How will AI change the world?

翻译人员: Lixin Foo 校对人员: Grace Man

在将来的岁月里，人工智能

In the coming years, artificial intelligence

极有可能会改变 你的生活，甚至全世界。

is probably going to change your life, and likely the entire world.

但人们对于这种改变的 呈现方式结论不一。

But people have a hard time agreeing on exactly how.

以下的采访摘录自

The following are excerpts from an interview

where renowned computer science professor and AI expert Stuart Russell

为我们辟谣的著名计算机科学教授

helps separate the sense from the nonsense.

兼人工智能专家， 斯图尔特·罗素（Stuart Russell）。

There’s a big difference between asking a human to do something

要求一个人做某件事与 将其作为目标交给人工智能系统

and giving that as the objective to an AI system.

是有很大的区别的。

When you ask a human to get you a cup of coffee,

当你拜托一个人帮你拿杯咖啡时，

you don’t mean this should be their life’s mission,

你并不在命令这个人 奉它为人生使命，

and nothing else in the universe matters.

以致宇宙里再也没有更重要的事了。

Even if they have to kill everybody else in Starbucks

就算除掉了星巴克里的其他人

to get you the coffee before it closes— they should do that.

都得在店铺关门之前买到你的咖啡。

No, that’s not what you mean.

不，你不是那个意思。

All the other things that we mutually care about,

其他我们共同在意的事物，

they should factor into your behavior as well.

都应当影响你的所作所为。

And the problem with the way we build AI systems now

而我们现在建造的 人工智能系统的问题在于，

is we give them a fixed objective.

我们给了它们一个固定目标。

The algorithms require us to specify everything in the objective.

算法要求我们规定目标里的一切。

And if you say, can we fix the acidification of the oceans?

比如说，“能修正 海洋的酸化问题吗？”

Yeah, you could have a catalytic reaction that does that extremely efficiently,

没问题，可以形成一道 非常有效率的催化反应，

but it consumes a quarter of the oxygen in the atmosphere,

但这将会吞噬大气层里 四分之一的氧气，

which would apparently cause us to die fairly slowly and unpleasantly

好像会导致我们 全都慢慢地、不愉快地

over the course of several hours.

在几个小时后死去。

So, how do we avoid this problem?

那，我们该如何避免这种问题呢？

You might say, okay, well, just be more careful about specifying the objective—

你可能会说，好吧，那我们 就对目标更具体地说明一下——

don’t forget the atmospheric oxygen.

别忘了大气层里的氧气。

And then, of course, some side effect of the reaction in the ocean

然后海洋里某种效应的副作用

poisons all the fish.

将会毒死鱼儿们。

Okay, well I meant don’t kill the fish either.

好吧，那我就再定义一下， 也别杀了鱼。

And then, well, what about the seaweed?

那么，海藻呢？

Don’t do anything that’s going to cause all the seaweed to die.

也别做什么会让海藻都死光的事。

And on and on and on.

以此类推。

And the reason that we don’t have to do that with humans is that

而我们之所以对于人类 不需要这样做

humans often know that they don’t know all the things that we care about.

是因为人们大都明白自己并不可能 对每个人的爱好无不知晓。

If you ask a human to get you a cup of coffee,

如果拜托一个人去帮你买咖啡，

and you happen to be in the Hotel George Sand in Paris,

而你刚好人在一杯咖啡 13 欧元的 巴黎乔治圣德酒店，

where the coffee is 13 euros a cup,

it’s entirely reasonable to come back and say, well, it’s 13 euros,

你很有可能会再回去问一下： “喂，这里咖啡得 13 欧元，

are you sure you want it, or I could go next door and get one?

你还要吗？要不我去隔壁店里 帮你买杯去？” 这完全合情合理。

And it’s a perfectly normal thing for a person to do.

这对于人类来讲再也正常不过。

To ask, I’m going to repaint your house—

在帮人重涂房子时问起：

is it okay if I take off the drainpipes and then put them back?

“我先把排水管拆了 再装回去，可以吗？”

We don't think of this as a terribly sophisticated capability,

我们并不觉得 这是个特别复杂厉害的能力，

but AI systems don’t have it because the way we build them now,

但人工智能系统没有个能力， 因为在我们当下的建造方法里，

they have to know the full objective.

它们必须熟知全面目标。

If we build systems that know that they don’t know what the objective is,

如果我们建造的系统 明白它们并不了解目标，

then they start to exhibit these behaviors,

它们就会开始展示此类行动：

like asking permission before getting rid of all the oxygen in the atmosphere.

比如在除掉大气层里的氧气 之前先征求许可。

In all these senses, control over the AI system

在这种意义上， 对于人工智能系统的控制

comes from the machine’s uncertainty about what the true objective is.

源于机器对真正目标的不确定性。

And it’s when you build machines that believe with certainty

而只有在建造对目标自以为有着 绝对肯定性的机器时，

that they have the objective,

that’s when you get this sort of psychopathic behavior.

才会产生这种精神错乱的行为。

And I think we see the same thing in humans.

而我觉得对于人类， 也是相同的理念。

What happens when general purpose AI hits the real economy?

通用型人工智能闯进实体经济后， 会发生什么？

How do things change? Can we adapt?

事态会如何改变？ 我们能够调整适应吗？

This is a very old point.

这是个非常古老的讨论点。

Amazingly, Aristotle actually has a passage where he says,

令人惊讶的是，据记载， 亚里士多德（Aristotle）就曾说过：

look, if we had fully automated weaving machines

“看吧，如果我们有了 完全自动化的织布机

and plectrums that could pluck the lyre and produce music without any humans,

与无需人们撩动 就能弹琴奏乐的弦拨，

then we wouldn’t need any workers.

那我们就不需要工人了。”

That idea, which I think it was Keynes

这个主意，也就是 1930 年 在凯恩斯（Keynes）

who called it technological unemployment in 1930,

口中的技术性失业，

is very obvious to people.

对于人们来说，这显而易见。

They think, yeah, of course, if the machine does the work,

想着，嗯，当然了， 如果机器做了我的工，

then I'm going to be unemployed.

那我就要失业了。

You can think about the warehouses that companies are currently operating

我们能想到，如今半自动化运行的 公司电子商务仓库。

for e-commerce, they are half automated.

The way it works is that an old warehouse— where you’ve got tons of stuff piled up

运行方法如下—— 比之让人们在堆积如山的

all over the place and humans go and rummage around

旧形仓库里到处寻找货物，

and then bring it back and send it off—

再取回送出，

there’s a robot who goes and gets the shelving unit

现在由机器人行驶到 你需要的物件的货架单元上，

that contains the thing that you need,

but the human has to pick the object out of the bin or off the shelf,

但仍然需要人把物品 从货架取出，

because that’s still too difficult.

因为那对机器人还是太难了。

But, at the same time,

但与此同时，

would you make a robot that is accurate enough to be able to pick

我们有可能创造出 一个能够在众多可购类别中

pretty much any object within a very wide variety of objects that you can buy?

精准挑选出 任何一件目标物品的机器人吗？

That would, at a stroke, eliminate 3 or 4 million jobs?

以此一举，或会削减 三、四百万有余的工作岗位？

There's an interesting story that E.M. Forster wrote,

E·M·福斯特（E.M. Forster） 写过一篇引人深思的故事：

where everyone is entirely machine dependent.

故事里的人们都完全依赖机器。

The story is really about the fact that if you hand over

其中寓意是， 如果你把文明的管理权

the management of your civilization to machines,

交给了机器，

you then lose the incentive to understand it yourself

那你将会失去自身了解文明、

or to teach the next generation how to understand it.

把文明传承于下一代的动力。

You can see “WALL-E” actually as a modern version,

我们可以将《机器人总动员》 视为现代版：

where everyone is enfeebled and infantilized by the machine,

由于机器，人们变得衰弱与幼儿化，

and that hasn’t been possible up to now.

到目前为止，这还不可能。

We put a lot of our civilization into books,

我们把大量文明注入书籍，

but the books can’t run it for us.

但书籍无法为我们管理文明。

And so we always have to teach the next generation.

所以我们必须一直指导下一代。

If you work it out, it’s about a trillion person years of teaching and learning

计算下来，这是一个往回一万亿年、

and an unbroken chain that goes back tens of thousands of generations.

数以万计的世代之间 绵延不绝的教导与学习的链条。

What happens if that chain breaks?

这条链如果断了，将会如何？

I think that’s something we have to understand as AI moves forward.

随着人工智能的进展， 我认为这是我们必须了解的事情。

The actual date of arrival of general purpose AI—

我们将无法精准地确认 通用型人工智真正来临的时日，

you’re not going to be able to pinpoint, it isn’t a single day.

因为那并不会是一日之劳。

It’s also not the case that it’s all or nothing.

也并不是存在或不存在的两项极端。

The impact is going to be increasing.

这方面的影响力将是与日俱进的。

So with every advance in AI,

所以随着人工智能的进步，

it significantly expands the range of tasks.

它所能完成的任务将显著地扩展。

So in that sense, I think most experts say by the end of the century,

这样一看，我觉得大部分的专家都说

we’re very, very likely to have general purpose AI.

我们极有可能在世纪之末前 生产通用型人工智能。

The median is something around 2045.

中位数位置在 2045 年左右。

I'm a little more on the conservative side.

我对此偏于保守派。

I think the problem is harder than we think.

我认为问题比我们想象的还要难。

I like what John McAfee, he was one of the founders of AI,

我喜欢人工智能的发明家之一 约翰·麦卡菲（John McAfee）

when he was asked this question, he said, somewhere between five and 500 years.

对这个问题的答案： 他说，应该在 5 到 500 年之间。

And we're going to need, I think, several Einsteins to make it happen.

而我觉得，这得要 几位爱因斯坦才能实现。