The Exciting, Perilous Journey Toward AGI | Ilya Sutskever | TED

翻译人员: Chunxiang Lu 校对人员: Lening Xu

You've all experienced the progress of artificial intelligence.

你们都经历了人工智能的进步。

Many of you may have spoken with a computer,

有许多人可能已经与计算机对过话，

and a computer understood you

而且它听懂了你说什么

and spoke back to you.

并对你做出回应。

With the rate of progress being that it is,

照目前的进步速度而言，

it's not difficult to imagine that at some point in the future,

不难想象，在未来的某个时点，

our intelligent computers will become as smart or smarter than people.

我们的智能计算机会变得和 我们一样或更加聪明。

And it's also not difficult to imagine that when that happens,

而且不难想象，当这种情况发生时，

the impact of such artificial intelligence

这样的人工智能的影响

is going to be truly, truly vast.

将变得非常、非常巨大。

And you may wonder:

那么，你也许会想：

Is it going to be OK when technology is so impactful?

当技术如此有影响力时，会依旧安全吗？

And here my goal is to point out the existence of a force

至此，我想引申出

that many of you may have not noticed,

一种你们中的大多数还未曾注意到的力量，

that gives me hope that indeed,

这着实给了我希望，

we will be happy with the result.

我想我们会对结果满意。

So artificial intelligence.

那么人工智能，

What is it, and how does it work?

它是什么，它是如何工作的？

Well, it turns out that it's very easy to explain

事实上，要想解释AI的工作原理很容易。

how artificial intelligence works.

Just one sentence.

只需要一句话。

Artificial intelligence is nothing but digital brains

人工智能

不过是大型计算机中的数字大脑。

inside large computers.

That's what artificial intelligence is.

这就是人工智能。

Every single interesting AI that you've seen

每一个你所见过的有趣的人工智能

is based on this idea.

都是基于这个原理。

Over the decades,

几十年来，

scientists and engineers have been figuring out

科学家和工程师一直在研究

how such digital brains should work

这种数字大脑应该如何运作，

and how to build them, how to engineer them.

应该如何建造和设计他们。

Now, I find it interesting

现在，我觉得有趣的是

that the seat of intelligence in human beings is our biological brain.

人类智慧的所在即是我们的生物大脑。

It is fitting that the seat of intelligence

人工智能中智能的所在是人工大脑， 这是恰当的。

in artificial intelligence

is an artificial brain.

Here, I'd like to take a digression and tell you about how I got into AI.

此刻，我题外话一下， 告诉你们我是如何步入人工智能的。

There were three forces that pulled me into it.

有三股力量把我拉进去。

The first one was that when I was a little child,

第一个问题是，当我还是个 大约五六岁的孩子时。

at around the age of five or six,

I was very struck by my own conscious experience,

我对自己的意识体验，

by the fact that I am me and I am experiencing things.

我就是我以及我正在经历一些事情 感到印象深刻。

That when I look at things, I see them.

当我看东西时，我就能看见它们。

Like, this feeling over time went away, though,

就好像，随着时间的推移， 这种感觉消失了，

by simply mentioning it to you right now it comes back.

但现在只要向你提一下，它就会回来。

But this this feeling of ...

但是这样的感觉……

That I am me, that you are you,

我是我，你是你。

I found it very strange and very disturbing almost.

让我觉得很奇怪， 也几乎很令人不安。

And so when I learned about artificial intelligence,

因此，当我了解了人工智能。

I thought, wow, if we could build a computer

我想，哇，如果我们能建造一台

that is intelligent,

智能计算机，

maybe we will learn something about ourselves,

也许我们会学到一些关于我们自己、

about our own consciousness.

关于我们自己意识的东西。

That was my first motivation that pulled me towards AI.

这是推动我走向 AI的第一个动机。

The second motivation was more pedestrian in a way.

从某种意义上说， 第二个动机更平淡无奇。

I was simply curious about how intelligence works.

我只是好奇智能是如何运作的。     

And when I was a teenager, an early teenager in the late '90s,

当我还是个青少年， 一个 90 年代末的少年时，

the sense that I got is that science simply did not know

我感觉到科学根本不知道智能

how intelligence worked.

是如何运作的。

There was also a third reason,

还有第三个原因，

which is that it was clear to me back then

那就是当时我很清楚，

that artificial intelligence,

如果人工智能奏效，

if it worked, it would be incredibly impactful.

它将产生惊人的影响。

Now, it wasn't at all obvious

现在，在人工智能领域取得

that it will be possible to make progress in artificial intelligence,

进展的可能性一点都不明显，

but if it were possible to make progress in artificial intelligence,

但是如果有可能在 人工智能领域取得进展，

that would be incredibly impactful.

那将产生难以置信的影响。

So these were the three reasons that pulled me towards AI.

因此，这是促使我走向AI的三个原因 。

That's why I thought that's a great area to spend all my efforts on.

这就是为什么我认为人工智能是一个 值得我全力以赴的好领域。

So now let's come back to our artificial intelligence,

那么，现在我们回到我们的人工智能，

the digital brains.

也就是数字大脑。

Today, these digital brains

如今，这些数字大脑

are far less smart than our biological brains.

远不如我们的生物大脑聪明。

When you speak to an AI chat bot,

当你与人工智能聊天机器人交谈时，

you very quickly see that it's not all there,

你很快就会发现，它不能都做出回应。

that it's, you know,

你懂的，

it understands mostly, sort of.

某种程度上，它只是片面地理解了。

But you can clearly see that there are so many things it cannot do

但是你可以清楚地意识到， 它做不到的事情太多了，

and that there are some strange gaps.

还有一些奇怪的差距。

But this situation, I claim, is temporary.

但我相信，这种情况是暂时的。

As researchers and engineers continue to work on AI,

随着研究人员和工程师 继续研究人工智能，

the day will come when the digital brains that live inside our computers

这一天终将到来， 生活在我们计算机中的数字大脑

will become as good and even better than our own biological brains.

将变得和我们自己的生物大脑 一样好，甚至更好。

Computers will become smarter than us.

计算机将变得比我们更聪明。

We call such an AI an “AGI,” artificial general intelligence.

我们将这样的人工智能称为 “AGI”， 即通用人工智能。

When we can say that the level at which we can teach the AI to do anything

当我们可以这样说时， 我们可以教人工智能做任何事情，

that, for example, I can do, or someone else.

例如，我或者他人能做的事情。

So although AGI does not exist today,

因此，尽管今天AGI还不存在，

we can still gain a little bit of an insight

但我们仍能察觉一旦AGI建成，

into the impact of AGI once it's built.

其带来的影响。

It is completely obvious

显而易见，

that such an AGI will have a dramatic impact

这样的AGI 将对 生活、人类活动和社会的各个领域

on every area of life, of human activity and society.

产生巨大影响。

And I want to go over a quick case study.

我想简要介绍一个案例研究。

This is a narrow example of a very, very broad technology.

这是一项非常非常广泛技术的一个狭义例子。

The example I want to present is health care.

我想举的例子是医疗保健。

Many of you may have had the experience of trying to go to a doctor.

你们中的许多人可能有过 尝试去看医生的经历。

You need to wait for many months sometimes,

有时候你需要等好几个月，

and then when you do get to see a doctor, you get a small,

然后当你去看医生时，你只有一小段，

very limited amount of time with the doctor

非常有限的时间与医生共处。

and furthermore, the doctor, being only human,

而且，医生也是人，

can have only limited knowledge of all this,

对所有这些、

all the medical knowledge that exists.

所有现存的医学知识的了解有限。

And then by the end of it, you get a very large bill.

最后，你会得到一份相当高昂的账单。

(Laughter)

（笑声）

Well, if you have an intelligent computer, an AGI, that is built to be a doctor,

好吧，如果你有一台智能计算机， 一台打造为医生的AGI，

it will have complete and exhaustive knowledge of all medical literature.

它将对所有医学文献有 完整而详尽的了解。

It will have billions of hours of clinical experience,

它将拥有数十亿小时的临床经验，

and it will be always available and extremely cheap.

随时可用还非常便宜。

When this happens,

当实现这一想法，

we will look back at today's health care

我们将像看待 16 世纪的牙科一样

similarly to how we look at 16th century dentistry.

回顾当今的医疗保健。

(Laughter)

（笑声）

You know, when they tied people with belts and then have this drill,

你知道，当他们用皮带绑住人们 然后进行这种演习时，

that's how today's health care will look like.

今天的医疗保健就会变成这样。

And again, to emphasize, this is just one example.

再强调一遍， 这只是一个例子。

This is just one example.

这只是一个例子。

AGI will have dramatic and incredible impact

AGI 将对人类活动的每个领域

on every single area of human activity.

产生巨大而令人难以置信的影响。

But when you see impact this large, you may wonder, "Gosh,

但是，当你看到如此之大的冲击力时， 你可能会想：“天哪，

isn't this technology too impactful?"

这项技术是不是太过强大了？”

And indeed, for every positive application of AGI,

事实上，对于AGI的每一积极应用，

there will be a negative application as well.

也会存在负面的应用。

This technology is also going to be different

这项技术也将与我们习惯的技术不同，

from technologies that we are used to,

because it will have the ability to improve itself.

因为它将具有自我改进的能力。

It is possible to build an AGI

建立一个适用于 下一 代AGI的AGI 是有可能的。

that will work on the next generation of AGI.

The closest analogue we have to this kind of rapid technological improvement

我们与这种快速的技术进步 最接近的类似之处

is when the Industrial Revolution has taken place, where humans,

是工业革命发生的时候，

the material condition of human society, was very, very constant.

人类以及人类社会的物质条件， 是非常非常恒定的。

And then it was a rapid increase, rapid growth.

随后是快速增长，快速发展。

With AGI,

有了AGI，

something like this could happen again, but on a shorter timescale.

类似这样的事情可能会再次发生， 但时间更短。

And then furthermore,

此外，

there are concerns around

有人担心 AGI 是否会变得非常非常强大，

if an AGI ever becomes very, very powerful,

which is possible,

这是可能的，

maybe it will want to go rogue being that it is an agent.

既然它是一个特工， 也许它想成为捣蛋鬼呢。

So this is a concern that exists with this unprecedented,

因此，这是这项前所未有的、

not yet existing technology.

尚未存在的技术中存在的问题。

And indeed,

事实上，

you look at all the positive potential of AGI

你看看AGI所有积极的潜力

and all the concerning possibilities of AGI as well,

以及AGI所有令人不安的可能性，

and you may say, "Gosh, where is this all headed?"

你可能会说：“天哪， 这一切都将走向何方？”

One of my motivations in creating OpenAI was,

我创建 OpenAI的动机之一是，

in addition to developing this technology,

除了开发这项技术外，

was also to address the questions

还要解决 AGI 带来的问题，

that are posed by AGI, the difficult questions,

由我们带来的棘手的问题和担忧。

the concerns that we raised.

In addition to working with governments

除了与政府合作

and helping them understand what is coming and prepare for it,

并帮助他们了解即将 发生的事情并为之做好准备外，

we are also doing a lot of research

我们也在做很多关于解决 事物技术方面的研究，

on addressing the technological side of things,

so that the AI will never want to go rogue.

这样人工智能就永远不会变成恶棍。

And this is something which I’m working on as well.

这也是 我正在研究的问题。

But I think, the thing to note,

但我认为，需要注意的是，

because AI and AGI is really the only area of the economy

因为人工智能和通用人工智能确实

where there is a lot of excitement, a lot of investment,

是经济中唯一一个 充满激情、大量投资的领域，

everyone is working on it,

人人都在为之奋斗。

there's a huge number of labs in the world trying to build the same thing.

世界上有大量的实验室在 尝试打造同样的东西。

Even if OpenAI takes these desirable steps that I mentioned,

即使 OpenAI 采取了 我提到的这些理想步骤，

what about the rest of the companies and the rest of the world?

其他公司和世界其他地区呢？

And this is where I want to make my observation

这就是我想观察

about the force that exists.

存在力量的地方。

The observation is this: consider the world one year ago,

观察结果是这样的： 以一年前的世界为例，

as recently as one year ago.

也就是最近的一年前。

People weren't really talking about AI, not in the same way at all.

人们并没有真正谈论人工智能， 根本不是以同样的方式。

What happened?

发生了什么？

We all experienced what it's like to talk to a computer and to be understood.

我们都经历过与计算机 交谈和被理解的感觉。

The idea that computers will become really intelligent

计算机将变得真正智能

and eventually more intelligent than us is becoming widespread.

并最终比我们更智能的想法 正在变得越来越普遍。

It used to be a niche idea that only a few enthusiasts

它曾经是一个小众的想法， 只有少数狂热爱好者和业余爱好者

and hobbyists and people who were very into AI were thinking about.

以及非常喜欢人工智能的人有这种想法。

But now everyone is thinking about it.

但是现在每个人都这么想。

And as the AI continues to make progress,

随着人工智能不断进步，

as technology continues to advance,

随着技术不断进步，

as more and more people see what AI can do

随着越来越多的人看到人工智能能做什么

and where it is headed towards,

什么以及它前进的方向，

then it will become clear just how dramatic,

那么人们就会清楚地看到人工智能 是多么引人注目、

incredible and ...

多么令人难以置信以及……

almost fantastical AGI is going to be

AGI 会变得几乎是异想天开的，

and how much trepidation is appropriate.

多少不安才是合适的。

And what I claim will happen is that people will start to act

我声称会发生的情况是， 出于自身利益，

in unprecedentedly collaborative way out of their own self-interest.

人们将开始以前所未有的合作方式行事。

It's already happening right now.

这种情况现在已经发生了。

You see the leading AGI companies starting to collaborate

举一个具体的例子， 你可以看到领先的AGI公司

for a specific example, through the Frontier Model Forum.

开始通过前沿模型论坛合作。

And we will expect that companies that are competitors

而且我们预计， 互相竞争的公司

will share technical information to make their AIs safe.

将共享技术信息 来确保其人工智能的安全。

We may even see governments do this.

我们甚至可能会看到政府这样做。

For another example,

再举一个例子，

at OpenAI, we really believed in how dramatic AGI is going to be.

在OpenAI，我们真正相信 AGI会变得非常不同凡响。

So one of the ideas that we were operating by,

因此，我们赖以运营的想法之一，

and it's been written on our website for five years now,

已经写在我们的网站上长达五年之久。

that when technology gets such that we are very, very close to AGI,

当科技发展到我们非常非常接近 通用人工智能、

to computers smarter than humans,

比人类更智能的计算机时，

and if some other company is far ahead of us,

如果其他公司远远领先于我们，

then rather than compete with them,

那么从某种意义上说， 我们不会与他们竞争，

we will help them out, join them, in a sense.

而是帮助他们，加入他们的行列。

And why do that?

为什么要这样做？

Because we feel, we appreciate

因为我们认为，

how incredibly dramatic AGI is going to be.

我们懂得AGI将变得 分外不同凡响。

And my claim is that with each generation of capability advancements,

我的主张是， 随着每一代人能力提高

as AI gets better

随着人工智能变得更好。

and as all of you experience what AI can do,

随着你们所有人体验到 人工智能可以做什么，

as people who run AI efforts and AGI efforts

负责人工智能工作和 通用人工智能工作的人

and people who work on them will experience it as well,

以及从事这些工作的人也会体验到，

this will change the way we see AI and AGI,

这将改变我们看待 人工智能和通用人工智能的方式，

and that will change collective behavior.

这将改变集体行为。

And this is an important reason

这就是为什么

why I'm hopeful

尽管这项技术带来了巨大的挑战，

that despite the great challenges that's posed by this technology,

但我仍对我们能克服这些挑战

we will overcome them.

充满希望的重要原因。

Thank you.

谢谢。

(Applause)

（掌声）