马丁·什克雷利：大药厂的盲点、做空量子计算，以及监狱教给他的一切

他被称作美国最遭人恨的人，因为他把一种救命药的价格抬高了。

He's been called the most hated man in America because he jacked up the cost of a life-saving drug.

把一种救命药的价格提高了5000%。

Raised the price of a life-saving drug 5,000%.

他成了资本主义贪婪的典型代表。

He became the poster child for capitalistic greed.

马丁·什克雷利，入狱七年。

Martin Shkreli, seven years in prison.

我们肯定打算涨价，这是肯定的。

We definitely plan on raising the price, that's for sure.

如果你不最大化利润，可能会被起诉，你必须尽一切可能赚更多的钱。

You can be prosecuted for not maximizing profit, and you have to do everything in your power to make as much money as possible.

如果我是你的敌人，那很好。

And if I'm the enemy for you, good.

我很高兴，兄弟。

I'm happy, man.

我在满足你生活中的某种需求。

I'm fulfilling some need in your life.

我买下了传奇武当派专辑的唯一一张，但没打算马上听。

Purchased the only copy of a legendary Wu-Tang Clan album with no immediate plans to play it.

这张专辑其实不怎么样。

It's not a good record.

自负是所有进步的源泉。

Ego is the fountainhead of all progress.

让我们欢迎马丁·施克雷利。

Please welcome Martin Shkreli.

看起来不错。

Looking good.

欢迎你，马丁。

Welcome, Martin.

你好吗？

How you doing?

谢谢你能来。

Thanks for coming out.

哦，是的。

Oh, yeah.

你好吗？

How are you?

很高兴见到你。

Nice to meet you.

你今天穿了布拉德的衣服。

You wore Brad's outfit today.

没错。

Exactly.

好的。

All right.

欢迎，马丁。

Welcome, Martin.

我们有很多话题想跟你聊，但我想先从一个开始。

We got so many topics to go into with you, but I want to start with one.

你基本上把量子计算给推翻了。

You've basically taken down quantum computing.

你说你认为这是个骗局，甚至可能有人要坐牢。

You said you think that this is a fraud, that people might be going to jail.

所以这个问题分两部分。

So two-part question.

第一，你能不能给我们讲讲你的论点，为什么这是个骗局，或者说是假的，是个假货。

Number one, maybe you could give us your thesis on why this is a fraud or fugazi, a fugize.

随便你怎么说。

Whatever it is.

丢人现眼。

Disgrazziod.

还有，丢人现眼。

And a disgrazziod.

然后第二个问题，对于那些真的进监狱的CEO们，你最好的建议是什么？

And then two, your best advice for those CEOs when they do go to jail.

你在这方面有点经验。

You've got a little experience.

哦，对。

Oh, yeah.

最好的建议是，你看起来不像那种能在监狱里混得好的人，但据我所知，马丁，你基本上——

Best advice in terms of, you don't seem to be the kind of guy who would thrive in prison, but from what I understand, Martin, you basically-

你就像个监狱里的老大。

You're like a shot caller.

等等。

Hold on

你在监狱里是老大？

You're a shot caller in prison?

你变成了莱克斯·卢瑟，管理着整个监狱，我听说。

You turned into Lex Luthor and you were running the yard, I understand.

那给我们讲讲，你进监狱的时候，量子交易和最佳实践是怎么搞的？

So break us through the quantum trade and best practices when you go to the hoosegow.

对，我得先感谢我妻子切尔西。

Yeah, so I have to shout out my wife, Chelsea.

切尔西，很少有人知道她，但我觉得她在科技界是个大人物。

So Chelsea, very few people know her, but she's, I think, a big deal in technology.

她是OpenAI最早的一批员工之一，而且她父亲，你信不信，是NVIDIA最早的一批人之一。

She was one of the first people at OpenAI, and her father, believe it or not, was one of the first people at NVIDIA.

在我家里，我是最早进监狱的那批人之一。

In my family, I was one of the first people in prison.

所以——我想着得改变一下基因库，稍微调整一下方向。

So- I thought I'd switch the gene pool, shift it over a little bit.

但切尔西在麻省理工读书，还跟斯科特·阿伦森一起教过量子计算，那可是这个领域的顶尖人物——

But Chelsea studied at MIT and taught quantum computing with Scott Aaronson, the preeminent guy-

当然

Sure

在量子领域，她花了很多时间教我，终于让我这个榆木脑袋明白了量子计算到底是什么。

... in quantum, and she got to teach me over many hours, penetrated my thick head, what exactly quantum computing even is.

我觉得很多人把它和更快的计算混为一谈，听起来很棒。

And I think a lot of people sort of synonymize it with faster computing, which sounds great.

但实际上根本不是那么回事。

That's really not what it is at all.

对。

No.

我觉得量子计算的核心在于有一种算法。

And I think the basic point of quantum computing is that there's one algorithm.

实际上，整个领域是因为彼得·肖尔的思维实验而诞生的。

In fact, the whole field was invented because of Peter Shor's thought experiment.

彼得考虑了一个想法，就是有些算法可以在不同的媒介上运行得更快。

Peter considered this idea that there are kinds of algorithms that you can run faster in different media.

这在计算领域是个革命性的想法，但对商业来说并不算革命性，因为，当然这是我的论点，因为只有一种算法在量子计算机上运行得更快。

That was a revolutionary idea in computing, but it's not really a revolutionary idea for business because, and this is my thesis, of course, because there's one algorithm that runs faster in a quantum computer.

所以如果你问别人，哪台电脑能给一副牌排序，如果大家知道那本著名的算法书CLRS，前五页讲的就是排序算法，比如冒泡排序、归并排序之类的。

And so if you ask somebody which computer can sort a deck of cards, if anybody knows the famous algorithms book, CLRS, the first five pages are about sorting algorithms, like bubble sort and merge sort and things like that.

我问那些量子计算爱好者：“那哪台能排得更快呢？”

And I asked quantum fans, "Well, which one can sort it faster?"

他们说：“那当然是量子计算机了。”

And they said, "Well, of course, the quantum computer."

我说：“当然不是。”

And I said, "Of course not.

“不是那样的。”

That's not the case."

现代量子计算机的时钟速度只有几百赫兹或者几千赫兹，而不是我们习惯的千兆赫兹。

The modern quantum computers run at clock speed of hundreds of hertz or thousands of hertz, not at the gigahertz that we're used to.

它们是单核的，而且在执行像归并排序这样的任务时也是单线程的。

They're single core, and they're single threaded when it comes to doing a job like a merge sort or something like that.

所以它真正擅长的，其实是个魔术般的技巧，就是把指数时间算法变成多项式时间算法，比如整数分解或者椭圆曲线离散对数问题，这个我稍后想提一下。

So the only thing it's really good at, which indeed is a magic trick, is going from exponential time algorithm, which is integer factorization or elliptic curve discrete log problem, which I want to mention a little later.

这里得给ECDLP问题点个赞。

Shout out to the ECDLP problem.

量子计算机能在多项式时间内解决这些问题，这确实很了不起。

The quantum computers can do those in polynomial time, which is remarkable.

对技术宅来说，这简直是最酷、最神奇的魔法。

And to a nerd, that's the greatest, coolest, magic trick ever.

但对商人来说，除非你专门做因数分解的生意，否则这里真没什么值得兴奋的。

But to a business person, unless you were in the business of factoring numbers, there's not a whole lot to be excited about here.

量子能做的其他算法速度也没快多少，所以它们永远比不过NVIDIA的机器。

So the other algorithms that quantum could do are not much faster, so they'll never beat an NVIDIA machine.

它们永远比不过一百万台NVIDIA机器。

They'll never beat a million NVIDIA machines.

而且量子计算机没有存储。

And quantum doesn't have storage.

根本不存在什么量子内存。

There's no such thing as quantum RAM.

所以你得设计一个输入非常非常小的问题。

So you have to design a problem that has a really, really small input.

你不能把全世界的计算能力都塞进去。

You cannot put the world's compute into it.

每次有人向OpenAI推销量子计算机，据我所知，他们都会说：'不，我们不感兴趣。'

Every time somebody pitches OpenAI, from what I've heard, about a quantum computer, they say, "No, we're not interested."

我觉得在量子问题上我跟你的看法比较接近，但让我站在另一边说说。

I think I'm closer to you in my perspectives on quantum, but let me take the other side.

我们就从另一边来看吧。

Let's just take the other side.

另一边会说，你看，我们现在正处于真正构建晶体管的非常早期、原始阶段。

The other side would say, look, we're in this very early, primordial phase of actually constructing the transistor.

有了晶体管之后，我们要做的就是把一堆这样的东西组合起来。

And from the transistor, what we're going to do is put a bunch of these things together.

我们要搞清楚逻辑门。

We're going to figure out gates.

我们要搞清楚逻辑。

We're going to figure out logic.

然后我们就能在上面建一个像样的操作系统。

We'll then be able to build some reasonable operating system on top of it.

我们会有一些合理的办法来给这些设备写软件。

We'll have some reasonable mechanisms to write software to these things.

而这些现在都还不存在。

And none of that really exists.

那你觉得我们是不是真的还处在早期阶段，所以才觉得计算能力受限，只能用在很窄的用途上？

So do you buy that we're just at an early phase, and this is why we're frustrated that there's a computational limitation, except to a very narrow use case?

你觉得这有可能吗？

Do you believe that that could be possible?

对，当然有可能。

Yeah, no, of course.

但我觉得问题在于，你会不会长期停留在打孔卡和真空管的阶段，对吧？

I think the issue, though, is would you go long punch cards and vacuum tubes, right?

水银延迟线、磁芯存储器，这些都是消失并破产的东西。

Mercury delay lines, core memory, these are all things that disappeared and went bankrupt.

这些曾经也是股票。

Those were stocks at one point.

CTR，计算机制表机公司，后来并入了IBM。

CTR, Computer Tabulator Machine, got incorporated into IBM.

好，但你的批评是商业模式和时机的问题，还是技术上的问题？

Okay, but your critique is business model and timing, or is it technological?

我觉得两者都有。

I think it's both.

我觉得现在你去问谷歌、IBM或微软，问他们为什么每年在这上面花几亿美元，

I think when you talk to Google or IBM or Microsoft now and ask them, "Why are you spending hundreds of millions of dollars a year on this?"

他们会说，也许有一天会出点什么成果。

They say, "Well, someday there might be something that comes out of it."

对我来说，这不算什么好生意。

That's not a great business to me.

确实挺让人震惊的。

Well, it is very shocking.

你大概能讲讲这个故事，但IBM和谷歌到底有多少台机器？

You can probably tell this story, but how many machines does IBM and Google actually have?

非常少。

Very few.

谷歌有几台原型机。

So Google has a few prototypes.

IBM有一个小的云服务。

IBM has a little cloud.

我和Chelsea试过世界上所有的量子计算机。

Chelsea and I have tried every single quantum computer in the world.

我们在任何一台上面都没法算300乘以200。

We cannot multiply 300 by 200 on any of these.

那你在做空它们吗？

And you're shorting them?

是的，我一直在做空它们。

Yeah, I've been short them.

不幸的是，我们在做空量子计算公司上亏了几百万美元，但我跟我妻子说，我们获得了价值数百万的洞察。

Unfortunately, we lost millions of dollars shorting quantum computing companies, but I told my wife that we gained billions in insight.

把握市场时机确实不容易，是的。

It's not easy to time the markets, yes.

这绝对是——

It's definitely-

那么，自然计算的优势在哪里？

So where is the natural computational edge?

有人会说，优势在于加密和破解加密，因为现在我们可以利用它真正擅长的东西，来暴力破解那些我们今天无法解决的问题。

Some people would say it's going to be towards encryption and defeating encryption because now we can actually use what it's actually good at to essentially brute force solutions to things that we can't solve today.

我觉得想这个都疯狂，因为我们有一个要解决的计算问题，每次有人提到AI，我其实想到的是矩阵乘法。

I think it's crazy to even think about it because we have a computational problem that we're trying to solve, and every time somebody uses the word AI, I actually think of the word matmul.

这些就是矩阵乘法机器，而且我觉得当我们——

These are matmul machines, and I think that as we-

他说的就是矩阵乘法。

Matrix multiplication is what he's referring to.

所以，把两组数字相乘，就是GPU百分之九十九点九的时间在做的事情。

So multiplying two lists of numbers is what a GPU spends 99.9% of its time doing.

没错。

Sure.

但它的时间复杂度是N的三次方。

But it does it in N cubed time.

有趣的是，我觉得像复杂度算法这样的概念，在未来五到十年里，随着我们社会的发展，会变成常用语。

Now, what's funny, I think things like complexity algorithms are going to be parlance in the next five to 10 years as we move towards a society.

所以N的三次方意味着，如果你想在Nvidia的机器上把一个1000乘1000的方阵和另一个1000乘1000的方阵相乘，需要十亿次运算。

So N cubed means if you want to multiply a 1,000-element square matrix versus another 1,000-element square matrix on an Nvidia machine, it takes a billion operations.

那可是很多运算。

That's a lot of operations.

通过我亲爱的妻子了解到这些计算知识后，我翻看《自然》杂志——这是我们平时的爱好——看到了一篇来自微软的论文，叫做AOC，模拟光学计算。

As we learned about this computing stuff through my lovely wife, I was flipping through "Nature," which is what we do as a hobby, and I saw a paper from Microsoft that was called AOC, analog optical computing.

在纽约，AOC 的意思可不一样，但这里的 AOC 是指那个。

In New York, that means something different, but AOC in this case is.

我眼角余光看到的是，光里的矩阵乘法是免费的。

And what I saw in the corner of my eye was that matmuls in light are free.

它们瞬间就完成了，这真的很了不起。

They happen in instant time, which is a really remarkable thing.

所以我们正在研究，也许我们会开一家公司之类的。

So we're investigating, maybe we'll start a company or something like that.

但有几家初创公司，比如 Neuro-Phos、OptiCore、Lumi，它们想制造这些光子矩阵乘法核心，性能会远超英伟达。

But there's a couple of startups like Neuro-Phos, OptiCore, Lumi that want to make these photonic matmul cores, which would run circles around what Nvidia does.

当然，这还不是全部。

Now, of course, that's not the whole story.

还有很多细节，比如非线性、存储，还有——

There's a lot of details and nonlinearity and storage and-

所以，除了你对量子的厌恶和你那可能赢也可能输的激进做空之外，你以前是对冲基金经理，对吧？

So just beyond the hatred for quantum and your raging short that may or may not be winning, you were a hedge fund manager, correct?

那个著名的故事，为了让大家跟上进度，就是你买了一种药，然后抬高了药价。

And the famous episode, just to catch everybody up, was you bought a drug and then you inflated the price of the drug.

是这样吗？

Is that what happened?

你能不能给我们讲讲——

Can you just give us the-

当然可以。

Sure

——就是——

... just the-

好，我当时在——

Yeah, so I worked-

——那段历史，以及是什么让你现在走到……

... the history and what takes you now to...

对，我16岁就开始在吉姆·克莱默的对冲基金工作了。

Yeah, I started at 16 at Jim Cramer's hedge fund.

克莱默其实是个非常厉害、非常有技巧的交易员。

Cramer was actually a tremendously effective and skilled trader.

但克莱默的投资理念就是每天从早到晚不停地买卖各种东西，而且他居然还能从中赚到钱。

But Cramer's idea of investing was to buy and sell all day, every day, everything, and he somehow squeezed profits out of it.

在八九十年代，大家差不多都是这么干的。

That was the way you do it sort of in the '80s and '90s.

而且他扣除费用后，每年还能赚23%。

And he actually made 23% a year net of fees.

所以你16岁就给他打工了？

And so you worked for him as a 16-year-old?

对，我16岁就跟着他干了。

I worked for him at 16.

你是怎么得到那份工作的？

How did you get that job?

就是因为在曼哈顿，我嘴皮子利索，而且那是个很棒的城市。

Just being in Manhattan, having a big mouth and it's a great city.

而且我特别热爱股票，热爱投资。

And loving stocks, loving investing.

比尔·盖茨是我的偶像。

Bill Gates was my hero.

我找到了进入这个领域的方法。

And I found a way to get into that.

有一天，克莱默指着我说：“基因组学。”

And Cramer one day pointed at me and said, "Genomics.

“你要去研究基因组学。”

You're going to look at genomics."

我说：“好的。”

And I said, "Okay."

那真的很有趣，后来我去了一个老虎基金的分支机构。

And it was a lot of fun, and I went to a Tiger Cub.

我创办了自己的基金。

I started my own fund.

我意识到我更适合做企业家和私募股权，而不是交易员。

I realized I was a much better entrepreneur and private equity guy than a trader.

我们发现分子市场有巨大的超额收益空间，而在上市公司市场，我们可以买药——我从百时美施贵宝花两百万美元买了一种药，而这家公司今天还是上市公司。

We found that the marketplace for molecules had huge amounts of alpha, whereas the marketplace at public companies, we could buy a drug from-- I bought a drug from Bristol Myers for $2 million, and that same company is public today.

现在它值六十亿美元。

It's worth $6 billion.

所以那是我的第一个创业项目。

So that was my first startup.

那真是一个效率很低的市场。

And it was a really inefficient market.

辉瑞会以一百万美元的价格卖给你一种药，而这种药价值十亿美元。

Pfizer would sell you a drug for a million bucks that was worth a billion dollars.

他们根本不在乎。

They didn't care.

把药的权益卖给你。

Sell you the rights to the drug.

你可以把它产品化，定价。

You could productize it, price it.

对，完全正确。

Yeah, absolutely.

抱歉，你是什么意思？

I'm sorry, what do you mean?

你就直接去辉瑞敲门，他们就说：“随便挑，一百万美元。”

You just knock on Pfizer's door and they're like, "Take what you'd like for a million dollars."

我们看到了这一点，就建了两家非常相似的公司，我们最大的限制是找不到人来评估这些资产。

We saw this, and we built two very similar companies that our biggest constraint was finding people to vet these assets.

所以我开始了一个小型的自然语言处理——

So I started a small NLP-

临床方面的？

Clinically?

对，就是读论文、读专利，去大学、小药厂和大药厂，建立关系。

Yeah, just reading papers, reading patents, going to universities and small drug companies and big drug companies, making relationships.

在我们这个领域，这叫做业务拓展，而我们最大的瓶颈就是阅读这些论文。

It's called business development in our field, and our biggest bandwidth problem was reading these papers.

所以2015年我顺便创办了一家小型NLP公司。

So I started a small NLP company on the side in 2015.

我们有Watson。

We had Watson.

那挺好的。

It was great.

我们试图阅读所有——其实并不好。

And we tried to read every-- It wasn't great.

我们试图阅读医学领域发表的每一篇论文。

We tried to read every paper that was published in medicine.

有4000万篇论文，靠人力去做真的很难。

It was 40 million papers, and having humans do that is a really tough thing.

你是怎么把一个分子从2500万美元变成60亿的？在你拥有它的那六到九个月里，你做了什么？

How do you take a molecule and flip it from $25 million to 6 bill-- What are you doing in that six or nine months that you own it?

所以我针对司帕生坦做的，是我命名的药物，现在已经获得FDA批准，成为重磅药物，最初是百时美施贵宝发现并开发用于治疗高血压的，那是一个没什么人真正需要的、死气沉沉的市场。

So what I did with sparsentan, a drug that I named and is now FDA-approved, now a blockbuster drug, is Bristol discovered and developed it for hypertension, which is sort of a dead, sleepy market that nobody really needs.

高血压，就是血压高。

Hypertension, high blood pressure.

对。

Yep.

然后我们转向了一种叫FSGS的疾病，这种病从来没有药物获批过。

And so we moved into a disease called FSGS, which is a disease that never has had a drug approved for it.

所以你得想办法找到合适的切入点。

So you sort of try to angle it in a way that makes sense.

在这种情况下，这个药获得了FDA的批准。

And in that case, the drug got FDA approved.

有三万名患有局灶节段性肾小球硬化的患者，以前没有药可用，现在有药了。

30,000 people who have focal segmental glomerulosclerosis and never had a drug before now have a drug.

我也可以做个好人。

I can be a good guy, too.

我觉得这就是这个领域的艺术所在。

And I think it's the art of that field.

这是赚十亿美元的好办法。

It's a great way to make $1 billion.

这是一个——

It's a-

这就是Royalty Pharma做的事情吗？

Is this what Royalty Pharma does?

Royvent。

Royvent.

但这也是Royalty Pharma做的事情吗？

But is this also what Royalty Pharma does?

他们会介入——

They'll come in-

对，Royalty有点不同，他们当然也是天才团队。

Yeah, Royalty's a bit of a different player of genius guys, obviously.

他们最出名的事迹是去了纽约大学。

Their big claim to fame was they went to NYU.

他们从类克那里拿到了一大笔专利费，而且是以一年收益的价格买下的。

It was getting this big royalty from Remicaid, and they bought it for one times earnings.

然后那笔专利费涨了20倍。

And then that royalty 20 X'd.

他们以比市场定价更大的折扣买入现金流——

They buy the cash flows at a bigger discount than how they're being priced-

布拉德，我知道你想插话，那就说吧。

Brad, I know you want to get in here, so jump in.

马丁，现在用这个框架来分析。

Martin, take this framework now.

你知道现在市场上的主要杠杆是什么，对吧？

You know kind of the big levers in the market today, right?

在算力方面有什么预期，在资本支出建设方面有什么预期。

What's expected in terms of compute, what's expected in terms of this CapEx build-out.

你花很多时间研究这个吗？你看到什么潜在的风险了吗？

Are you spending much time looking at that, and do you see any boogeymen there?

你担心什么？

What are you worried about?

你会做空什么？

What would you be shorting?

在那个领域你会做多什么？

What would you be long in that complex?

对，我之前提到过光子计算，因为我确实认为20年后，这些H100、B100、B200都会变成废铁。

Yeah, I mentioned photonic computing before because I do think in 20 years, these H100s are in the scrap heap, B100s are in the scrap heap, B200s in the scrap heap.

我们会找到更好的方法来做这件事。

We're going to find ways to do this better.

我觉得我们现在考虑太空真是太疯狂了。

I think it's insane that we're thinking about space.

我觉得我们现在考虑核能也太疯狂了。

I think it's insane we're thinking about nuclear.

萨姆·奥尔特曼终于公开表示，他认为光子计算可能是未来，他是在Arrival播客上说的。

Sam Altman finally said publicly that he thinks photonic computing is on Arrival Podcast that photonic computing could be the future.

所以我就纳闷，为什么我们会觉得这是个靠谱的想法。

So I'm just wondering why we think this is a sane idea.

用电脑把地球铺满，而我们明明能造出更快的。

To cover the Earth with computers, when we can make a faster one.

我觉得这并不意味着英伟达不行，我现在还做多半导体呢。

I think that that doesn't mean NVIDIA's not along, and I'm long semis right now.

我觉得杰文斯定律最终总会赢。

I think Jevon always wins in the end.

大家都在说杰文斯悖论——

Everyone Jevons paradoxes-

——把一切都搞砸了。

... the crap out of everything.

亚历克斯·王告诉我，杰文斯会搞定这一切。

Alex Wang told me Jevon's going to take care of all this.

计算能力会提升一百万倍。

It'll be a million X compute improvement.

硅谷有人在讨论更好的Transformer，性能提升一千倍。

There are guys in the Valley talking about better transformers that are 1,000 times better.

如果Transformer效率提高一千倍，那对英伟达来说可不是好事。

If there's 1,000 times more efficient transformer, that's bad for NVIDIA.

这一点毫无疑问。

There's no doubt about that.

所以，我们不会有一百万倍的需求增长。

So, we're not going to have a million X demand.

也许从其他方面会有，但人口不会增加一百万倍。

Now we might in another way, but there aren't a million X more people.

至于我们是否需要一百万倍的计算能力，还得看情况。

Whether we have a million X more compute needs, we'll have to see.

目前，我的猜测和大家一样，是看涨的，但我在2000年可能也会猜雅虎和思科会涨。

For now, I think my guess, like anybody else's, is long, but I probably would've guessed in 2000 that Yahoo and Cisco longs, too.

所以这是个棘手的地方。

So it's a tricky place.

那你现在还管理对冲基金吗？

So do you run a hedge fund still?

不管理了。

I don't.

好的。

Okay.

所以你现在是在管理自己的钱，当你考虑资产配置的时候，你是按照以前管理基金时的方式去思考，还是专门寻找那些高阿尔法事件、颠覆性创新或者创业机会，然后坚定地做多，而不太在意跟市场比拼？

So you're managing your own money, and when you think about portfolio allocation, are you thinking about it in the way you did when you ran your fund, or are you looking for specific high alpha events or startup or disruption and really going long behind those and not really thinking about competing with the market?

对，我比较灵活。

Yeah, I'm sort of nimble.

我觉得这次达里奥的交易是个很好的做多理由，所以我就加仓了，但我不太——

I think this Dario trade was a great reason to go long, so I sort of piled in onto the sphere, but I don't-

你是说昨天的抛售其实只是虚晃一枪？

Meaning the sell-off yesterday was sort of just a head fake.

对，我觉得那太离谱了，是的。

Yeah, I think it was ridiculous, yeah.

你对他的文章怎么看？

What did you think about his essay?

我想对他创建了一家万亿美元公司表示敬意，但我有点讨厌这个人，而且我觉得他——创建一家万亿美元公司并不容易。

I want to pay deference to him creating a trillion-dollar company, but I kind of hate the guy, and I think that he it's not easy to make a trillion-dollar company.

你为什么讨厌他？

Why do you hate him?

我觉得他是个散布恐惧的人。

I think he's a fearmonger.

想象一下，你和我一起创业，我说：“我要给你看潘多拉的盒子。”

Imagine you and I started a business, and I said, "I'm going to show you Pandora's box.

“就是这个。”

Here it is."

然后你说：“哇，太神奇了。”

And you say, "Wow, that's amazing."

然后你说：“非常感谢，马丁。”

And you said, "Thanks a lot, Martin.

“我要去街那头，也开一个潘多拉魔盒。”

I'm going down the street, and I'm going to start Pandora's box, too."

我说：“你在干什么？”

I said, "What are you doing?"

我可能会以泄露商业机密之类的理由起诉你。

I would probably sue you for trade secrets or something.

这恰恰就是发生的事情。

This is exactly what happened.

所以你的意思是，因为他在OpenAI工作，然后他和五六个联合创始人离开了，你是说，就像——

So you're saying because he was at OpenAI, and then he and his five or six co-founders leave, you're saying that, like-

你会和那种人做朋友吗？

Would you be friends with that guy?

你会喜欢那样吗？

Would you like that?

我觉得硅谷那边这种事挺常见的。

Well, I think the thing with Silicon Valley is that it kind of happens a lot.

而且很难说清楚哪些算专利，哪些只是脑子里记住的知识。

And it's sort of hard to pinpoint what is a patent versus what is retained knowledge.

我不是在替谁开脱，也不是说我知道发生了什么，但是——

I'm not excusing it, and I'm not saying I know anything that happened, but-

那个Waymo的家伙就是因为这个进监狱的。

The Waymo guy went to jail for it.

我觉得那是因为他被发现确实偷了实体文件。

Well, I think that that was because he was found to have actually stolen physical documents.

好吧，所以你对Anthropic的起源故事有意见。

Okay, so you have an issue with sort of the origin story of Anthropic.

就是那些危言耸听，那些夸张得离谱的说法。

That, the fearmongering, the ridiculous hyperbole.

他说AI会帮助药物研发。

He says that AI's going to help drug development.

我们会治愈所有疾病——

We'll cure all diseases-

好吧，让我问你，马丁——

Well, let me ask you, Martin-

……在五年内。

... in five years.

完全。

Total

那有效利他主义呢？

What about effective altruism?

对，更糟。

Yeah, even worse.

为什么？

Why?

你觉得这群人有什么让你这么讨厌的？

What do you find so loathsome about this group of people?

我们来分析一下。

Let's unpack it.

所以我帮助推广了e/acc运动，甚至可以说是我发起的，这是一场反叛乱运动。

So I helped popularize, and I even think I would say start the e/acc movement, which was this counterinsurgency.

我认为——

I think that-

抱歉，有EA——

Sorry, there's EA-

对

Yes

然后还有e/acc。

... and then there's e/acc.

这个名字是故意取的，目的是——

Which was intentionally named to-

EA是有效利他主义。

EA is effective altruism.

利他主义。

Altruism.

e/acc 是什么？

E/acc is?

有效加速主义。

Effective accelerationism.

有效加速主义，好的。

Effective accelerationism, okay.

意思是——

Which means-

什么意思？

Which means?

这是个好问题。

That's a great question.

它的意思就是快速行动，打破常规。

It does mean sort of move fast and break things.

而且大家都知道，扎克伯格前几天公开说过：“我们在AI方面应该加速前进。”

And it's no secret that Zuckerberg came out the other day and said, "We should move fast in AI."

我觉得那些危言耸听有点夸张，将来确实会有需要恐慌的时候。

I think the fearmongering is a little ridiculous, and there'll be a time to panic.

但现在还不是时候。

It's just not now.

物理AI就提供了这样的机会。

Physical AI presents those opportunities.

目前来说，你完全可以关掉电脑。

For now, you can turn off the computer.

这没什么大不了的。

It's not that big of a deal.

我觉得那些技术宅们有点过于兴奋了。

I think nerds get carried away.

如果你曾经是个书呆子或者运动健将，我其实两者都沾点，信不信由你。

If you ever were a nerd or a jock, I was, believe it or not, a little bit of both.

那些运动健将知道在聚光灯下该怎么做。

The jocks know what to do when the spotlight's on them.

而那些书呆子不知道，他们会紧张得说不出话来，有些人不但不说一切都好，反而想吸引别人的注意。

The nerds don't, and they clam up, and some of them, instead of saying everything's fine, they want to draw attention to themselves.

他们想说：“嗯，我正在做的东西真的很了不起。”

They want to say, "Well, the thing I'm working on is really amazing."

但现实是，这不过是软件而已。

And the reality is that this is just software.

这不是世界末日。

It's not the end of the world.

正如你之前所说，这是有史以来最重要的软件，但它不会把我们全都杀死。

It is the most important kind of software ever made, as you said earlier, but it's not going to kill us all.

大约两年前，达里奥还说我们会在1到5年内完全失业。

Dario also said that we'd be fully unemployed within one to five years, about two years ago.

好吧，他还有三年时间。

Well, he's got three years left.

他还说我们能治好所有疾病。

He also said that we'd cure all diseases.

我觉得未来十年内不会出现任何与AI相关的药物。

I don't think there'll be a single AI-relevant drug in the next 10 years.

AI对药物研发没有帮助。

AI does not help drug discovery.

如果你做过药物研发，请举手，谁做出过药物，并且获得了FDA的批准？

If you've ever built drugs, and raise your hand if you've made a drug, gotten FDA approval for a drug.

好的，我确定就这些。

Okay, I'm pretty sure that's it.

这非常难。

It's pretty hard .

做这件事非常难。

It's pretty hard to do.

而且我们使用计算机已经很长时间了。

And we've had computers for a long time.

我建了一个Cray集群来分析药物。

I built a Cray cluster to analyze drugs.

现在我能用GPU了。

Now I can use GPUs.

这很棒，但在药物发现方面，AI能帮的忙不多，除了——

It's great, but there isn't much AI can help us with in drug discovery other than-

推动估值的收入周期是什么？

What is the revenue cycle that's driving valuation?

那就从对冲基金的角度来拆解一下。

So just unpack it from a hedge funder.

你坐在那里，心里想，嗯——

You're sitting there, you're like, well-

那是个对冲基金。

It's a hedge fund

……你看到这些运行率，对吧？

... you see these run rates, right?

我觉得昨天詹森提到过，有四千亿美元的风险投资被转投到了人工智能领域。

I think Jensen mentioned it yesterday, $400 billion have gone into venture that has been redirected into AI.

那你是不是觉得，风投是在用有限合伙人的钱去买代币？

So do you interpret that as VCs funneling LP money to buy tokens effectively?

对，我是这么想的，而且——

Yeah, I think so, and-

这就是你的论点。

That's your argument.

而且我觉得你会遇到一个——

And I think you're going to have one-

那你觉得这里面有估值陷阱吗？

So there's a valuation trap in there, do you think?

我不这么认为。

I don't think so.

我觉得你之前提到的生产力问题，对智能和工作的需求是无限的。

I think that what you mentioned earlier with productivity, there is infinite demand for intelligence and infinite demand for work.