迈克尔·尼尔森——科学是如何真正进步的

我想听听你对此的看法。

And I want to get your reaction to that.

是的，我认为这是一个非常有趣的问题。

Yeah, I think it's an incredibly interesting question.

我的意思是，这可能是一个关键性的问题。

I mean, maybe a really pivotal question.

如果你持一种非常经典的观点，你会希望拥有这些深层次的解释性原则，希望尽可能少的自由参数，希望用非常简单的模型解释大量现象。

The sense of, if you sort take a very classic point of view, you want these deep explanatory principles, you want sort of as few free parameters as you possibly can, you want very simple models which explain a lot.

但AlphaFold完全不像那样。

And AlphaFold doesn't look anything like that.

因此，你可能会直接说，它不错，也许作为一个模型有帮助，但它并不具备科学解释的性质。

And so you might just say it's nice, it's maybe helpful as model but it doesn't have, it's not a scientific explanation.

所以这是一种保守的观点，可以算是对这个问题的第一个回答。

So that's like a conservative point of view, that's sort of answer one to the question.

我认为第二个回答是：也许你不应该把AlphaFold当作经典意义上的解释，但它内部可能包含许多细微的解释，因此，可解释性研究的一部分价值，就是你可以深入AlphaFold，开始提取其中的某些内容。

I think answer two is to say something like maybe you shouldn't think about AlphaFold as an explanation in the classic sense But maybe it contains lots of little explanations inside it and so maybe part of what you can get out of interpretability work is you can go into AlphaFold and you can start to extract certain things.

也许通过对AlphaFold进行类似考古的研究，我们可以更深入地理解这些原理。

Maybe basically by doing sort of archaeology of AlphaFold, we can actually understand a great deal more about these principles.

你可以开始提取所有这些内容，这个电路做了这件有趣的事，我们因此学到了东西。

You can start to extract it all, that circuit does this interesting thing and we learn this.

所以我不知道这种研究在AlphaFold上已经进行了多少。

So I don't know to what extent that's been done with AlphaFold.

我知道在象棋模型之类的东西上已经做了一些这样的工作。

I know it's been done a little bit with something like the chess models.

我认为是AlphaZero。

I believe it's AlphaZero.

似乎有一些策略至少被马格努斯·卡尔森借鉴了，他似乎直接从AlphaZero那里学来了这些。

There seem to be some strategies which were certainly borrowed by Magnus Carlsen at least, which he seems to have just taken from AlphaZero.

我不认为对此有任何公开确认，但一些专家注意到，在关于AlphaZero工作原理的公开分析发布后，他的棋风发生了显著变化。

I I don't think there's any public confirmation of this but some experts have noticed that he changed his game quite radically after some public forensics were released on how AlphaZero worked.

因此，这是一个例子，表明人类正开始从这些模型中提取意义，也许这会促使我们把模型视为潜在的解释来源。

So that's kind of an example where I think human beings are starting to extract meaning out of these models and maybe that starts to lead to viewing the models as a potential source of explanations.

你需要做更多工作，因为它们一开始并不容易理解，但你有可能从中提取出信息，我认为这是一种有趣的中间状态——它们本身并不是解释，但你可以从中提取出有趣的解释，并将其用作一种来源。

You need to do more work because they're not very legible upfront but you can extract them potentially and I think that's kind of an interesting intermediate situation where they're not explanations but you can extract interesting explanations out of them, can use them as of a source.

我认为第三种、也是最有趣的可能性是，它们在某种意义上是一种新型对象。

I think the third and the most interesting possibility is they're a new type of object in some sense.

它们应当被认真视为解释，而过去我们根本没有能力真正利用它们，但现在我们将能够开展一些有趣的新操作。

They should be taken very seriously as explanations that where in the past we haven't had the ability to really do anything with them Now we're going to have interesting new actions which we can do.

我们可以将它们合并、提炼，做各种这样的事情，这将在科学哲学领域开启一个巨大的新机遇。

We can merge them, we can distill them, we can do all these kinds of things and there's going be almost a new, it's a big opportunity in the philosophy of science start to do that.

某种程度上，我感觉这种预期已经存在了，比如我知道一些数学家和物理学家，历史上如果你遇到一个长达一百页的方程——这种情况确实会出现——在十九世纪或二十世纪，你根本无能为力。

There's sort of an anticipation of this in some sense I think in the way I know some mathematicians and physicists who historically if you had like a 100 page equation, which is the kind of thing that does come up, there's just nothing you can do if it's nineteen, twenty.

你根本无能为力。

There is nothing you can do.

到了那一步，你就只能放弃这个问题。

At that point you give up on the problem.

而如今，有了像Mathematica这样的工具，你可以继续推进，于是它现在成了一个你可以处理的对象，也存在一些例子：人们曾经认为某些东西太复杂而无法处理，但最终却从中得出了简洁的答案，这只不过是一个中间的过渡状态。

And now today with tools like Mathematica, you can just keep going and so that's an object now, that's a thing that you can work with and there are examples where people work with these things that formerly were regarded as too complicated and sometimes they get simple answers out of the end, that's just an intermediate working state.

所以我不禁想，在这种情况下是否也会发生类似的事情：你可以把这些模型以某种类似人们使用Mathematica的方式加以利用，认真对待它们——虽然它们并非传统意义上的解释，但可以对它们进行一些有趣的操作。

And so I sort of wonder if something similar is going to happen in this particular case where you can take these models and sort of just use them in a little bit the same way people do with Mathematica and take them seriously as they're not explanations in the classic sense but there'll be something else which interesting operations can be done on.

我担心的是，假设你有1600个数据点，或者1500个，而你是在一个我们先发展出深度学习、后才出现天文学的奇怪历史背景下训练模型。

The thing I worry about is suppose that you it's 1,600 and you're trained or 1,500 and you're training a model on this is a weird history where we developed deep learning before we had before we had cosmology.

但假设我们生活在这种世界里，你观察到恒星似乎不动，而行星却表现出各种奇怪的行为。

But, so suppose we live live in that world and you're observing how there's the stars, they don't seem to move, the planets have all these weird behaviors.

然后你用这些数据训练一个模型，再进行某种解释分析，试图弄清楚：我们在这里观察到了哪些模式？

And then you train a model on that and then you do some kind of interp on it and trying to figure out, well, what are the patterns we see here?

你所看到的，只是不断在托勒密模型的基础上继续叠加。

What you'd see are just these you just keep be able to keep building on Ptolemy's model.

你会看到：哦，还有我们之前没注意到的更多本轮。

You'd see like, oh, there's more epicycles we didn't notice.

又来了另一个本轮。

There there's another epicycle.

参数无论是什么，都编码了本轮；参数无论是什么，又编码了下一个本轮。

It's the how parameters whatever to whatever encode epicycle this, parameters whatever encode the next epicycle.

所以，如果你仅凭观测数据试图弄清楚太阳系为何如此运行，你可能会不断叠加本轮，但真正整合这一切并指出整体上更合理的解释，却需要一个天才的头脑。

So if you were just trying to figure out why is the solar system the way it is from observational data, you could just keep adding epicycles upon epicycles, but it really took one mind to integrate it all in and say, here's what makes more sense overall.

我的观点是，我们其实还不清楚该如何处理这些模型，我们甚至还没有明确的‘动词’来描述它们。

So I mean, there, this is sort of to my point that we don't really understand what to do with the models, we don't have the verbs necessarily yet.

但确实很有趣去思考这样一个问题：你何时开始对模型施加约束，本质上是在问，最简单的解释是什么？你能简化它吗？你能给我一个九成的解释吗？

But it is certainly interesting to think about the question, where you start to apply constraints to the models, sort of essentially saying what's the simplest possible explanation or can you simplify, can you give me sort of the ninetyten explanation?

你能不断深入，将它不断精简吗？

Can you go further and further and further in boiling it down?

也许它们最初确实会提供一个极其复杂、参数众多的模型，但你可以强制简化，这就像脚手架，或许只是它们试图理解某事物的早期阶段，必须经过这个复杂过程才能达到更简单的理解。

So it might be that indeed they sort of start out by providing a very, very complicated many, many, many parameter model but you can just force the case and basically that's scaffolding, which maybe is sort of the very early days of their attempt to understand something, they're forced through that to much more simple understanding.

抱歉，我刚才理解错了，但听起来你的意思是，也许存在某种正则化机制？

So sorry for misunderstanding, but it sounds like you're saying maybe there's some sort of regularizer

是的。

Yeah.

没错。

Exactly.

或者说是知识蒸馏，没错。

Or distillation Exactly.

你可以通过一个非常复杂的模型，提炼出更真实、更简洁的理论。

You could do of a very complicated model that gets to a truer, more parsimonious theory.

但比如说，托勒密和哥白尼的对比。

But, yeah, just take, Ptolemy versus Copernicus.

对吧？

Right?

你一开始会用大量托勒密的本轮，然后尝试对这个模型进行提炼。

So you start off with lots of Ptolemaic epicycles, and then you try to distill this model.

也许在这个过程中，一些对轨道均方误差匹配作用越来越小的本轮就被去除了。

And maybe it gets rid of some of the epicycles that were are less and less sort of necessary to get the mean squared error of the orbits to match.

但在某个时刻，必须做这样一件事，就是彻底转变两个东西。

But at some point, has to do this thing, which is like switch two things.

是的。

Yeah.

对。

Yeah.

对。

Yeah.

而在局部范围内，它实际上并没有让事情更准确。

And it locally, it actually doesn't make things more accurate.

对。

Yeah.

对。

Yeah.

对。

Yeah.

它是在全局意义上才是一种更进步的理论。

It's sort of in a global sense that it's it's a more progressive theory.

对。

Yeah.

是的。

Yeah.

而且有一个过程，显然是人类在Spanwick上完成的，那就是正则化或进行了这种替换。

And there's some process which obviously humanity did over Spanwick did that regularization or did that swap.

但如果只是原始的梯度下降，我并不觉得它会做到这一点。

But if raw gradient descent, it seems like I don't really feel like it would do that.

我的意思是，想想从牛顿引力到爱因斯坦广义相对论的例子，这两种理论截然不同，问题是，是什么导致了这种转变？据我所了解的历史，情况是爱因斯坦首先发展了狭义相对论，并且几乎立刻就明白了——这其实是一个非常明显的观察。

I mean, think about the example of going from Newtonian gravity to Einstein's general theory of relativity and these are shockingly different theories and the question is like what causes that flip and as nearly as I understand the history what goes on is Einstein develops special relativity and pretty much straight away he understands, I mean it's a very obvious observation.

在狭义相对论中，影响的传播速度不能超过光速；而在牛顿引力中，作用是超距的。事实上，在狭义相对论成立的前提下，使用牛顿引力会导致超光速信号传递，你甚至可以向过去发送信息，做各种疯狂的事情。因此，意识到这里存在一个大问题并不难。

In special relativity, influences can't propagate faster than the speed of light and in Newtonian gravity, action is at a distance, in fact it's straight away in special relativity could use Newtonian gravity to do faster than light signaling, you could send information backwards in time, you could do all kinds of crazy stuff And so it's not a big leap to realize oh we have a big problem here.

因此，这就是推动这一转变的关键动力。

And so that's the forcing function there.

你意识到旧的解释已经不够用了，你需要一个新的理论，于是你从最简单的东西开始尝试，但结果发现很多简单方法都效果不佳，于是你被迫一步步推进——事实上，这很有趣，爱因斯坦是被逼着经历这些步骤的：理论逐渐变得复杂，过程中出现了各种错误，而最终的理论却显得异常简单而优美，但它经历了相当粗糙的中间阶段。

You've realized that your old explanation is not sufficient, you need something new and then you're going to start by doing the simplest possible stuff and it just turns out that a lot of that stuff doesn't work very well and so you're sort of forced, in fact it is interesting, he is sort of forced to go through these steps of gradually it gets quite more complicated and it's sort of wrong in a variety of ways and the final theory appears really shockingly simple and beautiful but it's gone through somewhat ugly intermediate stages.

所以，如果你在思考AI如何加速科学，那么在一些领域，比如我们只需要局部解决方案的领域（比如蛋白质如何折叠），情况可能就是这样。

So if you're thinking about what does it look like to have AI accelerate science, there's one for maybe well understood domains where we just want local solutions like how does this protein fold.

我们只是用梯度下降法训练一个原始模型。

We just train a raw model using gradient descent.

然后还有一些像广义相对论这样的情况，你不可能通过训练宇宙中每一个观测数据就指望广义相对论自己冒出来。

Then there's things like coming up with general relativity where you couldn't really just train on every single observation in the universe and hope that general relativity pops out.

那么这需要什么条件呢？

And so what would it require?

当然，它也不是立刻就被发现的。

Well, it also certainly wasn't immediately discovered.

对吧？

Right?

所以这经历了数十年的思考。

So it was a lot of decades of thought.

我想你需要独立的研究项目，人们从这些偏见出发，比如爱因斯坦最初只是被这样一个思想实验所驱动：你能区分引力效应和仅仅向上加速的效果吗？

And I guess you need independent research programs where people start off with these biases, where Einstein is just initially motivated by this thought experiment of, you know, can you distinguish the effect of gravity from just being accelerated upwards?

嗯。

Mhmm.

然后你需要不同的AI研究者带着这些初始偏见开始探索，看看能从中孕育出什么。

And then you just need different AI thinkers to start off with these initial biases and see what can germinate out of them.

而这一过程的验证周期可能非常漫长，你只需要

And then the verification loop for that might be quite long, you just

保持

need to

同时维持所有这些研究项目的生命力。

keep all those research programs alive at the same time.

是的，我的意思是，你提到的保持所有不同研究项目持续发展的观点，我认为非常重要，而且可以说是核心所在。

Yeah, I mean, I think there's like, this point that you make about sort of keeping all the different research programs alive, that I think is very important and somehow central.

一个很好的例子是，在某些情况下正确的答案，在其他情况下却是错误的。

A great example is situations where the same answer has been correct in some circumstances and wrong in other circumstances.

因此，天王星的位置并不完全准确，人们据此著名地预测了海王星的存在。

So the planet Uranus was not in quite the right spot and people very famously predicted the existence of Neptune on this basis.

这是牛顿引力理论的一次巨大成功。

Wonderful massive success for Newtonian gravity.

水星的位置也不太对。

The planet Mercury is not in quite the right spot.

你预测存在另一颗干扰性的行星。

You predict the existence of some other distorting planet.

结果发现那颗行星并不存在。

Turns out that doesn't exist.

实际上，水星位置不对的原因是你需要广义相对论。

Actually the reason Mercury is not in the right spot is because you need general relativity.

所以你追求了非常相似的想法，但在一种情况下非常成功，在另一种情况下却完全失败。

And so you've pursued very similar ideas and it's been very successful in one case and it's been completely and utterly unsuccessful in the other case.

我认为，先验地看，你无法判断哪种做法是正确的，实际上你需要同时进行这两种探索。

And I think a priori, you can't tell which of these is the thing to do and you actually need to do both.

因此，科学史上的确如此，这种多样性——让很多人去探索大量潜在有前景的想法——你需要长期支持这种做法。

And so this is certainly very true in the history of science that this kind of diversity where you just have lots of people go off and pursue lots of potentially promising ideas, you just need to support that for a long time.

由于各种原因，这样做很难，但它似乎确实极其重要。

It's hard to do that for a variety of reasons, but it does seem to be very very very important.

所以，天王星和水星的这个例子非常有趣。

So this example of Uranus versus Mercury is very interesting.

是的。

Yeah.

在一种情况下，我认为这说明了证伪主义的困难。

In one, I think it illustrates sort of the difficulty of falsificationism.

对。

Yeah.

天王星的轨道在某种意义上否定了牛顿力学，但接着你做出了一些辅助性预测，说这是因为存在另一颗行星在扰动它的轨道，我想这是1846年勒维耶提出的。

Like, the orbit of Uranus is in some sense falsifying Newtonian mechanics, but then you say you make some ancillary prediction that says, oh, the reason this is happening is there must be another planet which is effective perturbing, your nearest orbit, and you I think it's La Verrier in 1846.

把望远镜对准正确方向，你就能发现天王星。

Point a telescope in the right direction, you find Uranus.

海王星。

Neptune.

哦，是吗？

Oh, is there?

海王星。

Neptune.

是的。

Yes.

但对于水星，确实观测到它的轨道椭圆每世纪比牛顿力学预测的多旋转了四十三角秒。

But with Mercury, yeah, it's observed that it's the ellipse which forms its orbit is rotating forty three arc seconds more

对。

Yeah.

每世纪比牛顿力学预测的多旋转了四十三角秒。

Every century than Newtonian mechanics would imply.

所以人们认为水星轨道内一定还有一颗行星。

So people say that there must be a planet inside Mercury's orbit.

他们称之为祝融星。

They call it Vulcan.

然后用望远镜观测，却根本不存在。

And point telescopes, it's not there.

但如果你是一个正统的牛顿主义者，你会说，也许有一些宇宙尘埃遮挡了这颗行星。

But if you're a proper Newtonian, what you do is say, well, maybe there's some cosmic dust that's occluding this planet.

或者这颗行星太小了，我们看不到它。

Or maybe the planet is so small we can't see it.

或者也许需要更强大的望远镜，或者存在某种磁场干扰了我们的测量。

Or maybe there's even more powerful telescope or maybe there's some magnetic field which is sort of occluding our measurements.

这种情况一再发生，对吧？

And this happens over and over right?

你知道，有太多故事都与此如出一辙。

You know there's just so many stories which are exactly like this.

我的意思是，上世纪90年代，有人注意到先驱号探测器的位置与预期略有偏差，于是你可能会非常兴奋，天啊，广义相对论错了，也许我们即将发现下一代引力理论。

I mean an example I love from in the 1990s some people noticed that the Pioneer spacecraft weren't quite where they were supposed to be and so you can get very excited about this, oh my goodness, general relativity is wrong, maybe we're going to discover the next theory of gravity.

如今，被广泛接受的解释是：不，实际上只是探测器存在轻微的不对称性，结果发现热辐射在一个方向上略强于另一个方向，从而产生了一个微小的朝向太阳的加速度。

Today, the accepted explanation is that no, actually there's just a slight asymmetry in the spacecraft, it turns out that the thermal radiation is slightly larger in one direction than the other and that's causing a tiny little acceleration towards the sun.

而大多数时候，当出现这些看似异常的现象时，原因往往就是这样。

And most of the time when there's these apparent exceptions it's just something like that's going on.

这非常类似于水星沃坎案例，但偶尔又不是，而先验上你无法区分这些情况。

It's very much like the Mercury Vulcan case but every once in a while it's not and a priori you can't distinguish these.

科学中充满了这样的例子。

Science is just full of these.

有趣的是，我们讲述科学史的方式听起来如此简单，好像只要关注正确的异常现象，就会意识到必须抛弃旧理论，然后诺贝尔奖就在等着你。

It's funny too, the way we tell the history of science, it sounds so simple, like oh you just focus on the right exception and you realize that you need to throw out the old theory and lo and behold your Nobel Prize awaits.

但实际上，这些异常现象随处可见，99.9%的情况下，结果都只是类似这种热加速效应——就像先驱者号探测器的情况一样。

But in fact these exceptions are all over the place and 99.9% of the time it just turns out to be some effect like this thermal acceleration in the case of the Pioneer spacecraft.

不幸的是，这其中存在大量的选择偏差，而且进入这些

Unfortunately there's a lot of selection bias going And into those

关键是，没有任何一种启发式方法能告诉你你处于哪种情况。

the thing is there's no ex anti heuristic which tells you which case you're in.

我之所以认为这很重要，是因为有些人认为人工智能会在科学领域取得不成比例的进展，因为它在那些有紧密验证循环的领域中取得了不成比例的进展。

And just to spell out why I think this is important is because some people have this idea that AI is going to make disproportionate progress towards science, because it makes disproportionate progress towards domains where there's tight verification loops.

因此，它在编程方面非常出色，因为你能运行单元测试，而科学可能也类似，因为你能做实验。

And so it's really good at coding because you can run unit And science may be similar because you can run experiments.

我认为这一点没有被充分认识到：首先，实验实际上并不能排除无限多种与任何给定实验结果相容的理论。

I think what that doesn't appreciate, one, is that experiments actually don't there's an infinite number of theories that are compatible with any given experiment.

随着时间推移，我们会逐渐接受——至少事后看来——我们认为更正确的一种，正如我们在这次对话中讨论的那样，这其实很难清晰表达。

And over time, we glob onto the, at least in retrospect we think is a more correct one is, as we're discussing in this conversation, sort of hard to articulate.

拉克特在书中确实提到了许多关于这类长期存在的敌对验证循环的有趣例子。

Lactate does actually has all kinds of interesting examples in the book about these kinds of hostile verification loops that are extremely long lasting.

他提到的一个例子是普劳特或普劳特。

So one he talks about is Prout or Prout.

我不知道该怎么发音，但这是1815年一位化学家的名字。

I don't know how to pronounce it, but there's this chemist in 1815.

他假设所有原子核的质量都必须是整数，并且基本上都是由氢构成的。

He hypothesizes that all atomic nuclei must have whole number weights, and they're basically all made of hydrogen.

他之所以这么认为，是因为如果你观察所有元素的测量值，确实看起来几乎所有的元素质量都恰好是整数。

And it's the reason he thinks this is because if you look at the measure rates of all elements, it does seem that the almost all of them do happen to have whole number weights.

但随后出现了一些例外情况。

But then there's some exceptions.

比如，氯的原子量是35.5。

Like, for example, chlorine comes out at 35.5.

于是，这个学派的人不断提出各种临时性的理论，比如可能是化学杂质造成的。

And so then there's all these ad hoc theories that people in this school keep coming up with like, oh, maybe there's chemical impurities.

但你无论做什么样的化学反应，似乎都无法消除这种现象，也许它本身就是分数形式，所以是35.5。

But then there's no chemical reaction you can do which seems to get rid of Maybe it's fractions of whole numbers, so it's 35.5.

它可以是二分之一。

It can be halves.

但如果你更精确地测量氯，它的数值其实是35.46。

But, actually, if you measure chlorine even closer, it's 35.46.

所以它实际上离正确的修正值越来越远了。

So it's actually getting further away from

是的。

Mhmm.

正确的修正。

The correct correction.

后来发现，你实际测量的是不同的同位素，这些同位素在化学上无法区分。

And later on, what is discovered is what you're actually measuring is different isotopes, which cannot be chemically distinguished.

它们只能通过物理方法区分。

They can only be physically distinguished.

嗯。

Mhmm.

但直到八十五年后，我们才真正理解什么是同位素，是的。

But so then you just have eighty five years before we realize what an isotope is where Yeah.

但验证过程实际上对你、对正确理论是敌对的，你只能依靠这个残存的信念来坚持。

The verification loop is actually actively hostile against you, against the correct theory, and you just need this remnant to be defending.

没有理性上的依据表明它是一个更优的理论。

There's no extent to reason it's a preferred theory.

作为整个科学共同体，我们应当鼓励人们去捍卫并尝试整合新发现，即使这些发现似乎与他们所信奉的学派观点不符。

Just as a community, we should just have people defend try to integrate new observations even if they don't fit seem to fit their school of thought with what they believe.

而且，希望如果这样的情况足够多的话。

And, hopefully, if if that enough of that happens.

总之，我想表达的是自动化科学所面临的困难。

Anyways, yeah, guess the thing I'm trying to articulate is the difficulty with automating science.

是的，问题在于瓶颈到底在哪里，某种程度上，我们是主要受限于某一种因素，还是多种不同的因素？

Yeah, mean the question is where is the bottleneck at some level and sort of, you know, are we primarily bottlenecked on one thing or one type of thing or we bottlenecked on sort of multiple types of things.

当然，和结构生物学领域的人交流时，他们认为AlphaFold是一个巨大的突破，令人震惊，所以在某种程度上，AI确实似乎能够帮助我们加快科学进程。

Certainly talking to structural biology people, they seem to think that AlphaFold was an enormous advance, it was a shock, so at some level, yes, AI can, it seems certain can help us speed up science.

因此，它确实在解决某种类型的瓶颈方面起到了作用。

So it is helping with a certain type of bottleneck.

但这并不意味着，如你所说，它一定能帮助解决所有类型的瓶颈，我想你所指出的问题是：哪些类型的瓶颈仍然存在，以及我们突破它们的前景如何。

That doesn't mean though, as you're saying, that it's necessarily going to help with all kinds of bottlenecks and I suppose the question you're pointing out is what are the types of bottlenecks that remain and what are the prospects for getting past them.

我认为，即使在编程领域，和程序员朋友们交谈时也十分有趣，他们目前都处于震惊和高度兴奋的状态，和他们交流时感觉他们到处都在谈论这个话题。

I think even in the case of coding, it's really interesting talking to programmer friends, at the moment they're all in this state of shock and high excitement and they're all over the place actually kind of talking to them.

你不禁会想，瓶颈接下来会转移到哪里？

You do wonder where is the bottleneck going to move to?

因此，显然很多程序员目前的瓶颈在于缺乏有趣的想法，尤其是缺乏有创意的设计思路。

So certainly one thing that a lot of them seem to be bottlenecked on is now having interesting ideas and in particular having interesting design ideas.

因此，目前并没有一个有效的验证机制来判断某个设计想法是否非常出色。他们不再被编写代码的能力所限制，但仍被这一其他问题所制约。

So there's not really a verification loop for knowing oh that design idea is very interesting So they are no longer nearly as bottlenecked by their ability to produce code but they are still bottlenecked by this other thing.

过去一直如此，从形式上讲，他们并不受此限制，因为写代码本身就要耗费大量时间，他们本可以有很多想法，只是在实施原型时要花上三周，然后才能着手下一个版本。

Always were, formally they weren't bottlenecked on it because just writing code took so much of their time, could they sort of have lots of ideas, while they were, you know, they'd take three weeks to implement their prototype and then they would implement the next version.

你知道，现在他们只需三个小时就能实现原型，但之后从设计角度看，却想不出那么好的点子了。

You know, they're taking three hours to implement the prototype and they don't have, you know, as good ideas, sort of after that from a design point of view.

去年，我预测到2028年，AI就能像一位称职的总经理那样帮我报税。

Last year, I predicted that by 2028, AI would be able to prep my taxes about as well as a competent general manager.

但我们现在已经非常接近这个目标了。

But we're already getting pretty close.

正如我之前提到的，我同时用Mercury处理我的业务和个人银行业务。

As I shared before, I use Mercury both for my business and my personal banking.

因此，我最近通过Mercury的MCP，让一个大语言模型访问了我在两个账户中的交易历史。

So I recently gave an LLM access to my transaction history across both accounts through Mercury's MCP.

我让它检查我2025年所有的交易记录，标记出那些看起来本应计入公司支出的个人开销。

I asked it to go through all my twenty twenty five transactions and flag any personal expenses that seem like they should actually be charged to the business.

这效果出奇地好。

And this worked shockingly well.

Mercury 的 MCP 暴露了大量详细信息，比如备注、备忘录以及收据的 JPEG 图片和 PDF 附件。

Mercury's MCP exposes a bunch of detailed information things like notes and memos and any JPEGs of receipts and PDF attachments.

所以我的 LLM 有足够的信息可以处理。

So my LLM had plenty of contacts to work with.

我最喜欢的一个例子是有一笔消费，如果你只看商家名称，可能会误以为是个人支出。

Of my favorite examples happened with a charge to If you looked at the vendor alone, you would have had to assume that it's a personal expense.

但 LLM 查看了收据和 Mercury 中附带的备注，发现这实际上是上一次线下团建活动的团队建设支出。

But the LLM looked at the receipt and the attached note in Mercury and realized this was actually a team bonding exercise from our last in person retreat.

所以这是一笔正当的业务支出。

So a legitimate business expense.

我估计传统银行要过很久才会推出 MCP。

I imagine it'll be a while before traditional banks have MCP.

正是这类功能让我选择使用 Mercury。

Functionality like this is why I use Mercury.

前往 mercury.com 了解更多信息。

Go to mercury.com to learn more.

Mercury 是一家金融科技公司，而非受 FDIC 保险的银行。

Mercury is a fintech company, not an FDIC insured bank.

银行服务由 Choice Financial Group 和 Column NA 提供，均为 FDIC 成员。

Banking services provided through Choice Financial Group and Column NA, Members FDIC.

你的观点非常有趣。

You have a very interesting take.

我想那是在一个脚注里，我记得你提到过你的文章，但我又找不到了，大意是：如果我们遇到外星人，他们很可能拥有与我们完全不同的技术栈。

I think it was a footnote where I know where your essays and I couldn't find it again, which was that it's very possible that if we met aliens that they would have a totally different technological stack than us.

这与我曾经持有但从未质疑过的一种普遍误解相矛盾，那就是：科学是文明早期就发展起来的一种活动，你达到某个阶段后，只需花上几百年时间系统地掌握基础原理，理解宇宙如何运作等等，然后你就掌握了科学。

And that contradicts, I guess, a common subception I had that I never questioned, is that science is this thing you do very relatively early on in the history of civilization where you get to a point and you have a couple hundred years of just cranking through the basics, understanding how the universe works, etcetera, and you've got it.

你已经掌握了科学。

You've got science.

然后基本上所有人都会趋同于同一种所谓的科学。

And then basically everybody would converge on the same quote unquote science.

我觉得这个想法非常有趣，我想让你多谈谈这个观点。

And so I found that a very interesting idea and I want you to say more about it.

是的，我的意思是，至少我比较认同的一个观点是，技术树或科学与技术的发展路径可能比我们意识到的要广阔得多。

Yeah, I mean, I think probably the idea there that I'm at least somewhat attached to is the idea that sort of the tech tree or the science and tech tree is probably much larger than we realize.

我的意思是，我们目前处于一种很奇特的境地。

I mean we're sort of in this funny situation.

人们有时会把‘万有理论’作为物理学的终极目标，然后默认一旦达成这个目标，物理学就完成了。

People will sometimes talk about a theory of everything as a potential goal for physics and then there is this presumption somehow that physics is done once you get there.

但当然，这完全不是真的。

And of course this is not true at all.

如果你想想计算机科学，它其实早在20世纪30年代，当图灵、丘奇等人奠定了计算理论的基础时，就已经有了它的‘万有理论’。

If you think about computer science, computer science basically got started in the 1930s when Turing and Church and so on just laid down what the theory of everything was.

他们只是阐明了计算是如何运作的，而此后近九十年，我们一直在探索这些理论的后果，并逐步构建出越来越多有趣的想法。

They just said here's how computation works and then we've spent ninety odd years since then just exploring consequences of that and gradually building up more and more interesting ideas.

这些想法在某种程度上可以被视为技术，但在某种程度上，由于它们是计算理论内部被发现的基本原理，我认为它们更应被视为科学，甚至在某些情况下是极为基础的科学。

And those ideas are, to some extent, can just regard as technology but to some extent insofar as they're sort of discovered principles inside that theory of computation, I think they're best regarded as science and in some cases very fundamental science.

像公钥加密这样的想法，其深度非同寻常，非常不直观，但从某种意义上说，这些想法早在1930年代就已经潜藏在那里了。

Ideas like public key cryptography are, I mean they're just incredibly deep, very non obvious ideas which in some sense lay hidden already sort of in the 1930s.

因此，我预计人们会以不同的方式探索这个技术树，而我们目前还处于相对初级的阶段。

And so my expectation is that there will be different ways of exploring this tech tree and we're still relatively low down.

我们仍处于刚刚理解这些基本理论的阶段，尚未充分探索它们。

We're still at the point where we're just understanding these basic fundamental theories and we haven't yet explored them.

我觉得挺有趣的一点是，如果你看看物质的相态。

A thing which I think is quite fun is if you look at just the phases of matter.

我上学的时候，学到的是物质有三种相态，有时是四种或五种，具体取决于你包含哪些。

When I was in school, we'd get taught that there are three phases of matter or sometimes four phases of matter or five phases of matter depending a little bit on what you included.

但作为成年人、作为物理学家，你开始意识到：哦，我们一直在扩充这个列表。

And then as an adult, as a physicist, you start to realize, oh, we've been adding to this list.

我们有了超导体、超流体，可能还有不同类型的超导体、玻色-爱因斯坦凝聚态、量子霍尔系统、分数量子霍尔系统，结果发现，实际上还有大量物质相态有待发现，而且我们会发现更多。

We've got superconductors and superfluids and maybe different types of superconductors and Bose Einstein condensates and the quantum Hall systems and fractional quantum Hall systems and it's starting to turn out it looks like actually there's a lot of phases of matter to discover and we're going to discover a lot more of them.

事实上，我们将来甚至能在某种程度上设计这些相态。

And in fact, we're going be able to start to design them in some sense.

我们仍然受物理定律的约束，但其中蕴含着巨大的自由度。

We'll still be subject to the laws of physics, but there is this tremendous freedom in there.

在我看来，这就像我们还处在技术树的最底层。

And this looks to me like, oh, we're down at sort of the bottom of the tech tree.

我们才刚刚起步，我预计这种情况在整体上也是如此。

We've barely gotten started there and I expect that to be the case sort of broadly.

当然，我认为编程是一个非常自然的观察领域。

Certainly in terms of, I think programming is a very natural place to look.

认为我们已经发现了编程中所有深刻的思想，这在我看来显然荒谬至极。

The idea that we've discovered all the deep ideas in programming just seems to me obviously ludicrous.

我们不断发现看似深刻而全新的基本思想，而我们的能力却非常有限。

We keep discovering what seems like deep new fundamental ideas and I mean we're very limited.

我们本质上只是稍微进化了一点的黑猩猩，所以进展缓慢，需要时间。

We're basically slightly jumped up chimpanzees so we're slow and it's taking us time.

但再过一百万年，从人类在如何操控计算机、操控信息方面所产生的一切想法来看，我们可能会发现，其实还有大量极其深刻的思想尚未被发现。

But what do we look like sort of another million years in the future in terms of all of the different ideas which people have had around how to manipulate computers, how to manipulate information, I think we're likely to discover that actually there are a lot of very deep ideas still to be discovered.

有句话说得很好，我忘了是谁说的了，我想想，

It's a nice, who was it, I think

是——

it was

高德纳在《计算机程序设计艺术》的前言里提到过类似的事：他早在60年代就开始撰写这本书，当时有一位数学家对计算机科学颇有些不屑，说计算机科学现在还算不上一门真正的学问。

Knuth in the preface to the art of computer programming, said something like he started this book back in the 60s and he talked to a mathematician who was a bit contemptuous and said, look, computer science isn't really a thing yet.

等这门领域里能拿出一千条深刻的定理再来找我谈吧。

Come back to me when there's a thousand deep theorems.

而克努思在几十年后撰写该书的序言时提到，如今显然已经诞生了上千条高深的定理。

And Knuth remarks, and he's writing this now decades later, the preface, there clearly are a thousand deep theorems now.

这就意味着，如果我们去畅想登上科技树更高层后的长远未来，比如我们要选择往哪个方向发展、要如何探索，会是一件十分有意思的事。我认为不同的文明或是不同的选择，很可能会让我们最终抵达科技树上完全不同的分支。

And that means, like it's really interesting to sort of think of it, like what's 's the long term future as you get higher and higher up in the tech tree, like choices about which direction we go and sort of how we choose to explore, I think it's potentially the case that different civilizations or different choices mean that we end up in different parts of that tree.

具体来说，就拿一些非常基础的事情举例：人类是高度依赖视觉的生物，而其他一些动物则更依赖听觉感知，这种差异会不会影响我们产生的思维类型？如果把这个逻辑延伸到更奇特的外星文明，也许它们感知世界、改造世界的天生偏向和我们截然不同，这就可能给它们探索科技树的方式带来重大改变。

And in particular just things, mean sort of very basic things about we're very visual creatures, certain other animals are much more orally based, does that bias the types of thoughts that you have and then you extend it to much more exotic kinds of civilizations where maybe just sort of their biases in terms of how they perceive and how they manipulate the world are maybe quite different than ours and that might make some significant changes in terms of how they do that exploration of the tech tree.

显然，这些都只是推测。

It's all speculation obviously.

这个观点太有趣了。

This is such an interesting take.

我想更好地理解它。

Want to better understand it.

理解这一点的一种方式是，有些东西可能非常基础，并且与现实有广泛的交集，因此它们迟早会被发现，比如广义相对论。

One way to understand it is that there might be some things which are so fundamental and have such a wide collision area against reality that they're inevitably going discover like general relativity.

数字，是的是的。

Numbers, yeah yeah.

在银河系所有的智慧生命中，也许这个数字只有一个，实际上，从某种意义上说，我们已经把这个数字提高了。

Of all of the intelligences in the Milky Way galaxy, maybe that number is one, actually arguably we've already increased the number.

但在所有这些智慧生命中，有多少比例发现了计数这个概念呢？

But of all of those, what fraction of the concept of counting?

这看起来确实非常自然。

And it does seem very natural.

有多少比例发现了类似十进制位值系统的概念呢？

What fraction have discovered the idea of some kind of decimal place system?

有趣的问题。

Interesting question.

也许我们忽略了某些非常简单而明显的东西，而那些东西实际上可能比这更好。

And maybe we're missing something really simple and obvious that's actually way better than that.

有多少人一开始就达到了这一点？

What fraction got there immediately?

有多少人必须经过某些中间阶段？

What fraction had to go through some other intermediate state?

有多少人使用了线性表示，而不是二维或三维表示？

What fraction used linear representations versus a two dimensional or three-dimensional representation.

我认为这些问题的答案根本不是显而易见的。

I think the answers to these questions are just not at all obvious.

这有大量设计上的自由度。

It's a lot of design freedom.

在理论计算机科学方面，这将会极其天真和傲慢。

On theoretical computer science, this is is gonna be extremely naive and arrogant.

但我选了斯科特·阿罗森的复杂性理论课，那可能是他教过最差的学生。

But I took Scott Aronson's, you know, class on complexity theory, and that was by far the worst student he's ever had.

但我记得的是，你曾是这一领域的先驱之一，当时我们弄清楚了量子计算机能解决的问题类别，以及它们与经典计算机能解决的问题之间的关系。

But I what I remember is, like, there was this period that you were one of the pioneers of where we figured out here's the class of problems that quantum computers can solve and how it relates to problems that a classical computer can solve.

这太突破性了，哦，原来这样行得通。

It's groundbreaking, oh, this works.

从那以后，就出现了一个被称为‘复杂性动物园’的网站，上面列出了所有复杂性类。

And then since then, it's been this literally, called Complexity Zoo, this website, which lists out here's all the complexity classes.

如果你有一个带有这种类型预言机的竞争性类，它在某种程度上等同于另一个类。

And if you have this competitive class with this kind of oracle, it's sort of equivalent to this other class.

这感觉就像我们在构建这种分类体系。

And that it feels like we're building out that taxonomy.

是的。

Yeah.

所以，理解你所说的话有几种方式。

And so there's a couple of ways to understand what you're saying.

第一，也许你只是不同意我的观点，认为这个领域并没有发生这样的变化。

One, maybe you just disagree with me that this is actually what's happened with this field.

另一种可能是，尽管这种事可能发生在任何一个领域，但在1880年，有多少人会认为计算机科学——除了巴贝奇之类的人——真的会成为一个学科呢？

Another is that while that might happen to any one field, the amount of fields who would have thought in 1880 that computer science, other than Babbage or something, that computer science was going to be a thing in the first place?

所以，就领域数量而言，我们低估了未来还可能涌现多少新领域。

So, amount of field, we're underestimating how many more fields there could be.

是的，对的，

Yeah, yeah,

当然。

for sure.

或者你可能觉得两者都有，也可能还有第三种你没说出来的想法，但我很好奇。

Or maybe you think both or maybe a third secret thing, but I'd be curious.

这里一个非常常见的论点是所谓的‘低垂的果实’论点，认为回报应该会递减。

I mean, a very common argument here is sort of the low hanging fruit argument, the argument that says there should be diminishing returns.

而且在

And in

事实上，我们从经验上看到这一点，对吧？

fact empirically we see this, right?

世界上科学家的数量呈指数级增长。

The amount of scientists in the world has just exponentially increased.

我认为值得思考的是：你为什么预期回报会递减？这个论点在实践中究竟有多适用？

And I think it's worth thinking about like why do you expect diminishing returns and how well does that argument actually apply, in practice.

我喜欢的一个类比是，想象参加某个活动，比如婚礼之类的，然后你去甜点自助区，那里摆了30种甜点。

An analogy I like, is actually thinking about sort of going to some event, going to a wedding or whatever and you go to the dessert buffet and they've put out 30 desserts.

当然，人们自然会先拿最好的甜点。

And of course naturally what people do, the best desserts go first.

我的意思是，我们并没有一个明确的偏好排序，但人类的喜好相当接近，所以最好的甜点会被优先拿走。

I mean we don't quite have a well ordered preference there so maybe there is some difference human beings are fairly similar so the best desserts will go first.

这就是为什么在许多不同领域中，你预期会出现回报递减的原因：如果什么是可以获得的相对容易判断，而且人们的偏好相似，那么最好的东西就会先被拿走，之后就越来越差。

And this is an argument for why you expect diminishing returns in a lot of different fields if it's relatively easy to see what's available and people have similar preferences then the best stuff goes first and it just gets worse and worse after that.

从一个静态的、即时的视角来看科学进步，也许这个说法有一定道理。

A very static snapshot in time of scientific progress, Maybe there's some truth to that.

但如果有人坐在甜点桌后面，不断补充和补货，可能会在稍后出现更好吃的甜点，于是人们就会去吃那些新的甜点。

But if somebody is sitting behind the dessert table and is replenishing, restocking the desserts and keeps adding new ones in, it may turn out that a little bit later, much better desserts appear so you're going to go and eat those instead.

科学进步带有一点这种味道。

And scientific progress has a little bit of that flavor.

我们会经历一些奇特的时期，计算机科学就是一个很好的例子，它本质上是数学哲学和逻辑中一些非常抽象问题的副产品。

We go through these sort of funny time periods, computer science is a great example, where computer science basically arose as sort of a side effect of some pretty abstruse questions in the philosophy of mathematics and logic.

于是，这些人在试图解决一些相当深奥的问题——这些问题在某种意义上属于高度探索性的、极其深奥的领域——却意外发现了一个全新的基础领域，随即爆发式发展。

And so you've got these people trying to attack these rather esoteric questions that seem quite high up in some sense in sort of exploration, quite esoteric, and they discover this fundamental new field and all of a sudden there's an explosion there.

在这种情况下，边际收益递减的论点根本不起作用。

Sort of the diminishing returns argument just didn't apply there.

我们只是没能看到那里有什么，而这种情况一再发生。

We just weren't able to see what was there and this has been the case over and over and over again.

新领域不断涌现，突然之间，取得进展变得容易了，年轻人蜂拥而入，因为21岁就能做出重大突破，而无需花25年掌握所有已有的知识——这显然极具吸引力。

New fields arrive and all of a sudden boom, it's actually easy to make progress again, young people flood in because you can be 21 and make major breakthroughs rather than having to spend 25 mastering everything that's been done before, it's obviously very attractive.

我不太明白，也不确定是否有人真正理解这种动态机制，比如为什么知识的结构会是这样，为什么新领域会不断涌现，但至少从经验上看，这确实是事实。

And I don't understand, I'm not sure anybody understands very well sort of the dynamics of that, like how to think about why the structure of knowledge is that way, that these new fields keep opening up, but it does seem empirically at least to be the case.

尽管如此，以深度学习为例，这显然是一个年轻人能够取得进展的新领域，大约十五年前，它重新开始快速发展。

Despite the fact that that is the case, take deep learning, Obviously this is an example of a new field where the 21 year olds can make progress and it's relatively new fifteen years or so when it sort of gets back into high gear.

但我们现在已进入一个阶段，需要数十亿、数百亿甚至上千亿美元的资金才能在前沿持续取得进展。

But already we're in a stage where you need billions or tens of billions or hundreds of billions of dollars to keep making progress at the frontier.

对此有几种理解方式。

And there's a couple of ways to understand that.

一种是，这实际上比古人所做的事情更困难，或者至少需要更多的资源投入。

One is that it actually is harder than the kinds of things the ancients had to do or requires more is more intensive at least.

另一种是，它本身可能并不更难，但由于我们的文明资源极其庞大，人口众多，资金充足，我们可以几乎立即实现古人需要花费数百年才能完成的进展。

Second is it might not have been, but because our civilizational resources are so large, the amount of people is so large, the amount of money is so large, that we can basically make the kind of progress it would have taken the ancients forever to make almost immediately.

我们一旦发现某件事有成效，就会立即投入所有资源。

We notice something is productive, immediately dump in all the resources.

但奇怪的是，这样的人并不多。

But it's also weird that there's not

这样的人没那么多。

that many of them.

我觉得深度学习之所以突出，是因为它是少数几个难以想到其他类似例子的领域之一

I feel like deep learning is notable because it is one big exception to the fact that it's hard to think of other I

我认为这是注意力机制架构的结果，对吧？

think that's a consequence of sort of the architecture of attention, right?

在任何给定时刻，总有一种最成功的技术。

Like at any given time there's always sort of a most successful thing.

如果深度学习不存在，也许你会谈论CRISPR，也许你会谈论别的东西；我们可能不会把蛋白质结构预测视为人工智能的成功，而可能会通过更广义的曲线拟合方法解决它，然后我们会说：哇，我们投入了大量计算资源，但蛋白质结构预测可能是一个极其重要的成就。

Maybe if deep learning wasn't a thing maybe you'd be talking about CRISPR, maybe you'd be talking about whatever it is, maybe we wouldn't think about solving sort of the protein structure prediction problem as a a success of AI, maybe we would have figured out how to doing it with sort of curve fitting like more broadly construed and we'd just be like, oh wow, we took a lot of computing resources but protein structure prediction might be an enormously important thing.

所以总有一个我们最关注的核心领域，我认为你指出的更多是注意力机制如何被集中化的结果。

So there is always sort of our biggest thing and I think what you're pointing out is more a consequence of the way in which attention gets centralized.

我其实想说，这本质上是一种潮流。

It's basically fashion is sort of what I'm saying.

这不仅仅是潮流，但其中确实存在某种动态机制。

It's not just fashion but there is some dynamic there.

这个观点有一个非常有趣且重要的含义：由于分支路径如此广泛、如此偶然且高度依赖路径，不同的文明会偶然发现完全不同的技术体系。

There's a very interesting and important implication of this idea that the branching is so wide and so contingent and so path dependent that different civilizations would stumble on entirely different technology sets.

是的

Yeah.

有一个非常

There's a very

一个有趣的含义，那就是将会产生贸易收益，是的。

interesting implication that there will will be gains from trade Yeah.

延伸到遥远的未来。

Into the far, far future.

是的。

Yeah.

这很有趣。

It's interesting.

这可能是关于遥远未来最重要的事实之一，涉及文明如何构建、如何协调、如何互动——并不是简单地去开拓和掠夺。

Which might actually be one of the most important facts about the far future in terms of how civilizations are set up, how they can coordinate, how they interface with like there's not this like go forth and exploit.

实际上，相邻殖民地之间或类似实体之间存在着巨大的贸易收益。

It's actually there are humongous gains to trade from adjacent colonies or whatever.

是的，有点吧。

Yeah, sort of.

真正困难的是什么，这是一个问题。

There's a question of what's actually hard.

如果只是想法的问题，那么这些想法传播得相对很快，分享想法是相对容易的。

If it's a question of if it's just the ideas, well, those spread relatively quickly, it's relatively easy to share ideas.

如果是更深层的东西，那就有点像王丹的那种观点，即实际上存在某种能力问题，你需要所有合适的技术、所有合适的制造能力等等。

If it's something more, it's almost sort of a Dan Wang kind of an idea where it's actually sort of, there's some notion of capacity, you need all the right techs, you need all of the right manufacturing capacity and so on.

所以，你知道，文明A的制造能力非常不同，即使文明B相对更先进，要在文明B中建造起来也并不容易，我认为这种情况确实存在。

And so, you know, Civilization A has very different kind of manufacturing capacity and it's just not so easy to build in civilization B even if civilization B is kind of ahead, I think that becomes true.

实际上存在比较优势，这确实非常值得重视，因为它将为双向贸易带来巨大的好处。

There is actually comparative advantage which is really worth, I mean it's going to provide massive benefits to trade in both directions.

最终，你将会预期一些创新的扩散。

Eventually you're going to expect some diffusion of innovation.

思考那里存在的障碍是什么，挺有趣的。

It is funny to think about what the barriers are there.

我喜欢思考的一个思想实验是，针对外星人版本的GitHub。

A thought experiment I like to think about is of GitHub but for aliens.

有人向你展示了某个外星文明的所有代码——我甚至不知道‘代码’在这里意味着什么，但大致上是他们的算法规范。

So somebody presents you with all of the code from some alien civilization and I don't even know what code means there but sort of their specification of algorithms.

其中会包含许多有趣的新想法，而人类要花上无穷无尽的时间去挖掘并试图提取所有这些内容。

It would have many interesting new ideas in there and it would take forever for human beings to dig through and to try and extract all of those.

其中一个原因，对我来说，最初是源于对自然界中蛋白质的思考。

One reason, I mean the origin of this for me was actually thinking about proteins in nature.

大自然赐予了我们无数奇妙的机器，而我们根本无法真正理解它们，只能逐个去尝试理解。

We've been gifted just this incredible variety of machines which we don't understand really at all and we just have to go and sort of try and understand them on a one by one basis.

我们至今仍在研究血红蛋白、胰岛素这类物质，毫无疑问，已知的蛋白质有数亿种之多。

We're still understanding hemoglobin and insulin and things like this and no doubt and there's hundreds of millions of proteins known.

所以，这有点像那种情况。

So it is a little bit like that.

生物学赐予了我们一个庞大的机器库，其中无疑蕴含着海量的精彩创意，而我们才刚刚开始触及它的皮毛。

We've been gifted by biology just this immense library of machines, no doubt containing an enormous number of very interesting ideas and we're just at the very, very, very beginning of understanding it.

所以实际上，这大概就是你的观点了，我需要稍微重新表述一下你的论点：你把这看作是来自外星文明的馈赠，虽然这显然不是，但你却这样看待它，心想天啊，这里面有这么多东西，我们要去研究它，天知道我们还要研究多久。

So actually that's I suppose kind of your point actually is I need to relabel your argument slightly but you sort of think of that as a gift from an alien civilization which obviously it isn't but you think of it that way and it's like oh my goodness there's so much in there and we're going to study it and goodness knows how long we could continue to study it.

关于血红蛋白之类的论文就有数以万计，我们仍然没有完全理解它们，但已经从中获得了这么多成果。

There's tens of thousands of papers about the hemoglobin and things like that and we still don't understand them and yet we're getting so much out of it.

想想胰岛素吧，它如此重要。

Think about insulin alone, it's such an important thing.

你在地球上拥有这种极其有用的直觉问题。

That's an incredibly useful intuition problem that you have on earth.

假设你有一个关于生命如何起源的理论，比如某种类似DNA的理论，经历了四十亿年，这时一个外星文明来到地球，对你说：这里有大量值得学习的东西，关于材料科学，关于任何你想到的领域，对吧？

Had I on where he had this theory about how life emerged, like, whatever theory you have, basically something like DNA, four billion years, and you have an alien civilization coming here and be like, there's all these interesting things to learn about material science, about you name it, right?

关于思考

About Think

想象一下胰岛素蛋白在行走，我们对这些蛋白质几乎一无所知，但仅有的那一点点已知事实就已经令人惊叹。

about Chenissen walking along, and we know almost nothing about these proteins and yet the tiny few facts we do know are just incredible.

核糖体，另一个例子，这个奇迹般的工程师般的装置，一个小工厂。

The ribosome, another example, this miraculous engineer, sort of device, little factory.

这一切都源于地球上的特定化学反应，即核酸和碳基生命形式，这种化学反应催生了所有这些有趣的事物，而外星文明会认为这些非常有趣。

And all seeded by just like there's this particular chemistry on earth, with nucleic acids and carbon based life forms that that chemistry gives rise to all of these interesting things which in the alien civilization would find very interesting.

因此，这种种子必然只是数万亿种可能的种子之一，我是说，仅仅是各种智力想法的种子，是的。

And so that that that very that seed which must be one among, you know, trillions of possible seeds of, I mean, just of general intellectual ideas Yeah.

至少这种丰富的多样性。

At least all this fecundity.

这为形成一种非常有趣的切入点。

That that's a very interesting intro to form.

我想深入思考一下这种互利交换的概念，因为我感觉这个想法中确实有些非常有趣的东西——如果你对技术如何进步、以及它在不同文明中可能如何不同有某种愿景，那么这对不同文明之间如何互动有着重要的影响。

I I wanna meditate on this gains for trade thing because I feel like I think there's something actually very interesting about this idea that if you have this vision of what techno how how technology progresses and how it may be different from in different civilizations, It has important implications about how different civilizations might interact with each other.

比如，将存在巨大的贸易收益。

Like the fact that there are going be these huge gains from trade.

这使得友好关系变得更加有价值。

It makes friendliness much more rewarding.

对。

Yes.

对吧？

Right?

是的。

Yeah.

这是一个非常重要的观察。

That's a very important observation.

对。

Yeah.

我根本没想过这一点。

I hadn't thought about that at all.

这是一个非常有趣的观察。

That is a very interesting observation.

很有趣，我的意思是，比较优势是人们喜欢引用的概念，这显然是一个非常美妙的想法。

It is funny, I mean, comparative advantage is something that people love to invoke and it's a very beautiful idea, obviously.

但它也有局限性。

There are limits to it.

这是一个特殊的限量型号。

It's a special limited model.

黑猩猩能做些有趣的事情。

Chimpanzees can do interesting things.

我们不和它们进行交易。

We don't trade with them.

我觉得思考一下背后的原因挺有意思的。

And I think it's sort of interesting to think about the reasons why.

部分原因只是权力，想想看。

Part of it is just power, think.

一旦权力差距足够大，通常——总是，但很常见的是，人群会转向另一种模式，即只想支配他人。

Once there's a sufficiently large power imbalance, very often, always, but very often groups of people seem to shift into this other mode where they just seek to dominate.

也许人类有一些特殊之处，但也许这也是一种更普遍的现象。

Maybe there's something special about human beings but maybe it's also sort of a more general sort of a thing.

它们不再如此，你需要所有这些特定条件成立，群体才会进行交易，而这并不一定是显而易见的。

They're no longer, they give up, you need all these special things to be true before groups will trade and it's not necessarily obvious.

我认为这里发生的大事是：交易成本和比较优势并不能说明，对于任何一个生产者来说，贸易的条件都能高于其生存水平。

I think the big thing going on here is one transaction costs and two comparative advantage does not tell you that the terms on which the trade happens are above subsistence for any given one producer.

因此，人们经常在这样的背景下提出：即使在后AGI世界中，人类也会被雇佣，因为具有巨大优势。

So people often bring this up in the context of, well, humans will be employed even in post AGI world because of a great advantage.

这个论点有五种不同的崩溃方式，但最容易理解的是：为什么我们不在路上到处看到马？毕竟汽车和马之间存在某种比较优势。

There's big there's there's like five different ways that argument breaks down, but the easiest ways to understand are why why don't we have horses all around on the roads because there's some competitive advantage between cars and horses.

很好的例子。

Good example.

同样地，建造同时兼容马和汽车的道路，其交易成本非常高。

Well, there's there's huge, transaction cost of building roads that are compatible with horses, and cars at the same time.

类似地，AI的思考速度是人类的千倍，它们能相互传递潜在状态，因此在供应链中保留人类参与的代价，远超过与人类互动所带来的收益。

In a similar way, AI is sort thinking at 1,000 times the speed and can sort of shoot their latent states again at each other are gonna find it way more costly than the benefit in just in terms of interacting with you to have a human being in the supply chain.

其次，仅仅因为存在数学上的比较优势，并不意味着值得花每年十万美元或类似的费用，来维持一头在旧金山的马。

And second, that, just because there's a have a comparative advantage mathematically does not mean that it is worth paying a 100 k a year or whatever cost to sustain a horse in San Francisco.

维持马匹的生存成本，将超过你从它身上获得的收益。

That subsistence is gonna be worth the benefit you get out of the horse.

我确实觉得这很有趣，因为我的预期和直觉显然与你的大不相同——技术树的大部分分支都永远不会被探索。

I I do think it's interesting like that that just the the sheer fact that, my expectation and my intuition obviously differs a great deal from yours on this is that most parts of the tech tree are never going to be explored.

可组合的方式太多了，有待发现的深刻想法也太多了，不仅是我们，任何人都不可能发现其中的大多数。

There's just too many interesting ways of combining things, there's too many sort of deep ideas waiting to be discovered and not only we but nobody ever is going to discover most of them.

因此，如何进行探索的选择实际上相当重要。

So choices about how to do the exploration actually matter quite a bit.

很有趣。

Interesting.

这正是我特别反感技术决定论论点的地方。

It's something I really dislike about sort of technological determinist arguments.

我愿意在进步相对简单的时候接受这种观点，但到了更高层次，你开始能够塑造探索的方式，而我们正在以一些有趣的方式塑造它。

I'm willing to buy it sort of low enough down when progress is relatively simple but higher up you start to get to shape the way in which you do the exploration and we are starting to shape it in interesting ways.

比如，有些技术实际上已被禁止，想想滴滴涕，想想氯氟碳化物，想想对核武器使用的限制，以及《不扩散核武器条约》。

I mean there's various technologies that have been essentially banned, think about DDT, you think about chlorofluorocarbons, you think about restrictions on the use of nuclear weapons, the Nuclear Non Proliferation Treaty.

这些事情以前并不是事先决定的，但在某些情况下，我们已经开始接近这种模式——直接预先决定不走那条路。

Those kinds of things are, they weren't done before the fact but it's starting to get pretty close in some cases where we just sort of preemptively decide we're not going to go down that path.

因此，这开始看起来像一套制度，我们正在实际影响如何探索技术树。

So that starts to look like a set of institutions where we are actually influencing how we explore the tech tree.

是的，你可能会看到贸易收益最明显的地方是在纯粹的信息可以来回传输的时候，因为这些信息的生产成本高昂，但验证和传输成本低廉。

Yeah, on where you would see these gains from trade obviously would be you'd see the most where it's pure information that can be sent back and forth because the information of the squalidly where it is expensive to produce but cheap to verify and cheap to send.

因此，未来生产力或类似的东西有多少能被提炼为信息，这将很有趣。

And so it'll be interesting how much of future productivity or whatever can be distilled down to information.

目前这还很难做到，因为你无法真正转移像中国在制造某物方面非常擅长的那种知识——这种过程知识分散在数亿名参与中国制造业的人脑海中，但未来如果由人工智能来执行，可能会更容易。

Right now it's kind of hard to do because you can't really transfer like if China is really good at manufacturing something, whether there's this process knowledge that's in the heads of 100,000,000 people involved in the manufacturing sector in China, but in the future it might be easier if AIs are doing

我的意思是，关于制造在多大程度上会变得非常标准化并真正商品化，这个问题一直存在，比如3D打印机至少二十年来都被视为下一个重大突破。

I mean the question about sort of to what extent does fabrication get sort of very uniform and get really commoditized like three d printers have been the next big thing for at least twenty years now.

为什么它们至今仍然不够好用？

Why do they still not work all that well?

为什么它们仍未真正成为制造业的核心，以及之后会怎样？

Why are they still not actually at the center of manufacturing and sort of what comes after that.

相比之下，看看核糖体就很有意思，它在生物学中确实处于核心地位，以许多非常有趣的方式发挥作用，而制造的未来是否也会如此简单——所有东西都通过某种途径流通，我不知道，也许是生物反应器之类的东西，你发送信息，然后生长出东西，或者你有一个真正有效的3D打印机。

It is funny to look at say the ribosome by contrast, it really is at the center of biology in a whole lot of really interesting ways and whether or not that's the future of manufacturing is something very simple, sort of where everything goes sort of as throughput through, I don't know, maybe it's a bioreactor or something like that so you send the information and then you grow stuff or you have some three d printer that actually works.

如果它们足够优秀，那么这确实会变成一个纯粹的信息问题，而一些这样的流程知识就会变得不那么重要。

If they're good enough, then actually it does become much more a pure information problem and some of this process knowledge becomes much less important.

Jane Street 拥有大量的计算资源，但 GPU 非常昂贵。

Jane Street has a lot of compute, but GPUs are very expensive.

因此，即使对 GPU 利用率只有相对较小影响的优化，也依然极其有价值。

And so even optimizations that have a relatively small effect on GPU utilization are still extremely valuable.

Jane Street 的两位机器学习工程师 Corwin 和 Sylvain 在 GTC 上详细介绍了他们的优化流程。

Two of Jane Street's ML engineers, Corwin and Sylvain, walked through some of their optimization workflow at GTC.

你并不是被网络速度太慢所限制。

You're not bottlenecked on the network being too slow.

你被限制的原因是等待训练中另一个进程完成工作。

You're bottlenecked on waiting for a different rank in your training, not having completed the work.

他们谈到了 Jane Street 如何分析追踪记录并诊断瓶颈，以及如何使用 CUDA 图、CUDA 流和自定义内核等技术来解决这些问题。

They talked about how Jane Street profiles traces and diagnoses bottlenecks, and then how they solve them using techniques like CUDA graphs and CUDA streams and custom kernels.

通过这些优化，Corwin 和 Sylvain 将他们的训练步骤时间从四百毫秒降低到了三百七十五毫秒。

With these sorts of optimizations, Coremen and Sylvain were able to get their training steps down from four hundred milliseconds to three seventy five milliseconds each.

这二十五毫秒的差异听起来可能很小，但考虑到Jane Street车队的规模，这一改进可能释放出数千张B200芯片。

This twenty five millisecond difference might sound small, but given the size of Jane Street's fleet, that improvement could free up thousands of B200s.

Jane Street已经将所有相关代码开源。

Jane Street open sourced all the relevant code.

如果你想了解，我在下方描述中提供了GitHub仓库和演讲的链接。

If you want to check it out, I've linked the GitHub repo and the talk in the description below.

如果你对这些内容感兴趣，Jane Street正在招聘研究员和工程师。

And if you find this stuff exciting, Jane Street is hiring researchers and engineers.

前往 janestreet.com/thwarkash 了解更多信息。

Go to janestreet.com/thwarkash to learn more.

我可以

Can I

问一个表达得非常笨拙的问题吗？

ask a very clumsily phrased question?

所以我们发现了几个深层原理。

So there's there's these deep principles that we've discovered a couple of.

其中一个观点是，如果某个维度上存在对称性，那么它就对应着一个守恒量。

One is this idea that, hey, if there's a symmetry across a dimension, it corresponds to a conserved quantity.

这是一个非常深刻的思想。

It's a very deep idea.

还有另一个观点，你写了很多相关内容，实际上还写过一本关于这个主题的教科书，即我们可以通过其他物理系统来理解可以计算哪些东西、哪些物理系统可以被理解，以及通用计算机是什么样子等等。

There's another which you've written a lot about written a textbook about in fact about there's we there's ways to understand the saying of what kinds of things you can compute, what kinds of physical systems you can understand with other physical systems, what a universal computer looks like, etcetera.

那么，你的观点是，如果深入到诺特定理或丘奇-图灵原理这样的层面，是否意味着存在无穷多条如此深刻的原理？

And is your view that if you go down to this level of idea of Noether CRM or the Church Turing principle, that there's an infinite number of extremely deep such principles?

因为我觉得，让这些原理特别之处在于，它们本身涵盖了世界可能存在的多种不同方式，但事实上，世界必须与这些少数几条极其深刻的原理相容。

Because I feel like what makes them special is that they themselves encompass so many different possible ways the world could be but no it has the the world has to be compatible with actually a couple of these very deep principles.

我不知道。

I don't know.

我的意思是，我这里所有的只是推测和某种直觉。

I I mean, you know, All I have here is speculation and sort of instinct.

我的直觉是，我们不断发现新的、非常根本的东西。

My instinct is we keep finding very fundamental new things.

对我来说，理解这些思想是非常具有塑造性的，正如我之前举的例子，丘奇和图灵以及其他人的那些关于通用可编程设备的精彩理念。后来你才明白，这些理念其实也包含了公钥加密的思想；再后来你又发现，它还蕴含着更深层的理念——人们称之为加密货币之类的，但其中确实有一套非常深刻的思想，关于如何集体维护一个基于此的共识账本，而其中可能还有许多其他深刻的思想。实际上，要找到这些思想的正确规范形式花了许多年时间。

It was very, I mean for me anyway, quite formative to understand, as I say, I gave the example before, these wonderful ideas of Church and Turing and these other people, ideas about universal programmable devices and then you understand later, oh, this also contains within it the ideas of public key cryptography and then you understand later, oh, that also contains within it the ideas, mean people refer to it as cryptocurrency or whatever but there's a very deep set of ideas there about the ability to collectively maintain an agreed upon ledger which is built upon this and there's probably many deep ideas to sort of it actually took whatever, it's taken many years really to figure out the right canonical form of those.

因此，正是这种不断发现看似深刻的新基本原理的事实，让我获得了非常重要的直觉突破。这种模式不仅体现在我刚才举的例子中，我认为你在许多其他领域也能看到同样的模式。

And so just this fact that you keep finding what seem like deep new fundamental primitives, I find very, for me that has been a very important intuition bump and it's across, I mean I've given that particular example but I think you see that same pattern in a lot of different areas.

那么，你如何解释这种经验现象：无论你把什么输入到科学或技术过程中，经济学家已经对此研究了无数次。

What is your interpretation then of this empirical phenomenon where ideas like, whatever input you consider into the scientific process or technological process, economists have studied this a million in a 100 ways.

它似乎始终以一种非常稳定的速度，每年需要增加x%的研究人员。

It just seems to require even at actually a very consistent rate, x percent more researchers per year.

几年前，尼古拉斯·布卢姆等人发表了一篇著名论文，他们问：半导体行业有多少人工作？这个数字随时间如何变化？

So there's this famous paper from a couple years ago, by Nicholas Bloom and others where they say, how many people are working in the semiconductor industry, And how does it increase over time

是的。

Yeah.

在摩尔定律的历史进程中？

Through the history of Moore's Law?

我认为他们发现，摩尔定律意味着计算能力每年增长40%，或者晶体管密度每年提升40%。

And I think they find, like, Moore's Law means computing increases 40% a year or transistor density increases 40% a year.

但为了维持这种趋势，科学家的数量每年增加了9%。抱歉。

But to keep that going, the amount of scientists has increased 9% a year Sorry.

半导体行业，他们对一个又一个行业都做出了这样的观察。

The semiconductor industry, And and they go through industry after industry with this observation.

所以你的观点是，这些深刻的想法确实存在，但它们变得越来越难发现吗？

And so is your view that there are these deep ideas, but they keep getting harder to find?

还是说，有另一种方式可以解释这些经验性观察背后的现象？

Or that, no, there's there's there's another way to think about what's happening with these empirical observations?

他们的所有例子都很狭窄，他们选择某个特定事物，然后观察某个特定指标。

Mean, of all, all of their examples are narrow, They pick a particular thing and then they look at some particular metric.

在这些研究中，根本没有出现GPU这样的东西，对吧？

Nowhere in that shows up, like GPUs don't show up there, right?

比如，突然间你获得了并行计算的能力，这非常有趣。

Like in the sense of all of a sudden you get this ability to parallelize and that's really interesting.

因此，有很多外部影响只是偶然出现的，因为他们采用了简单的定量测量方法，从农业生产力到各种其他领域都进行了研究，但你确实需要聚焦于狭窄的范围。正如我所说，我确实对这种不断出现的新类型进展感到非常感兴趣。

So there's sort of a lot of external consequences that are just delighted from basically they have these simple quantitative measures, they look at it in agricultural productivity, they look at it in a whole lot of different ways, but you do have to focus narrowly and I suppose I'm certainly interested, as I say, in this fact that just new types of progress keep becoming possible.

但即便如此，我认为仍然存在某种收益递减的现象。

But there is still, I think even there, there does seem to be some phenomenon of diminishing returns.

这是内在的吗？

Is that intrinsic?

这是世界结构本身的问题吗？

Is that something about the structure of the world?

这究竟是什么？

What is it?

有一件事变化不大，那就是从事这类工作的人类个体思维，或许我们也应该改善这些思维，或者建立某种反馈机制。

Well, one thing which hasn't changed that much is sort of the individual minds which are doing this kind of work and maybe those should be sort of being improved as well or some sort of feedback process going on there.

而这可能会改变事物的本质。

And maybe that changes the nature of things.

假设我回顾截至大约1700年之前的科学进展，那时进展非常缓慢，而且极不规律。

Suppose I look at scientific progress up until let's say 1700, something like that, and it was very slow and also it was very irregular.

早在基督诞生前五个世纪，爱奥尼亚人就已做出了相当了不起的成就，但如此多的知识却会丢失，然后又被重新发现，接着又再次遗失，你不得不承认，当时的进展极其缓慢。

Had the Ionians back sort of five centuries before Christ doing these quite remarkable things and so much knowledge would get lost and then it would be rediscovered and then it would be lost again and you'd have to say that progress was very slow.

那里部分原因只是因为我们当时还没有一些非常好的想法。

There it's partially just bound up with the fact that there were some very good ideas that we just didn't have.

即使你已经有了这些想法，你仍需要围绕它们建立制度，真正解决大量关于培训、资本配置以及这类问题，比如确保研究人员的基本安全，让他们不必担心宗教裁判所之类的事情。

Even once you've had the ideas, then you need to build institutions around them, actually need to solve a whole lot of different problems about training, about allocation of capital, about all these kinds of things, just about basic sort of security for researchers so they're not worried about the inquisition or things like that.

所以这些都是一些复杂的难题。

So there's all these kind of complicated problems.

当你解决了所有这些复杂的问题后，突然之间，科学进步就会迎来一次巨大的爆发。

You solve all those complicated problems and then all of a sudden boom, there's a massive sort of burst of scientific progress.

如果你不改变这些外部环境，如果这些外部条件陷入停滞，那么你确实可能会再次遇到收益递减的情况。

If you're not changing it, if there's some kind of stagnation there, if you're not changing those external sort of circumstances, yes, you may start to get sort of diminishing returns again.

但这并不意味着这种情况本身有什么内在必然性。

But that doesn't mean there's anything intrinsic about the situation.

也许某些外部因素又需要发生变化。

Maybe something external needs to change again.

显然，很多人认为人工智能可能成为推动因素。

Obviously a lot of people think AI is potentially going to be a driver.

我的意思是，从某种层面来看，它确实会如此，因为你可以把许多现代科学仪器看作本质上就是某种机器人。

I mean, it certainly will at some level, fact to the extent you can think of a lot of modern scientific instrumentation as really, I mean, some level kind of robots.

詹姆斯·韦伯太空望远镜是什么？

What is the James Webb Space Telescope?

也许把这描述为机器人显得有些非传统，但也不算完全不合理。

Well, it's unconventional maybe to describe it as a robot but it's not completely unreasonable either.

它是一个高度自动化、极其复杂的系统，配备了电子介导的传感器和执行器，事实上正在使用机器学习来处理数据。

It is an example of a highly automated, very sophisticated system with electronically mediated sensors and actuators where machine learning in fact is being used to process the data.

因此，从这个意义上说，我们已经开始看到这种转变，这种趋势已经持续了几十年。

So in that sense we're already starting to sort of see that transition, we've been seeing it for decades.

我有个想法，就是抽根大麻，吸一口。

I have this smoke a joint and take a puff thought which I think

我们已经讨论过几次了。

we've had a few.

是的，我想我们正说到这个部分了，你可以帮我把脚从嘴里拿出来，想出一个更具体的方式来思考这个问题。

Yeah, well I think we're getting to that part of the conversation and you can help me get foot out of my mouth and figure out a more concrete way to think about it.

所以你的观点是，AI会带来一场初始的革命——启蒙时代，而现在是AI时代，每一场变革可能以不同的速度或不同的方式改变科学的发展。

So your point that AI would there's an initial revolution, the enlightenment, and now there's AI, and each might be a different pace or a different way in which science happens.

如果你思考一下这些变革发生的速度，就可以看出，在人类历史的长河中，增长速率呈现出双曲线式的上升趋势。

If you think about the pace of how fast such transitions have been happening, you can draw over the long span of human history that's hyperbolic of the rate of growth is increasing.

所以，是的，十万年前，人类还处于石器时代。

So, yeah, a hundred thousand years ago, you had the stone age.

再往回追溯更久远的时期，灵长类动物已经存在了数百万年，十万年前是石器时代，一万年前是农业革命，三百年前是工业革命，每一次都标志着这种指数级增长速率的跃升。

You go back even much further how long a primate's been around, it would be like, let's say millions of years and a hundred thousand years ago, the stone age, then ten thousand years ago, the agricultural revolution, then three hundred year three hundred years ago, the industrial revolution, each marked by this exponent this increase in the rate of exponential growth.

然后人们认为AI将再次引发这样的变革，而且可能发生的速度会更快。

And then people think it's gonna happen again with AI, but that would happen potentially even faster.

在工业革命初期，没有人会想到，下一个里程碑会是人工智能。

It would not have occurred to somebody at the beginning of the industrial revolution that the next demarcation in this trend will be artificial intelligence.

如果事物的发展越来越快，而我们又难以预测下一次变革是什么，那我们或许只能把现在到AI之间的这个转折点视为‘奇点’，这正是过去与未来最根本的区别。

And so if things are getting faster and it's hard to anticipate what the next transition will be, I guess we just think of this singularity between now and AI and that's really what distinguishes the past from the future.

但如果我们沿用过去许多人所持有的那种思维模式，也许‘智能时代’也会非常短暂。

But just applying the same heuristic that many people in the past would have had, Maybe the intelligence age is also quite short.

而那之后的下一步是，我们甚至没有足够的本体论来描述它是什么，但未来的人不会把过去看作是‘前智能AI’和‘后AI’的分界。

And then the next thing after that is we don't even have the ontology to describe what it is, but the future will not think of the past as like there was pre intelligent AI and post AI.

不。

No.

这看起来——当然我们无法证明，但确实相当合理。

That seems, I mean obviously we can't prove this, but it certainly seems quite plausible.

我的意思是，当然，问题的一部分在于我们用来构想的载体似乎完全不对劲。

I mean part of the issue of course is just the substrate we have available to conceive seems all wrong.

你不能用一群黑猩猩去推测拥有语言会是什么样子。

You can't speculate with a bunch of chimpanzees about what it would be like to have language.

仅仅以过去的一个重大转折为例，转折本身才是关键，这似乎很有可能。

Just to sort of pick a major transition in the past, the transition itself is the thing and it seems likely.

如果我们谈论一些天马行空的想法，我确实觉得，未来会出现一种涉及使用经典计算机的通用人工智能的转折，但量子计算机也可能带来一场有趣的变革。

If we're talking about taking a puff kind of thoughts, I'm certainly amused by the idea that there's going to be some transition involving artificial general intelligence using classical computers, but actually there'll be an interesting transition with quantum computers as well.

它们可能具备一类更广泛、潜在更有趣的计算能力。

They're probably capable of a strictly larger class of potentially interesting computations.

所以，也许AQGI（或无论它该叫什么）的特性实际上是质的不同。

So maybe actually the character of AQGI or whatever it should be called is actually qualitatively different.

所以，也许在这两件事之间会有一个短暂的过渡期。

So maybe there's sort of a brief period between those two things.

有意思。

Interesting.

我的意思是，正如我所说，这仅仅是推测，但确实很有趣。

I mean, as I say, is just speculation, but it's certainly amusing.

有什么理由这么认为吗？

Is there a reason to think that?

据我所知，几十年来，像你这样的人已经为量子计算机能做的事情设定了相当严格的界限，它只会在一定程度上加速搜索。

From what I understand there's been, for decades, people like you have put pretty tight bounds on the kinds of things quantum computers can do and so it'll speed up search somewhat.

它确实能加速某些事情，比如肖尔算法，但也许这正印证了你的观点——我们无法提前预测技术树上会有什么，但至少从目前来看，你只是破解了加密，那除此之外你还用它做什么？

It will do and the kinds of things that extremely speeds up like Schirle's algorithm, it seems like it again maybe this is to your point that we can't predict in advance what's down the tech tree, but at least from here it seems like you break encryption, but what else are you using?

肖尔算法

Shor's algorithms

做。

do.

是的，我的意思是，我们只思考了三十年左右，或者四十来年，时间并不长，某种程度上，人类文明并没有真正深入地思考过这个问题。

Yeah, mean we've only been thinking about it for thirty years or whatever, forty or so years, not for very long and we sort of haven't in some sense thought that hard about it as a civilisation.

结果会不会发现它非常狭窄？

Does it turn out that it's very narrow?

也许吧。

Maybe.

结果会不会发现它非常宽广？

Does it turn out that it's very broad?

那也是一种极其激进的扩展。

That's also like a really radical expansion.

这显然是有可能的。

That seems distinctly possible.

还要记住，我们一直以来都是在没有相关设备的情况下进行探索的。

Keep in mind as well, we've been doing it without the benefit of having the devices.

对，这确实是个相当大的瓶颈。

Right, like that's a pretty big bottleneck to have.

如果你在18世纪考虑计算机科学，想着‘与’、‘或’这些操作，你到底该有什么

If you're thinking about computer science in the seventeen hundred's and you're like, okay, do and and or, what do have

要做什么？

to do?

你无法预见到比特币，也无法预见到深度学习。

You can't anticipate Bitcoin, you can't anticipate deep learning.

不，我的意思是，也许如果你足够聪明可以做到，但这确实是个非常困难的局面。

No, I mean maybe you could if you're sufficiently bright but it is a pretty hard situation.

你曾在90年代和2000年代参与并推动量子信息和量子计算，从内部视角来看，你是怎么看的？

What is your inside view having been in and contributing to quantum information quantum computing back in the 90s and 2000s?

你如何讲述这段历史？

What is your telling of the history?

当时的瓶颈是什么？

What was the bottleneck?

是什么关键的转变让量子计算成为了一个真正的领域？你如何评价费曼、德utsch以及其他后来者的贡献？

What was the key transition that made it a real field and how do you rank the contributions for Feynman to Deutsch to everybody else that came along?

是的，我的意思是，我们直接聚焦于真正发生了什么变化的问题。

Yeah so I mean let's just focus on the question about what actually changed.

那么，为什么量子计算在20世纪50年代还不存在？

So why was quantum computing not a thing in the 1950s?

本可以有的。

Could have been.

比如约翰·冯·诺伊曼，他是计算领域的先驱，也写过一本关于量子力学的重要著作，并且对量子力学深感兴趣。

Somebody like, I don't know, John von Neumann, good example, absolutely pioneering computation, also wrote a very important book about quantum mechanics and was deeply interested in quantum mechanics.

他本可以在那时就发明量子计算，我认为当时有不少人其实都有这个潜力。

Like he could have invented quantum computing at that time and I think there were quite a number of people who potentially could have.

那么，为什么到了80年代，像费曼和德utsch这样的人才发表了这些论文？

So why do we have these papers by people like Feynman and Deutsch in the 80s?

我认为，这些论文被广泛视为该领域的奠基之作。

Those are, I think, fairly regarded as the foundation of the field.

之前确实有一些零星的预示，但远未达到如此全面和深入的程度。

There are some partial anticipations a little bit earlier, but they were nowhere near as comprehensive and nowhere near as deep.

你应该去问大卫。

And well, you should ask David.

你可以去问费曼，可惜他已经不在了，但他对这个问题的理解肯定比我深刻得多。

You can ask Feynman unfortunately, but he'll know much better than I do.

我觉得有几件事很有趣，其中之一是，计算在70年代末、80年代初变得显著重要起来。

A couple of things that I think are interesting, one is that of course computation became far more salient sort of late 70s, early 80s.

它变成了一件吸引更多人关注的事情，部分原因非常平凡。

It just became a thing which many more people were interested in, partially for very banal reasons.

你可以去买一台个人电脑，买一台苹果II，买一台康懋达64，买所有这些设备。

You could go and buy a PC, you could buy an Apple II, you could buy a Commodore 64, you could buy all these kinds of things.

人们逐渐意识到，这些设备是非常强大且值得深入思考的。

It became apparent to people that these were very powerful devices, very interesting to think about.

与此同时，在量子领域，那也正是离子阱技术出现的时期，人们开始能够囚禁单个离子等，而在此之前，我们实际上还无法操控单个量子态。

At the same time, in the quantum case, that was also the time of the ball trap and the ability to trap single ions and so on and up to that point we hadn't really had the ability to manipulate single quantum states.

所以，这两件事情因为历史上的偶然原因，大约在1980年左右都趋于成熟了，像冯·迪门这样的人本可以更早提出这个想法，但我觉得这确实是个很有趣的现象。

So you kind of got these two separate things that just for historically contingent reasons had both matured around let's say 1980 or so and somebody like von Dymen could have had the idea earlier but it is I think quite an interesting fact.

我记得关于理查德·费曼的这个故事。

I remember the story about Richard Feynman.

他在1980年或1981年买了其中第一批个人电脑之一，显然对这个设备兴奋不已。

He went and got one of the first PCs around 1980, 1981 and he was apparently just so excited with this device.

他因为搬运自己新买的计算设备而摔倒，受了很重的伤。

He actually tripped and hurt himself quite badly carrying his brand new computing device.

这只是一个历史偶然的巧合，但能有这样一位在量子力学方面极为出色、同时又对这些新机器充满热情的人，确实难得。

That's a very historically contingent coincidence but having somebody who is very, very talented and understanding of quantum mechanics also just very excited about these new machines.

他因此产生这样的想法，或许并不令人意外。

It's not so surprising perhaps that he's thinking then.

十年前，你能讲出类似的故事吗？

What similar story could you have told ten years earlier?

根本不可能，当时的条件还不具备。

Like there is just no, the conditions don't exist for it.