什么是TechBio？——阿尼尔·马拉瓦拉普

非常感谢大家前来。在座各位大多是我、Anil、Lanay或Brady的朋友，甚至是我们所有人的朋友，能现场进行这次活动真是莫大的荣幸。我叫Paolo，正致力于投资那些我认为近乎疯狂的科学家们，帮助他们走出实验室创立公司。因为初创企业确实是推动创新、将新技术和发明带入现实的高效机制。所以我运营着一个名为'深度未来'的基金，我们专注做的就是发掘这类最杰出的项目。

Thanks a ton for coming out. Most of you guys are friends of either me or Anil or Lanay or Brady or all of us, and it's a real privilege to get to do this live. My name is Paolo, and I am trying to find ways to invest in what I think of as mostly like mad scientists and get more of them out of the lab and into a startup. And the reason for that is that startups have actually been a highly functional machinery for doing new things and bringing new technologies and inventions to life. And so I run a fund called Deep Future, and that's all we do is the best of these kinds of things.

其中涉及大量机器人、激光和核反应堆。但我明确不碰生命科学和生物领域——因为我喜欢可以重启的机器人。所有生物相关都超出我的能力范围。今晚与Aneel共同录制的'深度未来'播客，让我有机会向精通我不熟悉领域的智者取经，并带大家一同参与。顺便提一下，我上周刚出版了畅销书《深度未来》，内容正是关于这类技术及其实现过程，书中充满实例。

And it's a lot of robots and lasers and nuclear reactors. But one of the things I specifically don't do is life sciences and biological stuff because I like that you can reboot a robot. But all the biological stuff is out of my jurisdiction. So on the Deep Future podcast, which is what we're recording live tonight with Aneel, it's a chance for me to pick the brains of people who are really smart about things I'm knowing about and bring you guys along with me. I should mention, I wrote a bestseller last week that came out called Deep Future, which is all about the same kinds of stuff and how we bring these things to life, loaded with examples.

我带的书不够人手一册。这本书零售价约800美元，但今晚从我这里只需30美元就能买到——只要在售罄前从那堆书里拿一册。结束后我很乐意为各位签名。目前这本书反响很好，听到读者的反馈让我非常欣喜，这真是件乐事。

And I brought not enough for everybody. The book retails for like $800, but you get it from me for $30 if you get one off that stack before they're gone. And then I'll be happy to sign them at the end. So far, it's been really well received and delightful for me to hear how it's going from people who read it. So it's a real treat.

我并非想当作家，只是迫切想让更多人了解这些即将改变世界的技术。所有这些活动都是为了与大家分享这些内容。Linea和Neil有他们自己专注投资我所不涉足领域的风投基金。再次感谢各位的到来。

I'm not trying to be an author. I'm just desperately trying to get the word out about these kinds of technologies that we see coming that can really make a big difference in the world. And so all of this is a chance to share that stuff with you. Linea and Neil have their own venture fund really focused on the things that I don't do. Thanks for coming out.

我从未见过像他这样的人。

I've never met anyone like him.

他能在几分钟内黑进任何人的手机。这到底是谁？谁能做到这种事？Palos简直是世界上最有趣的人。

He could hack anybody's phone in a matter of minutes. Who is that guy? Who can do that? Palos is the most interesting man in the world.

他是我最喜欢的人之一

He's one of my favorite people on

这个星球的发明家，聪明绝顶，最擅长投机取巧。

the planet. Inventor. Wicked smart. The master of cutting corners.

如果你

If you

有一具尸体，还想给手机编程。他肯定会在列队指认时立刻被捕。

have a dead body and you're trying to code your phone. He would get arrested immediately in a lineup.

我说的是，爱。

I said, love.

给我们签个名吧。要知道，他的嘴根本跟不上脑子的速度。其实我也不知道帕布罗是做什么的。

To give us an autograph. You know, his mouth can barely keep up his brain. I don't actually know what Pablo does.

黑客魔术师。

A hacker magician.

未来之人。未来。未来。未来。未来。

Man of the future. Future. Future. Future. Future.

你

You

你有一个独特的视角，称之为‘技术生物学’。我们都听说过生物技术，但我想最好从你如何定义它以及为何认为有必要区分这两者开始。

have this unique angle on it is what you call tech bio. We've all heard of biotech, but I think the best place to start would be with how you define that and why you think it's important to make a distinction.

是的，这是个很好的问题。听起来像是同一回事，某种程度上是。人们会想，这不就是生物技术吗？我是说，技术和生物的结合。

Yeah, that's a great question. Sounds like the same thing. Kind of. And people think, Isn't that just biotech? I mean, it's tech and bio.

生物学中一直存在技术应用，但当前正发生重大变革，这主要归功于人工智能和大规模计算能力。同时，各类科学技术、化学知识和专业技能的积累也很关键，但真正的转折点在于我们现在拥有了AI。而生物技术是什么？我认为它真正兴起于上世纪七八十年代的学术实验室。生物学这个词，本质上是关于生命的研究。

And there's always been technology in biology when we've But done there's a big change that's happening right now, and that's largely because of AI and compute at scale. There's also been the accumulation of all kinds of scientific techniques and chemistry and all kinds of know how, but the real inflection point that's happening now is we have AI. What was biotech, though? Biotech was really something that I think of as emerging from the academic labs in the '70s and '80s. Biology, in the word, is about the study of life.

我会用稍显宽泛的笔触来描述这一特征，总体上它是成立的，当然边缘地带总能找到反例。但生物技术曾是门发现科学——我们探索生物学如何运作，寻找可以干预疾病相关蛋白质的切入点。我们掌握了操控DNA、理解蛋白质和细胞的技术，能开展大量研究。这正是生物技术在生物制药行业扮演的重要角色。

I will paint with a little bit of a broad brush in making this characterization as generally true, and then on the edges you can always find counter examples. But biotech was a discovery science where we were figuring out how does biology work, where are the places that we can intervene to touch a protein that's involved in a disease process. We have the tools for manipulating DNA, understanding proteins and cells, and we could do a lot of research. That's a lot of the role that biotech played in the biopharma industry.

因为我们不得不构建起那整套工具包。

Because we had to build that whole toolkit.

比如你怎么

Like how do you

观察这些蛋白质，如何分解分子，如何重组它们，如何测试它们的相互作用，这是其中很重要的一部分。

look at the proteins, how do you take the molecules apart, how do you put them back together, how do you test what they interact with, so that was a big part of it.

这是很重要的一部分，通常的互动方式是生物技术公司确定目标位置，然后将这些目标交给制药或医药公司，由他们做制药行业长期从事的工作——制造影响生物系统的小分子。只不过现在我们知道了目标。多年来，生物技术主要扮演着这样的角色。而现在这种情况正在改变，因为生物学正在成为一个新的工程平台。这是一个巨大的变革。

That was a big part of it, the way that interaction typically worked is that the biotech companies would figure out where do we target something, and then they would throw these targets over to pharma or to medicinal and they would do something that pharma has been doing for a long time, which is making small molecules that affect some piece of biology. It's just that now we knew the targets. And so that was largely the role of biotech for many, many years. And that that now is changing because biology is becoming a new engineering platform. And that is a huge change.

我认为，这甚至让我们都难以想象其意义，因为当我们看着自己的手或一株植物时，我们不会将其视为技术或工程。我们甚至无法想象创造这样的东西。

It's a thing that's even hard for us, I think, to conceive of what that means because if we look at our hands or we look at a plant, we don't think of that as technology. Engineering. Or engineering. We can't even conceive of making something

那么通过生物工程，是不是意味着我会变成六指人，或者长出匿名的第三只手臂在我睡觉时处理事情？我不知道。就像，什么...

So by engineering biology, does that mean I'm gonna get, you know, six fingered man and the anonymous third limb that just takes care of things while I'm asleep? I don't know. Like, what

你想要什么？

you want?

我想要什么？我们可以从零开始制造发光植物之类的东西。但你暗示的是，生物技术在试图改造实际生物系统时会发生转变。

What do I want? So we make things from scratch, glowing plants and stuff. But what you're suggesting is that biotech swaps out at the point where we start trying to engineer the actual biological system.

我们已经在做部分这样的工作。我们可以将蛋白质片段拼接到其他蛋白质上，获得复合功能。我们做过类似的事情，但现在我们正在酶的水平上创造全新的功能。

And we were doing some of that. We could paste bits of proteins onto other proteins and get composite functions. We were doing things like that, but now we're generating entirely new functions that we want at the enzymatic level.

CRISPR是

CRISPR is

可能是我们现在能够以某种方式调整生物学本身的最佳例证。这样说公平吗？

probably the prime example of now we can go and tune the biology itself in some way. Is that fair to say?

CRISPR是我们在生物体中实现这一点的例子之一。另一个方面是从头设计具有全新功能的蛋白质，进行前所未有的新反应。而这仅仅是最微小的起点。

CRISPR is one example where we can do it in organisms. And the other aspect of it is designing proteins de novo with new functions that have never existed before, doing new reactions. And that's just the tiniest starting point.

那么让我来谈谈这个发展轨迹。我认为这意味着，就像现在，我们可以观察蛋白质，试图了解它在自然界或实验室中的行为，但也许我们可以调整它的行为。然后我们有了像AlphaFold这样的工具，基本上是在预测这些蛋白质可能或将会做什么，准确率相当高。我们的目标是能够达到在实际构建之前就理解我们正在构建的东西。

So let me ask you about that trajectory. So I think what that means is, like right now, can, you know, protein, we can observe it, we can try to see what it's doing in the wild or some lab, but maybe we can go tune its behavior or something. But then we have these things like AlphaFold where we're looking to guess what these proteins could or would do, essentially. And with wild bars on accuracy, I guess. Where we'd go is we should be able to get to a point where we can actually understand what we're building before we build it.

我们应该能够建模一个蛋白质，在软件中测试它，然后在现实世界中部署它。这就是我们的方向吗？

We should be able to model a protein, test it in software, and then deploy it in the real world. Is that where we go?

我认为这是梦想。有两种方式来看待这个问题。我认为我们现在所处的世界是一个我们不会理解那个过程中所有细节的世界。我们已经在药物设计部分尝试过这样做，这被称为理性药物设计，你试图模拟蛋白质中的所有相互作用，从第一原理出发机械地进行，这真的很难。

I think that is the dream. And there are two ways of looking at that. And I think that the world that we're in now is one where we will not understand all of the details in that process. We have tried to do that on the drug design part, that's called rational drug design, where you try to model all of the interactions in a protein and you do it mechanistically from first principles and it's really hard.

就像我们甚至对阿司匹林都做不到这一点。

Like we can't even do it for aspirin.

我们做不到。真的不行。是的。而且我们无法那样做。人工智能的机遇在于我们输入大量数据。

We can't. Can't. Yes. And we can't do that. The opportunity with AI is we feed lots of data.

我们说，我们不会进行所有量子计算，那确实很难，但有些模式是我们用现有科学无法理解的。嗯。而AI正在识别并利用这些模式。这让我们能够触及新功能的领域。所以它就像一个编译器，但人类并未编写那个编译器，它是通过学习获得的。

We say, we are not gonna do all the quantum calculations, are really hard, but there are patterns that we haven't been able to understand Mhmm. With our science that AI is picking up and it's making use of. So that allows us to get to this area of new functions. And so it's like a compiler, but human beings haven't written that compiler. It's just been learned.

所以这与我们在机器学习其他应用中得出的结论非常相似：因果关系与相关性的关系发生了反转。获得更好的答案，关键在于系统处理所有数据。它运行着人类永远无法解析的庞大算法，给出更优解，但我们却不知道原因。

So it's very analogous to what we're basically figured out in every other use of machine learning, which is that your relationship between causation and correlation flips. Getting better answers, the thing is consuming all the data. It's running giant algorithms no human could ever parse, and it's giving us better answers, but we don't know why.

没错。

That's right.

那么如果这不是理性药物设计，该怎么称呼它？非理性药物设计？迷幻药物设计？AI药物设计。或者机器学习药物设计。最终我们达到了一个能制造出以前无法制造之物的阶段。

And so what's the name for that if it's not rational drug design? It's like irrational drug design or psychedelic drug design? AI drug design. AI So machine learning drug design. And then you end up in a position where now we can make a thing that we wouldn't have been able to make before.

我们仍然无法解释原理，但离命中目标更近了一大步。

We still can't explain it, but we're a lot closer to being able to hit the target.

是的，正是这样。我认为这个领域的机遇在于，我们将能制造出以前无法实现的东西，这可能是新化学技术与AI的结合，让我们进入药物化学家难以应对的复杂领域。对。所以这是个唾手可得的典型案例。

Yeah, that's right. I think that the opportunities in this space are we're going be able to make things that we weren't able to make before, and that might be a combination of new chemical techniques combined with AI that get us into places where medicinal chemists really just can't handle the complexity. Right. And so that's an interesting very low hanging fruit as an example.

好的。

Okay.

那些非常微小。它们只是药物分子，不是这些大蛋白质，不是细胞，也不是相互作用的蛋白质系统。

Where those are very small. They're just drug molecules. They're not these big proteins. They're not cells. They're not systems of interacting proteins.

它们还不是器官。不过我认为那些东西将会出现，那将是我们的下一步。

They're not organs yet. Though those things, I think, are going to come, and those will be our next steps.

哇。所以你认为最终我们会制造出我们目前没有的辅助器官？辅助器官？你首选会制造什么辅助器官？哦。

Wow. So you think eventually we'll make auxiliary organs that we haven't got already? Auxiliary organs? What would your first auxiliary organ choice be? Oh.

我想现在确实有人在尝试获取辅助器官，但据我所知他们是从其他人身上移植的。稍等一下。如果这部分内容不合适我们会剪掉。其他哺乳动物有哪些器官是我们没有的？

I guess there are people trying to get auxiliary organs, now that think about it. But they're grafting them from other humans, I guess. All just one second here. We'll cut this out if it's really bad. What organs do other mammals have that we don't?

作为一只蝙蝠是什么感觉？

What does it feel like to be a bat?

哦，经典问题。

Oh yeah, classic question.

我们能拥有回声定位能力吗？

Could we have echolocation?

它们能做到回声定位。

They can do echolocation.

气味。有什么特别有趣的吗？

Smells. What is so interesting?

我们可以有类似缓解恶心的滴剂。

We could have like a drop on nausea.

在消防栓上对狗狗来说。

On the fire hydrant with dogs.

是啊，就像我们对消防栓毫无用处。狗狗显然知道那里有些特别的东西。

Yeah, like we have no use for fire hydrants. Dogs obviously know there's something special there.

我是说我们缺失了很多感官能力。

I mean there are a lot of senses that we are missing.

好的，很酷。你回到网站上要展示的科技生物（tech bio）简明定义，那是什么？

Okay, cool. You go back to the succinct definition of tech bio that would go on the website, what is that?

科技生物是技术驱动的生物学。我们认为，这些是技术平台而非发现平台。在我们看来，科技生物公司正在开发一种新方法，用于制造新型化合物。这种能力可以做到传统药物化学家无法实现的事情。与生物技术公司不同——我认为约60%的生物技术初创企业只有单一资产。

Tech bio is tech driven biology. As we look at it, these are technology platforms rather than discovery platforms. Tech bio companies, as we look at them, are ones which are developing a new approach to making, say, a new class of compounds. And that's an ability to do things that, say, the medicinal chemists haven't been able to do. Instead of in biotech, 60%, I think, of biotech startups have a single asset.

它们有生物学理论依据，进行大量验证工作，确定要靶向的蛋白质和干预的途径。这些生物学家在该领域具有专长，采用广泛的研究方法。希望最终能取得成果并被制药公司收购——这种情况下就是孤注一掷，因为它们只有单一项目。而技术平台则拥有规模化高效运作的方法，团队规模较小，更接近硅谷初创企业的模式。创始人具有独特视野并付诸实践。这与波士顿那种完善的特定公司发展模式不同——那里有专门寻找生物学洞见的风险工作室，Third Rock就是这类公司的典范。它们非常成功，通常先寻找生物学洞见，然后组建专业制药团队，接着...

And they have a biological thesis, they do a lot of validation, they figure out, Okay, this is the protein we're going after, this is the pathway we're going to perturb, and they're biologists who have an expertise in the space, painting with a broad brush. Hopefully, that gets someplace and it's acquired by pharma, and it's an all or nothing shot in that case where they just have one. And tech platforms have an approach where they can do things at scale very efficiently, and they tend to look a little bit more Silicon Valley startups in that the teams are smaller, and then the founders are different in that they have a vision for an approach, and they implement that. And it's not something that lends itself to a certain style of company development that was really perfected in Boston where you have these venture studios essentially that go hunting for biological insights and Third Rock is a great example of this type of company. They're very successful, they go hunting typically for biological insights then they put a professional pharma team around it then they go

这部分我们接下来会处理。是的，我们...

and We'll take get the from here. Yeah, we

并推进完成临床试验。

And get it through clinical trials.

能给我们举个典型例子吗？什么样的公司是原型级的科技生物企业？有没有具体案例可以讨论？

Can you give us a we an example. What's a prototypical tech bio company? Is there one we can talk about?

有的，我们首笔投资的公司Unnatural Products就是典型代表。这些来自圣克鲁斯实验室的科学家致力于研究大环肽类分子（这个术语有点拗口），这类分子有时被称为'金发姑娘药物'，因为它们的尺寸恰到好处。

Yeah, there's this one company that was our first investment called Unnatural Products. Okay. And they're really very emblematic of this space. So these are some scientists who came out of a lab in Santa Cruz, and they were working on this class of molecules called macrocyclic peptides, which is a mouthful. They're sometimes called Goldilocks drugs because they're just the right size.

它们不同于小分子药物，后者通常是药丸形式服用的。小分子药物与这类新型注射化合物截然不同。GLP-1属于注射类药物，但更庞大的类别是抗体药物。这些抗体实际上由细胞产生，是生物体免疫反应的一部分。自然界还存在另一类被称为大环肽的化合物，比如环孢素这类'金发姑娘药物'，它源自真菌，用于防止组织排斥反应。

They're not like small molecules, which are the things that you typically take when you take a pill. Those are small molecules and they're not like this new class of compounds which people inject. GLP-one is an injectable but the much larger class are these antibodies. So those are actually produced by cells, they're produced by organisms and part of the immune response. But there's another class of compounds produced in nature called these macrocyclic peptides, these Goldilocks drugs, and cyclosporine is an example, it comes from a fungus and it's used to avoid tissue rejection.

它们有何特别之处？体积更大。更大的体积意味着它们与蛋白质相互作用时能实现更高特异性。它们拥有更多接触点，作用极其精准，不会像小分子药物那样与不该作用的靶点发生反应。

What's so special about them? They're bigger. Bigger means that they can be much more specific in the way that they interact with a protein. They have many more contact points and they can be highly precise, and they don't do what small molecules do, which is interact with things they're not supposed

去错地方了。明白了。

Go to the wrong place. Got it.

因此你能获得非常特异且紧密的相互作用，但这些分子设计起来极其困难。事实上，辉瑞首席化学家克里斯托弗·利平斯基评估所有项目后曾说：别尝试制造这些，我们总是失败。它们体积太大——为此他还制定了一系列规则。

And so you have a very specific and tight interaction, but these are incredibly hard to design. In fact, Christopher Lipinski, who is Pfizer's top chemist, looked at all of their projects and said, Don't try to make these because we always fail. They're too big, and he had a bunch of rules

不，这就像1980年代数码相机的分辨率问题，随着复杂度增长。但最终我们应该能做到，因为只要获得更先进的乐高套件，就能搭建出来。

No, about this just the a resolution problem, 1980s digital cameras, where the complexity grows. And so eventually we should be able to do it because once we got a more advanced LEGO kit, we could build them.

问题在于存在多个难点，我稍微深入些技术细节让你们理解为何如此困难。假设用10个氨基酸组成大环结构，其中还可以包含非天然氨基酸——这类氨基酸约有10万种。那么10的10万次方就是10的50次方种可能组合，相当于地球上的原子总数。

The problem, there are a few different problems and I'll just nerd out into the details just a little bit, just to give you a sense of why this is so hard. If you take, say, 10 amino acids and you link them up into one of these cycles, these macrocycles, you can also have unnatural amino acids. There are maybe a 100,000 of them. And so 10 or a 100,000 to the tenth power is 10 to the 50 different combinations that you could have. That's as many atoms as there are in our planet.

懂了。

Got it.

那是一个巨大的搜索空间，很难理解哪些会起作用，哪些会结合。而且不仅仅是结合。这些分子太大了，通常会发生的情况是它们直接从溶液中析出。它们会形成晶体，然后直接通过消化道，无法进入体内发挥预期作用。那么它们能溶解吗？

That's a humongous search space and it's very hard to understand what's gonna work, what's gonna bind. And then it's not just binding. These molecules are so big that typically what happens is they just crash out of solution. So they just form crystals and then they just go through your gut and they don't get into your body where they're supposed to do something. So can they be soluble?

好，现在它们能溶解了。太棒了，可以溶解。但它们现在又大又溶于水，无法穿过脂肪构成的细胞膜。所以它们不溶于脂肪。因此需要在胃液中可溶，同时又要能溶于脂质膜才能穿透。

Okay, now they're soluble. Great, it's soluble. Now they're so big and soluble that they can't get through your membranes, which are fat. So they're not soluble in fat. So you need to be soluble in the liquid in your stomach, but then also soluble to the lipid membrane so they can get through.

这是个难题。如何设计出这样的东西——既要能结合，又不会在肠道被消化？这是化学家们做不到的。

That's a hard problem. How do you engineer something like that and it's supposed to bind and it's not supposed to get digested in your gut? That's what the chemists can't do.

是啊，但它们确实存在，比如你和我，不过它们是在内部合成的，没有那些问题。

Yeah, but they existed, you and me, but they're made on the inside where they don't have those problems.

确实存在，或者说是生物体经过数十亿年自然选择形成的。

They do or they're made through billions of years of natural selection by organisms.

对，对，好吧，它们是这样被制造出来的，是在需要的地方直接产生，而不是在培养皿中合成后再植入。

Yeah, yeah, okay, that's how they were made, then they're produced in the spot where they need to be instead of in a petri dish and then put in later.

正是如此。

Exactly.

那你具体要怎么做呢？

So how would you even do that?

对，你的意思是等我们明确需求后，或许能找到方法在体内设计并生产它们。

Right, so you're suggesting maybe we can find a way to engineer them to be produced inside your body once we know what we need.

理论上可以这么做，但目前我们完全无需动用CRISPR技术对人体引入各种复杂分子和基因进行基因编辑——这绝非我的本意，尽管谁知道呢，或许遥远的未来会实现。我真正指的是研发一种新型药物，这就是我所说的‘低垂果实’。但即便如此，这仍是个极其棘手的难题。

You could do that, but in this case, this is very close to what we can do today without using CRISPR on people introducing all kinds of crazy molecules and genes. Doing gene engineering on people. That's not what I'm suggesting at all, though who knows, we might do that in the distant future. This is really just producing a new type of drug, and this is what I mean by low hanging fruit. That is a super hard problem.

事实上，默克公司一直在推进类似项目。

In fact, Merck has been working on a project like this.

你为何称其为‘低垂果实’？圣克鲁兹那些团队——或者他们来自哪里？他们有何独特之处？

What do you say it's low hanging fruit? What are these guys in Santa Cruz or where are they from? What are they doing that's different?

此前从未有人能成功制造这类物质。

Nobody's been able to make these before.

所以他们现在能造出来了？

So they can make them?

他们能做到，为了让你理解这有多难，默克公司已经尝试了十五年，终于成功研发出一款，那个分子——那款药物被提名并赢得了年度分子奖。确实有这么一个奖项叫做

They can make them and to give you a sense of how hard this is, Merck has been trying for fifteen years, they finally made one and that molecule got, that drug got nominated and won Molecule of the Year. There's actually an award called

怎么

How do

当你不在时，那些书呆子们是怎么做的

the nerds do when you're

我买了礼服但没收到邀请。年度分子还是分子

I got my dress but I wasn't invited. Molecule of the year or Molecule

的

of the

年度大奖，这需要一个庞大的化学家团队不断调试，最终才成功。而这家公司凭借其技术平台，可以用更小的团队并行生产许多这类分子。

year and it took a huge team of chemists and they were tinkering away at it and they finally got it. This company can produce many of these molecules in parallel with a much smaller team with their tech platform.

好的，真酷。

Okay, cool.

这就是我们对此的思考方式，你有一个分子，希望它能降低失败率，这在生物技术领域是个大问题。临床试验的失败率高达90%到95%，而这是我们拥有的伟大产业之一。想象一下，如果我们能把失败率降到80%，那将是生产力的巨大提升。如果我们能设计这些东西精确执行我们的指令，或者理解患者的基因组成，从而准确靶向药物，或者设计它们的组织特异性，那会怎样？

And that's like the way that we think about this, that you've got a molecule that's potentially the hope is that you're going to reduce the failure rate, which is a huge issue in biotech. We've got a ninety to ninety five percent failure rate in clinical trials and this is one of the great industries that we have. Imagine if we just reduce that to 80%. That is a massive gain in productivity. What if we can engineer these things to do exactly what we want or understand, say, the genetic complement in the patient and be able to target those drugs correctly or be able to engineer their tissue specificity.

我们可以开始做所有这些事情，这些东西几乎开始像小型智能机器人了。

We can start doing all of those things and these things are almost starting to resemble little smart robots.

所以你对此感到兴奋，仅仅因为它们现在至少能可靠地生产这些东西。我们正处在能够开始测试它们是否能到达目标位置的起点。

So you're excited about it just because they're at the point where they can at least produce these things reliably. We're at the beginning of being able to start testing them for could they get into position.

我们现在已经有了候选药物，这花了十五年时间——在默克公司。我是那家公司的董事会成员，见证了它们从零开始，现在几乎有了一个关于减脂增肌分子的候选药物。这太疯狂了。这类公司的运作方式通常是与制药公司合作建立伙伴关系，因此它们在这方面也不像传统生物技术公司，因为它们能赚钱。

We're at the point where we have drug candidates and like that took fifteen years Wow, in Merck's okay. I'm on the board of that company and I saw them go from zero to they're pretty much a drug candidate on one of these molecules that is fat loss and muscle gain. That's the kind of, That's crazy. And the way these companies work too is they typically work with the pharma companies and they do partnerships. And so they're not like biotech that way either because they can make money.

它们确实能赚钱。这是一项重大创新。

They can actually make money. This is a huge innovation.

这是我们在谈论的

This is a biotech we're

生物科技。

talking about.

生物科技 别对我发疯。

Biotech Don't get crazy on me.

没错。生物科技研发这些需要大量资金。而科技生物平台更轻量、更精简。它们建立合作伙伴关系，能更早获得收入来源。

Exactly. So biotech costs a lot of money to develop these. And with tech bio platforms, they're lighter, leaner. They do partnerships. They get revenue streams early.

它们通常能获得少量特许权使用费，随着信心增强，这部分收益会随时间增长。此外它们还拥有完全自主的资产，最终成为药物研发公司，在分子生产数量和研发成功率方面可能更具生产力。希望如此——这正是我们期待看到的。我相信我们将降低临床试验的失败率，届时我们将进入一个全新的世界。我们将能够设计出所需的药物。

They usually get a little bit of royalty, and that royalty component grows over time as there's more confidence And in the then they have wholly owned assets, and they become drug discovery companies that may be more productive just in terms of the number of molecules they can produce and their ability to get through. The hope is. That's what we're going to see. I believe we are going to reduce that rate of failure in clinical trials and we're in a fundamentally new world then. We are going to be able to engineer the drugs we need.

我们将能够设计出所需的细胞疗法。我们正处在这个领域的开端

We're going be able to engineer cell therapies that we need. We're right at the beginning of In this

在你看来，科技生物的发展空间是否足够广阔，以至于你们可以成立专注于这类平台的基金，在不同领域找到足够多的目标对象，围绕这一概念构建完整投资组合？是这个意思吗？

your mind, the aperture on tech bio is big enough that you guys can make a fund focused on these types of platforms, find enough of them in different targeting different areas that you could build a whole portfolio around this notion? Is that the idea?

这是生物学。

It's biology.

空间足够大。好吧。你并不担心

It's big enough. Okay. You're not worried about The

从机会空间来看，这个术语非常庞大。

term is ginormous in terms of the opportunity space.

有一个可能符合你论文定义的案例：这些人建造了一个肠道模拟器。他们可以从你体内提取粪便样本，在这台机器中培养你的微生物组，然后在这个设备上对你进行药物或饮食测试，而无需直接作用于你本人。关键在于，每次测试都会为机器学习模型提供数据。经过足够次数的测试后，我们最终可以在软件模型上测试药物和饮食方案，然后再应用于人体。

Have one that might fit your thesis as far the way you've defined it, is these guys built a gut simulator. And so they can take a scoop of shit out of you, grow your microbiome in this machine, and then test drugs on you or diets on you in the device before they ever do it to you. But then the whole point is where this goes is every time they do that, it feeds a machine learning model. They do it enough times. Eventually, we can test the drugs on you and software and the model, the diet on you and the model before we do it to you.

这就是发展方向。他们已经在你的肠道中发现了几种超级细菌，其中一些正与不同制药公司合作测试。有一种细菌在抗抑郁效果上优于SSRI类药物，简直能做出抗抑郁酸奶。

That's where this goes. That's the track they're on. But they've already discovered a few of these superbugs in your gut. Some of them are testing with different pharma companies. One of them is outperforming SSRIs for depression, like they could make antidepressant yogurt.

其中一种针对压力，另一种针对疼痛。总之他们正在发现肠道中与这些症状相关的微生物。说实话，我从未称它为技术生物平台。

One of them is for, I think, stress and one's for pain. But anyway, they're discovering these microbes in your gut that correlate to these things. You know, I've never called it a tech bio platform.

这就是技术生物。接下来将会实现个性化定制，也就是你所说的

That is tech bio. So there's gonna be personalization, which is what you're

哦对说到这个，我有个外行（甚至算不上外行）的猜想想请你验证：在我看来，在能收集足够数据的情况下（比如微生物组数据），我们完全可以进行数据挖掘并逆向工程对照组。只要数据足够，我们就能找到所有与你相似的人——吃同样数量的芝士汉堡、抽同样数量的香烟。数据越丰富，我们就能越精细地颠覆传统科学方法。

Oh talking yeah, let me ask you this. I have a layman, not even a layman, just a made up philosophy about this that you could validate for me, which is, it seems to me like in these cases where we could collect enough data, like in the case of the microbiome, if we just had enough data, we could just go fishing and reverse engineer control groups. We could just find here all the people who are like you, had the same number of cheeseburgers, smoked the same number of cigarettes. We could start to get real granular if we have enough data, and we flip the scientific method on its head.

完全正确。

100%.

你认为那就是我们要去的地方，还是

You think that's where we're going or is

这个某物

this something

100。但最大的挑战在于获取数据。这确实是该领域的重大难题。有几种实现途径。你指的是人类数据。

100. But the biggest challenge there is getting the data. And that is the big challenge in this field. And so there are a couple of ways that happens. You're talking about human data.

当然。显然23andMe解决了部分问题，其初始模型本应扩展至涵盖各类数据。

Sure. Obviously 23andMe solved part of that was the initial model that was supposed to grow into being all kinds of data.

因为那样你将拥有海量基因数据，基本上可以进行常规的大数据分析，找到所有人。

Because then you would have so much genetic data you could just go do essentially even normal big data analytics go find all people.

血液、反应、生活方式等诸多不同要素，若能掌握所有这些，就能为高效的机器学习模型提供数据。还有理解免疫系统的努力，其中存在巨大变数。这些系统还具有历史依赖性，复杂度极高。这正是AI擅长的——在复杂数据中发现有价值模式。

Blood, response, lifestyle, so many different components that if you knew all of those things, you'd be able to feed a very effective machine learning model. There are also efforts to understand the immune system, which has so much variability in it. Then there's history dependence in these systems. There's a lot of complexity. That's what AI is good at, is finding valuable patterns in complex data.

没有哪种数据比生物学更复杂。确实如此。另一个问题——也是我们思考的——是收集所有这些数据的公司商业模式是什么？因为目前我们虽有数据，却处于孤岛状态。

There is no form of data more complex than biology. So, that's right. And then the other question, and this is the other thing that we think about, is what is the model of the company that is going to collect all of that data? Because right now, we have this data. We have data that's siloed.

在医疗系统中，这是个巨大的问题。人们一直在讨论它。甚至有公司的整个宣传口号就是解决数据孤岛问题。而现在数字健康领域也面临同样困境——比如有家微生物组公司掌握一批患者数据，另一家公司又拥有不同的患者群体，而你真正需要的是整合这些人的基因数据、微生物组数据、Oura Ring智能戒指数据以及其他所有参数，这才是构建真正强大临床解决方案的关键。

In the medical system, this is a huge problem. They talk about it all the time. There are even companies whose entire tagline is solving the data siloing problem. And then now we've got that same problem in the digital health space because you've got a microbiome company, that's a pool of patients, and then you've got a different pool of patients over here, and you really want to know how this person's genetics and their microbiome and their Oura Ring data and their all of these other parameters come together, that's what's going to allow you to build Totally, powerful and clinical so

你觉得我们有可能...我知道我们身在美国，但你认为这些方面存在整合或改善的可能性吗？我是说...

do you think that we're I know we're in The US and all, but do you think there's some chance of these things coming together or improving? I is this

我认为有希望。我们持乐观态度。但我们现在不投资医疗系统，部分是因为激励结构问题，部分是因为这个古老而庞大的行业已经...（这里就不展开说了）

think so. We're hopeful. Yeah. But we don't invest in the medical system because we think that's lost hope partly because of the incentive structure, partly because it's just a very old, high mountain industry that's also won't get into I

明白

get

这说来话长

this is a whole thing.

是啊，这话题够另开一集节目了

Yeah, know. It's another episode.

那要如何构建呢？很可能会形成一个预防性健康平台。未来可能会出现一个获得消费者认可的预防健康平台，他们提供越来越多的检测服务。这是一种可行模式。另一种则是从根本上重构激励体系，因为存在这样的矛盾：我们支付了保险费用后，人们总觉得不该再自掏腰包支付医疗费用。

And how is that going to be built? It will probably be a preventive health platform. There's probably going to be a preventive health platform that gains consumer adoption where they offer more and more tests. So that's one model that works. And another is really to think about the incentive system itself because there's this question of we pay for insurance and then people feel like I shouldn't be paying for healthcare costs out of my own pocket.

这意味着我们的钱默认流向了疾病治疗系统，而这个系统并没有动力让我们保持长期健康。我们在雇主那里工作三年半后，之后的时间就由我自己支配了。

And that means that our dollars are by default going to the sick care system and they're not incentivized for keeping us healthy for a long and healthy life. We have three and a half years when we're with an employer and then outside of that, I'm giving

我们让你得到了一些。是的，你在这里获得了宝贵的大脑时间。

We you some get all that. Yeah, you got to your valuable brain time here.

我要去个地方。

I'm gonna go someplace.

好的，没问题。

Okay, all right.

我们还在考虑一种新的保险模式，因为这是该领域非常重要的一部分，尤其是在消费者健康领域。人们可以去看看，就在我的Substack上，叫做'寿命保险'。这是一种保险模式，保险会跟随患者一生，并激励医疗服务提供者为患者提供长期健康的生活。

So what we've also been thinking about is a new model of insurance because this is a very important part of this area, especially in the consumer health space. People can go and look at it. It's on my Substack. And it's called lifespan insurance. It's a model where there's insurance that follows the patient for their entire life and incentivizes the providers to provide a long and healthy life.

超级

Super

酷。那现在世界上有没有人在尝试类似的东西？

cool. And is anybody in the world trying something like that by now?

或许我们可以在这里试试。

Maybe we'll try it here.

也许在新加坡或中东之类的地方？他们似乎能领先解决这类问题。

Maybe in Singapore or The Middle East or something? It seems like they could get ahead of those things.

我认为这不必是一个政府主导的体系。重点不在于此。这是一个推动市场创新的体系。

I don't think that this has to be a government driven system. That's not what this is about. Is a system for innovation in the market.

是啊，感觉如果我们要彻底改革医疗体系，就得开辟一条独立路径，先私下构建出更优方案，然后让他们花一个世纪来追赶？

Yeah, it really feels if we're gonna overhaul healthcare, we're gonna have to create a separate track and just build something privately that's proving to be better and then they can catch up in a century?

我认为这才是正确方式——像初创企业那样运作。不要直接解决他们正在处理的问题，而要攻克他们无法解决的领域，比如疾病预防和延长寿命。

That I think is the right way and you do it the way startups do it. You don't go and attack the problems they're solving, you attack problems that they can't solve which are prevention and longevity.

人们常常没意识到，硅谷的重大胜利不在于修复任何行业，而是直接取代它们。我认为这正是必须采取的策略：开创更优方案，在现行体系外验证其价值，然后用初创企业颠覆整个行业。我们在出租车、媒体和酒店业已见证这种模式，为何不将其应用到更重要的领域？

One of the things I think people don't often realize is Silicon Valley's big wins is not they're not fixing any business or any industry. We're just replacing them. And I think that's what you have to do. Have to pioneer a better way, show it outside of the status quo system, and then replace the industry with a startup. And we've seen that in taxis and media and hotels, but why don't we do it to things that matter?

完全正确。

Exactly right.

好的，明白了。你走的是独特路线。你本可能在某个地方做着移液之类的琐事，对吧？所以你选择这条独特道路，不仅要思考这些事物如何发展，还要投资它们。

Okay, cool. Understood. You're on a unique track. You could be pipetting shit somewhere, right? So you're on a unique track trying to not only think about how these things could evolve, but invest in them.

你是怎么最终决定要把时间浪费在这件事上的？

How did you end up deciding this is what you want to waste your time on?

知道吗，因为我当时在做移液工作，比如把微量液体从试管转移到试管，我觉得这简直疯了。我大学毕业后的第一份工作是在千年制药公司，那时正值人类基因组时代，他们有机器人，我想天啊，这正是我期待的——远离实验台。千年开创了高通量生物学，某种程度上就是科技生物学的雏形，他们向各大药企提供发现服务。他们会做我跟你说的那些事：锁定靶点、验证，然后抛给其他公司去尝试靶向那个受体什么的。他们从测序到验证再到细胞实验，全程规模化运作。

Know, it's because I was pipetting things, like small amounts of liquid from tube to tube, and I thought that this was madness. My first job out of college was to go to this company called Millennium Pharmaceuticals out of grad school and they had robots and this was during the human genome era and I thought, my god, this is exactly what I've been waiting for, getting away from the lab bench. And Millennium pioneered high throughput biology and that in a way was tech bio and what Millennium did was sold discovery services to all of the big pharma companies. So they would do exactly this thing that I was I telling you telling We've got a target and we validated it and then they'd throw it over the transom and then the companies would go and try to target that receptor or whatever it was. So they were doing everything from sequencing all the way up to validation and doing cell assays and they were doing this all at scale.

于是我参与构建了许多核心技术，这些技术后来转让给了拜耳、安万特、辉瑞、礼来等公司，这让我真正尝到了甜头。我们必须引入机器人，必须规模化，我还思考了知识管理。这引发了一个问题：当你有了零件并开始理解它们如何连接后，接下来就要思考它们如何协同工作？后来我在哈佛基因组研究中心参与讨论时，开始推动科学家们往返于这个新成立的系统生物学系（哈佛三十年来首个新系），并闭口不谈'生物学必须可计算化'这种话了。

So I ended up building a lot of those technologies in their process technology core and those got tech transferred to Bayer and Aventis and Pfizer and Lilly and all of these companies and so that's really where I got my taste for. We've got to get robots involved, we've got to get scale involved and I thought about knowledge management there. That led to thinking about once you have the parts and you start understanding how the parts are connected, the next question is how do the parts work together? And I ended up then at the proposed, like, we need to start thinking about this and drug discovery And that led to conversations at Harvard Center for Genomics Research where I started getting scientists going back and forth between there and at this nascent thing which was the first new department at Harvard in thirty years or some long time, which was the systems biology department. And I shut off my mouth about how biology needed to be computable.

我们刚开始建立生物学的数学模型——是机械模型而非AI模型，比如描述这个蛋白质与那个相互作用，所有这些互动构成细胞这个庞大的鲁布·戈德堡机械，行为由此涌现。当时人们开始构建这类复杂精细的机械模型，我在想：怎样才能像物理学家那样做科研？能直接说'这是我的方程，这是你的方程，现在让我们共同理解这些'。

And we were just starting to write mathematical models of biology. You know, are mechanistic models, not AI models, where we would say, Here, this protein's interacting with this and all of these things are interacting together and you get this giant Rube Goldberg device that is the cell and somehow behavior emerges from it. And so people were starting to write models at that time, mechanistic models like this, but they're really complicated and very fiddly. I thought, how do we actually do science the way physicists do it? Where you can say, here's my equation, here's your equation, and now let's understand these things together.

于是我设计了一种语言，可以像软件工程师那样编写模块、共享包，它能编译这些模块并模拟所有变量。这让我真正开始思考——也是我开始大放厥词的时候：'对，我们就该这么做'。后来我意识到自己根本不懂行，必须成为计算机科学家。这使我深入研究AI数学和生物学的形式化表达，对知识表示和生物学的数学建模产生了根深蒂固的兴趣。

And so I wrote a language where you could write these things like a software engineer does, you just write modules and you share a package and it would compile those together and then develop all the variables and simulate it. And so that's how I really started thinking, that's when I really shot off my mouth saying, saying, Oh yeah, we should do this. And I realized I had no idea what I was doing and I had to become a computer scientist. And that led me to thinking about AI math and representing biology formally And so I got very deep into thinking about knowledge representation and mathematical representation of biology. So that's always been very deep in my DNA.

那么现在最前沿进展如何？如果要建模细胞但我们并不完全了解它，你们就直接注释掉吗？

And so what's the state of the art there? If we're trying to model a cell but we don't actually know everything about the cell, then you just comment out.

编造点东西出来。

Make stuff up.

随便编点东西就行。

Just make stuff up

这里。我们当时在猜测。

here. We were taking guesses.

去问大语言模型。

Ask an LLM.

你在猜测速率函数。

You take guesses at rate functions.

所以有很多未设定的变量。

So lots of unset variables.

这真的很难。真的很难。然后由于所有未知因素，这个领域实际上并没有那样发展。圆锥体。这就是为什么人工智能

It was really hard. It was really hard. Then the field didn't actually evolve that way because of all the unknowns. Cones. And that is why AI

所以你在考虑可以再次攻击同样的目标，但这次不是尝试构建这种形式化语言，而是让每个模块或对象类什么的都成为这个事物的机器学习模型。

So you're thinking you could attack the same thing again, but instead of trying to make this formal language, each of these modules or object classes or whatever would be just a machine learning model of this thing.

完全正确，其实是反过来的。百分之百是反过来的。你可以从底层开始机械地建模，或者直接取其表象然后输入给AI。

Totally, it's the other way around. It's 100% the other way around. You can mechanistically model from ground up or you can just take the outside of it and then feed that into an AI.

就像AI学习的方式一样。

The same way in AI learning.

它会做什么？但我们毫无头绪。这样做时我们往往缺乏理解。不过这种情况可能会改变，因为我们会有可解释AI。或许通过可解释AI我们真能有所发现。

What is it going to do? But we don't have any understanding. We tend not to have understanding when we do that. That may change too though because we'll have explainable AI. So we may actually discover with explainable AI.

我不确定那是否真的算可解释AI。巴尼，你想说说吗？解释那些我们无法解释的事物，还是仅仅解释它是如何得出结论的

I don't know if that really explainable AI. Barney, you want to tell us? Explain Can things we can't explain or just explain how it got to

这就是关键所在。可解释AI并没有具体的技术手段。传统专家系统逻辑推理非常出色，因为它们源自人们能理解使用的规则和知识。如果你解释这是导致结论的推理链条，或许还能理解。但随着模型越来越复杂，涉及大量搜索和推理，就越来越难以

the That's the idea. There's no actual specific technology for explainable AI. Logical inference systems in the classic expert system sense were very good because they were derived from rules and knowledge people would use and could understand. So if you then explained, this is the chain of reasoning that led to this, then maybe that makes some sense. As the models get more and more complicated and involve lots of search and inference, it becomes harder and harder to

推理链条最终还是会变得过于冗长复杂。

The chain of reasoning still ends up being too long and complicated.

我举个例子，如果你有一个必胜的棋局位置，为什么说它必胜？因为它存在一步棋，使得其他所有走法最终都会导向这步棋，大约50步之后就能见分晓。这其实没有复杂的解释，因为只要稍有不同，这个策略就会失效。所以有时你能解释清楚，有时则不能。说到这个话题，我们正面临这样的权衡：如果系统能给出好解释，我们就能更信任它们。

I give an example that if you've got, this is a winning chess position, position, and why is it a win? It's a win because there is a move for which all other moves lead to there is a move for which all the moves, and sort of 50 moves later, And it's a there's no actually other complex explanation because if some little thing had been different, it just wouldn't work. So sometimes you can explain, sometimes you can't. And one thing while I just I'm on that topic is we're gonna be faced with this kind of trade off. We can trust things better if they give good explanations.

因此我们将面临这样的选择空间：可能是性能较弱但我们感觉更可信的系统，与性能更强但我们可能不太信任的系统，我们该选哪个？

So there's gonna be the space of things which are possibly weaker performing systems that we feel like we can trust more, and better performing systems we likely trust less and which are we going to choose

完全正确。明白了。是的，这非常有帮助。

and Exactly. Got it. Yeah. That's super helpful.

应该说，当我思考人工智能时，我指的不是大语言模型。我说的不是LLM，甚至不是模型，连Transformer模型都不是。虽然有时确实是Transformer模型，但我们在这些公司看到的更多是传统的卷积神经网络之类。但当我谈论可解释AI时，我在大胆推测并构想新的架构，基于生物学原理的新型神经网络架构，能在网络内部计算更高层次、更有意义的函数。

I should say when I think about AI, I'm not thinking about LLMs. So when I say not LLMs and it's not It's not models. It's not even transformer models. Like sometimes it is transformer models but the models that we're seeing in these companies are some of the older convolutional neural networks, things like that. But when I talk about explainable AI, I'm speculating wildly and imagining new architectures, new neural network architectures that are based on new principles that are derived from biology that will compute higher level, meaningful functions in the network itself.

我认为这种架构会出现的原因是：当你将电极插入大脑时，出乎意料地经常能读取到与外部可测量且有意义的参数对应的信号。但如果在Transformer内部测量某些变量，得到的几乎全是无用数据。这说明我们构建AI的方式虽然有效，但计算效率非常低下。

And the reason I think that something like that is going to emerge is that when you put an electrode into a brain, you way more often than you would think you would get a readout that corresponds to some external measurable and meaningful parameter. You do that with some variable inside transformer, you get garbage. Almost every single one is going be garbage. And so what that is telling us is that the way we structured these AIs, well it works, but it's a very inefficient means of doing the compute.

我们之所以知道这点，是因为人脑功耗只有20瓦。

And we know that because your brain's 20 watts.

完全正确。

Exactly right.

它的表现超出预期，能够以某种方式说英语——我确信要让ChatGPT说英语，其推理功耗肯定超过20瓦。

It's outperforming and it can speak English in a way that like I'm sure it takes more than 20 watts of inference to get ChatGPT to speak English.

完全正确，这确实非常具有推测性和想象力，但已有公司在研究这类方向。我们投资了这样一家公司，部分理念是生物学也将帮助我们构建更好的人工智能，因为我们可以从中获取原理——尽管这不是我们基金的核心关注点，但人工智能正是源于神经元连接方式的启发。不过我们现有的神经网络对神经元功能的呈现非常肤浅，因为本质上那些细胞只是0到1之间的数值，而真实神经元会产生动作电位，当它们放电时会出现高频的快速动作电位随后衰减。实际上这种时域信息能高效表征历史，有推测认为可以计算更高阶的函数来实现这类表征功能。

Exactly right and so that's very speculative and imaginative but there are companies that are working on things like this. We invest in one company like this with the idea that part of it is that biology is going to also help us build better AI because we'll get principles from it, though that's not really the core focus of our fund, but that's where AI came from as it was inspired by, the connectivity of neurons. But we're only taking we have a very superficial representation of what neurons do in our neural networks because all of those cells, basically, they're like values from zero to one, but that's not what neurons do. Neurons produce action potentials and then when they fire, they have rapid action potentials where they have a high frequency and then they decline. And there's actually information in that time domain and that information can be used represent history very efficiently and there's some speculation that you can calculate much higher level functions that do this kind of stuff of representing

我们还有很长的

We have a long

路要走。是的，这涉及整套理论。我就不深入

way to Yeah, there's a whole thing. I won't go

讨论了——我马上要开放提问环节，还有什么我们没谈到你想分享的内容吗？我想了解对你们最有帮助的信息。你们这支基金进展如何？已经进行投资了吗？

into I'm gonna open up to questions in a minute but anything that we didn't get into that you wanna share. I would like to understand what's helpful for you guys. How far are you along with this fund? Have you made investments already?

已经投了。我们完成了四笔投资，其中UNP现在正进行B轮融资。

We did. We made four investments and UNP is now going for their B round.

UNP是什么？

What's UNP?

那些是非天然产品。哦，好吧。而且它们确实表现不错。签了不少好协议，我们还有另一家做补充剂的公司。这又是一个绝佳机会。

That's unnatural products. Oh, okay. And yeah, they've been doing well. Signed a bunch of great deals and we have another company that's in the supplement space. Here's another great opportunity.

补充剂可能很有帮助。这个领域现在很混乱，有家叫'激进科学'的公司正在为补充剂行业建立私立的FDA认证体系，他们先验证产品，然后打上品牌标签。这就像...

Supplements can be helpful. It's a wild west out there and this company called Radical Science is building a private FDA for the supplements industry where they validate it and then brand label on that. More much needed like

类似补充剂界的'能源之星'认证之类的。

energy star for supplements or something.

完全正确。实际上他们做得更多——参与他们消费者临床试验的受试者数量，比FDA过去几年里...

Exactly right. And so they have done more, they've had more participants in their trials, their consumer clinical trials than the entire FDA has over the last couple a

来罐Athletic Greens试试看效果如何。

can of athletic greens and see what happens.

我们会把它送检的，回头再考虑。

We'll put it through there, think later.

太酷了。我想确保大家有机会向你请教，所以接下来会把麦克风传给提问的人。如果没人提问，我就再编几个问题。有人想和阿尼尔辩论点什么吗？

That's super cool. So, I want to make sure that folks get a chance to pick your brain. So, we're going to pass this mic around to whoever has questions. If you don't have any, then I'll make up a bunch more. But anybody have something they want to argue with Anil about?

是啊，Poppy。

Yeah, Poppy.

嘿，感谢你的到来。这很棒。快速提问一下，像Varda这样的公司展示了微重力环境下药物开发的巨大机遇——更少的沉淀、更清晰的视觉呈现以及蛋白质层状结构的一致性。你认为在非地球环境下的技术生物平台最大机遇在哪里？对于非地球环境下的开发，你如何看待这些优化方向？

Hey, thanks for being here. It's great. Quick question. With companies like Varda that are showing really great opportunities at microgravity for consistency in drug development, less sedimentation or lambda vision, showing consistency in protein laminar structures, where do you think the biggest opportunities are for tech bio platforms in terms of non terrestrial opportunity, non terrestrial development, and how do you think about those optimizations?

我刚开始关注太空生物技术领域，我知道人们正在构建用于结晶实验的实验室平台，这是经典案例。我们通常不投资研究技术公司是有原因的。在硅谷，你开发一个SaaS平台，产出的是被其他企业调用的API接口——这是门好生意。但之所以是好生意，是因为你的API会在消费者触发的每个环节被调用。

I've only started looking at space biotech, and I know that people are building lab platforms for doing crystallization which is the classic example and we tend not to invest in research technology companies and there's a reason for that. In Silicon Valley, you make a SaaS platform and you're producing an API that is used and consumed by another business. That's a great business. But there's a reason that's a great business. And that is that your API gets called on every turn of the crank that starts at a consumer.

它能随需求扩展。消费者在前端操作，你们公司在后端响应。在生命科学领域，如果你不研发药物，那就入错行了，因为——

It scales with that. The consumer does something and then your company way on the back end does something. In life science, if you're not making a drug, you're in the wrong business because

我正想说，Varda就是在研发药物。

I was just going say, Varda is making drugs.

没错，当你以研发药物为目标时，这才是正道。如果你的技术能直接推动自主药物研发并获得优势，那就是正确方向。这类项目蕴藏着巨大机遇，属于理性药物设计范畴——他们先解析结构，获得精确构型后，就能通过观察蛋白质活性位点等方式进行理性药物设计。这确实是——

Yes, when you're making drugs and that is the path. If you're using your technology to directly enable your own drug discovery efforts and you get an advantage, that is the way to go. And then there's a great opportunity in that kind of, that's a rational drug design play because they're doing structure and then they get an accurate structure and then they can do rational drug design by taking a look at the active site of the protein or some such thing. And that's

很精彩也很有道理。但你的观点是，许多太空实验项目终将停留在研究阶段。

great and valid. But your point is that a lot of things that you might do in space are going to be research projects.

他们可能

They might

超出了风险基金所能覆盖的范围。

be outside the window of what you could do with a venture fund.

这只是我们的偏好。这些都是实实在在的企业，优秀的企业。我们认为最大的机遇在于拥有平台，尤其是新型模式——正如我之前提到的，我们90%的失败率很可能源于我们开发的产品类型。因此我们需要全新品类的产品，比如具有新型结构的药物，这些产品过去难以构建或难以论证，这正是我们倚重计算技术来填补这一缺口的原因。

That's just our taste. Those are real businesses. They're great businesses. We to we think that the biggest opportunity is when somebody owns the platform and especially a new modality because of the same issue that I talked about is that we fail 90% the time and that's probably because of the type of product that we're building. So it's new types of products, like new types of drugs that have new structures that have been hard to build or hard to reason about and that's why we lean into computation to solve that gap.

你现在听到的是非常主观的观点。我相信生物科技领域的人会指出我的错误，我确实没有深入钻研过这个领域——因为这些企业更多属于研究技术范畴，我们尚未深入涉足。我们真正深入思考的是人工智能驱动领域在多个维度的发展。

You're getting a highly opinionated view. I'm sure somebody who's in the space biotech space would tell me why I'm wrong and I haven't dug really deeply into that but it's because those tend to be on the research technology side that I haven't dug into them and we've really thought much more about the AI driven space in a bunch of different dimensions.

科技生物平台公司曾风靡一时，随后又失宠，尤其在当前这样的融资周期下。生物科技风险投资金额跌至多年最低点。因此你们面临着投资者意愿的瓶颈，他们只愿意投资那些从科技生物平台公司产出、且需经过临床开发的产品。你们将如何应对投资组合必然面临的这一挑战？

Tech bio platform companies have been in vogue and then they fall out of vogue, especially now with the funding cycle the way it is. Lowest amount of biotech venture capital raised in years. So you've a you have a kind of bottleneck of investors that are then willing to invest in that required clinical development of products that are coming out of your tech bio platform companies. How do you reconcile with that inevitable challenge that your portfolio is going to face?

我认为部分原因在于生态系统确实需要发展，而这将通过那些优秀公司证明自身能力的过程逐步实现。最重要的证明不仅是流程某个环节的自动化，而是他们能生产出前所未有的优质产品。这将带来翻天覆地的变化，人们已经开始关注。至于资金来源？这对广义上的科技平台一直是个难题，但这些科技平台略有不同。

I think part of it is the ecosystem does need to develop and I think that's going to happen partly through these proofs of the right companies demonstrating what they can do. I think that the biggest demonstration is not just we're automating some part of the process, it's that they will produce superior products that haven't been able to be produced before. That is going to be a sea change and people are starting to take notice already. Now, where does the funding come from? I think this has been a big problem for what people have broadly called tech platforms, but these tech platforms are a little bit different.

它们通常是发现平台，被称为科技生物平台。其中一些在硅谷支持下获得了极高估值，位于湾区。从我们的视角看，发现平台本身是需要解决的问题——我们并不需要大量新的或更有效的靶点，我们需要更好的产品。如果能生产出这些更好的产品，就能掌握更多主动权，与制药公司达成更有利的合作，而这正是制药公司所追求的。

Often they are discovery platforms and they're called tech bio platforms. There are some that have had really big valuations that have been supported by Silicon Valley that are here in the Bay Area. And they're discovery platforms and from our way of thinking, that's problem to solve because we don't really need a lot of new or better validated targets. We need better products. And those better products are going to haveand if you can produce the products, then you're much closer to the driver's seat, you can drive better deals with pharma, that's what pharma wants.

那里的资金结构主要来自制药领域，这类投资者通常类似于风险投资公司或成长型公司。他们提供资金是因为你的产品已准备好进入临床试验阶段，这正是我们在目标公司中看到的情况。在发现阶段运用AI技术，实现显微镜自动化并大规模理解生物学机制——这正是千禧年公司当年的做法，他们曾有过辉煌时期，完成了当时生物科技领域最大规模的交易。但从中得到的教训是，这类公司最终会陷入资源黑洞，将所有精力都投入单一产品的市场化。

And the funding structure there is pharma, which tend to be like venture firms or growth firms. They're providing capital because you've got a product that is ready to go into clinical trials and that's what we're seeing in the companies that we're targeting. There's AI for discovery where you're automating microscopes and understanding the biology at scale. That's what Millennium did and Millennium had a really great run. They did the biggest deals in biotech of the time, in history at the time, but I think the lesson from that is there ends up being this giant sucking sound with companies like that where all of the resources of the company go into just getting that one product into market and that's what you're talking about.

这就是为什么这类科技公司往往最终沦为单一药物公司。而我们的核心理念是：对于这类以产品为导向的生物科技公司，过去我们无法做到的事情现在可以实现了。我们开创性地提出：看，这是一类全新分子。我们不会面临相同的临床试验挑战，还能更快证明其具备正确特性。这类产品的专利生命周期也更短。

That's why those tech companies tend to collapse and just become single drug companies. Whereas our thesis is that with this type of product focused tech bio companies, we weren't able to do that before. So, that has not been done before where we say, Hey, have a new class of molecule. We're not going to have the same challenges of getting into clinical trials and we're going to be able to prove that it has the correct properties faster. You've got a shorter patent life on those things.

你们在产品调整和优化方面拥有更大灵活性，而这正是传统药物化学路线难以企及的。这就是为什么这类...

You've got much more flexibility to pivot and tune the product, which is generally harder with a medicinal chemistry route. So that's the thesis why this type of it's very

复杂 是否可以用软件行业的演变来类比？就像AWS、Stripe和众多数据库供应商的出现，各种基础服务由不同平台提供，你只需接入就能获得所需资源，不必像九十年代那样从头构建一切。您展望的未来图景中，是否会出现数十甚至上百个这样的生物科技平台，让研发者能专注于核心创新，同时利用这些先进工具包？这个类比是否恰当？

complex Is it fair to say an analogy is the way the software industry evolved to have AWS and Stripe and a bunch of different database vendors and all the fundamentals being provided by different platforms where you could just tap in, get the thing that you need from them. You don't have to build it all from scratch the way we did in the nineties. Is that a fair thing where you project out your vision for where this goes and there's dozens of these or hundreds of these tech bio platforms out there, I can just focus on the thing I'm trying to make and I can get help from all these other very advanced tool kits essentially. Not a fair analogy?

这个领域存在一个困境：有些对研究过程至关重要的技术公司本应分享成果收益，但如果它们提供的只是可被替代的商品化服务或见解，通常难以真正参与利润分配环节。必须建立消费者购买产品与价值创造之间的直接联系。Illumina就是个成功案例——作为纯粹的DNA测序技术平台，他们为每位癌症患者、每个人类个体运行测序仪，这种模式直接连接着终端消费场景。

This is a struggle in this space because there are incredibly important technology companies that important for the research process and they actually need to be able to participate in the outcomes. But if they're some sort of a commodity or they provide some insight which could be supplanted by something else, they usually can't meaningfully participate in where the money is made, where the consumer And buys a so you have to have that direct line of connection between the consumer buying the product and the turn on the crank. And Illumina is a great example of a company that did do that, right? Like, they're a pure tech play platform that sequences DNA but they run their sequencers for every cancer and they run it for every human being and that it's directly connected to a consumer interaction at the

最终

end of

这就是

the day and so that's how

我倾向于认为这是个问题。你必须靠近价值提取点。

I It is tend to think a problem. You've got to be close to the value extraction point.

在我看来，至少在我们一直讨论的生物制药领域，企业确实会开发新的治疗模式，并因其主导新品类的能力而成为巨头。

I see it as companies, at least in this particular space where we've been talking about biopharma the whole time, are really going to develop new modalities and then going to become giants because of their ability to dominate that new category.

酷，是啊。

Cool, yeah.

我有两个问题。首先，你刚才谈到生物制药，我想听听你对当前以技术平台为基础、以多种方式出现的非侵入性疗法的看法。其次，你提到的许多直觉理念也存在于整合医学中，比如中医讲究整体调理或植物药关注完整化学结构而非单一分子——这种复杂相互作用的方式。你如何看待这些？

I have two questions. The first is, you were talking about biopharma. I'd like your thoughts on the sort of non invasive therapeutics that we're seeing emerge in a variety of ways that have a technology platform basis, one. And two, a lot of the intuitions that you're drawing from are intuitions that also come in integrative medicine, whether it's Chinese medicine paying for wellness or botanicals looking at the full chemical structure as opposed to a particular molecule as the complexity of the way it interacts and so how are you thinking about those things?

这个问题有几个不同层面的回答。首先，确实存在非侵入性治疗模式，超声波就是个很有趣的例子。我们与神经科技公司交流时，他们使用聚焦超声波技术通过声波相长干涉来精准定位大脑区域，用于治疗阿尔茨海默症、帕金森症和脑部炎症。至于中医方面，你知道的，已有相关企业，我认为这些将会以数字健康平台的形式崛起。

There are a couple of different answers to that. One is yes, there are non invasive modalities. Ultrasound is a really interesting one. We talk to neurotech companies that are using focused ultrasound where they get constructive interference of sound waves so that they can target individual parts of the brain, targeting things like Alzheimer's or Parkinson's and inflammation in the brain through that. On the Chinese medicine side, that's something, you know, there are companies and I think those are going to emerge as digital health platform plays.

我的导师是位资深细胞生物学家，他在哈佛医学院创立了治疗创新项目。后来他对中医产生了浓厚兴趣并希望理解这些理念。从技术生物学的角度来看，我们现在要完全理解并设计应对方案还为时过早，但我们会通过研究来理解其作用机制。未来我们将能检测更多代谢物和分析物，从而获得更深入的认知。不过商业核心在于提供医疗服务，这需要通过扩大认知规模和获取更多数据来实现——这也是我将其视为消费者端的原因。

My advisor who's a hardcore cell biologist and started the therapeutics innovation program at Harvard Medical School, He got very interested in Chinese medicine and wants to understand these things. I think that there's going to be, from a tech bio perspective, it's early for us to understand all of those things and be able to engineer against that but we will have research understanding of what are they doing. We'll be able to measure many more metabolites and analytes and be able to get more insight into it but the business is around providing care and that is going to be by scaling up our understanding of it and getting more data around it and that, I think, is why I think of that as the consumer side of it.

我个人很期待能理解中医。我的中国女友总给我各种看不懂标签的奇怪东西，我怀疑她是在训练我，好为将来毒死我做准备。但我还是照单全收，根本不知道自己在吃些什么。

I, for one, am looking forward to understanding the Chinese medicine. My Chinese girlfriend gives me all kinds of weird shit I can't read the labels on. I think she's just training me to make it easy to poison me someday. I just take it. I don't know what I'm getting into.

而且这完全是不透明的，但她觉得对我好就行。

And it's, completely opaque, but whatever she thinks is good for me.

嘿，谢谢分享。在我看来，我们对人体生理学的理解在理性药物设计这个概念上还处于非常初级的阶段。能否详细描述一下还需要多少创新才能实现？完全成熟的理性药物设计会是什么样子？

Hey, thanks for this. It seems to me that our understanding of human physiology is still kind of very early on with this notion of rational drug design. Much more, paint a picture for us of how much more innovation needs to happen. What does fully realized rational drug design look like?

理性药物设计意味着曾有一家名为Vertex的公司专注于此，他们的核心技术是分子动力学——能够模拟蛋白质及其所有相互作用。这些是非常复杂的量子计算，帮助他们理解单个分子与蛋白质口袋的详细结合方式。传统上这非常困难。我是这样理解你的观点的。但你还提到了生理学，这其实是另一个维度，涉及药物分布和药代动力学，实际上人们已经为此建立了非常优秀的模型，可以从基本原理出发进行理性设计。

Rational drug design means that there was a company called Vertex that was all about rational drug design and their technology was molecular dynamics, which is being able to simulate the protein and all of its interactions. Are very complex calculations, quantum calculations, that also help them understand the detailed binding of an individual molecule into the pocket of a protein. That has traditionally been very hard. That's how I'm interpreting what you're saying. But you also talked about physiology, which is a kind of a separate thing where you have distribution and pharmacokinetics and there are actually really good models that people have built for that, that can be done rationally, that is from first principles.

人工智能不等于理性药物设计。AI是机器学习，属于黑箱系统。我认为在实现真正的理性药物设计之前，我们将长期处于基于黑箱的开发阶段，因为那些量子计算极其困难。虽然可能有人会突破这个难题，但这可能需要多年的研究。大规模计算会有所帮助——实际上我们考察过一家公司，他们就是单纯利用大规模计算进行量子化学计算，这是整个领域中被忽视的英雄。

AI is not rational drug design. AI is machine learning and they are black boxes. And so I think that we are on a long road of black box based development before we get to real rational drug design because those quantum calculations are incredibly hard and I don't know what the answer like maybe somebody will crack that but it will have to be that's probably years of research. Compute at scale is going to help and actually we have company that we've looked at that is simply using compute at scale to do quantum chemistry calculations. And so that's an unsung hero in all of this.

这听起来非常基础，无非就是大规模计算，架设服务器。但这确实推动了进展，因为现在你可以批量完成某类量子计算，多次运算确实能带来改变。所以我想这应该算作理性药物设计的一部分。可以说我们已经实现了这个目标。

It seems really so pedestrian, just compute at scale, you have servers. But that has actually also moved the needle because now you can do a certain class of quantum calculations at scale and do multiple of them and that does make a difference. And so I guess that counts as rational drug design. So I think we're we're there already.

你也可以用华尔街的黑箱理论来类比描述这种情况。我永远无法窥见其他人的黑箱内部，但我们都在与之博弈。每个黑箱执行着不同策略，做着不同的事。它们不可解释，但都在同一个沙盘里运作。你必须在这种不完全理解的环境中竞争并取得好结果。

You could use the black box analogy from Wall Street as a way of describing this too. I don't ever get to see into everybody else's black boxes, but we're all trading against them. And so every one of them is executing a different strategy, doing a thing. They're unexplained, but they're all playing in the same sandbox. And so you have to compete against that and you have to get good results in a world where you don't actually understand everything.

你没有全部数据，看不到全局，无法预测他们会做出什么决策。但不知怎的，这就是我们的股票市场。现在唯一的玩法就是运行自己的黑箱。除非是买卖特斯拉、Robinhood和GameStop，否则人类在这个领域已无法进行有意义的交易。

You don't have all the data, you don't see everything, you have no way of predicting what decisions they're going to make. And somehow, that's our stock market. And the only way to play now is to run the black box. No human can trade meaningfully in that space unless they're buying Tesla and Robin Hood and GameStop.

关键在于立足市场。在生物学领域，这意味着你需要具备数据生成与积累的能力——无论是通过与人或消费者互动，还是通过科学流程来提升能力规模，因为这些数据并非现成存在于世界上。这正是这些公司与我们所知AI企业的本质区别。虽然它们显然也在获取新数据的采集方法，但更多是依托科学方法论和测量体系——通过内部实验获得专属模型与数据。因此这类公司确实展现出我们前所未见的独特形态。

It's all about being in market. And in biology, that means that you have a way of producing data and accumulating it and that could be by interacting with people, with consumers, or it means that you have a scientific process where you can scale up your ability because that data is not just sitting out there in the world. And so that's another way that these companies are very different from the AI companies we know. Though they're obviously getting collection methods for getting new data as well, but it's often about scientific methodology and a measurement methodology where they're just running those experiments internally and then those models are entirely theirs and that data is entirely theirs. And so these companies really have a different profile from what we're used to.

确实如此。有多少

They do. Yeah. How many

这类黑箱现象你观察到？从商业机遇角度看，是几十家还是数百家？你预见到行业整合趋势吗？通用人工智能是否会彻底颠覆这一切？就像'你们的所有黑箱都归我们所有'那种局面？

of these black boxes do you see? Is this like from a commercial opportunity perspective? Is it dozens or hundreds and then do you see some sort consolidation happening? And then does AGI just wipe it all out? They're all like, all your black boxes belong to us kind of thing?

我们很难精准预测通用人工智能的意义。若它能通过深邃的科学洞察力解决理性药物设计问题，那么确实可能出现一家能理性处理一切的公司。但以当前领域发展水平，我难以预见这种可能性，因为这属于通用人工智能才能掌握的'魔法'。我只是个普通人，我会

It's hard for us to predict exactly what AGI means. AGI can just solve rational drug design by brilliant deep scientific insight then yeah it's possible that there's a company that just can do everything rationally. But right now, where the field is, and it's hard for me to see that because this is some magic that the AGI will know. I'm just a regular human. I Will

让巫师来处理？

wizard do it?

是的，巫师或许能做到。我们目前处于黑箱世界，这意味着每个掌握数据池的玩家都将在其专注领域获得优势，成为细分赛道赢家。随后将是不同模式间的竞争——哪些更有效？行业整合很可能随之发生。但至少在我看来，这呈现出独特的业态。我预计会出现数百家这类企业，因为生物学极其复杂，即便在分子设计领域也存在大量差异化的干预机会。

Yeah. A wizard might be able to do it. We're in black box world for the time and what black box world means is that everybody who is sitting on their pool of data is going to have an advantage over whatever modality they are going to be the category winner. Then it's going to be a competition between what are these modalities, which work better and probably be consolidation amongst those companies as well but it just has a different flavor I think for me at least right now. I expect that there will be hundreds of those and I say that because biology is extremely complex and there are many different opportunities for intervening in different places even within the molecular design space.

你们有多少种基础分子设计策略？大环肽是一种，还有诱导接近策略——通过双价化合物靶向蛋白质隐秘位点实现特殊功能。此外还有可编程细胞疗法等丰富方向。可以预见某家巨头将构建核心技术来主导某个领域。现有生物技术行业就是现成的参照。

How many different fundamental molecular design strategies do you have? Macrocyclic peptides are one, there's another approach of induced proximity where you have bivalent compounds and then they maybe target a little cryptic site on one protein and they bind on another and they do special things like that. You might have other kinds, you might have people are working on programmable cells which are therapeutics, that's a very rich space. You can imagine one behemoth company really building the core technology on that and dominating that space. We can just take a look at the existing biotech industry.

当前存在大量机遇，我认为多家公司将在这个领域开发多种不同方法。有一家干细胞治疗公司正在使用诱导多能干细胞（iPSCs），这些细胞被逆转回胚胎阶段后重新分化，从而获得用于治疗帕金森病的定制细胞——这些细胞将取代黑质，就像用新细胞替换病变细胞一样，我们现在就能设计这种治疗方案。

There is lots of opportunity to go around and I think that multiple companies will develop many different approaches across that space. There's a stem cell therapy company that is taking these iPSCs, these induced pluripotent stem cells, where they're regressed back to the embryonic stage and then they re differentiate them so that we can get custom cells for treating Parkinson's and so you just get here are cells that are gonna replace the substantia nigra and you can just like supplant the disease cells with some new ones and we can design them and that's happening right now.

如果大脑功率是20瓦，而我们已知的最强大超级计算机与之相比，我们距离让大脑成为服务器的差距还有多远？这是否会催生原生超级智能，届时可能无法区分是我的思想还是运行在这个系统上的人工智能？

So if the brain is at 20 watts, the greatest supercomputer that we know of, how close are we to spanning the gap to where the brain becomes the server? And then does that translate into native superintelligence where it may be indiscernible, whether it's my thought or the artificial intelligence that's now operating on this system?

问题在于我们是否会对人类进行种系工程改造，以优化大脑结构？

Is the question whether we're going to do engineering, germline engineering on humans to improve the structure of our brains?

还要多久我就能成为承载这些计算的服务器，而不是...

How long until I'm hosting, this is the server hosting the computations instead of

哦，你是说我们下载软件并直接在脑中运行？

Oh, that we download software and run it actually in our brains?

你们现在就在做类似的事——训练自己的模型并进行推理，我们每个人都这样做过。现在的情况是，我们可能找到人工调整你模型中权重的方法。等等，这其实每当你刷TikTok时就在发生——你正在调整模型权重，持续训练着它。

So you're doing a version of that now, but you train your own model, and you're doing inference on your own model, and we each have done that. And so you're at a point where, you know, so the ways it could go are that we could find a way to artificially tune the weights in your model. Like, wait, that happens every time you scroll TikTok right now. You're tuning the weights in your model. You're constantly training it.

我们能在脑中安装新程序吗？我们什么时候能...

Can we install new programs in our brain? Can we When are we

要向宿主推销这个吗？我们会不会开始出租大脑空间？最终变成广告位。没错。

gonna sell to host that? Will we start to will we start to rent out space in our brain? And then eventually Ads. Right.

这是一场广告游戏。

It's an ad play.

然后开始游戏。现在我们知道了未来的商业模式。

And then play. Now we know the business model of the future.

但最终，这会不会变得难以分辨？我们会不会开始用超级智能来破解AI模型？我们会不会开始像'How'一样思考？

But eventually, does that become something that's indiscernible, and then do we start to use that super intelligence to figure out the AI model? So do we begin to think as if if we are the How

我们距离实现这种桥梁还有多远？目前我们处于非常低分辨率的交互界面。问题在于我们受限于这个为人类交流设计的低带宽界面。最多就是挖掉眼球插根网线，才能获得更高带宽的大脑接口。但离真正实现还很远。我们可以关闭麦克风，但更有趣的角度是探索其他向大脑输入数据的通道有哪些？

far away from bridging are we between I the we're doing now, but we're at a very low res interface. And so the problem is we're limited because we have this interface we created for communicating with other humans, and it's pretty low bandwidth. The best we could do is chop off an eyeball, and then we can stick an ethernet cable in there, and we can start to get a higher bandwidth interface to your brain. But we're really far from it. We could shut off the mics, but an interesting angle to go into this is what are the ways, what are the other channels of getting data into a brain?

我认为这个领域最有趣的问题是：什么是大脑？什么是脑机接口？这个领域长期属于哲学范畴，但我相信意识将不再只是哲学话题，它正在成为真正的科学领域，这将是科学创新和新科学基础的巨大突破。

Most interesting question I think in this space is what is the brain and what is the brain mind interface and this has been an area that has been solidly in the domain of philosophy and things I believe are going to change in that space where consciousness is not going to be a philosophical topic, it's emerging as a true scientific area and that is a huge area of scientific innovation and new foundations of science I think.

我们才刚刚起步。

Where we're just at the beginning.

我们甚至还没开始，我们处于炼金术世界，这不正常

We're not even at the beginning, we're alchemy world Not right

现在有风投公司愿意投资吗？

investable with a venture firm yet?

确实存在这样的情况，当你与神经科技创始人交谈时，你会发现这是我们很少讨论但非常感兴趣的领域——自从接触Brain Mind后。这些创始人要么研发医疗设备，要么开发消费级产品，深入交流后你会发现他们几乎都对开悟境界感兴趣。他们真正思考的是如何通过调节意识本身来达到不同的脑状态。目前主流观点仍停留在市场参与层面，但越来越多人开始认真研究意识基础，将开悟状态视为对世界本质的真实探索，而非仅仅是宗教体验。关于这个话题的讨论非常热烈。

There are definitely, like you talk to neurotech founders and this is another area we didn't talk about it very much but an area that we've gotten very interested in after going to Brain Mind. And you talk to neurotech founders and they have their medical device or they have their consumer device and you sidebar with them and you find out almost all of them are interested in enlightenment. And they are really thinking about how to get these other brain states that modulate consciousness itself. The normie thesis is there because that's how you can participate in the markets as far as we know right now, But there's an emerging interest in really understanding the basis of consciousness for people reaching enlightenment states and taking them seriously not just as an interesting religious experience but actually something more akin to real exploration of what the nature of our world is. There's a lot of conversation about that.

我最近经常思考这个问题，但这足够另开一集节目了。

I've been thinking about that a fair bit, but that's a whole other episode.

得让这些人行动起来才行，不过非常感谢。结束前还有什么问题要问我吗？

Gotta let these people get off their butts, but thanks a ton. Any questions for me before we wrap this up?

你是怎么做到的，Pablo？我该怎么做到——这就是你的专长所在，太不可思议了。

How do you do it, Pablo? How do I do This is just your whole thing. It's so amazing.

我不知道。

I don't know.

你接触过那么多我们恰好遇到的杰出创始人。当我们发现他们时，就会觉得帕布罗必须认识这个人，结果发现他们早就认识帕布罗了。

Just You've plugged in to so many amazing founders that we have happened to have met. Then when we find them, we're like, Pablo's has gotta know about this person. We find out they already know Pablo's.

关键在于你找不到他们。你得让他们主动找到你。至少对我来说是这样，对吧？因为我要找的是那些埋头在车库里、头发乱糟糟、开着德罗宁汽车的家伙。他们可不会去参加TED演讲

The trick is you can't find them. Like, you've got to make them find you. And at least for me, that's true, right? Because I'm trying to find these people who are heads down in a garage with crazy hair and a DeLorean. They're not going to Ted

或者什么脑力激荡会。我说的就是这类人

or Brain Mind or whatever. Those are the people I'm talking

没错。所以要找到他们，你得成为避雷针。我自己就是个吸引这些疯狂科学家的边缘怪人。就算不是直接吸引，这些人大多也是被引荐给我的。

about. Yeah. So to find them, you've got to be a lightning rod. And I'm like a fringe wacko myself who attracts a lot of these kind of mad scientists. And if not them, most of these folks are introduced to me.

我甚至收到过这样的电话：'我侄子正在车库里捣鼓危险的东西，我们完全看不懂他在做什么，但担心他会送命。你能和他聊聊吗？'我就说：'太好了，让他来找我吧。这就是我要找的创始人。'

I've deputized I get calls as my nephew's in the garage and it looks dangerous. We don't totally understand what he's doing, but we're afraid for his life. But can you just talk to him? I'm like, Gold, send him my way. That's my founder.

所以我走的是不同路线。我想尽早结识这些人，尽力帮助他们并找出最优秀的。这就是我的方式。顺便说，跳莎莎舞也有帮助。记下了。

So I'm on a different track. I do wanna meet these folks and I wanna meet them early and I wanna try and help them and find the best ones. That's how I do it. And salsa dancing helps too. Noted.

好的，太棒了。非常感谢。我觉得这次对话真的很有收获，我对你的愿景有了更清晰的认识，非常感谢你的分享。谢谢，谢谢

Okay, cool. Thanks a ton. I think this is really a great conversation. I'm getting a little bit closer to congealing with what your vision is here, and I really appreciate you speaking. Thanks, thanks

太棒了，这超级有趣。

so much. This was super fun.

我们会和任何想参加的人一起玩。阿尼尔？

And we'll hang out for whoever wants. Anil?

还有

And

如果你们想互相掰手腕的话，书足够多，大概够十几个人每人一本，我很乐意留下来签名。

then if you guys wanna arm wrestle each other, there's enough books for, I think, like, a dozen of you to get one, and I'm happy to hang out and sign them.