How will AI change the world?

譯者: Lilian Chiu 審譯者: Helen Chang

在不久的將來，人工智慧 可能會改變你的人生，

In the coming years, artificial intelligence

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

還很可能改變全世界。

但對於改變的方式， 大家很難取得共識。

But people have a hard time agreeing on exactly how.

接下來的片段取自一場訪談，

The following are excerpts from an interview

where renowned computer science professor and AI expert Stuart Russell

知名資訊科學教授 及人工智慧專家史都華‧羅素

helps separate the sense from the nonsense.

要來協助大家區分 什麼有理，什麼是胡說。

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

要求一個人類做某件事情，很不同於

and giving that as the objective to an AI system.

把那件事情設定為 人工智慧系統的目標。

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

當你請一個人類幫你弄一杯咖啡時，

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

你並不是要他把這件事 當作人生的使命，

and nothing else in the universe matters.

且宇宙中其他一切都無所謂，

Even if they have to kill everybody else in Starbucks

就算他得要把星巴克裡 所有人的殺光才能

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

在關門前幫你弄到 一杯咖啡，也應該去做。

不，那不是你的意思。

No, that’s not what you mean.

All the other things that we mutually care about,

我們互相都在乎的所有其他因素 都應該納入行為時的考量。

they should factor into your behavior as well.

而我們現在建造人工智慧 系統的方式，問題在於

And the problem with the way we build AI systems now

我們給系統固定的目標。

is we give them a fixed objective.

演算法需要我們把目標中的 一切都明確定義清楚。

The algorithms require us to specify everything in the objective.

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

如果你說，我們能否處理 海洋酸化的問題？

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

可以，可以找到一種 極有效的催化反應來達成，

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

但那會消耗掉大氣中 四分之一的氧氣，

which would apparently cause us to die fairly slowly and unpleasantly

很顯然那會讓我們死亡，

且過程長達七小時， 是緩慢又不愉快的死法。

over the course of several hours.

So, how do we avoid this problem?

所以我們要如何避免這個問題？

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

你可能會說，好，那就在訂 明確目標的時候更謹慎點——

don’t forget the atmospheric oxygen.

附註別忘了考量大氣中的氧氣。

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

接著，當然，這個反應 在海洋中的一些副作用

poisons all the fish.

毒死了所有的魚類。

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

好，我的意思是也不能害死魚類。

And then, well, what about the seaweed?

那，水草呢？

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

別做任何會造成海草死亡的事。 以此一直類推下去。

And on and on and on.

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

對人類就不需要這麼做，原因是因為

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

人類通常知道

我們在乎之事他們不全都知道。

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

如果你請人類幫你弄一杯咖啡，

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

而你剛好住在巴黎的喬治·桑酒店，

where the coffee is 13 euros a cup,

在那裡的咖啡一杯要價十三歐元。

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

非常合理的行為是回來跟你說，

要十三歐元，你確定你要嗎？ 還是要我去隔壁買？

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

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

對人類來說，這樣做非常正常。

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

人類會問，我打算要 重新油漆你的房子——

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

我先把排水管拿下來， 之後再放回去可以嗎？

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

我們不認為這是 非常精密複雜的能力，

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

但人工智慧系統做不到，因為

我們建造它們的方式 必須要給它們完整的目標。

they have to know the full objective.

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

如果我們建立的系統能知道 它們不知道目標是什麼，

then they start to exhibit these behaviors,

那麼它們就會展現這些行為，

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

比如在把大氣中所有氧氣 都除掉之前會先問可不可以。

In all these senses, control over the AI system

在所有這些意義上， 對人工智慧系統的掌控

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

來自於機器無法完全確定 真正的目標是什麼。

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

當你建造出的機器 非常堅信它們有目標時，

that they have the objective,

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

就會出現這種精神錯亂的行為。

And I think we see the same thing in humans.

我想，在人類身上也能看到這現象。

What happens when general purpose AI hits the real economy?

當通用人工智慧和真實經濟 碰撞時，會發生什麼事？

How do things change? Can we adapt?

會有什麼改變？我們能適應嗎？

This is a very old point.

這是個古老的論點。

Amazingly, Aristotle actually has a passage where he says,

讓人驚奇的是，亞里斯多德 有段話是這樣說的：

look, if we had fully automated weaving machines

如果我們有完全自動化的編織機器

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

和弦撥，不用人類就可以 撥里拉琴和製作音樂，

then we wouldn’t need any workers.

那我們就不需要任何工人。

That idea, which I think it was Keynes

我認為那個想法是凱因斯的，

who called it technological unemployment in 1930,

他在 1930 年稱之為技術性失業，

is very obvious to people.

對大家來說這想法很明顯。

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

他們會想，當然， 如果機器能做這些事，

then I'm going to be unemployed.

那我就會失業。

You can think about the warehouses that companies are currently operating

可以想想目前企業為了 電子商務而經營的倉庫，

for e-commerce, they are half automated.

它們是半自動化的。

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

老式倉庫的運作方式是： 一大堆東西到處堆疊，

all over the place and humans go and rummage around

人類在倉庫中到處翻找， 再把貨拿去寄出，

and then bring it back and send it off—

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

是有機器人會去取得

你要找的東西所屬的那個儲存單位，

that contains the thing that you need,

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

但要由人類把物品 從箱子中取出或從架上取下，

because that’s still too difficult.

因為那仍然太困難。

But, at the same time,

但，同時，

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

你是否會做一個夠精確的機器人，

能夠從你能購買的各種物品中 挑選出任何物品？

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

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

那會不會一下子就消滅了 三、四百萬個工作？

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

艾德華‧摩根‧佛斯特 寫了一個有趣的故事，

where everyone is entirely machine dependent.

故事中，人人都完全仰賴機器。

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

故事要談的事實是，

如果你把文明的管理權交給機器，

the management of your civilization to machines,

you then lose the incentive to understand it yourself

就不會有動機刺激你

自己去了解它， 或教下一代如何了解它。

or to teach the next generation how to understand it.

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

可以看到《瓦力》其實就是現代版，

where everyone is enfeebled and infantilized by the machine,

機器讓所有人都很軟弱、 被當孩子對待，

and that hasn’t been possible up to now.

目前為止還不可能發生。

We put a lot of our civilization into books,

我們的文明有很多 都被我們放到書中，

but the books can’t run it for us.

但書無法為我們經營文明。

And so we always have to teach the next generation.

所以我們一直得要教導下一代。

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

算一算，是大約一兆人年的教與學，

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

以及一條完整的鏈， 可追溯到數萬個世代以前。

What happens if that chain breaks?

如果那條鏈斷了呢？

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

我認為那是隨著人工智慧 向前邁進時我們該去了解的。

The actual date of arrival of general purpose AI—

通用人工智慧出現的確切日期——

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

無法明確指出來， 那並不是單一個日子。

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

現實也不是全有或全無。

The impact is going to be increasing.

衝擊會漸漸增大。

So with every advance in AI,

每當人工智慧有所進展，

it significantly expands the range of tasks.

就會讓工作任務的範圍顯著再擴大。

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

就那方面來說，我想， 大部分的專家說

到了這個世紀末

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

我們非常非常有可能 會有通用人工智慧。

The median is something around 2045.

中位數大概會在 2045 年。

I'm a little more on the conservative side.

我比較偏向保守派一點， 我認為問題比我們想的更困難。

I think the problem is harder than we think.

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

我很喜歡人工智慧創造者 約翰‧麥克菲的說法，

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

被問到這個問題時，他說 是在五年和五百年之間。

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

我想，我們需要好幾個 愛因斯坦才能實現。