The Last 6 Decades of AI — and What Comes Next | Ray Kurzweil | TED

翻译人员: Alvin Lee 校对人员: Sue Lu

So we've heard a lot about artificial intelligence.

我们听到很多 有关人工智能（AI）的话题。

I've actually been involved with AI for 61 years, which is a record.

我干 AI 这行 61 年， 算是创纪录了。

And we've heard a lot about what people think about AI today.

我们已了解人们对 AI 的很多看法。

So I tried to figure out,

因此我想搞清楚，

what did we think about artificial intelligence

在 61 年前，我们是如何

61 years ago.

看待人工智能的。

So first of all, people asked, "What are you into?"

首先，人们问我，“你是做什么的？”

I'd say artificial intelligence.

我说人工智能。

And they'd say, "What's that?"

他们又问，“那是什么东西？”

So no one was really aware of it.

所以大家其实都不了解它。

I joined, in 1962,

我入行是在 1962 年，

1956 was the conference where artificial intelligence got its name.

“人工智能”在 1956 年 一次会议上得名。

So the views were quite different.

当时的观点很不一样。

People who were in computer science had heard of artificial intelligence.

从事计算机科学工作的人 才听说过人工智能。

Most people were quite skeptical.

大部分人对其持怀疑态度，

They thought it would never happen, or if they thought it would happen,

要么认为不可实现，要么认为即使实现，

maybe it would happen in a century or several centuries.

也还要等一个或几个世纪。

但那些参加了 1956 年 达特茅斯会议的人

But the people that actually came to that Dartmouth conference in 1956,

they were quite optimistic.

还是非常乐观的。

Some of them, including Minsky,

他们中有的人，包括明斯基在内，

thought it would take like, one semester to reach --

觉得再过一个学期……

(Laughter)

（笑声）

To reach the level of intelligence that humans had.

人工智能就能达到人类智力水平。

And in fact, that led to our first argument.

事实上这导致了我们的第一次争论。

He was my mentor for 50 years.

他当我的导师 50 年。

But we argued about that because I thought it would take decades,

我们争论是因为我认为还需要数十年，

but we would see it within our lifetime.

但在我们的有生之年是可以看到的。

So we're the only species

我们是唯一的物种

that actually creates tools that enhances our intelligence.

通过创造工具来增强自己的智能。

I mean, I'll bet almost everybody has one of these

我打赌，几乎每个人都拥有这些工具

that makes us more intelligent.

以变得更加智能。

This connects to the cloud.

这些工具与云端连接，

It gets more intelligent every year.

每年都会变得更智能。

Basically, the singularity is going to bring that into our minds.

基本上，奇点会将它带入我们的大脑，

We're going to become smarter.

我们会变得更聪明。

And there's two different things we have in our anatomy

从生理结构上看，我们有两样东西，

that enable us to do that.

保证我们能做到这一点。

One is our brain,

第一个是我们的大脑，

but we're not the only species that has a brain

但其他物种也有大脑，

or even a comparable brain.

哪怕仅仅是相似的大脑。

Elephants and whales actually have a brain that's larger than [ours].

大象和鲸的大脑比我们的还大。

But there's another aspect of their anatomy that they don't have

但在生理结构上，它们缺一样东西，

and that no one else has aside from humans

这东西只有人类有，

which is our thumb.

那就是大拇指。

So I can look at a tree and I can imagine, yeah,

因此我可以看着一棵树，心里想象着，

I could take those poles and create a tool

我可以用这些树枝做一个工具，

and then I can actually do it.

然后我真的可以做出来。

Now, monkeys, if you look at them they have a thumb,

而猴子，你观察一下，它们也有大拇指，

but it doesn't really work very well,

但不是很好用，

it's actually an inch down.

短了一英寸。

They can grab things without much force.

它们能不费吹灰之力就抓住东西，

They can create a first generation of tools,

它们可以制造第一代工具，

but they can't use that tool to create another set of tools.

但它们无法用其制造另一组工具。

So they really can't create a whole set of tools

因此，它们无法创造一整套工具

that will enhance their intelligence.

来增强它们的智能。

We're the only species that does that.

我们是唯一可以这么做的物种。

And that's what artificial intelligence is doing.

这就是人工智能在做的事情。

From the very first hominid

从最开始的人类祖先，

that created a very primitive tool

创造出非常原始的工具，

to Gemini and GPT-4 today,

到今天的 Gemini 和 GPT-4，

we create tools that make us smarter.

我们一直在创造让我们更加聪明的工具。

And so I've been actually monitoring

其实我一直在监测

the growth of computation,

算力的增长，

which is right here.

就是这张表。

I spent like, 45 years on this.

我花了大概 45 年时间在这上面。

And as you go up the chart, it represents exponential growth.

沿着表格往上看，是指数级的增长。

You might think that someone was in charge of this.

你也许会想，有人在控制这个。

Gee, we've done this much,

天哪，我们已经做到这个程度了，

it's in a straight line,

进展呈直线增长，

let's get our next computer to be right here.

让下一台电脑达到那个水平吧。

But no one was aware of it.

但没有人意识到这一点。

No one even knew that this was happening for the first 40 years.

在最初 40 年里， 没人预料到这种增长。

I discovered this 45 years ago.

我在 45 年前就发现了这点，

I had various reasons to feel it would continue at this pace.

我有很多理由相信它会如此增长。

In 1939

在 1939 年，

that represents

图表显示

0.000007

每恒定美元

calculations per second per constant dollar.

每秒可进行 0.000007 次计算。

At the upper right hand corner, you've got a Google computer,

在右上角是谷歌计算机，

which was 130 billion calculations per second.

每秒能进行 1300 亿次计算。

And recently Nvidia just came out with a chip

最近 英伟达（Nvidia）新出一款芯片，

which is half a trillion calculations per second.

每秒能计算 5000 亿次。

So this little chart represents a growth

所以这个小图表代表了

of 75 quadrillion-fold increase.

75 兆亿倍的增长。

That's why we didn't have large language models in 1939

这就是为什么我们在 1939 年 没有大型语言模型，

or even three years ago.

甚至在 3 年前都没有。

We did have something called large language models.

当时那个叫做大型语言模型的东西，

They didn't work very well three years ago,

在 3 年前并不好用，

began to work two years ago.

2 年前才开始能用了，

We've seen a tremendous progress that's happened in the last two years.

过去 2 年里出现了巨大进展。

In 1999, I was asked to make a prediction of when would we see AGI,

在 1999 年，有人让我预测 AGI，

artificial general intelligence.

即“通用人工智能”，何时能实现。

And so I figured that this chart would continue, which it has.

我指出进展将如图表所示，

And I figured we'd need about a trillion calculations per second to do AGI.

并且算力需要达到 1 万亿次每秒。

So I estimated 2029.

我预测 AGI 将在 2029 年实现，

That was met with a lot of skepticism.

这遭到了很多怀疑。

Stanford had actually been monitoring my predictions.

斯坦福大学一直在监测我的预测，

They called an international conference to talk about my prediction.

还为此召开了国际会议进行讨论。

And hundreds of AI scientists came from around the world.

几百名全球 AI 科学家齐聚一堂，

And they agreed that it would happen.

一致认为我的预测会实现，

We would achieve AGI, but not within 30 years.

AGI 会实现，但不是在 30 年内,

The estimate was 100 years.

他们的估计是 100 年。

And I've talked actually to some of the people who were there

我其实和部分与会科学家讨论过，

who said 100 years then

他们在会议中认为要 100 年，

and they're basically agreeing it's going to happen very soon.

但现在基本同意这一天会很快到来。

Musk says it's going to happen in two years.

马斯克说 2 年内就会实现,

It's not an unreasonable position.

这不是毫无理由的。

Other people saying three or four years,

还有人说 3 至 4 年，

I'm sticking with five years.

我还是坚持说 5 年，

But it could happen soon.

但有可能更快。

But everybody agrees now, AGI is very soon.

现在大家都同意 AGI 会很快到来。

So I have another book coming out,

于是我写了另一本书，

"The Singularity is Nearer."

《奇点临近》。

(Laughter)

（笑声）

And I've got about 50 graphs in there.

书里有大概 50 张图表。

I can't explain it right now, but if you talk to me later,

现在没时间细讲，如果过后你来找我，

I can explain these charts,

我可以解释这些图表，

but it basically shows that artificial intelligence

它们讲的基本就是 AI

is going to take over everything.

会接管一切。

The amount of --

数量……

(Laughter)

（笑声）

The amount of money that we make right now

我们今天创造的财富数量

is ten times greater in constant dollars

用恒定美元计算的话，

than it was 100 years ago.

比 100 年前的 10 倍还多。

We were very, very poor 100 years ago,

100 年前我们非常穷，

there was no government programs.

政府没有开展这些项目。

So we're much richer than we were then.

我们现在比那时候富有多了。

And the movement,

论及发展，

not only of computation,

不仅仅是在算力方面，

but every single technology,

也体现在每一项技术上，

is done by creating,

其产生都是以

taking the latest thing we've created and making the next one.

最新技术为基础而开发出来的。

We take the latest chip

比如，我们利用最新的芯片

and we use that to create the next one.

来开发下一代芯片。

We have greater wealth, as I said, that leads to better education,

我们有了更多财富，可以提供更好教育，

leads to better doctors,

更好的医生，

leads to healthy people,

更健康的人民，

leads to more global wealth.

并创造更多全球财富。

All of these things work together.

所有这些都是相互协作的，

AI supercharges everything.

而 AI 会加强这一切。

So I could talk about each thing

我可以讲述每一件

as being actually revolutionized.

被实际革新的事。

I think the most interesting thing is actually medicine.

我认为最有趣的就是医药。

There are a lot of people who are experts in AI who are against what's happening,

有很多 AI 专家反对目前的发展，

and they're very nervous about it, and they think it's going to wipe us out.

他们非常紧张，觉得 AI 会消灭我们。

But people tend to get diseases which are threatening to them.

但人会生病，疾病也是个威胁。

And what's going to happen?

接下来会发生什么呢？

People are going to get diseases and AI is going to come up with a cure,

人会生病，而 AI 会提供治疗方案，

very soon, which will lead to a great deal of appreciation.

这很快就能实现了， 到时我们就会感激 AI 的帮助。

People say that AI is not creative.

有人说 AI 不够有创意。

It's very creative.

但实际它很有创意。

You can actually put together possibilities that might work.

你可以尝试很多可能性。

For example, Moderna was trying to create their COVID vaccine.

比如，莫德纳公司尝试研制新冠疫苗，

They actually put together a list of different mRNA sequences.

列出了一系列不同的 mRNA 序列。

Now, what would we do in the past?

过去没有 AI 时，我们会怎么做？

有人会说，“好，这儿有几十亿种选择，

Someone would come in and say, "OK, there's several billion.

Let's try this one."

让我们来试试这一种。”

Or maybe they'd pick three.

或者他们可能挑选三种。

You can't ...

但你不可能……

do a clinical test on billions of different possibilities.

对数十亿种可能性进行临床测试。

But that's exactly what they did by simulating the reaction.

但现在他们就这么做了，通过模拟反应，

And that took two days.

仅仅花了两天时间。

So in two days, they created the Moderna vaccine.

两天内，他们就开发出了莫德纳疫苗。

And that is still on the market.

目前，这种疫苗依然在销售，

It's been the best vaccine.

是最好的疫苗。

It was done in two days.

而开发它只用了两天。

And we're going to do that with every other thing.

我们也可以这么来开发其他的。

There's some very promising cancer cures that are out there, which AI produced,

目前有些很有前景的癌症疗法， 由 AI 生成的，

and they're looking very promising.

看起来很有希望。

The next few years is going to be remarkable for medicine.

对于医药领域而言， 接下来几年将非比寻常。

We had 190,000 proteins done by people in 2022.

在 2022 年， 人工解析了19 万蛋白质。

2023, AlphaFold 2 did 200 million,

而在 2023 年， AlphaFold 2 解析了 2 亿个，

基本涵盖所有蛋白质及其折叠方式，

basically every protein and how they fold.

Every protein that's used in humans

涵盖人类和地球上所有其他物种

and every other species on Earth

使用的每一种蛋白质。

done in a few months.

这一切仅花了几个月时间。

And we're going to be able to go through

我们将以同样的速度

cures for diseases at the same rate.

找到疾病的治疗方法。

So we're going to simulate trials digitally.

我们将用数字的方式进行模拟，

It'll be much safer.

这样会更安全，

It'll be a million times faster.

并快上百万倍。

And by the end of this decade, as we go into the 2030s,

在我们进入 2030 年代时，

we're going to achieve a new milestone.

我们将达到新的里程碑，

It's called longevity escape velocity.

称为“寿命逃逸速度”。

Let me say that again,

我再重复一遍，

because you're going to be hearing a lot about that.

因为你们将会听到很多这方面的信息。

寿命逃逸速度。

Longevity escape velocity.

Right now you go through a year

现在你度过一年，

and you use up a year of your longevity.

就减寿一年。

However, scientific progress is also progressing,

但是，随着科学的进步，

which is actually bringing us back.

它将带我们返老还童。

It's giving us cures for diseases, new forms of treatment.

科学提供治愈疾病的新疗法。

So right now you're getting back about four months.

现在你能得到约 4 个月的返还；

So you lose a year, you get back four months,

度过 1 年，返还 4 个月，

so you're losing eight months.

所以你只失去了 8 个月。

However, the scientific progress is on an exponential.

但是，科学进程是呈指数增长的，

It's going to get faster and faster.

会越来越快。

And as we get to the early 2030s,

到了 2030 年代初期，

let's say between 2029 and 2035,

也就是 2029 年到 2035 年间，

depending on how diligent you are,

具体时间取决于你们多勤奋，

you're going to get back a full year.

你将得到一整年的返还。

So you lose a year, you get back a year.

即度过一年，就被返还一年。

As we actually go past that point,

当我们逾越那个临界点后，

you'll actually get back more than a year and you'll go backwards in time,

被返还时间甚至会超过一年， 你将返老还童，

which would be cool.

这会很酷。

(Laughter)

（笑声）

Now some people are concerned we’re going to run out of resources.

有些人担心资源会耗尽。

And actually, if we just went ahead and didn't make any new resources,

的确，如果我们只是一味发展， 而不创造新的资源，

we would run out of resources like energy, for example.

资源确实会耗尽，比如能源。

But this is not happening in a vacuum.

但一切并非发生在真空中。

AI is revolutionizing everything.

AI 正在革新一切。

For example, we only have to connect one part in 10,000

例如，我们只需要利用万分之一

of the sunlight that falls on the Earth to meet all of our energy needs.

照射在地球上的太阳光， 就可以满足所有能源需求。

It's plenty of headroom and that's growing exponentially

空间还很大，增长速度也很快，

and will achieve that within ten years.

在 10 年内就可以做到。

And that's also growing exponentially.

增长确实很快。

So ...

因此，

We will have plenty of resources.

我们将会拥有充足的资源。

And when we get to the 2030s,

到 2030 年代，

nanobots will connect our brains to the cloud

纳米机器人将会连接我们的大脑和云端，

就像你的手机一样。

just the way your phone does.

It'll expand intelligence a million-fold by 2045.

到 2045 年，智能将提升一百万倍。

That is the singularity.

这就是奇点。

We will be funnier.

我们会变得更有趣、

(Laughter)

（笑声）

Sexier, smarter, more creative,

更性感、更聪明、更有创造力，

free from biological limitations.

摆脱生物学上的限制。

We'll be able to choose our appearance.

我们将可以选择自己的外貌，

We'll be able to do things we can't do today,

做今天无法做到的事，

like visualize objects in 11 dimensions.

比如将 11 个维度中的物体可视化。

We can speak all languages.

我们会讲所有的语言，

We'll be able to expand consciousness in ways we can barely imagine.

以想象不到的方式扩展我们的意识，

We will experience richer culture with our extra years.

利用延长的生命去感受更丰富的文化。

So I've recently become a grandfather

最近我当爷爷了，

我很期待，

I'm very much looking forward to that,

spending more time with family, friends,

与我的家人和朋友度过更多时光，

loving and being loved,

爱和被爱，

all enhanced by AI.

借助 AI，一切都将更加美好。

I believe this gives life its greatest meaning.

我相信这才是生命最重要的意义。

Thank you very much.

非常感谢各位。

(Applause)

（掌声）