over time.

Host

你能对此做出任何时间线预测吗?

Can you make any timeline predictions around that?

Max

未来几年,随着这些新一波药物开始进入临床,我们应该会看到这种变化。它们应该能够以新的方式应对疾病。我们想问问——业内其他人也在努力攻克这些新靶点。当我们有了更好的方法来设计分子和设计新药,这将开辟新的前沿领域。所以我认为,甚至在未来 10 年内,我们应该会看到患者可用的药物类型发生巨大转变。是的,所以在这个十年里,我认为我们将真正开始看到这种转变。

We should be seeing this change now over the next couple of years as these new waves of medicines start coming into the clinics. They should be able to tackle diseases in new ways. We want to ask—other people right in the industry are trying to go after these new targets. As we have this better way to engineer molecules and engineer new medicines, that's going to open up new frontiers of things to work on. So I think even over the next 10 years we should be seeing a big shift in the types of medicines that are available for patients. And yeah, so over this decade I think we're going to really start seeing that shift.

预期寿命与瓶颈 Life Expectancy and Bottlenecks

Host

这很吸引人,我想知道我们的预期寿命会发生什么变化,下一个瓶颈会是什么?

That's fascinating and I wonder what's going to happen to our life expectancy and what will be the next bottleneck?

Max

是的,你可以想想——显然有些非常晚期的疾病是我们想要纠正的,但也有很多疾病只是与衰老相关,影响着我们的生活质量。这还没谈到试图逆转衰老、延长寿命。所以我认为有很多近期的机会来真正改善我们的生活质量。这还没涉及长寿和试图逆转器官衰老等具体事情。

Yeah, you can think about—there's obviously very advanced stages of disease that we want to rectify, but then there's a lot of disease that's just associated with getting older and that's afflicting our quality of life. That's before you even talk about trying to reverse and live longer. So I think there's a lot of near-term opportunities to really improve our quality of life. And that's before you get into things like longevity and trying to reverse age of organs and specific things.

给AI for Science的建议 Advice for AI for Science

Host

给那些未来五年真想从事 AI for science 的人一些建议。

Give advice to someone who really wants to work on AI for science in the next five years.

Max

我认为不要忽视学习科学本身。即使 AI 变得更好,能够产生许多假设,我们把算力聚焦在哪里?我们的算力仍然有限。我们需要思考你所在领域的核心科学问题是什么。我们究竟把即将到来的所有这些惊人的智能和智能爆发聚焦在哪里?我们聚焦在哪里?我们问什么问题?所以我认为,真正理解科学的基本原理及其背后的机制将至关重要。然后,在这个领域,要非常有创造力且积极地采用智能体和所有 AI 技术。

I think don't neglect learning science itself. Even as AI gets better and is able to produce so many hypotheses, where do we focus this compute? We still have a finite amount of compute. We need to think about what are the core scientific questions for your respective field to ask. Where do we actually focus all of this amazing intelligence and explosion of intelligence with compute that's coming? Where do we actually focus this? What questions do we ask? And so I think still really truly understanding the fundamentals of science and the mechanisms behind that is going to be imperative. And then just be really creative and aggressive with adopting agents and all of AI, I think, around the space.

Host

是的,归根结底是问对问题,这确实反映了研究是如何进行的,对吧?就是知道该研究哪个问题。

Yeah, it comes down to asking the right questions and it really reflects how research works, right? It's knowing which question to work on.

Max

是的,回答这些问题的过程将会改变。随着新技术出现,这一直在变。但该问什么问题,把有限资源聚焦在哪里,无论是人力时间、智能体时间还是湿实验室工作——总得聚焦在某个地方。

Yeah, the process of answering those questions—that's going to change. That's always been changing as new technology. But what questions to ask, where to focus your finite resources, whether that's human time, agent time, wet lab work—it's going to have to be focused somewhere.

研究品味与通才思维 Research Taste and Generalist Thinking

Host

你认为这取决于你获得的知识量吗?取决于你的品味、你所提问题的质量吗?

Do you think it depends on the amount of knowledge that you have acquired? Does it depend on your taste, the quality of questions that you're asking?

Max

是的,我认为品味是真实存在的。我实际上在内部经常谈论这个——研究品味——我认为那是一个函数。你的研究品味是你筛选该问什么问题、快速筛选出可能解决这些问题好方法的能力。我认为研究品味来自经验,经验帮助你建立直觉和对领域的理解。选择你最喜欢的科学领域或数学、工程、科学领域——这些都非常非常复杂。很多复杂性几乎太高维,无法以其真实状态在头脑中把握。所以你会想出心智框架和直觉,帮助作为科学家或工程师的你驾驭它。我认为这需要通过亲身经验和接触这些领域来建立。而且我认为保持足够通才也很重要。我是一个通才科学家-工程师。我喜欢这样,我认为你可以从数学和科学的其他领域学到很多东西并应用到别处。往往是那些交叉想法效果非常好。

Yeah, I think taste is a real thing. I actually talk a lot about this internally—research taste—and I think that is a function. Your research taste is your ability to filter what questions to ask and filter quickly to what could be some good ways to tackle answering those questions. And I think that research taste comes from experience, and that experience helps you build your intuition and your understanding of the domain. Choose your favorite scientific domain or area of maths, engineering, science—these are really, really complex. A lot of that complexity is almost too high-dimensional to hold in its true state in your head. So you come up with mental frameworks and intuitions to help you navigate that as a scientist or an engineer. I think that's the thing that needs to be built up through hands-on experience and just exposure to these fields. And also I think there's a lot to be said for keeping generalist enough. I'm a generalist scientist-engineer. I love that, and I think there's a lot that you can learn and take from other areas of maths and science and apply elsewhere. And it's often those crossover ideas that work really well.

Host

构建框架,我认为它只是教你的大脑理解正在发生的事情,并用简单的语言表达出来。我认为这对 AI 世界来说是一项非常棒的技能。

Building frameworks, I think it just teaches your brain to understand what's going on and put it into simple words. And I think it's a really great skill for the AI world.

Max

是的,是的。

Yeah, yeah.

结语 Closing

Host

非常感谢你,Max。这远不止是治愈疾病。我为你的使命祈祷,希望未来几年我们将看到 AI 对医疗保健的巨大影响。

Thank you so much, Max. That was so much more than just curing diseases. And I'm keeping fingers crossed for your mission and hopefully in the next few years we're going to see a great effect of AI on healthcare.

Max

非常感谢。非常荣幸。

Thank you so much. Been such a pleasure.

Host

如果你曾向 ChatGPT 问过健康问题,请观看我与斯坦福医生 Jonathan Chan 的对话。我们讨论了你提问的方式如何影响答案,以及 AI 在哪里有用、哪里可能出错。点击这里。

If you've ever asked ChatGPT a health question, watch my conversation with a Stanford doctor Jonathan Chan. We talk about how the way you ask can shape the answer and where AI can be useful and where it can go wrong. Click here.

互动版:逐字朗读 + 针对本期提问 →