2020-05-08 Ilya Sutskever.Deep Learning

Refer To：《Ilya Sutskever: Deep Learning | Lex Fridman Podcast #94》。

Lex Fridman:

The following is a conversation with Ilya Sutskever, co-founder and chief scientist of OpenAI, one of the most cited computer scientists in history with over 165,000 citations, and to me, one of the most brilliant and insightful minds ever in the field of deep learning.

亚历克斯·弗里德曼：

以下是与伊利亚·苏茨克维尔的对话。他是OpenAI的联合创始人兼首席科学家，历史上被引用次数超过16.5万次的计算机科学家之一。在我看来，他是深度学习领域最聪明、最有洞察力的人之一。

There are very few people in this world who I would rather talk to and brainstorm with about deep learning, intelligence, and life in general than Ilya, on and off the mic. This was an honor and a pleasure.

在这个世界上，几乎没有几个人我会比伊利亚更愿意与之交流、探讨深度学习、智能以及生活，无论是在麦克风前还是私下。这是一种荣幸和乐趣。

This conversation was recorded before the outbreak of the pandemic. For everyone feeling the medical, psychological, and financial burden of this crisis, I'm sending love your way. Stay strong. We're in this together. We'll beat this thing.

这次对话是在疫情爆发前录制的。对于那些在这场危机中感受到医疗、心理和经济负担的人们，我向你们送上关爱。保持坚强。我们同舟共济，一定会战胜这一切。

This is the Artificial Intelligence Podcast. If you enjoy it, subscribe on YouTube, review it with Five Stars on Apple Podcast, support it on Patreon, or simply connect with me on Twitter at Lex Fridman, spelled F-R-I-D-M-A-N.

这是《人工智能播客》。如果你喜欢，请在YouTube上订阅，在Apple Podcast上给予五星评价，通过Patreon支持，或者在Twitter上与我联系，我的名字是Lex Fridman，拼写为F-R-I-D-M-A-N。

As usual, I'll do a few minutes of ads now and never any ads in the middle that can break the flow of the conversation. I hope that works for you and doesn't hurt the listening experience.

像往常一样，我现在会做几分钟的广告，并且绝不会在对话中插入广告打断节奏。我希望这样对你们来说可以接受，不会影响听感。

This show is presented by Cash App, the number one finance app in the App Store. When you get it, use code LEXPODCAST. Cash App lets you send money to friends, buy Bitcoin, invest in the stock market with as little as $1.

本节目由Cash App赞助，这款应用是App Store上的第一大金融应用。当你下载时，使用代码LEXPODCAST。Cash App让你可以向朋友转账，购买比特币，并以最低1美元的金额投资股市。

Since Cash App allows you to buy Bitcoin, let me mention that cryptocurrency in the context of the history of money is fascinating. I recommend Ascent of Money as a great book on this history. Both the book and audiobook are great.

既然Cash App可以购买比特币，那么让我提一下货币历史背景下的加密货币是多么令人着迷。我推荐《货币崛起》这本关于货币历史的好书，书本和有声版都很棒。

Debits and credits on ledgers started around 30,000 years ago. The US dollar created over 200 years ago, and Bitcoin, the first decentralized cryptocurrency, released just over 10 years ago.

账本上的借贷记录始于大约三万年前。美元诞生于两百多年前，而比特币这一首个去中心化的加密货币问世仅仅十多年。

So given that history, cryptocurrency is still very much in its early days of development, but it's still aiming to, and just might, redefine the nature of money.

因此，鉴于这一历史，加密货币仍处于发展的早期阶段，但它的目标是，而且很可能，会重新定义金钱的本质。

So again, if you get Cash App from the App Store or Google Play and use the code LEXPODCAST, you get $10. And Cash App will also donate $10 to FIRST, an organization that is helping advance robotics and STEM education for young people around the world. And now, here's my conversation with Ilya Sutskever.

所以，再次提醒，如果你从App Store或Google Play下载Cash App并使用代码LEXPODCAST，你将获得10美元。Cash App还将捐赠10美元给FIRST，一个致力于推动全球青少年机器人技术和STEM教育的组织。现在，请听我与伊利亚·苏茨克维尔的对话。

You were one of the three authors with Alex Grishevsky, Jeff Hinton of the famed AlexNet paper that is arguably the paper that marked the big catalytic moment that launched the deep learning revolution.

你是与亚历克斯·克里舍夫斯基和杰夫·辛顿共同撰写著名的AlexNet论文的三位作者之一。这篇论文可以说是引发深度学习革命的重要催化剂。

At that time, take us back to that time, what was your intuition about neural networks, about the representational power of neural networks?

当时，请带我们回到那个时期，你对神经网络及其表征能力有什么直觉？

And maybe you can mention how did that evolve over the next few years up to today, over the 10 years?

你也可以谈谈这些直觉在接下来的几年到如今的十年间是如何发展的？

Ilya Sutskever:

Yeah, I can answer that question. At some point in about 2010 or 2011, I connected two facts in my mind. Basically, the realization was this.

伊利亚·苏茨克维尔：

好的，我可以回答这个问题。在大约2010或2011年，我将脑海中的两个事实联系在一起。基本上，我的领悟是这样的。

At some point we realized that we can train very large, I shouldn't say very, you know, they were tiny by today's standards, but large and deep neural networks end-to-end with backpropagation.

我们意识到可以用反向传播端到端训练非常大的神经网络。按今天的标准来说，这些网络非常小，但当时它们已经算是大型和深层的神经网络了。

At some point, different people obtained this result. I obtained this result.

在某个时刻，不同的人都得到了这个结果。我也得到了这个结果。

The first moment in which I realized that Deep neural networks are powerful was when James Martens invented the Hessian Free Optimizer in 2010, and he trained a 10-layer neural network end-to-end without pre-training from scratch.

我第一次意识到深度神经网络的强大是在2010年，当时詹姆斯·马滕斯发明了Hessian-Free优化器，并且他从头开始端到端训练了一个10层的神经网络，无需预训练。

And when that happened, I thought this is it. Because if you can train a big neural network, a big neural network can represent very complicated function.

当时我就觉得，这就是关键。如果能训练一个大规模的神经网络，那么这个大网络就能够表征非常复杂的函数。

Because if you have a neural network with 10 layers, it's as though you allow the human brain to run for some number of milliseconds. Neuron firings are slow. And so in maybe 100 milliseconds, your neurons only fire 10 times.

因为如果有一个10层的神经网络，那就像让人脑运行几个毫秒。神经元的放电速度很慢，所以可能在100毫秒内，神经元只放电10次。

So it's also kind of like 10 layers. And in 100 milliseconds, you can perfectly recognize any object. So I thought, so I already had the idea then that we need to train a very big neural network on lots of supervised data.

这也就相当于10层。而在100毫秒内，人类可以完美识别任何物体。所以我当时的想法是，我们需要在大量的监督数据上训练一个非常大的神经网络。

And then it must succeed because we can find the best neural network. And then there's also theory that if you have more data than parameters, you won't overfit.

然后它一定会成功，因为我们能够找到最好的神经网络。当时还有一种理论认为，如果数据量大于参数数量，就不会过拟合。

Today we know that actually this theory is very incomplete and you won't overfit even if you have less data than parameters, but definitely if you have more data than parameters, you won't overfit.

今天我们知道，这种理论其实非常不完整，即使数据量少于参数数量也不会过拟合，但如果数据量大于参数数量，肯定不会过拟合。

Lex Fridman:

So the fact that neural networks were heavily over-parameterized wasn't discouraging to you? So you were thinking about the theory that the number of parameters, the fact there's a huge number of parameters is okay, it's going to be okay?

莱克斯·弗里德曼：

所以，神经网络的参数严重过多这一点并没有让你感到气馁吗？你认为参数数量巨大的事实是可以接受的吗？它会没问题，对吗？

Ilya Sutskever:

I mean, there was some evidence before that it was okay-ish, but the theory was that if you had a big data set and a big neural net, it was going to work. The over-parameterization just didn't really figure much as a problem.

伊利亚·苏茨克维尔：

我是说，之前有一些证据表明这是大致可行的。但理论上，如果你有一个大数据集和一个大的神经网络，它就会奏效。参数过多并没有被视为一个大问题。

I thought, well, with images, you're just going to add some data augmentation and it's going to be okay.

我当时认为，对于图像问题，你只需要增加一些数据增强就会没问题。

Lex Fridman:

So where was any doubt coming from?

莱克斯·弗里德曼：

那么疑虑来自哪里呢？

Ilya Sutskever:

The main doubt was can we train a bigger, will we have enough compute to train a big enough neural net?

伊利亚·苏茨克维尔：

主要的疑虑是，我们能否训练更大的神经网络，我们是否有足够的计算能力来训练一个足够大的网络？

Lex Fridman:

With backpropagation.

莱克斯·弗里德曼：

用反向传播？

Ilya Sutskever:

Backpropagation I thought would work. The thing which wasn't clear was whether there would be enough compute to get a very convincing result.

伊利亚·苏茨克维尔：

我认为反向传播是可行的。唯一不确定的是是否有足够的计算能力来获得一个非常有说服力的结果。

And then at some point Alex Kerzhevsky wrote these insanely fast CUDA kernels for training convolutional neural nets and that was BAM. Let's do this. Let's get ImageNet and it's going to be the greatest thing.

后来在某个时刻，亚历克斯·克里舍夫斯基编写了这些极其快速的CUDA内核，用于训练卷积神经网络，这就是关键。我们决定这么干了。我们拿下了ImageNet，这将是最伟大的成就。

Lex Fridman:

Was most of your intuition from empirical results by you and by others? So like just actually demonstrating that a piece of program can train a 10-layer neural network? Or was there some pen and paper or marker and whiteboard thinking intuition? Because you just connected a 10-layer large neural network to the brain. So you just mentioned the brain.

莱克斯·弗里德曼：

你的大部分直觉来自你和其他人的实验结果吗？比如，实际上证明一段程序可以训练一个10层的神经网络？还是有一些笔和纸、记号笔和白板上的直觉推导？因为你刚刚把一个10层的大型神经网络与大脑联系了起来。你刚刚提到了大脑。

So in your intuition about neural networks, does the human brain come into play as an intuition builder.

那么在你对神经网络的直觉中，人类大脑是否起到了直觉构建的作用？

Ilya Sutskever:

Definitely. I mean, you know, you got to be precise with these analogies between artificial neural networks and the brain, but there is no question that the brain is a huge source of intuition and inspiration for deep learning researchers since all the way from Rosenblatt in the 60s.

伊利亚·苏茨克维尔：

毫无疑问。虽然在人工神经网络与大脑之间的类比上需要谨慎，但不可否认，自20世纪60年代罗森布拉特以来，大脑一直是深度学习研究者的重要直觉和灵感来源。

Like, if you look at the whole idea of a neural network is directly inspired by the brain. You had people like McCallum and Pitts who were saying, hey, you got these neurons in the brain.

比如，神经网络的整体概念直接受到大脑的启发。像麦卡洛克和皮茨这样的人说：“看，大脑中有这些神经元。”

And hey, we recently learned about the computer and automata. Can we use some ideas from the computer and automata to design some kind of computational object that's going to be simple, computational, and kind of like the brain?

“而且我们最近学到了关于计算机和自动机的知识。我们能否从计算机和自动机中借鉴一些想法，设计出一种简单、计算型，并有点类似于大脑的计算对象？”

And they invented the neuron. So they were inspired by it back then. Then you had the convolutional neural network from Fukushima.

于是他们发明了神经元。他们当时就是受到了这种启发。后来又有了福岛提出的卷积神经网络。

And then later, Jan LeCun, who said, hey, if you limit the receptive fields of a neural network, it's going to be especially suitable for images, as it turned out to be true.

再后来，扬·勒昆说：“如果限制神经网络的感受野，它将特别适合处理图像。”事实证明，他的想法是正确的。

So there was a very small number of examples where analogies to the brain were successful. And I thought, well, probably an artificial neuron is not that different from the brain if you squint hard enough.

所以，确实有少数成功的例子将神经网络与大脑类比。我当时想，如果足够仔细观察，人工神经元可能与大脑并没有那么大的不同。

So let's just assume it is and roll with it.

所以，不如就假设是这样，并继续探索。

Lex Fridman:

So we're now at a time where deep learning is very successful. So let us squint less and say, let's open our eyes and say, what to you is an interesting difference between the human brain? Now I know you're probably not an expert, neither a neuroscientist nor a biologist, but loosely speaking, what's the difference between the human brain and artificial neural networks that's interesting to you for the next decade or two?

莱克斯·弗里德曼：

我们现在正处于深度学习非常成功的时期。所以，让我们少些猜测，睁大眼睛来谈一谈。对你来说，人类大脑与人工神经网络之间有哪些有趣的区别？我知道你可能不是专家，既不是神经科学家也不是生物学家，但从广义上讲，在接下来的十到二十年间，这种差异中有哪些是你感兴趣的？

Ilya Sutskever:

That's a good question to ask. What is an interesting difference between the brain and our artificial neural networks? So I feel like today artificial neural networks.

伊利亚·苏茨克维尔：

这是个好问题。大脑与我们的人工神经网络之间有哪些有趣的差异呢？我觉得在今天，人工神经网络的表现如何。

So we all agree that there are certain dimensions in which the human brain vastly outperforms our models.

我们都同意，在某些方面，人类大脑远远优于我们的模型。

But I also think that there are some ways in which our artificial neural networks have a number of very important advantages over the brain.

但我也认为，在某些方面，我们的人工神经网络相较于大脑也具有许多非常重要的优势。

Looking at the advantages versus disadvantages is a good way to figure out what is the important difference. So the brain uses spikes, which may or may not be important.

比较优势与劣势是找出重要差异的一个好方法。比如，大脑使用脉冲，这可能重要，也可能不重要。

Lex Fridman:

That's a really interesting question. Do you think it's important or not? That's one big architectural difference between artificial neural networks.

莱克斯·弗里德曼：

这是一个非常有趣的问题。你觉得这重要吗？这是人工神经网络的一大架构差异。

Ilya Sutskever:

It's hard to tell, but my prior is not very high, and I can say why.

伊利亚·苏茨克维尔：

很难说，但我对它的重要性预期不高，我可以解释原因。

You know, there are people who are interested in spiking neural networks, and basically what they figured out is that they need to simulate the non-spiking neural networks in spikes. And that's how they're going to make them work.

你知道，有些人对脉冲神经网络感兴趣，基本上他们发现需要用脉冲来模拟非脉冲神经网络。这是让它们工作的方法。

If you don't simulate the non-spiking neural networks in spikes, it's not going to work because the question is, why should it work? And that connects to questions around backpropagation and questions around deep learning.

如果不这样做，它就不会有效，因为问题是：为什么它应该有效？这与反向传播和深度学习的问题有关。

You've got this giant neural network. Why should it work at all? Why should the learning rule work at all? It's not a self-evident question, especially if you,

你有一个庞大的神经网络。为什么它应该起作用？为什么学习规则应该起作用？这并不是一个显而易见的问题，尤其是如果你，

let's say if you were just starting in the field and you read the very early papers, you can say, hey, people are saying, let's build neural networks.

比如说，你刚进入这个领域并阅读了早期的论文，你会发现人们在说：“让我们来构建神经网络。”

That's a great idea because the brain is a neural network, so it would be useful to build neural networks. Now, let's figure out how to train them. It should be possible to train them probably, but how?

这是个好主意，因为大脑是一个神经网络，所以构建神经网络应该很有用。那么，我们来研究如何训练它们。这可能是可行的，但要怎么做？

And so the big idea is the cost function. That's the big idea. The cost function is a way of measuring the performance of the system according to some measure.

于是，大的突破点就是成本函数。这是关键概念。成本函数是一种根据某种指标衡量系统性能的方法。

Lex Fridman:

By the way, that is a big... Actually, let me think. Is that one, a difficult idea to arrive at? And how big of an idea is that? That there's a single cost function? Sorry, let me take a pause.

莱克斯·弗里德曼：

顺便说一句，这真是一个重大……让我想想。这是一个难以得出的概念吗？它有多重要？一个单一的成本函数？抱歉，我稍作停顿。

Is supervised learning a difficult concept to come to?

监督学习是一个难以得出的概念吗？

Ilya Sutskever:

I don't know. All concepts are very easy in retrospect.

伊利亚·苏茨克维尔：

我不知道。所有概念在回顾时都显得很简单。

Lex Fridman:

Yeah, that's what it seems trivial now. But I—so, because the reason I asked that—and we'll talk about it—is, are there other things? Are there things that don't necessarily have a cost function?

莱克斯·弗里德曼：

是的，现在看起来这些似乎很简单。但是我问这个问题是因为——我们可以讨论——有没有其他可能？有没有一些东西不一定有一个成本函数？

Maybe have many cost functions or maybe have dynamic cost functions or maybe a totally different kind of architecture? Because we have to think like that in order to arrive at something new, right?

也许有多个成本函数，或者是动态的成本函数，或者完全不同类型的架构？因为我们必须这样思考，才能得出一些新东西，对吧？

Ilya Sutskever:

So the only good examples of things which don't have clear cost functions are GANs. And again, you have a game.

伊利亚·苏茨克维尔：

目前唯一没有明确成本函数的好例子是生成对抗网络（GANs）。在GAN中，你实际上是在玩一个博弈。

So instead of thinking of a cost function, where you know that you have an algorithm—gradient descent—which will optimize the cost function, and then you can reason about the behavior of your system in terms of what it optimizes.

所以，与其说是成本函数——你知道可以用梯度下降算法来优化成本函数，然后通过它优化的内容来推导系统行为——

With GANs, you say, I have a game and I'll reason about the behavior of the system in terms of the equilibrium of the game. But it's all about coming up with these mathematical objects that help us reason about the behavior of our system.

在GAN中，你会说：“我有一个博弈，我会通过博弈的均衡来推导系统的行为。”但本质上，这是在寻找能帮助我们推导系统行为的数学对象。

Lex Fridman:

Right, that's really interesting. Yeah, so GAN is the only one. The cost function is emergent from the comparison.

莱克斯·弗里德曼：

对，这确实很有趣。是的，所以GAN是唯一的例子。它的成本函数是通过比较产生的。

Ilya Sutskever:

I don't know if it has a cost function. I don't know if it's meaningful to talk about the cost function of a GAN. It's kind of like the cost function of biological evolution or the cost function of the economy.

伊利亚·苏茨克维尔：

我不确定GAN是否有成本函数。我也不确定谈论GAN的成本函数是否有意义。这有点像谈论生物进化的成本函数或经济的成本函数。

You can talk about regions to which it will go towards, but I don't think the cost function analogy is the most useful.

你可以讨论它可能趋向的区域，但我不认为成本函数的类比是最有用的。

Lex Fridman:

So evolution doesn't—that's really interesting. So if evolution doesn't really have a cost function, like a cost function based on its something akin to our mathematical conception of a cost function.

莱克斯·弗里德曼：

所以进化没有一个真正的成本函数，这确实很有趣。那么，如果进化没有一个类似于我们数学概念中的成本函数的东西——

Then do you think cost functions in deep learning are holding us back? Yeah, so you just kind of mentioned that cost function is a nice first profound idea. Do you think that's a good idea? Do you think it's an idea we'll go past?

那么你认为深度学习中的成本函数是否限制了我们？你刚才提到成本函数是一个很好的、深刻的初步想法。你觉得这是个好主意吗？你认为我们会超越这个想法吗？

So self-play starts to touch on that a little bit in reinforcement learning systems.

自我博弈在强化学习系统中开始稍微涉及到这一点了。

Ilya Sutskever:

That's right. Self-play and also ideas around exploration where you're trying to take action that surprises a predictor. I'm a big fan of cost functions. I think cost functions are great and they serve us really well.

伊利亚·苏茨克维尔：

没错。自我博弈，以及关于探索的想法，比如尝试采取会让预测器感到意外的行动。我是成本函数的忠实粉丝。我认为成本函数非常棒，它们对我们帮助很大。

And I think that whenever we can do things with cost functions, we should.

我认为，只要我们能用成本函数解决问题，就应该这样做。

And you know, maybe there is a chance that we will come up with some yet another profound way of looking at things that will involve cost functions in a less central way. But I don't know.

你知道，也许我们有可能找到另一种深刻的视角，用一种不那么核心依赖成本函数的方式来看待事物。但我不确定。

I think cost functions are—I mean, I would not bet against cost functions.

我认为成本函数——我的意思是，我不会对成本函数持反对态度。

Lex Fridman:

Is there other things about the brain that pop into your mind that might be different and interesting for us to consider in designing artificial neural networks? So we talked about spiking a little bit.

莱克斯·弗里德曼：

你能想到其他与大脑有关的事情吗？它们可能不同于人工神经网络，但对设计人工神经网络有趣且值得考虑？我们之前谈到了脉冲。

Ilya Sutskever:

I mean, one thing which may potentially be useful, I think people, neuroscientists have figured out something about the learning rule of the brain, or I'm talking about spike-time-dependent plasticity,

伊利亚·苏茨克维尔：

我认为可能有用的一件事是，神经科学家们似乎已经弄清楚了大脑的一些学习规则，比如我说的是“脉冲时序依赖可塑性”（STDP），

and it would be nice if some people were to study that in simulation.

如果有人能在模拟中研究这个，那就太好了。

Lex Fridman:

Wait, sorry, spike-time-independent plasticity?

莱克斯·弗里德曼：

等等，抱歉，是“脉冲时序非依赖性可塑性”吗？

Ilya Sutskever:

Yeah, that's right. What's that? STDP. It's a particular learning rule that uses spike timing to figure out how to determine how to update the synapses.

伊利亚·苏茨克维尔：

对，是这个。是什么呢？STDP（脉冲时序依赖可塑性）。这是一种利用脉冲时序来决定如何更新突触的学习规则。

So it's kind of like if the synapse fires into the neuron before the neuron fires, then it strengthens the synapse. And if the synapse fires into the neuron shortly after the neuron fired, then it weakens the synapse.

它大致是这样的：如果突触在神经元发放脉冲之前就发放脉冲，那么突触会得到增强；而如果突触在神经元发放脉冲之后短时间内发放脉冲，那么突触会减弱。

Something along this line. I'm 90% sure it's right. So if I said something wrong here, don't get too angry.

大致是这样的。我有90%的把握这是对的。所以，如果我说错了，不要太生气。

Lex Fridman:

But you sounded brilliant while saying it. But the timing, that's one thing that's missing. The temporal dynamics is not captured. I think that's like a fundamental property of the brain, is the timing of the signals.

莱克斯·弗里德曼：

但你说得听起来很棒。不过时序，这是我们遗漏的一点。时序动态没有被捕捉到。我认为信号的时序是大脑的一个基本特性。

Ilya Sutskever:

Well, you have recurrent neural networks.

伊利亚·苏茨克维尔：

嗯，我们有循环神经网络（RNN）。

Lex Fridman:

But you think of that as, I mean, that's a very crude, simplified, what's that called? There's a clock I guess to recurrent neural networks.

莱克斯·弗里德曼：

但你认为那就是……我的意思是，那是一个非常粗略、简化的版本，叫什么来着？我猜循环神经网络有一个“时钟”。

It seems like the brain is the general, the continuous version of that, the generalization where all possible timings are possible and then within those timings is contained some information.

看起来大脑是循环神经网络的更一般化、连续版本，所有可能的时序都可以实现，并且这些时序中包含了一些信息。

You think recurrent neural networks, the recurrence in recurrent neural networks can capture the same kind of phenomena as the timing of—that seems to be important in the firing of neurons in the brain.

你认为循环神经网络中的“循环”能否捕捉到类似于大脑中神经元发放时序的现象？那似乎对神经元的发放很重要。

Ilya Sutskever:

I mean, I think recurrent neural networks are amazing and I think they can do anything we'd want a system to do. Right now, recurrent neural networks have been superseded by transformers, but maybe one day they'll make a comeback.

伊利亚·苏茨克维尔：

我认为循环神经网络很了不起，我觉得它们可以做任何我们想让一个系统做的事情。目前，循环神经网络已被Transformer取代，但或许有一天它们会卷土重来。

Maybe they'll be back. We'll see.

也许它们会回来。我们拭目以待。

Lex Fridman:

Let me, on a small tangent, say, do you think they'll be back? So, so much of the breakthroughs recently that we'll talk about on natural language processing and language modeling has been with Transformers that don't emphasize recurrence.

莱克斯·弗里德曼：

让我稍微岔开话题，你觉得循环神经网络会回归吗？最近在自然语言处理和语言建模方面的许多突破，都是通过不强调循环的Transformer实现的。

Do you think recurrence will make a comeback?

你认为循环会重新崛起吗？

Ilya Sutskever:

Well, some kind of recurrence, I think very likely. Recurrent neural networks as they're typically thought of for processing sequences, I think it's also possible.

伊利亚·苏茨克维尔：

嗯，我认为某种形式的循环很可能会回归。传统上用于处理序列的循环神经网络，我觉得也是有可能的。

Lex Fridman:

What is, to you, a recurrent neural network? And generally speaking, I guess, what is a recurrent neural network?

莱克斯·弗里德曼：

在你看来，什么是循环神经网络？广义上来说，循环神经网络是什么？

Ilya Sutskever:

You have a neural network which maintains a high-dimensional hidden state. And then when an observation arrives, it updates its high-dimensional hidden state through its connections in some way.

伊利亚·苏茨克维尔：

循环神经网络是一个能够维护高维隐藏状态的神经网络。当一个观测值到来时，它通过某种方式利用连接更新其高维隐藏状态。

Lex Fridman:

So do you think—you know, that's what expert systems did, right? Symbolic AI, the knowledge base, growing a knowledge base is maintaining a hidden state, which is its knowledge base, and is growing it by sequentially processing.

莱克斯·弗里德曼：

那么，你认为这不就是专家系统的做法吗？符号式AI通过知识库来扩展，维护一个隐藏状态（即其知识库），并通过顺序处理来扩展它。

Do you think of it more generally in that way? Or is it simply, is it the more constrained form of a hidden state with certain kind of gating units that we think of as today with LSTMs and that?

你是否更广泛地这样看待它？或者说，它只是像我们今天所理解的LSTM那样，通过某种门控单元约束的隐藏状态形式？

Ilya Sutskever:

I mean, the hidden state is technically what you described there, the hidden state that goes inside the LSTM or the RNN or something like this.

伊利亚·苏茨克维尔：

从技术上讲，隐藏状态就是你所描述的那种隐藏状态，存在于LSTM或RNN中的那种。

But then what should be contained, you know, if you want to make the expert system analogy, I'm not... I mean, you could say that the knowledge is stored in the connections and then the short-term processing is done in the hidden state.

但至于其中应包含什么，如果你要做专家系统的类比，我觉得……你可以说知识存储在连接中，而短期处理则是在隐藏状态中完成的。

Lex Fridman:

Yes, could you say that?

莱克斯·弗里德曼：

是的，你能这样说吗？

Ilya Sutskever:

Yes.

伊利亚·苏茨克维尔：

是的。

Lex Fridman:

So sort of do you think there's a future of building large-scale knowledge bases within the neural networks?

莱克斯·弗里德曼：

你是否认为在神经网络中构建大规模知识库有未来？

Ilya Sutskever:

Definitely.

伊利亚·苏茨克维尔：

绝对有。

Lex Fridman:

So we're going to pause in that confidence because I want to explore that. Well, let me zoom back out and ask: back to the history of ImageNet, neural networks have been around for many decades, as you mentioned.

莱克斯·弗里德曼：

我们稍微暂停一下这个话题，因为我想深入探讨。但让我先回顾一下，谈谈ImageNet的历史。正如你提到的，神经网络已经存在了几十年。

What do you think were the key ideas that led to their success, that ImageNet moment and beyond, the success in the past 10 years?

你认为哪些关键思想促成了它们的成功，尤其是ImageNet时刻以及之后的过去10年的成功？

Ilya Sutskever:

Okay, so the question is, to make sure I didn't miss anything, the key ideas that led to the success of deep learning over the past 10 years.

伊利亚·苏茨克维尔：

好的，你的意思是，要确保我没漏掉任何内容，过去10年中促成深度学习成功的关键思想。

Lex Fridman:

Exactly, even though the fundamental thing behind deep learning has been around for much longer.

莱克斯·弗里德曼：

没错，尽管深度学习背后的基本原理已经存在很久了。

Ilya Sutskever:

The key idea about deep learning, or rather the key fact about deep learning before deep learning started to be successful, is that it was underestimated.

伊利亚·苏茨克维尔：

关于深度学习的关键思想，或者更确切地说，在深度学习取得成功之前的一个关键事实，是它被低估了。

People who worked in machine learning simply didn't think that neural networks could do much. People didn't believe that large neural networks could be trained.

从事机器学习的人普遍认为神经网络没什么用。他们不相信可以训练出大型神经网络。

People thought that, well, there was a lot of debate going on in machine learning about what are the right methods and so on. And people were arguing because there was no way to get hard facts.

人们认为，机器学习领域有很多关于正确方法的争论。这些争论持续着，因为当时没有办法得到确凿的证据。

And by that I mean, there were no benchmarks which were truly hard, that if you do really well on them, then you can say, look, here is my system. That's when you switch from...

我的意思是，没有真正艰难的基准测试。即便你在某个测试中表现很好，也很难说：“看，这就是我的系统。” 这是从理论转向实践工程的关键点。

That's when this field becomes a little bit more of an engineering field. So in terms of deep learning, to answer the question directly, the ideas were all there. The thing that was missing was a lot of supervised data and a lot of compute.

这就是这个领域开始更多地偏向工程领域的转折点。因此，就深度学习而言，直接回答问题就是：所有的思想早已存在，缺少的是大量的监督数据和计算能力。

Once you have a lot of supervised data and a lot of compute, then there is a third thing which is needed as well, and that is conviction.

当你有了大量监督数据和计算能力后，还需要第三样东西，那就是信念。

Conviction that if you take the right stuff, which already exists, and apply and mix it with a lot of data and a lot of compute, that it will in fact work. And so that was the missing piece.

这种信念是：如果你采用正确的方法（它们已经存在），并将其与大量数据和计算能力结合，那么它一定会奏效。这就是当时缺失的一环。

It was you had the, you needed the data, you needed the compute, which showed up in terms of GPUs, and you needed the conviction to realize that you need to mix them together.

你需要数据，需要计算能力（以GPU的形式出现），还需要意识到将它们结合起来的信念。

Lex Fridman:

So that's really interesting. So I guess the presence of compute and the presence of supervised data allowed the empirical evidence to do the convincing of the majority of the computer science community.

莱克斯·弗里德曼：

这真是有趣。所以我想，计算能力和监督数据的存在，使得实证结果能够说服计算机科学界的大多数人。

So I guess there's a key moment with Jitendra Malik and Alyosha Efros who were very skeptical, right? And then there's a Geoffrey Hinton that was the opposite of skeptical.

我想，有一个关键时刻是吉滕德拉·马利克和阿廖沙·埃夫罗斯对此非常怀疑，对吧？而杰弗里·辛顿则完全不怀疑。

And there was a convincing moment and I think ImageNet served as that moment.

当时有一个具有说服力的时刻，我认为ImageNet就是那个时刻。

Ilya Sutskever:

That's right.

伊利亚·苏茨克维尔：

没错。

Lex Fridman:

And that represented this kind of where the big pillars of the computer vision community kind of, the wizards got together, and then all of a sudden there was a shift. And it's not enough for the ideas to all be there and the compute to be there. It's for it to convince the cynicism that existed.

莱克斯·弗里德曼：

这标志着计算机视觉领域的核心人物聚集在一起，然后突然间出现了转变。光有思想和计算能力还不够，必须克服存在的怀疑态度。

That's interesting that people just didn't believe for a couple of decades.

有趣的是，人们在几十年里就是不相信。

Ilya Sutskever:

Yeah, well, but it's more than that. It's kind of—when put this way, it sounds like, well, you know, those silly people who didn't believe what was missing.

伊利亚·苏茨克维尔：

是的，但不止如此。这样说起来好像是那些不相信的人有点愚蠢，没能看出关键所在。

But in reality, things were confusing because neural networks really did not work on anything, and they were not the best method on pretty much anything as well. And it was pretty rational to say, yeah, this stuff doesn't have any traction.

但实际上，事情是混乱的，因为神经网络确实在任何任务上都没有奏效，它们几乎不是任何问题的最佳方法。说“这种东西没有发展潜力”是相当理性的。

And that's why you need to have these very hard tasks, which produce undeniable evidence. And that's how we make progress.

这就是为什么需要那些非常艰难的任务，它们能产生不可否认的证据。这样才能推动进步。

And that's why the field is making progress today, because we have these hard benchmarks, which represent true progress. And this is why we are able to avoid endless debate.

这也是为什么今天这个领域在不断进步，因为我们有了这些艰难的基准，它们代表了真正的进步。这也是为什么我们能够避免无休止的争论。

Lex Fridman:

So, incredibly, you've contributed some of the biggest recent ideas in AI, in computer vision, language, natural language processing, reinforcement learning, sort of everything in between. Maybe not GANs.

莱克斯·弗里德曼：

不可思议的是，你为AI的许多最新重大想法做出了贡献，包括计算机视觉、语言、自然语言处理、强化学习，以及它们之间的一切。也许除了GANs。

There may not be a topic you haven't touched. And of course, the fundamental science of deep learning. What is the difference to you between vision, language, and as in reinforcement learning, action, as learning problems?

可能没有你没有涉足过的主题。当然，还有深度学习的基础科学。对于你来说，视觉、语言和强化学习中的行动作为学习问题，有什么区别？

And what are the commonalities? Do you see them as all interconnected? Are they fundamentally different domains that require different approaches?

它们之间的共性是什么？你认为它们是相互关联的吗？还是它们是需要不同方法的根本不同的领域？

Ilya Sutskever:

Okay, that's a good question. Machine learning is a field with a lot of unity, a huge amount of unity.

伊利亚·苏茨克维尔：

好的，这是个好问题。机器学习是一个具有高度统一性的领域，非常统一。

Lex Fridman:

What do you mean by unity? Like overlap of ideas?

莱克斯·弗里德曼：

你所说的统一是什么意思？是指思想的重叠吗？

Ilya Sutskever:

Overlap of ideas, overlap of principles. In fact, there's only one or two or three principles, which are very, very simple.

伊利亚·苏茨克维尔：

是的，是思想的重叠，也是原则的重叠。实际上，只有一两个或者三个非常简单的原则。

And then they apply in almost the same way, in almost the same way to the different modalities of the different problems.

这些原则几乎以同样的方式适用于不同问题的不同模态。

And that's why today, when someone writes a paper on improving optimization of deep learning in vision, it improves the different NLP applications and it improves the different reinforcement learning applications.

这就是为什么今天有人写了一篇关于改进视觉中深度学习优化的论文，它不仅改善了不同的自然语言处理（NLP）应用，也改善了不同的强化学习（RL）应用。

So I would say that computer vision and NLP are very similar to each other. Today they differ in that they have slightly different architectures. We use transformers in NLP and we use convolutional neural networks in vision.

所以我会说，计算机视觉和NLP非常相似。它们的区别在于架构略有不同。在NLP中我们使用Transformer，而在视觉中我们使用卷积神经网络（CNN）。

But it's also possible that one day this will change and everything will be unified with a single architecture.

但也有可能有一天这会改变，一切都将以单一架构统一起来。

Because if you go back a few years ago in natural language processing, there were a huge number of architectures for every different tiny problem had its own architecture. Today, there's just one transformer for all those different tasks.

因为如果回顾几年前的自然语言处理，每个小问题都有自己独特的架构。而今天，所有这些不同的任务只用一种Transformer架构。

And if you go back in time even more, you had even more and more fragmentation and every little problem in AI had its own little subspecialization and sub—

you know, little set of collection of skills, people who would know how to engineer the features. Now it's all been subsumed by deep learning. You have this unification.

再往前看，AI领域中每个小问题都有自己专门化的子领域和技能集，比如如何设计特征。而现在，这一切都被深度学习取代，实现了统一。

And so I expect vision to become unified with natural language as well.

因此，我预计视觉将与自然语言也实现统一。

Or rather I shouldn’t say expect, I think it’s possible. I don’t want to be too sure because I think the convolutional neural net is very computationally efficient. RL is different.

或者说，我不应该说“预计”，而应该说“可能”。我不想太过肯定，因为卷积神经网络在计算上非常高效。而强化学习是不同的。

RL does require slightly different techniques because you really do need to take action. You really do need to do something about exploration. Your variance is much higher. But I think there is a lot of unity even there.

强化学习确实需要稍微不同的技术，因为你需要采取行动，你需要处理探索的问题，方差会更高。但即使在这里，我也认为存在很大的统一性。

And I would expect, for example, that at some point there will be some broader unification between RL and supervised learning where somehow the RL will be making decisions to make the supervised learning go better, and there will be—

我预计，比如，某个时候强化学习和监督学习之间会出现更广泛的统一，强化学习会以某种方式做出决策来改进监督学习，最终会有——

I imagine one big black box and you just shovel things into it and it just figures out what to do with whatever you shovel in it.

我设想一个大黑箱，你把东西放进去，它会自动决定如何处理你输入的任何东西。

Lex Fridman:

Reinforcement learning has some aspects of language and vision combined almost. There's elements of a long-term memory that you should be utilizing and there's elements of a really rich sensory space.

莱克斯·弗里德曼：

强化学习几乎结合了语言和视觉的一些方面。它包含了应该利用的长期记忆元素，以及一个非常丰富的感官空间。

So it seems like the—it's like the union of the two or something like that.

所以这看起来像是两者的结合，或者类似的东西。

Ilya Sutskever:

I'd say something slightly differently. I'd say that reinforcement learning is neither, but it naturally interfaces and integrates with the two of them.

伊利亚·苏茨克维尔：

我会稍微不同地表达。我会说，强化学习既不是语言，也不是视觉，但它自然地与二者交互并融合。

Lex Fridman:

You think action is fundamentally different? So yeah, what is interesting about, what is unique about policy of learning to act?

莱克斯·弗里德曼：

你认为行动本质上是不同的吗？那么，关于学习行动的策略，有什么有趣的或独特的地方？

Ilya Sutskever:

Well, so one example, for instance, is that when you learn to act, you are fundamentally in a non-stationary world. Because as your actions change, the things you see start changing. You experience the world in a different way.

伊利亚·苏茨克维尔：

举个例子，当你学习如何行动时，你本质上处于一个非平稳的世界中。因为随着你的行动变化，你所看到的东西也会开始变化。你以不同的方式体验这个世界。

And this is not the case for the more traditional static problem where you have some distribution and you just apply a model to that distribution.

而在更传统的静态问题中，情况并非如此。在那些问题中，你有一个分布，你只是将模型应用于该分布。

Lex Fridman:

You think it's a fundamentally different problem, or is it just a more difficult generalization of the problem of understanding?

莱克斯·弗里德曼：

你认为这是一个本质上不同的问题，还是只是理解问题的一个更难的泛化？

Ilya Sutskever:

I mean, it's a question of definitions almost. There is a huge amount of commonality for sure. You take gradients, you try to approximate gradients in both cases.

伊利亚·苏茨克维尔：

我的意思是，这几乎是一个定义的问题。当然，二者之间有大量的共性。你在两种情况下都会计算梯度，尝试近似梯度。

In the case of reinforcement learning, you have some tools to reduce the variance of the gradients. You do that. There's lots of commonality. You use the same neural net in both cases. You compute the gradient, you apply Adam in both cases.

在强化学习中，你有一些工具可以降低梯度的方差。你会这样做。二者之间有很多共性。你在两种情况下都使用相同的神经网络，都计算梯度，都使用Adam优化。

So, I mean, there's lots in common for sure, but there are some small... differences which are not completely insignificant.

所以，我的意思是，确实有很多共性，但也有一些不完全微不足道的小差异。

It's really just a matter of your point of view, what frame of reference, how much do you want to zoom in or out as you look at these problems.

这实际上取决于你的视角，你的参考框架，以及你在看这些问题时希望放大或缩小多少。

Lex Fridman:

Which problem do you think is harder? So people like Noam Chomsky believe that language is fundamental to everything. So it underlies everything. Do you think language understanding is harder than visual scene understanding or vice versa?

莱克斯·弗里德曼：

你认为哪个问题更难？像诺姆·乔姆斯基这样的人认为语言是万物的基础，支撑着一切。你认为语言理解比视觉场景理解更难，还是相反？

Ilya Sutskever:

I think that asking if a problem is hard is slightly wrong. I think the question is a little bit wrong, and I want to explain why. So what does it mean for a problem to be hard?

伊利亚·苏茨克维尔：

我认为问一个问题是否困难，这个问题本身有点不对。我觉得这个问题有点偏颇，我想解释一下为什么。那么，问题“困难”到底是什么意思呢？

Lex Fridman:

Okay, the non-interesting dumb answer to that is there's a benchmark, and there's a human-level performance on that benchmark, and how is the effort required to reach the human-level benchmark.

莱克斯·弗里德曼：

好吧，乏味且简单的答案是，有一个基准测试，有一个在基准测试上的人类水平表现，而达到这一人类水平需要多少努力。

Ilya Sutskever:

So from the perspective of how much until we get to human level on a very good benchmark. Yeah, I understand what you mean by that.

伊利亚·苏茨克维尔：

从达到某个优秀基准上人类水平的角度来看，我明白你指的是什么。

So what I was going to say is that a lot of it depends on, you know, once you solve a problem, it stops being hard. And that's always true.

我想说的是，这在很大程度上取决于，一旦你解决了一个问题，它就不再是难题了。这总是如此。

And so whether something is hard or not depends on what our tools can do today.

因此，某个问题是否困难，取决于我们今天的工具能够做到什么。

So, you know, you say today, through human level, language understanding and visual perception are hard in the sense that there is no way of solving the problem completely in the next three months. So I agree with that statement.

所以，今天你说，达到人类水平的语言理解和视觉感知是困难的，因为无法在接下来的三个月内完全解决这个问题。对此我同意。

Beyond that, my guess would be as good as yours. I don't know.

除此之外，我的猜测和你的差不多。我也不知道。

Lex Fridman:

Okay, so you don't have a fundamental intuition about how hard language understanding is.

莱克斯·弗里德曼：

好的，所以你对语言理解的难度没有根本的直觉？

Ilya Sutskever:

I know, I changed my mind. I'd say language is probably going to be harder. I mean, it depends on how you define it. Like if you mean absolute, top-notch, 100% language understanding, I'll go with language.

伊利亚·苏茨克维尔：

我改变主意了。我认为语言可能会更难。这取决于你如何定义。如果指的是绝对顶尖、完全无懈可击的语言理解，我会选择语言更难。

But then if I show you a piece of paper with letters on it, you see what I mean? You have a vision system. You say it's the best human-level vision system. I show you, I open a book, and I show you letters.

但是，如果我给你看一张写有字母的纸，你明白我的意思吗？你有一个视觉系统，你说这是最好的接近人类水平的视觉系统。我打开一本书，给你看字母。

Will it understand how these letters form into words and sentences and meaning? Is this part of the vision problem? Where does vision end and language begin?

它能理解这些字母如何组成单词、句子并表达意义吗？这算是视觉问题的一部分吗？视觉从哪里结束，语言从哪里开始？

Lex Fridman:

Yeah, so Chomsky would say it starts at language.

莱克斯·弗里德曼：

是的，乔姆斯基会说，这从语言开始。

So vision is just a little example of the kind of a structure and, you know, a fundamental hierarchy of ideas that's already represented in our brain somehow that's represented through language.

所以，视觉只是一个小例子，展示了一种结构，以及某种已在我们大脑中通过语言表达的基本思想层次。

But where does vision stop and language begin? That's a really interesting question.

但是，视觉在哪里停止，语言从哪里开始？这是一个非常有趣的问题。

So one possibility is that it's impossible to achieve really deep understanding in either images or language without basically using the same kind of system. So you're going to get the other for free.

一种可能性是，如果不使用相同类型的系统，就不可能真正深入理解图像或语言。因此，如果获得了一个，另一个可能也会随之得到。

Ilya Sutskever:

I think it's pretty likely that yes, if we can get one, our machine learning is probably that good that we can get the other. But I'm not 100% sure. And also, I think a lot of it really does depend on your definitions.

伊利亚·苏茨克维尔：

我认为很有可能是这样，如果我们能搞定一个，那么我们的机器学习可能已经足够好，能够搞定另一个。但我并不是百分之百确定。此外，我认为这很大程度上取决于你的定义。

...definitions of like perfect vision. Because, you know, reading is vision, but should it count?

……比如完美视觉的定义。因为你知道，阅读是视觉的一部分，但这应该算吗？

Lex Fridman:

Yeah, to me, so my definition is if a system looked at an image and then a system looked at a piece of text and then told me something about that and I was really impressed.

莱克斯·弗里德曼：

是的，对我来说，我的定义是，如果一个系统看了一张图像，然后又看了一段文本，接着告诉我一些关于它们的事情，并让我感到非常震撼。

Ilya Sutskever:

That's relative. You'll be impressed for half an hour, and then you're gonna say, well, I mean, all the systems do that, but here's the thing they don't do.

伊利亚·苏茨克维尔：

这很相对。你会感到震撼半个小时，然后你会说：“好吧，所有系统都能做到这一点，但有些事情它们还做不到。”

Lex Fridman:

Yeah, but I don't have that with humans. Humans continue to impress me.

莱克斯·弗里德曼：

是的，但我对人类不是这样。人类总是让我感到惊讶。

Ilya Sutskever:

Is that true?

伊利亚·苏茨克维尔：

真的吗？

Lex Fridman:

Okay, so I'm a fan of monogamy. So I like the idea of marrying somebody, being with them for several decades.

莱克斯·弗里德曼：

好吧，我是一夫一妻制的拥护者。我喜欢与某人结婚并相伴几十年的想法。

So I believe in the fact that yes, it's possible to have somebody continuously giving you pleasurable, interesting, witty, new ideas, friends. Yeah, I think so. They continue to surprise you.

所以我相信，确实可以有一个人持续为你带来愉悦、有趣、机智的新想法，成为朋友。是的，我认为这是可能的。他们会不断地让你感到惊喜。

The surprise, it's that injection of randomness.

这种惊喜，是随机性的注入。

It seems to be a nice source of, yeah, continued inspiration, like the wit, the humor. I think, yeah, that would be, it's a very subjective test, but I think if you have enough humans in the room.

这似乎是持续灵感的一个很好的来源，比如机智和幽默。我认为，是的，这是一个非常主观的测试，但如果你房间里有足够多的人类……

Ilya Sutskever:

Yeah, I understand what you mean. Yeah, I feel like I misunderstood what you meant by impressing you.

伊利亚·苏茨克维尔：

是的，我明白你的意思了。我觉得我之前误解了你所说的“让你感到震撼”的意思。

I thought you meant to impress you with its intelligence, with how well it understands an image.

我以为你是指用它的智能、它对图像的理解程度来让你感到震撼。

I thought you meant something like I'm gonna show it a really complicated image and it's gonna get it right and you're gonna say, wow, that's really cool. Our systems of, you know, January 2020 have not been doing that.

我以为你是指，我给它看一张非常复杂的图像，它能正确理解，你会说：“哇，这真酷。” 我们2020年1月的系统还没有做到这一点。

Lex Fridman:

Yeah. No, I think it all boils down to like the reason people click like on stuff on the internet, which is like it makes them laugh. So it's like humor or wit or insight.

莱克斯·弗里德曼：

是的，不。我认为这归根结底在于，人们在互联网上点击“赞”的原因——就是让他们发笑。所以这涉及幽默、机智或洞察力。

Ilya Sutskever:

I'm sure we'll get that as well.

伊利亚·苏茨克维尔：

我相信我们也会做到这一点。

Lex Fridman:

So forgive the romanticized question, but looking back to you, what is the most beautiful or surprising idea in deep learning or AI in general you've come across?

莱克斯·弗里德曼：

原谅我问一个浪漫化的问题，但回顾你的经历，你认为深度学习或人工智能领域中你遇到的最美或最令人惊讶的想法是什么？

Ilya Sutskever:

So I think the most beautiful thing about deep learning is that it actually works. And I mean it because you got these ideas, you got the little neural network, you got the backpropagation algorithm.

伊利亚·苏茨克维尔：

我认为深度学习最美妙的地方在于它确实有效。我的意思是，你有这些想法，有小型神经网络，有反向传播算法。

And then you got some theories as to, you know, this is kind of like the brain. So maybe if you make it large, if you make the neural network large and you train it on a lot of data, then it will do the same function that the brain does.

然后你会有一些理论，比如“这有点像大脑”。所以也许如果你把神经网络做大，并在大量数据上训练它，那么它就会执行与大脑相同的功能。

And it turns out to be true. That's crazy. And now we just train these neural networks and you make them larger and they keep getting better. And I find it unbelievable.

结果证明这是正确的。这太疯狂了。现在我们训练这些神经网络，只要把它们做得更大，它们就会变得更好。这让我觉得难以置信。

I find it unbelievable that this whole AI stuff with neural networks works.

我觉得整个神经网络的人工智能领域能奏效实在不可思议。

Lex Fridman:

Have you built up an intuition of why? Are there little bits and pieces of intuitions of insights of why this whole thing works?

莱克斯·弗里德曼：

你对为什么会奏效有直觉吗？是否有一些关于为什么整个事情奏效的零星直觉和洞察？

Ilya Sutskever:

I mean, some definitely, well, we know that optimization, we now have good—you know, we've had lots of empirical, you know, huge amounts of empirical reasons to believe that optimization should work on most problems we care about.

伊利亚·苏茨克维尔：

当然有一些。我们知道优化，现在我们有了很好的、经过大量实验证明的理由，相信优化在我们关心的大多数问题上应该是有效的。

Lex Fridman:

Do you have insights of what—so you just said empirical evidence—is most of your sort of empirical evidence kind of convinces you. It's like evolution is empirical.

莱克斯·弗里德曼：

你有没有关于这个的洞见——你刚才提到实验证据——你的大部分说服来自实验证据吗？这有点像进化是基于经验的。

It shows you that, look, this evolutionary process seems to be a good way to design organisms that survive in their environment. But it doesn't really get you to the insights of how the whole thing works.

它向你展示，这种进化过程似乎是一种设计能在环境中生存的生物体的好方法。但它并不能真正让你理解整个过程是如何运作的。

Ilya Sutskever:

I think a good analogy is physics. You know how you say, hey, let's do some physics calculation and come up with some new physics theory and make some prediction. But then you got to run the experiment.

伊利亚·苏茨克维尔：

我认为一个很好的类比是物理学。你知道，当你说，“嘿，让我们做一些物理计算，提出一些新的物理理论并做出一些预测。” 但随后你需要进行实验验证。

You know, you got to run the experiment. It's important. So it's a bit the same here, except that maybe sometimes the experiment came before the theory. But it still is the case.

你知道，必须进行实验验证，这很重要。这里有点类似，只不过有时实验可能先于理论提出。但情况仍然如此。

You know, you have some data and you come up with some prediction. You say, yeah, let's make a big neural network. Let's train it and it's going to work.

你有一些数据，然后你提出一些预测。你说：“是的，让我们做一个大神经网络，训练它，它就会奏效。”

Much better than anything before it and it will in fact continue to get better as you make it larger. And it turns out to be true. That's amazing when a theory is validated like this. It's not a mathematical theory.

比之前的任何东西都好，并且事实上它会随着规模的增大而变得更好。而结果证明这是对的。当一个理论像这样被验证时，实在令人惊叹。这并不是一个数学理论。

It's more of a biological theory almost. So I think there are not terrible analogies between deep learning and biology. I would say it's like the geometric mean of biology and physics. That's deep learning.

它更像是一个生物学理论。所以，我认为深度学习与生物学之间有不错的类比。我会说，深度学习是生物学和物理学的几何平均。

Lex Fridman:

The geometric mean of biology and physics. I think I'm gonna need a few hours to wrap my head around that. Just to find the geometric, just to find the set of what biology represents.

莱克斯·弗里德曼：

生物学和物理学的几何平均。我想我需要几个小时来消化这一点。仅仅是为了弄清楚生物学所代表的范围。

Ilya Sutskever:

Well, in biology, things are really complicated. It's really hard to have good predictive theory. In physics, the theories are too good.

伊利亚·苏茨克维尔：

嗯，在生物学中，事情非常复杂，很难有好的预测理论。而在物理学中，理论又太精准了。

In physics, people make these super precise theories, which make these amazing predictions. And in machine learning, we're kind of in between.

在物理学中，人们提出了这些超精准的理论，能做出惊人的预测。而在机器学习中，我们介于两者之间。

Lex Fridman:

Kind of in between, but it'd be nice if machine learning somehow helped us discover the unification of the two as opposed to sort of the in between. But you're right, you're kind of trying to juggle both.

莱克斯·弗里德曼：

确实介于两者之间，但如果机器学习能帮助我们发现生物学和物理学的统一，而不仅仅是停留在两者之间，那就太好了。但你说得对，我们确实在努力平衡这两方面。

So do you think there's still beautiful and mysterious properties in neural networks that are yet to be discovered?

那么，你认为神经网络中是否仍有美丽而神秘的属性尚待发现？

Ilya Sutskever:

Definitely. I think that we are still massively underestimating deep learning.

伊利亚·苏茨克维尔：

绝对如此。我认为我们仍然严重低估了深度学习。

Lex Fridman:

What do you think it'll look like? Like what...

莱克斯·弗里德曼：

你觉得它会是什么样子？比如……

Ilya Sutskever:

If I knew, I would have done it, yeah. But if you look at all the progress from the past 10 years, I would say there have been a few cases where things that felt like really new ideas showed up,

伊利亚·苏茨克维尔：

如果我知道，我早就做了，是的。但如果你看看过去十年的所有进展，我会说有一些案例看起来确实有全新的想法出现，

but by and large, it was every year we thought, okay, deep learning goes this far. Nope, it actually goes further.

但总体而言，每年我们都会觉得，“好吧，深度学习就到这里了。” 不，实际上它走得更远了。

And then the next year, okay, now this is big deep learning, we are really done. Nope, it goes further.

接着下一年，“好吧，现在深度学习已经很强大了，我们应该到头了。” 不，它又更进一步了。

It just keeps going further each year. So that means that we keep underestimating, we keep not understanding it. It has surprising properties all the time.

它每年都在继续前进。这意味着我们不断低估它，不断无法完全理解它。它总是有令人惊讶的特性。

Lex Fridman:

Do you think it's getting harder and harder?

莱克斯·弗里德曼：

你认为这变得越来越难了吗？

Ilya Sutskever:

To make progress?

伊利亚·苏茨克维尔：

取得进展？

Lex Fridman:

Need to make progress?

莱克斯·弗里德曼：

需要取得进展？

Ilya Sutskever:

It depends on what you mean. I think the field will continue to make very robust progress for quite a while.

伊利亚·苏茨克维尔：

这取决于你的意思。我认为这个领域在相当长一段时间内会继续取得稳健的进展。

I think for individual researchers, especially people who are doing research, it can be harder because there is a very large number of researchers right now.

我认为对个别研究人员来说，尤其是那些从事研究的人，这可能会更难，因为目前研究人员的数量非常多。

I think that if you have a lot of compute, then you can make a lot of very interesting discoveries, but then you have to deal with the challenge of managing a huge compute cluster to run your experiments. It's a little bit harder.

我认为如果你有大量的计算资源，那么你可以做出很多非常有趣的发现，但你也必须面对管理一个巨大的计算集群来运行实验的挑战。这有点更难了。

Lex Fridman:

So I'm asking all these questions that nobody knows the answer to, but you're one of the smartest people I know, so I'm going to keep asking.

莱克斯·弗里德曼：

我在问一些没人知道答案的问题，但你是我认识的最聪明的人之一，所以我会继续问下去。

So let's imagine all the breakthroughs that happen in the next 30 years in deep learning. Do you think most of those breakthroughs can be done by one person with one computer?

那么，让我们想象未来30年中深度学习的所有突破。你认为大多数这些突破是否可以由一个人用一台电脑完成？

Sort of in the space of breakthroughs, do you think compute and large efforts will be necessary?

在突破的领域中，你认为计算能力和大规模努力是必要的吗？

Ilya Sutskever:

I mean, I can't be sure. When you say one computer, you mean how large?

伊利亚·苏茨克维尔：

我的意思是，我不能确定。你说一台电脑，是指多大规模？

Lex Fridman:

You're clever. I mean, one GPU.

莱克斯·弗里德曼：

你真聪明。我是指一块GPU。

Ilya Sutskever:

I see. I think it's pretty unlikely. I think it's pretty unlikely. I think that the stack of deep learning is starting to be quite deep.

伊利亚·苏茨克维尔：

我明白了。我认为这非常不可能。我认为深度学习的技术栈已经开始变得相当深。

If you look at it, you've got all the way from the ideas, the systems to build the data sets, the distributed programming, the building the actual cluster, the GPU programming, putting it all together.

如果你仔细看，从想法、系统，到构建数据集、分布式编程、实际集群的搭建、GPU编程，将这一切整合在一起。

So now the stack is getting really deep and I think it becomes, it can be quite hard for a single person to become, to be world class in every single layer of the stack.

所以现在整个技术栈变得非常深，我认为一个人要在每一层都达到世界级水平可能相当困难。

Lex Fridman:

What about the, what like Vladimir Vapnik really insists on is taking MNIST and trying to learn from very few examples. So being able to learn more efficiently.

莱克斯·弗里德曼：

那么像弗拉基米尔·瓦普尼克特别坚持的，通过MNIST并尝试从很少的样本中学习。这种更高效的学习呢？

Do you think there'll be breakthroughs in that space that may not need a huge compute?

你认为在这一领域会有突破，而这些突破可能不需要大量的计算能力吗？

Ilya Sutskever:

I think there will be a large number of breakthroughs in general that will not need a huge amount of compute. So maybe I should clarify that.

伊利亚·苏茨克维尔：

我认为总体上会有许多突破，而这些突破不需要大量的计算能力。也许我应该澄清一下。

I think that some breakthroughs will require a lot of compute.

我认为有些突破确实需要大量计算能力。

And I think building systems which actually do things will require a huge amount of compute. That one is pretty obvious.

而且我认为构建实际能够执行任务的系统将需要大量的计算能力。这一点很明显。

If you want to do X and X requires a huge neural net, you gotta get a huge neural net.

如果你想做某个任务，而这个任务需要一个巨大的神经网络，那么你就得构建一个巨大的神经网络。

But I think there will be lots of—I think there is lots of room for very important work being done by small groups and individuals.

但我认为会有很多——我认为还有很多空间可以由小团队和个人完成非常重要的工作。

Lex Fridman:

Can you maybe sort of on the topic of the science of deep learning, talk about one of the recent papers that you've released, The Deep Double Descent, where bigger models and more data hurt? I think it's a really interesting paper.

莱克斯·弗里德曼：

关于深度学习科学的话题，你能否谈谈你最近发表的一篇论文《深度双重下降》，其中提到更大的模型和更多的数据有时会带来负面影响？我觉得这是一篇非常有趣的论文。

Can you describe the main idea?

你能描述一下主要思想吗？

Ilya Sutskever:

Yeah, definitely. So what happened is that over the years, some small number of researchers noticed that it is kind of weird that when you make the neural network larger, it works better, and it seems to go in contradiction with statistical ideas.

伊利亚·苏茨克维尔：

当然可以。事情是这样的，多年来，少数研究人员注意到，当你将神经网络做大时，它表现得更好，这似乎与统计学的观点相矛盾。

And then some people made an analysis showing that actually you got this double descent bump.

然后，有人进行了一些分析，显示实际上存在这种“双重下降”的峰值现象。

And what we've done was to show that double descent occurs for pretty much all practical deep learning systems.

而我们的研究表明，这种双重下降现象几乎出现在所有实际的深度学习系统中。

Lex Fridman:

What's the x-axis and the y-axis of a double descent plot?

莱克斯·弗里德曼：

双重下降图的x轴和y轴分别是什么？

Ilya Sutskever:

Okay, great. So, you can do things like, you can take a neural network and you can start increasing its size slowly while keeping your dataset fixed.

伊利亚·苏茨克维尔：

好的，很棒。比如，你可以取一个神经网络，并在保持数据集固定的情况下，慢慢增加它的规模。

So if you increase the size of the neural network slowly, and if you don't do early stopping—that's a pretty important detail—then when the neural network is really small, you make it larger, you get a very rapid increase in performance.

如果你慢慢增加神经网络的规模，并且不进行早停（这是一个非常重要的细节），那么当神经网络非常小时，你把它做大，会看到性能快速提升。

Then you continue to make it larger and at some point performance will get worse. And it gets the worst exactly at the point at which it achieves zero training error, precisely zero training loss.

然后你继续增大它的规模，性能在某个点会变差。而且性能最差的点恰好是训练误差为零、训练损失完全为零的时刻。

And then as you make it large, it starts to get better again. And it's kind of counterintuitive because you'd expect deep learning phenomena to be monotonic.

接着，当你进一步增大规模，性能又开始变好。这有点违反直觉，因为你会期望深度学习的现象是单调的。

Lex Fridman:

Do you have an intuition of why?

莱克斯·弗里德曼：

你对为什么会这样有直觉吗？

Ilya Sutskever:

It’s hard to be sure what it means, but it also occurs in the case of linear classifiers.

伊利亚·苏茨克维尔：

很难确定这意味着什么，但这种现象也出现在线性分类器中。

And the intuition basically boils down to the following: when you have a large dataset and a small model, then small, tiny, random...

直觉上，可以归结为以下几点：当你有一个大数据集和一个小模型时，一些微小的随机……

So, basically, what is overfitting? Overfitting is when your model is somehow very sensitive to the small, random, unimportant stuff in your dataset.

基本上，什么是过拟合？过拟合是指你的模型对数据集中那些微小的、随机的、不重要的内容非常敏感。

Lex Fridman:

In the training data.

莱克斯·弗里德曼：

在训练数据中。

Ilya Sutskever:

In the training data set, precisely. So, if you have a small model and you have a big data set, and there may be some random thing, you know, some training cases are randomly in the data set and others may not be there.

伊利亚·苏茨克维尔：

没错，在训练数据集中。如果你有一个小模型和一个大数据集，那么可能会有一些随机情况，比如某些训练样本随机地在数据集中，而其他样本可能不在其中。

But the small model is kind of insensitive to this randomness because there is pretty much no uncertainty about the model when the data set is large.

但小模型对这些随机性相对不敏感，因为当数据集很大时，模型几乎没有不确定性。

Lex Fridman:

So, okay, so at the very basic level, to me, it is the most surprising thing that neural networks don't overfit every time very quickly before ever being able to learn anything. There are a huge number of parameters.

莱克斯·弗里德曼：

所以，好的，从最基本的层面来说，我觉得最令人惊讶的是，神经网络没有每次都在还没学到任何东西之前就迅速过拟合。参数数量是巨大的。

Ilya Sutskever:

So here is—so there is one way, okay, so maybe, so let me try to give the explanation. Maybe that will work.

伊利亚·苏茨克维尔：

那么，这里有一种解释方式，好吧，让我试着解释一下，或许这会奏效。

So you got a huge neural network. Let's suppose you've got a—you have a huge neural network, you have a huge number of parameters. Now let's pretend everything is linear, which is not, let's just pretend.

你有一个巨大的神经网络。假设你有一个——一个大规模的神经网络，参数数量非常多。现在我们假设一切都是线性的（虽然实际不是），我们暂且这样假设。

Then there is this big subspace where your neural network achieves zero error. And SGD is going to find approximately the point, that's right, approximately the point with the smallest norm in that subspace.

那么，在这个大子空间中，你的神经网络可以达到零误差。而随机梯度下降（SGD）会大致找到这个子空间中具有最小范数的点。

And that can also be proven to be insensitive to the small randomness in the data when the dimensionality is high.

而且可以证明，当维度很高时，这个点对数据中的微小随机性是不敏感的。

But when the dimensionality of the data is equal to the dimensionality of the model, then there is a one-to-one correspondence between all the data sets and the models.

但是，当数据的维度等于模型的维度时，所有数据集和模型之间会存在一对一的对应关系。

So small changes in the data set actually lead to large changes in the model, and that's why performance gets worse. So this is the best explanation more or less.

所以，数据集中的微小变化实际上会导致模型的巨大变化，这就是为什么性能会变差。大致来说，这就是最佳解释。

Lex Fridman:

So then it would be good for the model to have more parameters, to be bigger than the data.

莱克斯·弗里德曼：

所以，模型的参数比数据量多、模型更大是好事，对吧？

Ilya Sutskever:

That's right. But only if you don't early stop. If you introduce early stopping in your regularization, you can make the double descent bump almost completely disappear.

伊利亚·苏茨克维尔：

没错。但前提是你不要使用早停（early stopping）。如果在正则化中引入早停，你几乎可以让“双重下降”的峰值完全消失。

Lex Fridman:

What is early stop?

莱克斯·弗里德曼：

什么是早停？

Ilya Sutskever:

Early stopping is when you train your model and you monitor your validation performance. And then if at some point validation performance starts to get worse, you say, okay, let's stop training. We're good. We're good enough.

伊利亚·苏茨克维尔：

早停是指在训练模型时监控验证集的性能。如果在某个时刻验证性能开始变差，你就会说：“好吧，停止训练了，够好了。”

Lex Fridman:

So the magic happens after that moment, so you don't want to do the early stopping.

莱克斯·弗里德曼：

所以神奇的事情发生在那个时刻之后，所以你不想使用早停。

Ilya Sutskever:

Well, if you don't do the early stopping, you get the very pronounced double descent.

伊利亚·苏茨克维尔：

嗯，如果你不早停，就会出现非常明显的“双重下降”现象。

Lex Fridman:

Do you have any intuition why this happens?

莱克斯·弗里德曼：

你对为什么会发生这种情况有直觉吗？

Ilya Sutskever:

Double descent? Oh, sorry, are you stopping?

伊利亚·苏茨克维尔：

双重下降？哦，抱歉，你还在问吗？

Lex Fridman:

No, the double descent, so the...

莱克斯·弗里德曼：

不是，我是问双重下降现象……

Ilya Sutskever:

Well, yeah, so I try, let's see. The intuition is basically this: when the data set has as many degrees of freedom as the model, then there is a one-to-one correspondence between them.

伊利亚·苏茨克维尔：

嗯，是的，我试着解释一下。直觉上，基本是这样的：当数据集的自由度和模型的自由度相等时，二者之间存在一对一的对应关系。

And so small changes to the data set lead to noticeable changes in the model. So your model is very sensitive to all the randomness. It is unable to discard it.

因此，数据集中的微小变化会导致模型的显著变化。你的模型对所有的随机性非常敏感，无法忽略这些随机性。

Whereas it turns out that when you have a lot more data than parameters or a lot more parameters than data, the resulting solution will be insensitive to small changes in the data set.

然而，当数据量远多于参数量，或者参数量远多于数据量时，最终的解对数据集中的微小变化就不再敏感了。

Lex Fridman:

So it's able to, let's nicely put, discard the small changes, the randomness.

莱克斯·弗里德曼：

所以它能够很好地去除微小变化和随机性。

Ilya Sutskever:

The spurious correlation which you don't want.

伊利亚·苏茨克维尔：

那些你不想要的虚假相关性。

Lex Fridman:

Jeff Hinton suggested we need to throw backpropagation away. We already kind of talked about this a little bit, but he suggested that we need to throw away backpropagation and start over.

莱克斯·弗里德曼：

杰夫·辛顿建议我们需要摒弃反向传播。我们已经稍微谈过这个了，但他提到我们需要抛弃反向传播，重新开始。

I mean, of course, some of that is a little bit wit and humor. But what do you think? What could be an alternative method of training neural networks?

我的意思是，当然，这其中有些只是机智和幽默。但你怎么看？是否有可能找到训练神经网络的替代方法？

Ilya Sutskever:

Well, the thing that he said precisely is that to the extent that you can't find backpropagation in the brain, it's worth seeing if we can learn something from how the brain learns.

伊利亚·苏茨克维尔：

他说的具体意思是：既然我们无法在大脑中找到反向传播，那么值得研究一下我们能否从大脑的学习方式中学到什么。

But backpropagation is very useful, and we should keep using it.

但反向传播非常有用，我们应该继续使用它。

Lex Fridman:

Oh, you're saying that once we discover the mechanism of learning in the brain or any aspects of that mechanism, we should also try to implement that in neural networks.

莱克斯·弗里德曼：

哦，你是说，一旦我们发现了大脑的学习机制或其中的某些方面，我们也应该尝试将其应用于神经网络。

Ilya Sutskever:

If it turns out that we can't find backpropagation in the brain.

伊利亚·苏茨克维尔：

如果最终证明我们无法在大脑中找到反向传播的话。

Lex Fridman:

If we can't find backpropagation in the brain. Well, so I guess your answer to that is backpropagation is pretty damn useful. So why are we complaining?

莱克斯·弗里德曼：

如果我们在大脑中找不到反向传播。那么，我猜你的答案是，反向传播非常有用。所以我们为什么要抱怨呢？

Ilya Sutskever:

I mean, I personally am a big fan of backpropagation. I think it's a great algorithm because it solves an extremely fundamental problem, which is finding a neural circuit subject to some constraints. And I don't see that problem going away.

伊利亚·苏茨克维尔：

我的意思是，我个人是反向传播的忠实粉丝。我认为这是一个伟大的算法，因为它解决了一个极其基础的问题，即在一些约束条件下找到一个神经回路。我认为这个问题不会消失。

So that's why I really, I think it's pretty unlikely that we'll have anything which is going to be dramatically different. It could happen, but I wouldn't bet on it right now.

所以，这就是为什么我认为我们很难有完全不同的东西。虽然有可能，但我现在不会赌这个。

Lex Fridman:

So let me ask a sort of big picture question. Do you think neural networks can be made to reason?

莱克斯·弗里德曼：

让我问一个宏观的问题。你认为神经网络可以被用来推理吗？

Ilya Sutskever:

Why not? Well, if you look, for example, at AlphaGo or AlphaZero, the neural network of AlphaZero plays Go, which we all agree is a game that requires reasoning, better than 99.9% of all humans.

伊利亚·苏茨克维尔：

为什么不呢？例如，如果你看AlphaGo或AlphaZero，AlphaZero的神经网络下围棋——这是一款我们都同意需要推理的游戏——比99.9%的人类都下得更好。

Just the neural network, without the search, just the neural network itself. Doesn't that give us an existence proof that neural networks can reason?

仅仅是神经网络，不带搜索，仅仅是神经网络本身。这难道不是神经网络能够推理的一个存在性证明吗？

Lex Fridman:

To push back and disagree a little bit, we all agree that Go is reasoning. I think I agree. I don't think it's trivial.

莱克斯·弗里德曼：

稍微反驳一下，我们都同意围棋涉及推理。我认为我同意这一点，我不觉得这很简单。

So obviously, reasoning like intelligence is a loose gray area term a little bit. Maybe you disagree with that.

显然，推理和智能一样，是一个略显模糊的灰色术语。也许你不同意这一点。

But yes, I think it has some of the same elements of reasoning. Reasoning is almost like akin to search, right?

但我认为它确实有一些推理的要素。推理几乎类似于搜索，对吧？

There's a sequential element of stepwise consideration of possibilities and sort of building on top of those possibilities in a sequential manner until you arrive at some insight.

推理具有一种顺序性的元素，逐步考虑可能性，并以这种顺序性的方式在这些可能性之上逐步构建，直到你获得某种洞察力。

So yeah, I guess playing Go is kind of like that. And when you have a single neural network doing that without search, it's kind of like that.

所以，是的，我想下围棋有点像这样。当你有一个单一的神经网络在没有搜索的情况下做到这一点时，也有点像这样。

So there's an existence proof in a particular constrained environment that a process akin to what many people call reasoning exists, but more general kind of reasoning. So off the board.

所以，在一个特定受限的环境中，有一个存在性证明，表明存在一种类似于许多人所称的推理的过程。但更广泛的推理呢？比如棋盘外的情况。

Ilya Sutskever:

There is one other existence proof.

伊利亚·苏茨克维尔：

还有另一个存在性证明。

Lex Fridman:

Oh boy, which one? Us humans?

莱克斯·弗里德曼：

哦天啊，哪个？我们人类吗？

Ilya Sutskever:

Yes. Okay.

伊利亚·苏茨克维尔：

是的。好吧。

Lex Fridman:

All right. Do you think the architecture that will allow neural networks to reason will look similar to the neural network architectures we have today?

莱克斯·弗里德曼：

好的。你认为能够让神经网络进行推理的架构会与我们今天的神经网络架构类似吗？

Ilya Sutskever:

I think it will. I think, well, I don't want to make overly definitive statements.

伊利亚·苏茨克维尔：

我认为会。我觉得……嗯，我不想下过于确定的结论。

I think it's definitely possible that the neural networks that will produce the reasoning breakthroughs of the future will be very similar to the architectures that exist today.

我认为，未来实现推理突破的神经网络很可能与今天的架构非常相似。

Maybe a little bit more recurrent, maybe a little bit deeper, but these neural nets are so insanely powerful. Why wouldn't they be able to learn to reason?

可能会更具循环性，可能会更深一些，但这些神经网络已经非常强大。为什么它们不能学会推理呢？

Humans can reason, so why can't neural networks?

人类可以推理，为什么神经网络不可以？

Lex Fridman:

So do you think the kind of stuff we've seen neural networks do is a kind of just weak reasoning? So it's not a fundamentally different process? Again, this is stuff nobody knows the answer to.

莱克斯·弗里德曼：

所以，你认为我们目前看到神经网络所做的事情只是某种弱推理吗？它不是一个完全不同的过程？当然，这些问题没人知道答案。

Ilya Sutskever:

So when it comes to our neural networks, the thing which I would say is that neural networks are capable of reasoning.

伊利亚·苏茨克维尔：

对于我们的神经网络，我想说的是，神经网络是有能力进行推理的。

But if you train a neural network on a task which doesn't require reasoning, it's not going to reason.

但如果你让神经网络训练一个不需要推理的任务，它就不会进行推理。

This is a well-known effect where the neural network will solve the problem that you pose in front of it in the easiest way possible.

这是一个众所周知的现象，即神经网络会以最简单的方式解决摆在它面前的问题。

Lex Fridman:

Right, that takes us to one of the brilliant sort of ways you've described neural networks, which is you've referred to neural networks as the search for small circuits and maybe general intelligence as the search for small programs.

莱克斯·弗里德曼：

没错，这让我想起了你对神经网络的一种精彩描述：你将神经网络比喻为寻找小型电路，将通用智能比喻为寻找小型程序。

Which I found as a metaphor very compelling. Can you elaborate on that difference?

我觉得这个比喻非常有吸引力。你能详细阐述一下两者的区别吗？

Ilya Sutskever:

Yeah, so the thing which I said precisely was that if you can find the shortest program that outputs the data at your disposal, then you will be able to use it to make the best prediction possible.

伊利亚·苏茨克维尔：

好的，我具体说的是，如果你能找到生成手头数据的最短程序，那么你就能够用它做出最好的预测。

And that's a theoretical statement which can be proved mathematically.

这是一个可以用数学证明的理论陈述。

Now, you can also prove mathematically that finding the shortest program which generates some data is not a computable operation. No finite amount of compute can do this.

此外，你也可以用数学证明，找到生成某些数据的最短程序并不是一个可计算的操作。没有有限的计算能力可以做到这一点。

So then, with neural networks, neural networks are the next best thing that actually works in practice.

因此，对于神经网络来说，神经网络是实践中实际可行的次优选择。

We are not able to find the best, the shortest program which generates our data, but we are able to find, you know, a small, but now that statement should be amended, even a large circuit which fits our data in some way.

我们无法找到生成数据的最佳、最短程序，但我们能够找到——现在应该修正这个说法——甚至是一个较大的电路，它以某种方式适配我们的数据。

Lex Fridman:

Well, I think what you meant by the small circuit is the smallest needed circuit.

莱克斯·弗里德曼：

嗯，我认为你所说的小型电路是指最小需要的电路。

Ilya Sutskever:

Well, the thing which I would change now, back then I really haven't fully internalized the over-parameterized results.

伊利亚·苏茨克维尔：

嗯，我现在会改变这种说法，那时候我还没有完全吸收过参数化的研究结果。

The things we know about over-parameterized neural nets, now I would phrase it as a large circuit whose weights contain a small amount of information, which I think is what's going on.

我们现在对过参数化神经网络的了解是，我会将其表述为：一个大电路，其权重包含少量信息，我认为这就是实际情况。

If you imagine the training process of a neural network as you slowly transmit entropy from the data set to the parameters, then somehow the amount of information in the weights ends up being not very large,

如果你将神经网络的训练过程想象为将数据集的熵逐渐传递到参数中，那么权重中的信息量最终并不会很大，

which would explain why they generalize so well.

这可以解释为什么它们的泛化性能如此之好。

Lex Fridman:

So the large circuit might be one that's helpful for the generalization.

莱克斯·弗里德曼：

所以，大型电路可能有助于泛化。

Ilya Sutskever:

Yeah, something like this.

伊利亚·苏茨克维尔：

是的，类似这样的情况。

Lex Fridman:

Do you see it important to be able to try to learn something like programs?

莱克斯·弗里德曼：

你认为尝试学习类似程序的东西重要吗？

Ilya Sutskever:

I mean, if we can, definitely. I think the answer is kind of yes, if we can do it. We should do things that we can do.

伊利亚·苏茨克维尔：

我的意思是，如果我们能做到，那当然重要。我认为答案是肯定的，如果我们能做到，我们就应该做。

The reason we are pushing on deep learning, the fundamental reason, the root cause, is that we are able to train them.

我们推动深度学习的原因，根本原因，是因为我们能够训练它们。

So in other words, training comes first. We've got our pillar, which is the training pillar. And now we are trying to contort our neural networks around the training pillar.

换句话说，训练是第一位的。我们有了训练这一支柱。现在我们正在围绕这一训练支柱调整我们的神经网络。

We gotta stay trainable. This is an invariant we cannot violate.

我们必须保持可训练性。这是一个不能违反的不变量。

Lex Fridman:

And so, being trainable means starting from scratch, knowing nothing, you can actually pretty quickly converge towards knowing a lot.

莱克斯·弗里德曼：

所以，可训练性意味着从零开始，一无所知，你实际上可以很快收敛到掌握许多知识。

Ilya Sutskever:

Or even slowly. But it means that given the resources at your disposal, you can train the neural net and get it to achieve useful performance.

伊利亚·苏茨克维尔：

甚至可以是缓慢的。但这意味着，鉴于你拥有的资源，你可以训练神经网络，使其达到有用的性能。

Lex Fridman:

Yeah, that's a pillar we can't move away from.

莱克斯·弗里德曼：

是的，这是我们不能偏离的一根支柱。

Ilya Sutskever:

That's right, because if you say, hey, let's find the shortest program. Well, we can't do that. So it doesn't matter how useful that would be. We can't do it. So we won't.

伊利亚·苏茨克维尔：

没错，因为如果你说：“我们来找最短的程序吧。” 但是，我们做不到。所以不管那有多有用，我们也做不到。因此，我们不会去做。

Lex Fridman:

So do you think, you kind of mentioned that the neural networks are good at finding small circuits or large circuits. Do you think then the matter of finding small programs is just the data?

莱克斯·弗里德曼：

所以你提到神经网络擅长找到小型电路或大型电路。你认为找到小型程序的问题仅仅是数据的问题吗？

Ilya Sutskever:

No.

伊利亚·苏茨克维尔：

不是。

Lex Fridman:

Sorry, not the size or character, the type of data. Sort of ask giving it programs.

莱克斯·弗里德曼：

抱歉，我的意思不是数据的大小或特性，而是数据的类型。比如提供程序数据。

Ilya Sutskever:

Well, I think the thing is that right now, there are no good precedents of people successfully finding programs really well.

伊利亚·苏茨克维尔：

嗯，我认为目前没有人成功找到程序的好先例。

And so the way you'd find programs is you'd train a deep neural network to do it, basically. Which is the right way to go about it.

所以找到程序的方法基本上是训练一个深度神经网络来完成。这是正确的方法。

Lex Fridman:

But there's not good illustrations of that.

莱克斯·弗里德曼：

但目前没有这方面的好例子。

Ilya Sutskever:

It hasn't been done yet. But in principle, it should be possible.

伊利亚·苏茨克维尔：

确实还没有人做到。但原则上，这应该是可能的。

Lex Fridman:

Can you elaborate a little bit? What's your insight in principle? Put another way, you don't see why it's not possible.

莱克斯·弗里德曼：

你能详细说明一下吗？你在原则上的见解是什么？换句话说，你不认为这有什么不可能。

Ilya Sutskever:

Well, it's kind of like more—it's more a statement of I think that it's unwise to bet against deep learning.

伊利亚·苏茨克维尔：

嗯，这更像是这样一种表述——我认为低估深度学习是不明智的。

And if it's a cognitive function that humans seem to be able to do, then it doesn't take too long for some deep neural net to pop up that can do it too.

如果这是人类似乎能够完成的一种认知功能，那么不久之后就会出现某种能够实现这一功能的深度神经网络。

Lex Fridman:

Yeah, I'm there with you. I've stopped betting against neural networks at this point because they continue to surprise us.

莱克斯·弗里德曼：

是的，我同意你的看法。我现在已经不再低估神经网络了，因为它们不断给我们带来惊喜。

What about long-term memory? Can neural networks have long-term memory or something like knowledge bases?

那么长期记忆呢？神经网络能否拥有长期记忆或类似知识库的东西？

So being able to aggregate important information over long periods of time that would then serve as useful sort of representations of state that you can make decisions by.

也就是说，能够在长时间内聚合重要信息，从而形成有用的状态表示，以便进行决策。

So have a long-term context based on what you make in the decision.

这样就可以基于长期上下文来进行决策。

Ilya Sutskever:

So in some sense the parameters already do that. The parameters are an aggregation of the entirety of the neural net's experience, and so they count as long-term knowledge.

伊利亚·苏茨克维尔：

从某种意义上说，参数已经在做这件事了。参数是神经网络全部经验的聚合，因此它们可以算作长期知识。

And people have trained various neural nets to act as knowledge bases, and people have investigated language models as knowledge bases, so there is work there.

人们已经训练了各种神经网络来充当知识库，也有人研究了语言模型作为知识库的可能性，所以这方面是有研究的。

Lex Fridman:

Yeah, but in some sense. Do you think in every sense?

莱克斯·弗里德曼：

是的，但这是某种意义上的。那么在所有意义上都成立吗？

Do you think it's all just a matter of coming up with a better mechanism of forgetting the useless stuff and remembering the useful stuff?

你认为这是否只是找到更好的机制来忘记无用的信息并记住有用的信息的问题？

Because right now, I mean, there's not been mechanisms that do remember really long-term information.

因为现在，还没有能真正记住长期信息的机制。

Ilya Sutskever:

What do you mean by that precisely?

伊利亚·苏茨克维尔：

你具体指的是什么？

Lex Fridman:

Precisely, I like the word precisely. I'm thinking of the kind of compression of information the knowledge bases represent, sort of creating a...

莱克斯·弗里德曼：

具体来说，我喜欢“具体”这个词。我指的是知识库所代表的信息压缩，类似于创建一种……

Now, I apologize for my sort of human-centric thinking about what knowledge is, because neural networks aren't interpretable necessarily with the kind of knowledge they have discovered.

现在，我为我以人类为中心的知识定义方式感到抱歉，因为神经网络发现的知识不一定是可解释的。

But a good example for me is knowledge bases, being able to build up over time something like the knowledge that Wikipedia represents.

但对我来说，一个很好的例子是知识库，它能够随着时间积累类似于维基百科所代表的知识。

It's a really compressed, structured, knowledge base.

这是一种高度压缩、结构化的知识库。

Obviously not the actual Wikipedia or the language, but like a semantic web, the dream that semantic web represented.

显然，这并不是实际的维基百科或语言，而是类似语义网——语义网所代表的梦想。

So it's a really nice compressed knowledge base or something akin to that in a non-interpretable sense as neural networks would have.

所以这是一种非常好的压缩知识库，或者是神经网络以不可解释的方式拥有的类似东西。

Ilya Sutskever:

Well, the neural networks would be non-interpretable if you look at their weights, but their outputs should be very interpretable.

伊利亚·苏茨克维尔：

嗯，如果你查看神经网络的权重，它们可能是不可解释的，但它们的输出应该是非常可解释的。

Lex Fridman:

Okay, so yeah, how do you make very smart neural networks, like language models, interpretable?

莱克斯·弗里德曼：

好吧，那么，如何让非常智能的神经网络，比如语言模型，可解释？

Ilya Sutskever:

Well, you ask them to generate some text and the text will generally be interpretable.

伊利亚·苏茨克维尔：

嗯，你可以让它们生成一些文本，而这些文本通常是可以解释的。

Lex Fridman:

Do you find that the epitome of interpretability? Like, can you do better?

莱克斯·弗里德曼：

你觉得这就是解释性的极致吗？有没有更好的方法？

Like, can you, because you can't, okay, I'd like to know what does it know and what doesn't it know.

比如，我想知道它知道什么，不知道什么。

I would like the neural network to come up with examples where it's completely dumb and examples where it's completely brilliant.

我希望神经网络能够提供一些它完全愚蠢的例子，以及一些它完全出色的例子。

And the only way I know how to do that now is to generate a lot of examples and use my human judgment.

目前我知道的唯一方法是生成大量例子并依靠我的人类判断。

But it would be nice if the neural net had some self-awareness about it.

但如果神经网络能对此有一些自我意识，那就太好了。

Ilya Sutskever:

Yeah, 100%. I'm a big believer in self-awareness, and I think neural net self-awareness will allow for things like the capabilities, like the ones you described,

伊利亚·苏茨克维尔：

是的，完全同意。我非常相信自我意识，我认为神经网络的自我意识将会带来你描述的那些能力，

like for them to know what they know and what they don't know, and for them to know where to invest to increase their skills most optimally.

比如让它们知道自己知道什么、不知道什么，并让它们知道应该在哪里投资以最优化地提升自己的技能。

And to your question of interpretability, there are actually two answers to that question.

至于你关于解释性的问题，其实有两个答案。

One answer is, you know, we have the neural net, so we can analyze the neurons and we can try to understand what the different neurons and different layers mean.

一个答案是，我们有神经网络，因此我们可以分析神经元，尝试理解不同神经元和不同层的意义。

And you can actually do that, and OpenAI has done some work on that.

实际上你可以这么做，OpenAI已经在这方面做了一些工作。

But there is a different answer, which is that, I would say that's the human-centric answer, where you say, you know, you look at a human being, you can't read...

但还有另一个答案，我会说这是以人为中心的答案，比如你看一个人，你无法读取……

How do you know what a human being is thinking? You ask them, you say, hey, what do you think about this? What do you think about that? And you get some answers.

你怎么知道一个人在想什么？你问他们，比如“你对这个怎么想？”“你对那个怎么看？”然后你得到一些答案。

Lex Fridman:

The answers you get are sticky in the sense you already have a mental model. You already have a mental model of that human being.

莱克斯·弗里德曼：

你得到的答案是“粘性”的，因为你已经有了一个关于那个人的心理模型。

You already have an understanding of, like, a big conception of that human being, how they think, what they know, how they see the world. And then everything you ask, you're adding on to that.

你已经对那个人有了一个总体的概念，比如他们如何思考、知道什么、如何看待世界。然后你所问的每件事都会增加到这个概念中。

And that stickiness seems to be—that's one of the really interesting qualities of the human being, is that information is sticky.

这种“粘性”似乎是人类一个非常有趣的特性，即信息具有粘性。

You don't—you seem to remember the useful stuff, aggregate it well, and forget most of the information that's not useful.

你似乎记住了有用的信息，并很好地进行整合，而忘记了大部分无用的信息。

That process, but that's also pretty similar to the process that neural networks do.

这一过程实际上与神经网络的处理过程相当相似。

It's just that neural networks are much crappier at this time.

只不过神经网络目前在这方面要差得多。

It doesn't seem to be fundamentally that different. But just to stick on reasoning for a little longer, you said, why not? Why can't it reason?

这似乎在本质上并没有那么大的不同。但回到推理的话题，你说为什么不？为什么它不能推理？

What's a good impressive feat benchmark to you of reasoning? That you'll be impressed by what neural networks are able to do. Is that something you already have in mind?

对于你来说，推理的一个令人印象深刻的基准是什么？什么样的能力会让你对神经网络的表现感到震撼？你已经有这样的想法了吗？

Ilya Sutskever:

Well, I think writing really good code. I think proving really hard theorems. Solving open-ended problems with out-of-the-box solutions.

伊利亚·苏茨克维尔：

嗯，我认为是编写非常好的代码。证明非常困难的定理。以及用创造性的方法解决开放性问题。

Lex Fridman:

And sort of theorem-type mathematical problems.

莱克斯·弗里德曼：

还有类似定理类型的数学问题。

Ilya Sutskever:

Yeah, I think those ones are a very natural example as well. You know, if you can prove an unproven theorem, then it's hard to argue that neural networks can't reason.

伊利亚·苏茨克维尔：

是的，我认为这些也是非常自然的例子。如果你能够证明一个未被证明的定理，那么很难否认神经网络具有推理能力。

And so, by the way, this comes back to the point about the hard results—you know, machine learning, deep learning as a field is very fortunate because we have the ability to sometimes produce these unambiguous results.

顺便说一下，这又回到了关于难点结果的问题——你知道，机器学习、深度学习这个领域非常幸运，因为我们有时能够产出这些明确的结果。

And when they happen, the debate changes, the conversation changes. We have the ability to produce conversation-changing results.

当这些结果出现时，争论就会改变，对话也会改变。我们有能力产出改变对话的结果。

Lex Fridman:

Conversation. And then, of course, just like you said, people kind of take that for granted and say, that wasn't actually a hard problem.

莱克斯·弗里德曼：

对话。然后，正如你所说，人们会理所当然地认为那并不是一个真正困难的问题。

Ilya Sutskever:

Well, I mean, at some point we'll probably run out of hard problems.

伊利亚·苏茨克维尔：

嗯，我是说，总有一天我们可能会没有难题可解。

Lex Fridman:

Yeah. That whole mortality thing is kind of a sticky problem that we haven't quite figured out. Maybe we'll solve that one.

莱克斯·弗里德曼：

是的。关于死亡的问题有点棘手，我们还没有完全解决。也许我们会解决那个问题。

Ilya Sutskever:

I think one of the fascinating things in your entire body of work, but also the work at OpenAI recently, one of the conversation changers has been in the world of language models.

伊利亚·苏茨克维尔：

我认为，在你所有的研究中，包括最近OpenAI的工作，语言模型领域的进展无疑是改变对话的重要因素之一。

Lex Fridman:

Can you briefly kind of try to describe the recent history of using neural networks in the domain of language and text?

莱克斯·弗里德曼：

你能简要描述一下最近在语言和文本领域中使用神经网络的发展历程吗？

Ilya Sutskever:

Well, there's been lots of history. I think the Elman network was a small, tiny recurrent neural network applied to language back in the 80s. So the history is really, you know, fairly long at least.

伊利亚·苏茨克维尔：

嗯，这方面的历史相当丰富。我认为Elman网络是上世纪80年代用于语言的一个小型循环神经网络。所以，这段历史确实相当悠久，至少相对而言。

And the thing that started, the thing that changed the trajectory of neural networks and language is the thing that changed the trajectory of all deep learning, and that's data and compute.

而改变神经网络和语言发展轨迹的事情，也是改变整个深度学习轨迹的事情，那就是数据和计算能力。

So suddenly you move from small language models, which learn a little bit, and with language models in particular, you can, there's a very clear explanation for why they need to be large to be good.

因此，你从能够学习一点点的小型语言模型，突然过渡到更大的模型。特别是对于语言模型，为什么它们需要更大才能表现更好，这有一个非常清晰的解释。

Because they're trying to predict the next word. So when you don't know anything, you'll notice very, very broad strokes, surface level patterns like sometimes there are characters and there is a space between those characters.

因为它们试图预测下一个单词。所以，当你一无所知时，你会注意到非常粗略的、表面层次的模式，比如有时字符之间有空格。

You'll notice this pattern. And you'll notice that sometimes there is a comma, and then the next character is a capital letter. You'll notice that pattern.

你会注意到这一模式。你还会注意到，有时逗号后面的字符是大写字母。这种模式也会被发现。

Eventually, you may start to notice that certain words occur often. You may notice that spellings are a thing. You may notice syntax.

最终，你可能会开始注意到某些单词经常出现。你可能会注意到拼写是一种规则。你可能会注意到句法。

And when you get really good at all these, you start to notice the semantics. You start to notice the facts. But for that to happen, the language model needs to be larger.

当你在这些方面变得非常出色时，你会开始注意到语义，甚至注意到事实。但要实现这一点，语言模型需要更大。

Lex Fridman:

So let's linger on that, because that's where you and Noam Chomsky disagree.

莱克斯·弗里德曼：

让我们在这一点上稍作停留，因为这是你和诺姆·乔姆斯基意见分歧的地方。

So you think we're actually taking incremental steps, so the larger network, larger compute will be able to get to the semantics,

所以，你认为我们实际上是在采取渐进的步骤，随着网络规模和计算能力的增加，能够最终达到语义层次，

be able to understand language without what Noam likes to sort of think of as a fundamental understanding of the structure of language,

能够理解语言，而不需要诺姆所强调的那种对语言结构的根本理解，

like imposing your theory of language onto the learning mechanism.

比如将你的语言理论强加到学习机制上。

So you're saying the learning—you can learn from raw data the mechanism that underlies language.

所以你是说，通过学习——可以从原始数据中学习到语言的内在机制。

Ilya Sutskever:

Well, I think it's pretty likely, but I also want to say that I don’t really—I don’t know precisely what Chomsky means when he talks about it. You said something about imposing your structure on language.

伊利亚·苏茨克维尔：

嗯，我认为这是很有可能的，但我也想说，我不太确切知道乔姆斯基在谈论这个问题时的具体含义。你提到了将结构强加于语言的观点。

I'm not 100% sure what he means, but empirically, it seems that when you inspect those larger language models, they exhibit signs of understanding the semantics, whereas the smaller language models do not.

我并不是百分之百确定他的意思，但从经验上看，当你观察那些更大的语言模型时，它们表现出理解语义的迹象，而较小的语言模型则没有。

We’ve seen that a few years ago when we did work on the sentiment neuron.

几年前，当我们研究情感神经元时，我们就看到了这一点。

We trained a small, you know, small LSTM to predict the next character in Amazon reviews.

我们训练了一个小型LSTM来预测亚马逊评论中的下一个字符。

And we noticed that when you increase the size of the LSTM from 500 LSTM cells to 4,000 LSTM cells, then one of the neurons starts to represent the sentiment of the article, sorry, of the review.

我们注意到，当你将LSTM的规模从500个LSTM单元增加到4000个时，其中一个神经元开始表示文章——不，应该是评论的情感。

Now, why is that? Sentiment is a pretty semantic attribute. It’s not a syntactic attribute.

为什么会这样呢？情感是一种非常语义化的属性，而不是句法属性。

Lex Fridman:

And for people who might not know, I don’t know if that’s a standard term, but sentiment is whether it’s a positive or negative review.

莱克斯·弗里德曼：

对于可能不太了解的人来说，我不知道这是不是一个标准术语，但情感指的是评论是正面的还是负面的。

Ilya Sutskever:

That’s right. Is the person happy with something or is the person unhappy with something?

伊利亚·苏茨克维尔：

没错。这指的是一个人对某事是否满意或不满意。

And so, here we had very clear evidence that a small neural net does not capture sentiment, while a large neural net does.

因此，我们有非常明确的证据表明，小型神经网络无法捕捉情感，而大型神经网络可以。

And why is that? Well, our theory is that at some point, you run out of syntax to model, so you start to focus on something else.

为什么会这样呢？我们的理论是，在某个时刻，你对句法的建模达到了极限，于是开始转向其他方面，比如语义。

Lex Fridman:

And with size, you quickly run out of syntax to model, and then you really start to focus on the semantics, would be the idea.

莱克斯·弗里德曼：

随着规模的扩大，你很快就会用尽句法的建模能力，然后真正开始专注于语义，这就是这个想法。

Ilya Sutskever:

That's right. And so I don't want to imply that our models have complete semantic understanding, because that's not true.

伊利亚·苏茨克维尔：

没错。不过我并不想暗示我们的模型拥有完整的语义理解，因为那是不真实的。

But they definitely are showing signs of semantic understanding, partial semantic understanding, but the smaller models do not show those signs.

但它们确实表现出了语义理解的迹象，部分语义理解，而较小的模型没有表现出这些迹象。

Lex Fridman:

Can you take a step back and say what is GPT-2, which is one of the big language models that was the conversation changer in the past couple years?

莱克斯·弗里德曼：

你能否退一步解释一下什么是GPT-2？它是过去几年改变对话模式的大型语言模型之一。

Ilya Sutskever:

Yes, so GPT-2 is a transformer. With one and a half billion parameters that was trained on about 40 billion tokens of text, which were obtained from web pages that were linked to from Reddit articles with more than three upvotes.

伊利亚·苏茨克维尔：

是的，GPT-2是一个Transformer模型。它有15亿参数，训练使用了大约400亿个文本标记，这些文本来自链接到获得超过三次点赞的Reddit文章的网页。

Lex Fridman:

And what's the transformer?

莱克斯·弗里德曼：

什么是Transformer？

Ilya Sutskever:

The transformer is the most important advance in neural network architectures in recent history.

伊利亚·苏茨克维尔：

Transformer是神经网络架构最近历史上最重要的进步。

Lex Fridman:

What is attention, maybe, too? Because I think that's an interesting idea, not necessarily sort of technically speaking, but the idea of attention versus maybe what recurrent neural networks represent.

莱克斯·弗里德曼：

那么，什么是注意力机制？因为我认为这是一个有趣的概念，不一定是技术层面上的，而是注意力机制与循环神经网络所代表的东西之间的比较。

Ilya Sutskever:

Yeah, so the thing is, the transformer is a combination of multiple ideas simultaneously, of which attention is one.

伊利亚·苏茨克维尔：

是的，Transformer是多个想法的结合，其中之一就是注意力机制。

Lex Fridman:

Do you think attention is the key?

莱克斯·弗里德曼：

你认为注意力机制是关键吗？

Ilya Sutskever:

No, it's a key, but it's not the key. The transformer is successful because it is the simultaneous combination of multiple ideas.

伊利亚·苏茨克维尔：

不，它是一个关键点，但不是唯一的关键点。Transformer之所以成功，是因为它同时结合了多个想法。

And if you were to remove either idea, it would be much less successful.

如果你去掉其中任何一个想法，它就会变得不那么成功。

So the transformer uses a lot of attention, but attention existed for a few years. So that can't be the main innovation.

Transformer用了大量的注意力机制，但注意力机制已经存在了好几年。所以这不可能是主要的创新。

The transformer is designed in such a way that it runs really fast on the GPU.

Transformer的设计使得它能在GPU上运行得非常快。

And that makes a huge amount of difference. This is one thing.

这带来了巨大的区别。这是一点。

The second thing is that transformer is not recurrent. And that is really important too, because it is more shallow and therefore much easier to optimize.

第二点是Transformer不是循环的。这也非常重要，因为它更浅，因此更容易优化。

So in other words, it uses attention. It is a really great fit to the GPU, and it is not recurrent, so therefore less deep and easier to optimize.

换句话说，它使用注意力机制，非常适合GPU，而且不是循环的，因此更浅、更容易优化。

And the combination of those factors makes it successful.

这些因素的结合造就了它的成功。

So now it makes great use of your GPU.

因此，现在它可以很好地利用你的GPU。

It allows you to achieve better results for the same amount of compute. And that's why it's successful.

它使你在相同的计算量下可以取得更好的结果。这就是它成功的原因。

Lex Fridman:

Were you surprised how well Transformers worked and GPT-2 worked?

莱克斯·弗里德曼：

你对Transformer和GPT-2的表现感到惊讶吗？

So you worked on language. You've had a lot of great ideas before Transformers came about in language.

你一直从事语言方面的工作。在Transformer出现之前，你在语言领域有很多很棒的想法。

So you got to see the whole set of revolutions before and after. Were you surprised?

所以你见证了前后的整个革命。你感到惊讶吗？

Ilya Sutskever:

Yeah, a little.

伊利亚·苏茨克维尔：

是的，有点。

Lex Fridman:

A little?

莱克斯·弗里德曼：

有点？

Ilya Sutskever:

Yeah. I mean, it's hard to remember because you adapt really quickly, but it definitely was surprising. It definitely was.

伊利亚·苏茨克维尔：

是的。我的意思是，这很难回忆起来，因为你适应得很快，但当时确实是令人惊讶的，确实是。

In fact, you know what? I'll retract my statement. It was pretty amazing.

实际上，你知道吗？我要收回我的说法。这确实非常惊人。

It was just amazing to see it generate this text.

看到它生成这样的文本，真是令人惊叹。

And you know, you got to keep in mind that we've seen at that time, we've seen all this progress in GANs, improving—you know, the samples produced by GANs were just amazing.

你要记住，那时我们已经看到了GAN的所有进展，改进的样本令人惊艳。

You have these realistic faces, but text hasn't really moved that much.

你有这些逼真的人脸，但文本领域并没有太大进展。

And suddenly we moved from, you know, whatever GANs were in 2015, to the best, most amazing GANs in one step.

然后突然间，我们从2015年的GAN水平直接跃升到最佳、最惊人的GAN水平。

And that was really stunning.

这确实让人震撼。

Even though theory predicted, yeah, you train a big language model, of course, you should get this.

尽管理论预测是，你训练一个大语言模型，当然应该得到这样的结果。

But then to see it with your own eyes, it's something else.

但亲眼看到它时，那是完全不同的感觉。

Lex Fridman:

And yet, we adapt really quickly. And now there's a sort of—some cognitive scientists write articles saying that GPT-2 models don't truly understand language.

莱克斯·弗里德曼：

然而，我们适应得非常快。现在有些认知科学家写文章说，GPT-2模型并没有真正理解语言。

So we adapt quickly to how amazing the fact that they're able to model the language so well is.

所以我们很快适应了它们能够如此出色地建模语言这一令人惊叹的事实。

So what do you think is the bar?

那么，你认为衡量标准是什么？

Ilya Sutskever:

For what?

伊利亚·苏茨克维尔：

什么的标准？

Lex Fridman:

For impressing us that...

莱克斯·弗里德曼：

让我们感到惊叹的……

Ilya Sutskever:

I don't know.

伊利亚·苏茨克维尔：

我不知道。

Lex Fridman:

Do you think that bar will continuously be moved?

莱克斯·弗里德曼：

你认为这个标准会不断被提高吗？

Ilya Sutskever:

Definitely. I think when you start to see really dramatic economic impact, that's when... I think that's in some sense the next barrier. Because right now, if you think about the work in AI, it's really confusing.

伊利亚·苏茨克维尔：

肯定会。我认为，当人工智能开始对经济产生显著影响时，那就是……在某种意义上，那将是下一个障碍。因为现在，如果你思考人工智能的研究，它真的很令人困惑。

It's really hard to know what to make of all these advances. It's kind of like, okay, you got an advance and now you can do more things and you got another improvement and you got another cool demo.

很难理解这些进展的意义。就像是，好吧，你取得了一个进展，现在可以做更多的事情，然后又有了另一个改进，再展示了一个很酷的演示。

At some point, I think people who are outside of AI, they can no longer distinguish this progress anymore.

我认为，在某个时候，对于那些不从事人工智能的人，他们将无法再分辨这些进步的差异。

Lex Fridman:

So we were talking offline about translating Russian to English and how there's a lot of brilliant work in Russian that the rest of the world doesn't know about. That's true for Chinese.

莱克斯·弗里德曼：

我们刚才私下聊到从俄语翻译成英语，以及很多俄语的出色作品，世界其他地方的人并不了解。对于中文来说也是如此。

It's true for a lot of scientists and just artistic work in general. Do you think translation is the place where we're going to see sort of economic big impact?

这对于许多科学家和一般的艺术作品都成立。你认为翻译会是我们看到重大经济影响的领域吗？

Ilya Sutskever:

I don't know. I think there is a huge number of applicants. I mean, first of all, I want to point out that translation already today is huge. I think billions of people interact with big chunks of the internet primarily through translation.

伊利亚·苏茨克维尔：

我不确定。我认为有大量的应用。首先，我想指出，翻译如今已经是一个庞大的领域。我认为数十亿人主要通过翻译与互联网的庞大内容进行交互。

So translation is already huge and it's hugely, hugely positive too. I think self-driving is going to be hugely impactful and that's... You know, it's unknown exactly when it happens.

所以翻译已经是一个巨大的领域，而且它带来了极大的积极影响。我认为自动驾驶将会产生巨大的影响，这……你知道的，具体什么时候发生还不确定。

But again, I would not bet against deep learning. So there's deep learning in general, but you think deep learning for self-driving?

但我还是不会对深度学习下注。我是说，深度学习整体上发展不错，你觉得深度学习在自动驾驶领域如何？

Lex Fridman:

Yes, deep learning for self-driving, but I was talking about sort of language models.

莱克斯·弗里德曼：

是的，深度学习在自动驾驶领域，但我其实是在谈语言模型。

Ilya Sutskever:

I see. I veered off a little bit.

伊利亚·苏茨克维尔：

明白了。我稍微跑题了。

Lex Fridman:

Just to check, you're not seeing a connection between driving and language?

莱克斯·弗里德曼：

只是确认一下，你没有看到驾驶和语言之间的联系？

Ilya Sutskever:

No, no.

伊利亚·苏茨克维尔：

没有，没有。

Lex Fridman:

Okay.

莱克斯·弗里德曼：

好的。

Ilya Sutskever:

Or rather, both use neural nets.

伊利亚·苏茨克维尔：

更准确地说，两者都使用神经网络。

Lex Fridman:

That'd be a poetic connection. I think there might be some, like you said, there might be some kind of unification towards a kind of multitask transformers that can take on both language and vision tasks.

莱克斯·弗里德曼：

那会是一种诗意的联系。我认为可能确实存在，就像你说的，可能会出现某种多任务的统一模型，比如可以同时处理语言和视觉任务的变换器。

That'd be an interesting unification. Now let's see, what can I ask about GPT-2 more?

那将是一个有趣的统一点。现在让我想想，我还能问些什么关于 GPT-2 的问题呢？

Ilya Sutskever:

It's simple, so not much to ask. You take a transform, you make it bigger, give it more data and suddenly it does all those amazing things.

伊利亚·苏茨克维尔：

它很简单，所以没什么好问的。你只需要一个变换器，把它做得更大，提供更多的数据，然后它就能做出所有那些惊人的事情。

Lex Fridman:

Yeah, one of the beautiful things is that GPT, the transformers are fundamentally simple to explain, to train. Do you think bigger will continue to show better results in language?

莱克斯·弗里德曼：

是的，GPT 的一个美妙之处在于，变换器从根本上讲易于解释、易于训练。你觉得更大的模型会继续在语言方面表现更好吗？

Ilya Sutskever:

Probably.

伊利亚·苏茨克维尔：

可能会。

Lex Fridman:

Sort of like, what are the next steps with GPT-2, do you think?

莱克斯·弗里德曼：

那么，你认为 GPT-2 的下一步是什么？

Ilya Sutskever:

I mean, I think for sure seeing what larger versions can do is one direction. Also, I mean, there are many questions. There's one question which I'm curious about, and that's the following.

伊利亚·苏茨克维尔：

我认为，毫无疑问，探索更大版本的能力是一个方向。另外，还有许多问题。其中有一个让我好奇的问题，那就是以下这个：

So right now, GPT-2, so we feed it all this data from the internet, which means that it needs to memorize all those random facts about everything in the internet. And it would be nice if...

目前，GPT-2 的数据全部来自互联网，这意味着它需要记住互联网上关于所有事物的随机信息。如果能做到……那就更好了。

The model could somehow use its own intelligence to decide what data it wants to accept and what data it wants to reject. Just like people. People don't learn all data indiscriminately. We are super selective about what we learn.

模型能以某种方式利用自身的智能，决定接受哪些数据，拒绝哪些数据。就像人类一样。人类并不会不加选择地学习所有数据，我们对学习内容非常挑剔。

And I think this kind of active learning, I think, would be very nice to have.

我认为，这种主动学习会是一个非常理想的能力。

Lex Fridman:

Yeah, listen, I love active learning. So let me ask, does the selection of data, can you just elaborate that a little bit more? Do you think the selection of data—I have this kind of sense that the optimization of how you select data,

莱克斯·弗里德曼：

是的，听着，我很喜欢主动学习。那么让我问一下，数据选择方面，你能再详细说明一下吗？你认为数据选择——我觉得优化数据选择的方式，

so the active learning process, is going to be a place for a lot of breakthroughs. Even in the near future, because there hasn't been many breakthroughs there that are public.

也就是主动学习的过程，将会是许多突破的领域。即便是在不远的将来，因为在这个领域，公开的突破并不多。

I feel like there might be private breakthroughs that companies keep to themselves, because the fundamental problem has to be solved if you want to solve self-driving, if you want to solve a particular task.

我感觉可能有一些公司私下的突破，因为如果你想解决自动驾驶或者某个特定任务的核心问题，就必须要解决这个问题。

What do you think about the space in general?

你对这个领域总体怎么看？

Ilya Sutskever:

Yeah, so I think that for something like active learning, or in fact for any kind of capability like active learning, the thing that it really needs is a problem. It needs a problem that requires it.

伊利亚·苏茨克维尔：

是的，我认为，对于主动学习这种技术，或者说任何类似主动学习的能力，它真正需要的是一个问题。它需要一个需要它来解决的问题。

It's very hard to do research about the capability if you don't have a task, because then what's going to happen is that you will come up with an artificial task, get good results, but not really convince anyone.

如果没有一个具体任务，很难进行关于这种能力的研究，因为那样的话，你会设计一个人工任务，得到不错的结果，但却无法真正让人信服。

Lex Fridman:

Right, like we're now past the stage where getting a result on MNIST, some clever formulation of MNIST will convince people.

莱克斯·弗里德曼：

对，就像我们已经过了仅凭在 MNIST 上取得结果，或者在 MNIST 上搞个巧妙的方案就能让人信服的阶段。

Ilya Sutskever:

That's right. In fact, you could quite easily come up with a simple active learning scheme on MNIST and get a 10x speed up. But then, so what?

伊利亚·苏茨克维尔：

没错。事实上，你可以很轻松地在 MNIST 上设计一个简单的主动学习方案，实现 10 倍的加速。但是，然后呢？

And I think that with active learning, active learning will naturally arise as problems that require it pop up. That's my take on it.

我认为，对于主动学习，当有需要它的实际问题出现时，主动学习就会自然地被应用。这是我的看法。

Lex Fridman:

There's another interesting thing that OpenAI has brought up with GPT-2, which is when you create a powerful artificial intelligence system, and it was unclear what kind of detrimental, once you release GPT-2,

莱克斯·弗里德曼：

OpenAI 在 GPT-2 上提出了另一个有趣的问题，那就是当你创造出一个强大的人工智能系统时，一旦你发布 GPT-2，会带来何种负面影响，这一点并不清楚。

what kind of detrimental effect it will have. Because if you have a model that can generate pretty realistic text, you can start to imagine that it would be used by bots in some way that we can't even imagine.

它可能带来怎样的负面影响？因为如果你有一个能够生成非常逼真文本的模型，可以想象它可能会被机器人以某种我们无法预见的方式使用。

So like there's this nervousness about what it's possible to do. So you did a really kind of brave and I think profound thing, which is started a conversation about this.

因此，对于它可能会带来的影响存在一种紧张情绪。而你们做了一件我认为非常勇敢且深刻的事情，那就是启动了一场关于此事的讨论。

Like how do we release powerful artificial intelligence models to the public? If we do it at all, how do we privately discuss with even competitors about how we manage the use of the systems and so on?

比如，我们该如何向公众发布强大的人工智能模型？如果要发布，我们如何与竞争对手私下讨论如何管理这些系统的使用等问题？

So from that, this whole experience, you released a report on it. But in general, are there any insights that you've gathered from just thinking about this, about how you release models like this?

从这一整个经历来看，你们为此发布了一份报告。但总的来说，从思考这件事中，你有没有获得关于如何发布此类模型的任何见解？

Ilya Sutskever:

I mean, I think that my take on this is that the field of AI has been in a state of childhood. And now it's exiting that state, and it's entering a state of maturity. What that means is that AI is very successful, and also very impactful.

伊利亚·苏茨克维尔：

我的观点是，我认为人工智能领域一直处于一种初级阶段。而现在，它正在走出这个阶段，进入一个成熟的阶段。这意味着人工智能已经非常成功，并且影响深远。

And its impact is not only large, but it's also growing. And so for that reason, it seems wise to start thinking about the impact of our systems before releasing them, maybe a little bit too soon, rather than a little bit too late.

它的影响不仅巨大，而且还在持续增长。因此，提前思考这些系统的影响似乎是明智之举，也许提前一点比晚一点更好。

And with the case of GPT-2, like I mentioned earlier, the results really were stunning. And it seemed plausible. It didn't seem certain. It seemed plausible that something like GPT-2 could easily use to reduce the cost of disinformation.

就像我之前提到的，GPT-2 的结果确实令人惊艳。并且似乎很有可能——虽然不是确定的，但看起来有可能像 GPT-2 这样的模型能轻易地降低传播虚假信息的成本。

And so there was a question of what's the best way to release it and a staged release seemed logical. A small model was released and there was time to see the many people use these models in lots of cool ways.

因此，问题在于什么是发布它的最佳方式，而分阶段发布似乎是合理的。先发布一个小型模型，然后有时间观察许多人用这些模型进行各种很酷的应用。

There have been lots of really cool applications. There haven't been any negative applications we know of, and so eventually it was released. But also other people replicated similar models.

确实出现了很多非常有趣的应用。我们所知并没有出现负面的应用，因此最终模型被发布了。不过，也有人复制出了类似的模型。

Lex Fridman:

That's an interesting question, though, that we know of. So, in your view, staged release is at least part of the answer to the question of what do we do once we create a system like this?

莱克斯·弗里德曼：

不过，这个问题很有趣，“我们所知的”。所以在你看来，分阶段发布至少是回答“当我们创造出这样的系统时该怎么办”问题的部分答案？

Ilya Sutskever:

It's part of the answer, yes.

伊利亚·苏茨克维尔：

是的，这是部分答案。

Lex Fridman:

Is there any other insights? Say you don't want to release the model at all because it's useful to you for whatever the business is.

莱克斯·弗里德曼：

还有其他见解吗？比如说，如果你完全不想发布这个模型，因为它对你的业务有用。

Ilya Sutskever:

Well, plenty of people don't release models already.

伊利亚·苏茨克维尔：

嗯，已经有很多人不发布模型了。

Lex Fridman:

Right, of course, but is there some moral, ethical responsibility when you have a very powerful model to communicate? Just as you said, when you had GPT-2, it was unclear how much it could be used for misinformation. It's an open question.

莱克斯·弗里德曼：

对，当然了，但当你拥有一个非常强大的模型时，是否存在某种道德或伦理责任需要沟通？正如你所说，当时你们有了 GPT-2，却不清楚它会被用来传播多少虚假信息。这是一个悬而未决的问题。

And getting an answer to that might require that you talk to other really smart people that are outside of your particular group.

而要找到答案，可能需要你与其他不在你团队中的聪明人进行交流。

Please tell me there's some optimistic pathway for people across the world to collaborate on these kinds of cases. Or is it still really difficult from one company to talk to another company?

请告诉我，人们能否通过某种乐观的途径，在这类问题上进行全球范围的合作？还是说一家公司与另一家公司之间的沟通仍然非常困难？

Ilya Sutskever:

So it's definitely possible. It's definitely possible to discuss these kind of models with colleagues elsewhere and to get their take on what to do.

伊利亚·苏茨克维尔：

这绝对是可能的。完全可以与其他地方的同事讨论这类模型，并听取他们对应该如何应对的看法。

Lex Fridman:

How hard is it though?

莱克斯·弗里德曼：

不过，这有多难呢？

Ilya Sutskever:

I mean.

伊利亚·苏茨克维尔：

我是说。

Lex Fridman:

Do you see that happening?

莱克斯·弗里德曼：

你认为那会发生吗？

Ilya Sutskever:

I think that's a place where it's important to gradually build trust between companies because ultimately, all the AI developers are building technology, which is going to be increasingly more powerful.

伊利亚·苏茨克维尔：

我认为，在这个领域，逐步建立公司之间的信任是很重要的，因为归根结底，所有人工智能开发者都在构建越来越强大的技术。

And so it's the way to think about it is that ultimately we're all in it together.

所以，应该这样看待这个问题：最终，我们大家都是一起在做这件事。

Lex Fridman:

Yeah, I tend to believe in the better angels of our nature, but I do hope that when you build a really powerful AI system in a particular domain,

莱克斯·弗里德曼：

是的，我倾向于相信人性中的善良天使，但我确实希望，当你在某个特定领域构建一个非常强大的人工智能系统时，

that you also think about the potential negative consequences of it. It's an interesting and scary possibility that there will be a race for AI development that would push people to close that development and not share ideas with others.

你也能考虑到它可能带来的负面后果。这是一种有趣又令人害怕的可能性，即人工智能开发的竞赛会促使人们封闭开发过程，不再与他人分享想法。

Ilya Sutskever:

I don't love this. I've been a pure academic for 10 years. I really like sharing ideas and it's fun. It's exciting.

伊利亚·苏茨克维尔：

我不喜欢这种情况。我当了 10 年的纯学术研究者。我真的很喜欢分享想法，这很有趣，也很令人兴奋。

Lex Fridman:

Let's talk about AGI a little bit. What do you think it takes to build a system of human-level intelligence? We talked about reasoning, we talked about long-term memory, but in general, what does it take, do you think?

莱克斯·弗里德曼：

我们来谈谈通用人工智能（AGI）。你认为打造一个具有人类水平智能的系统需要什么？我们谈到了推理能力，也谈到了长期记忆，但总的来说，你觉得需要哪些要素？

Ilya Sutskever:

Well, I can't be sure, but I think that deep learning plus maybe another small idea.

伊利亚·苏茨克维尔：

我不敢确定，但我认为可能是深度学习加上某个小的想法。

Lex Fridman:

Do you think self-play will be involved?

莱克斯·弗里德曼：

你觉得自我博弈会涉及其中吗？

Sort of like you've spoken about the powerful mechanism of self-play where systems learn by sort of exploring the world in a competitive setting against other entities that are similarly skilled as them and so incrementally improve in this way.

就像你谈到过的那种强大的自我博弈机制，系统通过在与其他相似技能实体的竞争环境中探索世界，从而以这种方式逐步改进。

Do you think self-play will be a component of building an AGI system?

你认为自我博弈会是构建 AGI 系统的一个组成部分吗？

Ilya Sutskever:

Yeah, so what I would say to build AGI I think is going to be deep learning plus some ideas and I think self-play will be one of those ideas. I think that that is a very...

伊利亚·苏茨克维尔：

是的，我认为构建 AGI 会是深度学习加上一些想法，而自我博弈会是其中一个。我认为这确实是一个非常……

Self-play has this amazing property that it can surprise us in truly novel ways. For example, like we, I mean, pretty much every self-play system, both our Dota bot,

自我博弈有一种神奇的特性，它能以真正新颖的方式给我们带来惊喜。比如说，我们的每一个自我博弈系统，无论是 Dota 机器人，

I don't know if OpenAI had a release about multi-agent where you had two little agents who were playing hide and seek. And of course, also AlphaZero. They were all produced surprising behaviors.

我不知道 OpenAI 是否发布过一个关于多智能体的例子，其中两个小智能体玩捉迷藏。当然，还有 AlphaZero。它们都产生了令人惊讶的行为。

They all produced behaviors that we didn't expect. They are creative solutions to problems. And that seems like an important part of AGI that our systems don't exhibit routinely right now.

它们都表现出我们意想不到的行为，是问题的创造性解决方案。而这似乎是 AGI 的一个重要组成部分，是我们现在的系统还没有常规展现的能力。

And so that's why I like this area, I like this direction, because of its ability to surprise us.

这就是我喜欢这个领域、喜欢这个方向的原因，因为它有能力让我们感到惊喜。

Lex Fridman:

To surprise us. And an AGI system would surprise us fundamentally.

莱克斯·弗里德曼：

让我们感到惊讶。而 AGI 系统会从根本上带来惊喜。

Ilya Sutskever:

Yes. And to be precise, not just a random surprise, but to find a surprising solution to a problem that's also useful.

伊利亚·苏茨克维尔：

是的。更准确地说，不只是随机的惊喜，而是找到一个令人意外又有用的解决方案。

Lex Fridman:

Right. Now, a lot of the self-play mechanisms have been used in the game context, or at least in the simulation context. How far along the path to AGI do you think will be done in simulation?

莱克斯·弗里德曼：

对。现在，许多自我博弈机制都被用在游戏场景中，或者至少是在模拟环境中。你认为在通向 AGI 的道路上，有多少工作可以在模拟中完成？

How much faith, promise do you have in simulation versus having to have a system that operates in the real world, whether it's the real world of digital real-world data or real world like actual physical world of robotics?

相比之下，你对模拟的信心有多大？还是说必须有一个能在现实世界中运行的系统，无论是数字化的现实数据世界，还是像机器人那样的物理现实世界？

Ilya Sutskever:

I don't think it's an either-or. I think simulation is a tool and it helps. It has certain strengths and certain weaknesses and we should use it.

伊利亚·苏茨克维尔：

我认为这不是非此即彼的选择。我认为模拟是一种工具，它有帮助。它有某些优势和劣势，我们应该利用它。

Lex Fridman:

Yeah, but okay. I understand that. That's... That's true. But one of the criticisms of self-play, one of the criticisms of reinforcement learning is one of the, the, its current power, its current results, while amazing, have been demonstrated in simulated environments, or very constrained physical environments.

莱克斯·弗里德曼：

是的，但好吧。我明白了，这确实没错。不过，自我博弈的一个批评点，也是强化学习的一个批评点在于，它目前的强大能力和惊人的成果，都是在模拟环境或非常受限的物理环境中展示出来的。

Do you think it's possible to escape them? Escape the simulated environments and be able to learn in non-simulated environments?

你认为有可能突破这些局限吗？突破模拟环境，在非模拟环境中进行学习？

Or do you think it's possible to also just simulate in a photo-realistic and physics-realistic way the real world in a way that we can solve real problems with self-play in simulation?

或者你认为，是否也可以通过照片级真实和物理真实的方式来模拟现实世界，从而在模拟中通过自我博弈解决实际问题？

Ilya Sutskever:

So I think that transfer from simulation to the real world is definitely possible and has been exhibited many times by many different groups. It's been especially successful in vision.

伊利亚·苏茨克维尔：

我认为，从模拟到现实世界的迁移绝对是可能的，而且许多不同的团队已经多次展示过。这在视觉领域尤其成功。

Also, OpenAI in the summer has demonstrated a robot hand which was trained entirely in simulation in a certain way that allowed for sim-to-real transfer to occur.

此外，OpenAI 在夏天展示了一只机械手，它完全是在模拟环境中训练的，并以某种方式实现了从模拟到现实的迁移。

Lex Fridman:

Is this for the Rubik's Cube?

莱克斯·弗里德曼：

这是用于魔方的吗？

Ilya Sutskever:

Yes, that's right.

伊利亚·苏茨克维尔：

是的，没错。

Lex Fridman:

I wasn't aware that was trained in simulation.

莱克斯·弗里德曼：

我不知道那是完全在模拟中训练的。

Ilya Sutskever:

It was trained in simulation entirely.

伊利亚·苏茨克维尔：

它完全是在模拟环境中训练的。

Lex Fridman:

Really, so it wasn't in the physical, the hand wasn't trained?

莱克斯·弗里德曼：

真的吗？所以机械手并没有在物理环境中训练？

Ilya Sutskever:

No. 100% of the training was done in simulation and the policy that was learned in simulation was trained to be very adaptive. So adaptive that when you transfer it, it could very quickly adapt to the physical world.

伊利亚·苏茨克维尔：

没错。100% 的训练都是在模拟环境中完成的，而且在模拟中学到的策略非常具有适应性。适应性强到当你将其迁移到物理世界时，它可以非常迅速地适应。

Lex Fridman:

So the kind of perturbations with the Giraffe or whatever the heck it was, were those part of the simulation?

莱克斯·弗里德曼：

那么，像长颈鹿玩偶之类的扰动，那些是模拟的一部分吗？

Ilya Sutskever:

Well, the simulation was generally, so the simulation was trained to be robust to many different things, but not the kind of perturbations we've had in the video. So it's never been trained with a glove.

伊利亚·苏茨克维尔：

嗯，模拟总体上是针对许多不同情况进行训练的，目的是增强其鲁棒性，但并没有包括我们视频中看到的那种扰动。所以它从未用手套训练过。

It's never been trained with a stuffed giraffe.

它也从未用长颈鹿玩偶进行过训练。

Lex Fridman:

So in theory, these are novel perturbations.

莱克斯·弗里德曼：

那么理论上，这些是新颖的扰动。

Ilya Sutskever:

Correct. It's not in theory, in practice, that those are novel perturbations.

伊利亚·苏茨克维尔：

没错。这不仅是理论上，在实践中这些确实是新颖的扰动。

Lex Fridman:

Well, that's okay. That's a clean, small-scale but clean example of a transfer from the simulated world to the physical world.

莱克斯·弗里德曼：

嗯，这没问题。这是一个从模拟世界到物理世界的简洁、小规模但清晰的例子。

Ilya Sutskever:

Yeah, and I will also say that I expect the transfer capabilities of deep learning to increase in general. And the better the transfer capabilities are, the more useful simulation will become.

伊利亚·苏茨克维尔：

是的，我还想说，我预计深度学习的迁移能力总体上会不断提高。而迁移能力越强，模拟就会变得越有用。

Because then you could take, you could experience something in simulation and then learn a moral of the story which you could then carry with you to the real world. As humans do all the time when they play computer games.

因为那样的话，你可以在模拟中体验某些事情，并从中学到某种教训或启示，然后将其应用到现实世界中。就像人类玩电子游戏时经常做的那样。

Lex Fridman:

So let me ask sort of an embodied question, staying on AGI for a sec. Do you think AGI says that we need to have a body?

莱克斯·弗里德曼：

让我问一个与“身体”有关的问题，继续聊 AGI。你认为 AGI 是否需要有一个“身体”？

We need to have some of those human elements of self-awareness, consciousness, fear of mortality, self-preservation in the physical space, which comes with having a body.

我们是否需要具备一些人类特质，比如自我意识、意识、对死亡的恐惧，以及在物理空间中的自我保护，这些都是有“身体”所带来的。

Ilya Sutskever:

I think having a body will be useful. I don't think it's necessary. But I think it's very useful to have a body for sure because you can learn things which cannot be learned without a body.

伊利亚·苏茨克维尔：

我认为有一个身体会很有用，但我不认为这是必要的。不过，毫无疑问，有一个身体确实非常有用，因为你可以学习到一些没有身体无法学习到的东西。

But at the same time, I think that if you don't have a body, you could compensate for it and still succeed.

但同时，我认为即使没有身体，你也可以通过某种方式弥补这一点，并依然取得成功。

Lex Fridman:

You think so?

莱克斯·弗里德曼：

你这么认为？

Ilya Sutskever:

Yes. Well, there is evidence for this. For example, there are many people who were born deaf and blind and they were able to compensate for the lack of modalities. I'm thinking about Helen Keller specifically.

伊利亚·苏茨克维尔：

是的。这方面是有证据的。例如，有许多人天生失聪失明，但他们能够弥补感官上的缺失。我特别想到海伦·凯勒。

Lex Fridman:

So even if you're not able to physically interact with the world, and if you're not able to, I mean, I actually was getting at—maybe let me ask on the more particular, I'm not sure if it's connected to having a body or not,

莱克斯·弗里德曼：

所以即便你无法与物理世界进行互动，如果你无法——我是说，我其实是在探讨，或许让我更具体地问，我不确定这是否与拥有一个身体有关，

but the idea of consciousness and a more constrained version of that is self-awareness. Do you think an AGI system should have consciousness? We can't define consciousness, whatever the heck you think consciousness is.

但意识的概念以及它的一个更受限的版本——自我意识。你认为 AGI 系统应该有意识吗？尽管我们无法定义意识，无论你认为意识究竟是什么。

Ilya Sutskever:

Yeah, hard question to answer given how hard it is to define it.

伊利亚·苏茨克维尔：

是啊，这是个很难回答的问题，因为“意识”本身很难定义。

Lex Fridman:

Do you think it's useful to think about?

莱克斯·弗里德曼：

你认为思考这个问题有意义吗？

Ilya Sutskever:

I mean, it's definitely interesting. It's fascinating. I think it's definitely possible that our systems will be conscious.

伊利亚·苏茨克维尔：

我觉得，这个问题绝对有趣，令人着迷。我认为，我们的系统很可能会具备意识。

Lex Fridman:

Do you think that's an emergent thing that just comes from, do you think consciousness could emerge from the representation that's stored within your networks?

莱克斯·弗里德曼：

你认为这是某种涌现现象吗？你觉得意识可以从神经网络中存储的表征中涌现出来吗？

So like that it naturally just emerges when you become more and more, you're able to represent more and more of the world.

就是说，当网络能够越来越多地表征世界时，意识会自然地涌现出来。

Ilya Sutskever:

Well, I'd say I'd make the following argument, which is humans are conscious and if you believe that artificial neural nets are sufficiently similar to the brain,

伊利亚·苏茨克维尔：

嗯，我会提出这样一个论点：人类是有意识的，如果你相信人工神经网络与大脑有足够的相似性，

then there should at least exist artificial neural nets you should be conscious to.

那么至少应该存在具备意识的人工神经网络。

Lex Fridman:

You're leaning on that existence proof pretty heavily, okay.

莱克斯·弗里德曼：

你对这个“存在性证明”依赖得很重啊，好吧。

Ilya Sutskever:

That's the best answer I can give.

伊利亚·苏茨克维尔：

这是我能给出的最好的答案了。

Lex Fridman:

I know, I know, I know. There's still an open question if there's not some magic in the brain that we're not.

莱克斯·弗里德曼：

我知道，我知道，我知道。仍然有一个未解的问题——大脑中是否存在某种我们未曾发现的“魔力”。

I mean, I don't mean a non-materialistic magic, but that the brain might be a lot more complicated and interesting than we give it credit for.

我的意思不是某种非物质的“魔力”，而是说，大脑可能比我们想象的要复杂得多，也有趣得多。

Ilya Sutskever:

If that's the case, then it should show up and at some point we will find out that we can't continue to make progress. But I think it's unlikely.

伊利亚·苏茨克维尔：

如果真是这样，那应该会显现出来，并且在某个时候我们会发现无法继续取得进展。但我认为这种可能性不大。

Lex Fridman:

So we talk about consciousness, but let me talk about another poorly defined concept of intelligence. Again, we've talked about reasoning. We've talked about memory. What do you think is a good test of intelligence for you?

莱克斯·弗里德曼：

我们谈到了意识，但让我再谈一个同样定义不清的概念——智能。我们已经谈过推理和记忆。那么在你看来，智能的一个好测试是什么？

Are you impressed by the test that Alan Turing formulated with the imitation game with natural language? Is there something in your mind that you will be deeply impressed by if a system was able to do?

图灵提出的模仿游戏（以自然语言为基础）测试是否令你印象深刻？在你看来，有什么事情是一个系统做到后会让你深感震撼的？

Ilya Sutskever:

I mean, lots of things. There is a certain frontier of capabilities today, and there exist things outside of that frontier, and I would be impressed by any such thing. For example,

伊利亚·苏茨克维尔：

很多事情都会令我印象深刻。目前能力上有一个边界，而在边界之外的事情，任何一个都能让我感到惊叹。例如，

I would be impressed by a deep learning system which solves a very pedestrian task like machine translation or computer vision task or something which never makes a mistake a human wouldn't make under any circumstances.

如果一个深度学习系统能完成一些非常常见的任务，比如机器翻译或计算机视觉任务，或者从不犯人类在任何情况下都不会犯的错误，那将令我印象深刻。

I think that is something which has not yet been demonstrated and I would find it very impressive.

我认为这是一件尚未实现的事情，而如果实现了，我会觉得非常了不起。

Lex Fridman:

Yeah, so right now they make mistakes in different, they might be more accurate than human beings, but they still, they make a different set of mistakes.

莱克斯·弗里德曼：

是的，目前这些系统会犯一些不同的错误，尽管它们可能比人类更准确，但它们仍然会犯与人类不同类型的错误。

Ilya Sutskever:

So my, I would guess that a lot of the skepticism that some people have about deep learning is when they look at their mistakes and they say, well, those mistakes, they make no sense.

伊利亚·苏茨克维尔：

所以，我猜很多人对深度学习的怀疑来自于他们看到这些系统犯的错误时，会觉得“这些错误完全没有意义”。

Like if you understood the concept, you wouldn't make that mistake. And I think that changing that would inspire me. That would be, yes, this is progress.

比如说，如果你理解了这个概念，你就不会犯这样的错误。我认为，如果能改变这一点，会让我深受启发。这就是，我会认为这是进步。

Lex Fridman:

Yeah, that's a really nice way to put it. But I also just don't like that human instinct to criticize a model is not intelligent. That's the same instinct as we do when we criticize any group of creatures as the other.

莱克斯·弗里德曼：

是的，这个说法很棒。但我也不喜欢人类本能地批评模型不够智能。这种本能就像我们批评任何其他生物群体时的态度一样。

Because it's very possible that GPT-2 is much smarter than human beings at many things.

因为很可能 GPT-2 在许多方面比人类聪明得多。

Ilya Sutskever:

That's definitely true. It has a lot more breadth of knowledge.

伊利亚·苏茨克维尔：

那肯定是真的。它拥有广泛得多的知识面。

Lex Fridman:

Yes, breadth of knowledge and even perhaps depth on certain topics.

莱克斯·弗里德曼：

是的，知识的广度，甚至在某些话题上可能还有深度。

Ilya Sutskever:

It's kind of hard to judge what depth means, but there's definitely a sense in which humans don't make mistakes. These models do.

伊利亚·苏茨克维尔：

深度的定义有点难判断，但确实存在一种感觉，即人类不会犯某些错误，而这些模型会。

Lex Fridman:

The same is applied to autonomous vehicles. The same is probably going to continue being applied to a lot of artificial intelligence systems.

莱克斯·弗里德曼：

这同样适用于自动驾驶汽车，也很可能继续适用于许多人工智能系统。

We find this is the annoying thing. This is the process of, in the 21st century, the process of analyzing the progress of AI is the search for one case where the system fails in a big way where humans would not, and then many people writing articles about it.

我们发现这是一件烦人的事情。在 21 世纪，分析人工智能进展的过程，就是寻找系统在某个重大方面失败的案例，而人类在那种情况下不会失败，然后很多人开始写相关文章。

And then broadly, the public generally gets convinced that the system is not intelligent. And we like pacify ourselves by thinking it's not intelligent because of this one anecdotal case. And this seems to continue happening.

然后公众普遍被说服认为系统不够智能。我们通过这种单一的偶发案例来安慰自己，认为它不够智能。这种情况似乎会不断发生。

Ilya Sutskever:

Yeah, I mean, there is truth to that. Although I'm sure that plenty of people are also extremely impressed by the system that exists today.

伊利亚·苏茨克维尔：

是的，我的意思是，这种说法确实有道理。不过我相信，今天已有的系统也让很多人印象深刻。

But I think this connects to the earlier point we discussed that it's just confusing to judge progress in AI. And you have a new robot demonstrating something. How impressed should you be?

但我认为，这与我们之前讨论的观点有关，那就是评价人工智能的进展确实很让人困惑。当你看到一个新机器人展示某些能力时，你应该感到多么惊讶呢？

And I think that people will start to be impressed once AI starts to really move the needle on the GDP.

我认为，一旦人工智能开始真正推动 GDP 的增长，人们就会开始感到震撼。

Lex Fridman:

So you're one of the people that might be able to create an AGI system here. Not you, but you and OpenAI. If you do create an AGI system and you get to spend sort of the evening with it, him, her, what would you talk about do you think?

莱克斯·弗里德曼：

所以，你是有可能在这里创造出 AGI 系统的人之一。不只是你，而是你和 OpenAI。如果你真的创造出一个 AGI 系统，并有机会与它（他、她）共度一个晚上，你觉得会聊些什么？

Ilya Sutskever:

The very first time?

伊利亚·苏茨克维尔：

第一次见面？

Lex Fridman:

First time.

莱克斯·弗里德曼：

第一次。

Ilya Sutskever:

Well, the first time I would just ask all kinds of questions and try to get it to make a mistake and I would be amazed that it doesn't make mistakes and just keep asking broad questions.

伊利亚·苏茨克维尔：

嗯，第一次我会问各种各样的问题，试图让它犯错，然后会对它没有犯错感到惊叹，并不断地问一些广泛的问题。

Lex Fridman:

What kind of questions do you think, would they be factual or would they be personal, emotional, psychological? What do you think?

莱克斯·弗里德曼：

你觉得会是什么样的问题呢？是事实类的，还是个人、情感、心理方面的？你怎么看？

Ilya Sutskever:

All of the above.

伊利亚·苏茨克维尔：

以上全部都会问。

Lex Fridman:

Would you ask for advice?

莱克斯·弗里德曼：

你会向它寻求建议吗？

Ilya Sutskever:

Definitely. I mean, why would I limit myself talking to a system like this?

伊利亚·苏茨克维尔：

当然会。我是说，为什么要限制自己与这样一个系统交流的机会呢？

Lex Fridman:

Now, again, let me emphasize the fact that you truly are one of the people that might be in the room where this happens. So let me ask sort of a profound question about, I just talked to a Stalin historian.

莱克斯·弗里德曼：

再次强调，你确实是可能参与其中的人之一。那么让我问一个比较深刻的问题，我刚刚和一位研究斯大林的历史学家聊过。

I've been talking to a lot of people who are studying power. Abraham Lincoln said, nearly all men can stand adversity. But if you want to test a man's character, give him power.

我最近与许多研究权力的人交谈过。亚伯拉罕·林肯说过，几乎所有人都能忍受逆境，但如果你想测试一个人的品格，就给他权力。

I would say the power of the 21st century, maybe the 22nd, but hopefully the 21st, would be the creation of an AGI system and the people who have control, direct possession and control of the AGI system.

我会说，21 世纪的权力，或许是 22 世纪，但希望是 21 世纪的权力，将是创造 AGI 系统，以及那些直接掌控和拥有 AGI 系统的人。

So what do you think, after spending that evening having a discussion with the AGI system, what do you think you would do?

那么，你觉得在与 AGI 系统共度一个晚上并进行讨论之后，你会做什么？

Ilya Sutskever:

Well, the ideal world I'd like to imagine is one where humanity are like the board members of a company where the AGI is the CEO.

伊利亚·苏茨克维尔：

嗯，我理想中的世界是这样的：人类就像公司的董事会成员，而 AGI 则是首席执行官。

So it would be, I would like, the picture which I would imagine is you have some kind of different entities, different countries or cities and the people that leave their vote for what the AGI that represents them should do and an AGI that represents them goes and does it.

所以，在我的设想中，会有不同的实体，不同的国家或城市，人们通过投票决定代表他们的 AGI 应该做什么，然后由这个 AGI 去执行。

I think a picture like that I find very appealing. You could have multiple, you would have an AGI for a city, for a country, and it would be trying to, in effect, take the democratic process to the next level.

我觉得这样的图景非常吸引人。可以有多个 AGI，分别为城市、国家服务，它们将努力实质性地将民主进程提升到一个新的水平。

Lex Fridman:

And the board can always fire the CEO.

莱克斯·弗里德曼：

董事会可以随时解雇 CEO。

Ilya Sutskever:

Essentially, press the reset button, say. Rerandomize the parameters.

伊利亚·苏茨克维尔：

基本上就是按下重置按钮，比如重新随机化参数。

Lex Fridman:

Well, let me sort of, that's actually, okay, that's a beautiful vision, I think, as long as it's possible to press the reset button. Do you think it will always be possible to press the reset button?

莱克斯·弗里德曼：

嗯，让我说一下，我觉得这确实是一个美好的愿景，只要可以按下重置按钮。你觉得重置按钮会始终存在吗？

Ilya Sutskever:

So I think that it's definitely will be possible to build. So you're talking, so the question that I really understand from you is will humans or humans people have control over the AI systems that they build? Yes.

伊利亚·苏茨克维尔：

我认为这绝对是可以实现的。所以你问的问题，我真正理解的是，人类是否能对他们所构建的 AI 系统拥有控制权？答案是肯定的。

And my answer is it's definitely possible to build AI systems which will want to be controlled by their humans.

我的回答是，绝对可以构建出希望被人类控制的 AI 系统。

Lex Fridman:

Wow, that's part of their, so it's not that just they can't help but be controlled, but that's, one of the objectives of their existence is to be controlled.

莱克斯·弗里德曼：

哇，这是它们的一部分，不只是无法避免被控制，而是说，它们存在的目标之一就是被控制。

Ilya Sutskever:

In the same way that human parents generally want to help their children. They want their children to succeed. It's not a burden for them. They are excited to help the children and to feed them and to dress them and to take care of them.

伊利亚·苏茨克维尔：

就像人类父母通常希望帮助他们的孩子一样。他们希望孩子成功。这对他们来说不是负担。他们乐于帮助孩子，为他们提供食物、穿衣并照顾他们。

And I believe with high conviction that the same will be possible for an AGI. It will be possible to program an AGI, to design it in such a way that it will have a similar deep drive,

我非常相信，这对 AGI 来说也是可能的。我们可以编程并设计出一个 AGI，使其拥有类似的深层驱动力，

that it will be delighted to fulfill and the drive will be to help humans flourish.

它会乐于实现这种驱动力，而这种驱动力将是帮助人类繁荣发展。

Lex Fridman:

But let me take a step back to that moment where you create the AGI system. I think this is a really crucial moment. And between that moment and the democratic board members with the AGI at the head, there has to be a relinquishing of power.

莱克斯·弗里德曼：

让我退一步，回到你创造 AGI 系统的那一刻。我认为这是一个非常关键的时刻。而在那一刻与让 AGI 成为民主董事会首席之间，必须存在一个放权的过程。

So as George Washington, despite all the bad things he did, one of the big things he did is he relinquished power. He, first of all, didn't want to be president.

就像乔治·华盛顿，尽管他做过一些不好的事情，但他所做的一件大事就是放弃了权力。他起初甚至不想成为总统。

And even when he became president, he gave, he didn't keep just serving as most dictators do indefinitely.

即使当了总统，他也没有像大多数独裁者那样无限期地执政。

Do you see yourself being able to relinquish control over an AGI system, given how much power you can have over the world?

考虑到你可以通过 AGI 系统掌握巨大的世界影响力，你认为自己能够放弃对它的控制吗？

At first financial, just make a lot of money, right? And then control by having possession of this AGI system.

首先是经济上的收益，可以赚很多钱，对吧？然后通过拥有这个 AGI 系统来获得控制权。

Ilya Sutskever:

I'd find it trivial to do that. I'd find it trivial to relinquish this kind of power. I mean, you know, the kind of scenario you are describing sounds terrifying to me. That's all. I would absolutely not want to be in that position.

伊利亚·苏茨克维尔：

我觉得放弃这种权力对我来说毫无难度。我觉得放弃这种权力非常容易。我的意思是，你描述的那种场景对我来说听起来太可怕了。就是这样。我绝对不想处于那种位置。

Lex Fridman:

Do you think you represent the majority or the minority of people in the AI community?

莱克斯·弗里德曼：

你认为自己代表了人工智能社区的大多数人还是少数人？

Ilya Sutskever:

Well, I mean...

伊利亚·苏茨克维尔：

嗯，我是说……

Lex Fridman:

It's an open question, an important one. Are most people good is another way to ask it.

莱克斯·弗里德曼：

这是一个悬而未决的问题，也是一个重要的问题。换种方式问，大多数人是善良的吗？

Ilya Sutskever:

So, I don't know if most people are good, but I think that when it really counts, people can be better than we think.

伊利亚·苏茨克维尔：

嗯，我不知道大多数人是否是善良的，但我认为在真正关键的时刻，人们可能会比我们想象的更好。

Lex Fridman:

That's beautifully put. Are there specific mechanisms you can think of for aligning AI values to human values? Do you think about these problems of continued alignment as we develop the AI systems?

莱克斯·弗里德曼：

说得真好。你能想到一些将 AI 价值观与人类价值观对齐的具体机制吗？在我们开发 AI 系统时，你是否会思考这些持续对齐的问题？

Ilya Sutskever:

Yeah, definitely. In some sense, the kind of question which you are asking is, so if I were to translate the question to today's terms,

伊利亚·苏茨克维尔：

是的，绝对会。从某种意义上说，你的问题可以翻译成今天的术语，

it would be a question about how to get an RL agent that's optimizing a value function which itself is learned.

也就是如何让一个强化学习（RL）代理优化一个自身学习出来的价值函数。

And if you look at humans, humans are like that because the reward function, the value function of humans is not external, it is internal. That's right.

如果你看人类，人类就是这样的，因为人类的奖励函数、价值函数不是外在的，而是内在的。确实如此。

And there are definite ideas of how to train a value function, basically an objective, you know, as objective as possible perception system that will be trained separately.

而且，确实有一些方法可以训练价值函数，基本上是一个尽可能客观的感知系统，可以单独训练。

...to recognize, to internalize human judgments on different situations. And then that component would then be integrated as the base value function for some more capable RL system. You could imagine a process like this.

……去识别并内化人类对不同情境的判断。然后，这个组件可以被整合为更强大 RL 系统的基础价值函数。你可以想象这样的一个过程。

I'm not saying this is the process, I'm saying this is an example of the kind of thing you could do.

我不是说这就是唯一的过程，我是说这是你可以尝试的一种方式。

Lex Fridman:

So, on that topic of the objective functions of human existence, what do you think is the objective function that's implicit in human existence? What's the meaning of life?

莱克斯·弗里德曼：

关于人类存在的目标函数这个话题，你认为人类存在中隐含的目标函数是什么？生命的意义是什么？

Ilya Sutskever:

I think the question is wrong in some way. I think that the question implies that there is an objective answer, which is an external answer, you know, your meaning of life is X. I think what's going on is that we exist and that's amazing.

伊利亚·苏茨克维尔：

我认为这个问题本身有点不对。我觉得这个问题暗示着有一个客观的答案，一个外在的答案，比如“你的生命意义是 X”。但我认为实际上，我们存在本身就是一种奇迹。

And we should try to make the most of it and try to maximize our own value and enjoyment of our very short time while we do exist.

我们应该努力充分利用它，尽力在我们有限的时间内最大化我们的价值和享受。

Lex Fridman:

It's funny because action does require an objective function. It's definitely there in some form, but it's difficult to make it explicit and maybe impossible to make it explicit, I guess is what you're getting at.

莱克斯·弗里德曼：

有趣的是，行动确实需要一个目标函数。它肯定以某种形式存在，但很难明确表达出来，也许根本无法明确表达，我猜这就是你的意思。

And that's an interesting fact of an RL environment.

这也是强化学习环境中的一个有趣事实。

Ilya Sutskever:

Well, what I was making a slightly different point is that humans want things and their wants create the drives that cause them to, you know, our wants are our objective functions, our individual objective functions.

伊利亚·苏茨克维尔：

嗯，我的观点稍有不同。人类有各种需求，而这些需求驱动了他们的行动。我们的需求就是我们的目标函数，我们每个人的个人目标函数。

We can later decide that we want to change, that what we wanted before is no longer good and we want something else.

我们之后可以决定改变，觉得之前想要的东西不再合适，而我们想要别的东西。

Lex Fridman:

Yes, but they're so dynamic. There's got to be some underlying sort of Freud. There's the things, there's like sexual stuff.

莱克斯·弗里德曼：

是的，但这些目标函数非常动态。一定存在某种弗洛伊德式的潜在因素，比如性方面的东西。

There's people who think it's the fear of death and there's also the desire for knowledge and, you know, all these kinds of things. Procreation, sort of all the evolutionary arguments.

有人认为是对死亡的恐惧，也有人认为是对知识的渴望，还有各种类似的东西。生育、所有进化论的论点。

It seems to be, there might be some kind of fundamental objective function from which everything else emerges. But it seems like that's very difficult.

似乎可能存在某种基本的目标函数，从中衍生出一切。但这似乎非常难以确定。

Ilya Sutskever:

I think that probably is an evolutionary objective function, which is to survive and procreate and make your children succeed. That would be my guess, but it doesn't give an answer to the question, what's the meaning of life?

伊利亚·苏茨克维尔：

我认为很可能存在一个进化目标函数，那就是生存、繁衍并让你的孩子成功。这是我的猜测，但这并没有回答“生命的意义是什么”这个问题。

I think you can see how humans are part of this big process, this ancient process. We exist on a small planet and that's it. So given that we exist, try to make the most of it and try to enjoy more and suffer less as much as we can.

我觉得你可以看到，人类是这个庞大过程的一部分，这个古老的过程。我们存在于一个小星球上，仅此而已。所以既然我们存在，就应该尽量充分利用，尽可能多享受，少受苦。

Lex Fridman:

Let me ask two silly questions about life. One, do you have regrets? Moments that if you went back you would do differently?

莱克斯·弗里德曼：

让我问两个关于人生的愚蠢问题。第一，你有没有遗憾？如果回到过去，有哪些时刻你会选择不同的做法？

And two, are there moments that you're especially proud of that made you truly happy?

第二，有哪些让你特别自豪、真正感到快乐的时刻吗？

Ilya Sutskever:

So I can answer both questions. Of course, there's a huge number of choices and decisions that I've made that, with the benefit of hindsight, I wouldn't have made them.

伊利亚·苏茨克维尔：

我可以回答这两个问题。当然，有许多我做出的选择和决定，在事后看来，我本不会那样做。

And I do experience some regret, but, you know, I try to take solace in the knowledge that at the time I did the best I could.

我确实有些遗憾，但我试图从这样一种认识中得到安慰：在当时的情况下，我已经尽力了。

And in terms of things that I'm proud of, I'm very fortunate to have done things I'm proud of.

至于让我自豪的事情，我很幸运能够做出一些让我感到自豪的事情。

And they made me happy for some time, but I don't think that that is the source of happiness.

它们确实让我感到快乐了一段时间，但我并不认为那是幸福的根源。

Lex Fridman:

So your academic accomplishments, all the papers, you're one of the most cited people in the world. All the breakthroughs I mentioned in computer vision and language and so on. What is the source of happiness and pride for you?

莱克斯·弗里德曼：

你在学术上的成就，所有的论文，你是世界上被引用次数最多的人之一。我提到的计算机视觉、语言等领域的所有突破。对你来说，幸福和自豪的来源是什么？

Ilya Sutskever:

I mean all those things are a source of pride for sure. I'm very grateful for having done all those things and it was very fun to do them. My current view is that happiness comes to a very large degree from the way we look at things.

伊利亚·苏茨克维尔：

所有这些事情当然是自豪的来源。我非常感激能够完成这些事情，而且做它们的过程也非常有趣。我目前的看法是，幸福在很大程度上来源于我们看待事物的方式。

You can have a simple meal and be quite happy as a result, or you can talk to someone and be happy as a result as well. Or conversely, you can have a meal and be disappointed that the meal wasn't a better meal.

你可以吃一顿简单的饭就感到很幸福，也可以通过和某人交谈而感到幸福。相反地，你也可能因为饭菜不够好而感到失望。

So I think a lot of happiness comes from that. But I'm not sure. I don't want to be too confident.

所以我认为，很多幸福源于此。但我不确定，我不想过于自信。

Lex Fridman:

Being humble in the face of the uncertainty seems to be also part of this whole happiness thing.

莱克斯·弗里德曼：

在不确定性面前保持谦逊，似乎也是幸福的一部分。

Well, I don't think there's a better way to end it than meaning of life and discussions of happiness. So, Ilya, thank you so much.

嗯，我认为没有比谈论生命的意义和幸福更好的结束方式了。那么，伊利亚，非常感谢你。

You've given me a few incredible ideas. You've given the world many incredible ideas. I really appreciate it and thanks for talking today.

你给了我一些令人难以置信的想法，也给了世界许多不可思议的创意。我真的很感激，感谢你今天的分享。

Ilya Sutskever:

Yeah, thanks for stopping by. I really enjoyed it.

伊利亚·苏茨克维尔：

嗯，感谢你的到访。我真的很享受这次对话。

Lex Fridman:

Thanks for listening to this conversation with Ilya Sutskever, and thank you to our presenting sponsor, Cash App. Please consider supporting the podcast by downloading Cash App and using the code LEXPODCAST.

莱克斯·弗里德曼：

感谢您收听我与伊利亚·苏茨克维尔的对话，同时感谢我们的主赞助商 Cash App。请通过下载 Cash App 并使用代码 LEXPODCAST 来支持本播客。

If you enjoy this podcast, subscribe on YouTube, review it with 5 stars on Apple Podcasts, support on Patreon, or simply connect with me on Twitter at Lex Fridman.

如果您喜欢本播客，请在 YouTube 上订阅，在 Apple Podcasts 上给予五星好评，或者通过 Patreon 支持，亦或是在 Twitter 上与我联系，账号是 Lex Fridman。

And now, let me leave you with some words from Alan Turing on machine learning. Instead of trying to produce a program to simulate the adult mind, why not rather try to produce one which simulates the child?

现在，让我用艾伦·图灵关于机器学习的一段话结束今天的节目。他曾说：“与其尝试编写一个模拟成人思维的程序，不如尝试编写一个模拟儿童思维的程序。”

If this were then subjected to an appropriate course of education, one would obtain the adult brain. Thank you for listening and hope to see you next time.

如果这样的程序接受适当的教育课程，就能获得成人的大脑。感谢您的收听，我们下次再见。

2020-05-08 Ilya Sutskever.Deep Learning

2020-05-08 Ilya Sutskever.Deep Learning

Article Comments Update

热门标签

Related Articles

2024-05-29 Microsoft Corporation (MSFT) Jefferies Software Conference (Transcript)

2023-03-27 Ilya Sutskever.Why Next-Token Prediction Could Surpass Human Intelligence

2024-04-25 Microsoft Corporation (MSFT) Q3 2024 Earnings Call Transcript

2023-03 Ilya Sutskever.AI Today and Vision of the Future

2024-06-06 Microsoft Corporation (MSFT) BofA Securities 2024 Global Technology Conference (Transcript)