Can a computer write poetry? | Oscar Schwartz

譯者: 易帆 余 審譯者: Tim Lan

I have a question.

我有一個問題，

Can a computer write poetry?

電腦可以寫詩嗎？

This is a provocative question.

這是個有爭議的問題。

You think about it for a minute,

你稍微想一下，

and you suddenly have a bunch of other questions like:

腦海裡突然就會浮現出 很多其他的問題：

What is a computer?

例如，甚麼是電腦？

What is poetry?

甚麼是詩？

What is creativity?

甚麼是創造力？

But these are questions

但這些問題，

that people spend their entire lifetime trying to answer,

很多人窮盡一生才能試著給出答案，

not in a single TED Talk.

單單一場ＴＥＤ演說並不能回答。

So we're going to have to try a different approach.

所以，我們必須用不一樣的方法，

So up here, we have two poems.

上面這裡有兩首詩，

One of them is written by a human,

其中一首是人類寫的，

and the other one's written by a computer.

另一首是電腦寫的。

I'm going to ask you to tell me which one's which.

我會讓各位來分辨哪首是誰寫的，

Have a go:

我們開始吧：

Poem 1: Little Fly / Thy summer's play, / My thoughtless hand / Has brush'd away.

１號詩：小蒼蠅，夏天的嘻戲， 我輕率的手，已揮走。

Am I not / A fly like thee? / Or art not thou / A man like me?

難道我，不是像你一樣的蒼蠅， 抑或妳，是像我一樣的人？

Poem 2: We can feel / Activist through your life's / morning /

２號詩：我們可以感受到， 激進派在妳每日生活的清晨出沒

Pauses to see, pope I hate the / Non all the night to start a / great otherwise (...)

暫且停下感受，那我憎惡的教皇 並非每晚都能開始，一個偉大的其他可能...

好的，時間到。

Alright, time's up.

Hands up if you think Poem 1 was written by a human.

認為１號詩是人寫的請舉手，

好的，你們大部分都是。

OK, most of you.

Hands up if you think Poem 2 was written by a human.

認為２號詩是人寫的請舉手，

你們很勇敢，

Very brave of you,

because the first one was written by the human poet William Blake.

因為第一首詩是由詩人William Blake所寫，

The second one was written by an algorithm

第二首詩是由一個演算法所寫出來的，

that took all the language from my Facebook feed on one day

這裡所有的文法是從我 臉書裡一天灌進去的，

and then regenerated it algorithmically,

然後，用演算法重新製作出來的，

according to methods that I'll describe a little bit later on.

關於方法我稍後會提到一些。

So let's try another test.

我們來做另一個測驗，

Again, you haven't got ages to read this,

我再次說明， 你不用花太多時間去讀它，

so just trust your gut.

所以，相信你的直覺。

Poem 1: A lion roars and a dog barks. It is interesting / and fascinating

１號詩：獅吼，狗吠， 鳥飛，卻不吼也不吠，這真迷人且有趣吶

that a bird will fly and not / roar or bark. Enthralling stories about animals

我夢裡有著關於動物的迷人故事

are in my dreams and I will sing them all if I / am not exhausted or weary.

如果我不筋疲力盡或疲憊不堪 我會為他們歌頌。

Poem 2: Oh! kangaroos, sequins, chocolate sodas! / You are really beautiful!

２號詩：喔！袋鼠、亮片、 巧克力蘇打！你們真漂亮！

Pearls, / harmonicas, jujubes, aspirins! All / the stuff they've always talked about (...)

珍珠、口琴、棗子、阿斯匹林！ 全是他們一直提到的東西（...)

Alright, time's up.

好的，時間到。

So if you think the first poem was written by a human,

如果你認為第一首詩是人寫的，

put your hand up.

請舉手。

OK.

好的。

And if you think the second poem was written by a human,

如果你認為第二首詩是人寫的，

put your hand up.

請舉手。

We have, more or less, a 50/50 split here.

我們這裡大約是50／50比例，

這題比較難一點。

It was much harder.

The answer is,

答案是，

the first poem was generated by an algorithm called Racter,

第一首詩是一個名叫Racter的 電腦演算法

that was created back in the 1970s,

在1970年所創造的，

and the second poem was written by a guy called Frank O'Hara,

第二首詩是一位叫 Frank O'Hara的傢伙寫的，

who happens to be one of my favorite human poets.

他意外地成為我最喜歡 的“ 人類詩人”其中之一，

(Laughter)

（笑聲)

So what we've just done now is a Turing test for poetry.

所以，我們為這首詩 做了「圖靈測試」。

The Turing test was first proposed by this guy, Alan Turing, in 1950,

「圖靈測試」在1950年， 由Alan Turing做第一次發表，

in order to answer the question,

是為了回答一個問題：

can computers think?

「電腦會思考嗎？」

Alan Turing believed that if a computer was able

Alan Turing相信，如果電腦能夠

和人類進行一場流暢的以文字交流，

to have a to have a text-based conversation with a human,

結果讓人無法分辨

with such proficiency such that the human couldn't tell

whether they are talking to a computer or a human,

對方是人還是一台電腦，

then the computer can be said to have intelligence.

那麼這台電腦可以被稱呼為 擁有人工智慧。

So in 2013, my friend Benjamin Laird and I,

所以在2013年，我的朋友 Benjamin Laird和我，

we created a Turing test for poetry online.

我們創造了一個 詩的線上圖靈測試程式，

It's called bot or not,

叫做「bot or not」（是不是機器人），

and you can go and play it for yourselves.

你可以上線自己玩玩看。

But basically, it's the game we just played.

但基本上，它就是我們剛剛玩的遊戲，

You're presented with a poem,

你會看到一首詩，

you don't know whether it was written by a human or a computer

你不知道它是人寫的還是電腦寫的，

and you have to guess.

然後你必須猜一猜。

So thousands and thousands of people have taken this test online,

好幾千人已經在線上做測驗，

so we have results.

所以，我們有一個結論，

And what are the results?

那結論是甚麼呢？

Well, Turing said that if a computer could fool a human

Turing說如果電腦可以騙過30%的人，

30 percent of the time that it was a human,

那它就可以被當作人，

then it passes the Turing test for intelligence.

它就通過了圖靈測試。

We have poems on the bot or not database

我們在 bot or not 資料庫裡的詩集

that have fooled 65 percent of human readers into thinking

已經騙過65% 的人，

認為裡面的詩是人寫的。

it was written by a human.

So, I think we have an answer to our question.

所以，我認為我們的問題有答案了，

According to the logic of the Turing test,

根據圖靈測試的邏輯，

can a computer write poetry?

電腦可以寫詩嗎？

Well, yes, absolutely it can.

是的，它絕對可以。

But if you're feeling a little bit uncomfortable

但，如果你覺得對這答案 有點讓你不太舒服，

with this answer, that's OK.

也沒關係，

If you're having a bunch of gut reactions to it,

如果你花了很多時間與它互動，

that's also OK because this isn't the end of the story.

這也沒關係，因為這還沒完。

Let's play our third and final test.

我們來玩第三個 最後一個測驗，

我再說明一下，你們要讀完後，

Again, you're going to have to read

and tell me which you think is human.

告訴我哪一個是人寫的。

Poem 1: Red flags the reason for pretty flags. / And ribbons.

１號詩：紅旗之所以漂亮 除了紅色，還有緞帶

Ribbons of flags / And wearing material / Reasons for wearing material. (...)

旗上的緞帶及耐磨的材質 耐磨材料之所以（．．．）

Poem 2: A wounded deer leaps highest, / I've heard the daffodil

２號詩：受傷的鹿跳最高， 我聽見水仙在訴說，

I've heard the flag to-day / I've heard the hunter tell; /

我今天聽旗子說、 我聽到獵人說；

'Tis but the ecstasy of death, / And then the brake is almost done (...)

這是對死亡的狂喜， 而傷害幾乎已經造成（．．．）

OK, time is up.

好的，時間到。

So hands up if you think Poem 1 was written by a human.

認為１號詩是人寫的請舉手，

Hands up if you think Poem 2 was written by a human.

認為２號詩是人寫的請舉手，

Whoa, that's a lot more people.

哇！多很多人！

So you'd be surprised to find that Poem 1

你會很驚訝地發現，

was written by the very human poet Gertrude Stein.

１號詩由一位純正的人類詩人Gertrude Stein所寫的，

And Poem 2 was generated by an algorithm called RKCP.

而２號詩是一個叫 RKCP演算法所創造的，

Now before we go on, let me describe very quickly and simply,

在我們要繼續以前， 讓我簡單快速描述一下

how RKCP works.

RKCP是如何運作的。

So RKCP is an algorithm designed by Ray Kurzweil,

RKCP是Ray Kurzweil 所設計的演算法，

who's a director of engineering at Google

他是一位谷歌的工程師主管，

and a firm believer in artificial intelligence.

也是一位人工智慧的堅定支持者。

So, you give RKCP a source text,

那麼，你給 RKCP一個來源文字，

it analyzes the source text in order to find out how it uses language,

為了找出要如何使用這個語言， 它會分析來源文字，

and then it regenerates language

然後，它會重新創造一段話來模仿源文字。

that emulates that first text.

So in the poem we just saw before,

所以，我們剛剛看到的詩，

Poem 2, the one that you all thought was human,

你們認為是人類寫的２號詩，

it was fed a bunch of poems

它被灌入了很多一位名叫 Emily Dickinson詩人的詩，

by a poet called Emily Dickinson

it looked at the way she used language,

它取用了這位詩人的語言，

learned the model,

學習她的模式，

and then it regenerated a model according to that same structure.

然後它依據同樣的結構 重製一首詩出來。

但我們對RKCP最需要了解的是，

But the important thing to know about RKCP

is that it doesn't know the meaning of the words it's using.

它不明白它自己用的文字意義，

The language is just raw material,

語言只是它的原料，

it could be Chinese, it could be in Swedish,

它可以是中文，瑞典文，

it could be the collected language from your Facebook feed for one day.

它可以是你臉書上一天的文字。

It's just raw material.

它就只是個原料而已。

And nevertheless, it's able to create a poem

除此之外，它還有辦法寫一首

that seems more human than Gertrude Stein's poem,

比Gertrude Stein寫的還要更有人味的詩，

and Gertrude Stein is a human.

但Gertrude Stein才是人啊...

So what we've done here is, more or less, a reverse Turing test.

所以，我們剛剛做的 差不多就是，反向圖靈測試。

So Gertrude Stein, who's a human, is able to write a poem

所以Gertrude Stein這位人類，

可以寫出讓大部分人 誤認為是電腦寫出來的詩。

that fools a majority of human judges into thinking

that it was written by a computer.

Therefore, according to the logic of the reverse Turing test,

所以，根據圖靈測試的邏輯，

Gertrude Stein is a computer.

Gertrude Stein這人是個電腦．．．（笑聲）

(Laughter)

感覺很困惑嗎？

Feeling confused?

I think that's fair enough.

我認為這情有可原。

So far we've had humans that write like humans,

目前為止，我們有人可以寫出 像是人寫出的詩、

we have computers that write like computers,

我們有電腦可以寫出 像是電腦寫出的詩、

we have computers that write like humans,

我們有電腦可以寫出 像是人寫出的詩，

but we also have, perhaps most confusingly,

但我們同時也有會讓我們搞混 寫詩像電腦的人。

humans that write like computers.

So what do we take from all of this?

所以，我們從這裏面了解到甚麼呢？

Do we take that William Blake is somehow more of a human

我們會認為William Blake

比Gertrude Stein更像是個人嗎？

than Gertrude Stein?

Or that Gertrude Stein is more of a computer than William Blake?

或者Gertrude Stein比 William Blake更像是個電腦？

(Laughter)

（笑聲）

These are questions I've been asking myself

這兩年來，

我一直在問我自己，

for around two years now,

and I don't have any answers.

但我沒有任何答案，

But what I do have are a bunch of insights

但我真的有領悟到很多有關於

about our relationship with technology.

我們與科技的關係。

So my first insight is that, for some reason,

所以，我的第一個領悟是，

we associate poetry with being human.

為了一些原因，我們把 人與詩結合一起，

So that when we ask, "Can a computer write poetry?"

所以，當我們問，＂電腦會寫詩嗎？＂

we're also asking,

我們也在問，

"What does it mean to be human

人的定義是什麼？

and how do we put boundaries around this category?

我們要如何界定、分類呢？

How do we say who or what can be part of this category?"

我們要如何分辨誰或是東西 是歸於哪一類？＂

This is an essentially philosophical question, I believe,

我相信，本質上這是一道哲學的問題，

and it can't be answered with a yes or no test,

而且，這不是像圖靈測試是個

like the Turing test.

對或錯的測試，

I also believe that Alan Turing understood this,

我也相信，

Alan Turing在1950年發明這個理論時， 也了解這一點，

and that when he devised his test back in 1950,

he was doing it as a philosophical provocation.

他當時引發了一個哲學上的爭議。

So my second insight is that, when we take the Turing test for poetry,

我的第二個領悟是， 當我們在為詩做圖靈測試時，

we're not really testing the capacity of the computers

我們並不是真的在測試電腦的能力，

because poetry-generating algorithms,

因為用演算法作詩相當簡單，

they're pretty simple and have existed, more or less, since the 1950s.

而且它們大約在1950年代 早就已經存在了。

我們現在為詩做的圖靈測試，

What we are doing with the Turing test for poetry, rather,

反而，比較像是在收集 甚麼是構成人性的條件。

is collecting opinions about what constitutes humanness.

So, what I've figured out,

所以，我發現，

we've seen this when earlier today,

稍早我們今天看到的，

we say that William Blake is more of a human

我們說William Blake

than Gertrude Stein.

比Gertrude Stein更像個人，

Of course, this doesn't mean that William Blake

當然，這不代表

was actually more human

William Blake比較有人性

or that Gertrude Stein was more of a computer.

或者Gertrude Stein比較像電腦。

It simply means that the category of the human is unstable.

這只能單純的說明， 對人類的界定是不穩定的。

This has led me to understand

這讓我明白了一件事，

that the human is not a cold, hard fact.

就是人性不是冷的、死板的事實，

Rather, it is something that's constructed with our opinions

反倒是一種由我們 的意見所構成的東西，

and something that changes over time.

而這個東西會隨著時間而改變。

So my final insight is that the computer, more or less,

所以我最後的領悟是，

電腦，或多或少只是

works like a mirror that reflects any idea of a human

一面反映我們輸入進去的人類思想的鏡子。

that we show it.

We show it Emily Dickinson,

我們向它展示Emily Dickinson，

it gives Emily Dickinson back to us.

它僅是模仿Emily Dickinson給我們，

We show it William Blake,

我們向它展示William Blake，

that's what it reflects back to us.

它就回應William Blake給我們的，

We show it Gertrude Stein,

我們向它展示Gertrude Stein，

what we get back is Gertrude Stein.

我們得到的回應僅是Gertrude Stein。

還有其他更多的科技也是，

More than any other bit of technology,

the computer is a mirror that reflects any idea of the human we teach it.

電腦只是我們教它甚麼 它就反應甚麼的一面鏡子。

所以，我確定你們大部分人都曾聽過

So I'm sure a lot of you have been hearing

a lot about artificial intelligence recently.

很多有關人工智慧的事情。

And much of the conversation is,

而大部分的對話就類似：

「我們該建造它嗎？」

can we build it?

Can we build an intelligent computer?

「我們可以建立一個智慧型電腦嗎？」

Can we build a creative computer?

「我們可以建立一個創造型電腦嗎？」

What we seem to be asking over and over

我們一次又一次的被問到，

is can we build a human-like computer?

我們可以建立一個 類似人類的電腦嗎？

But what we've seen just now

但就我們剛剛看到的，

is that the human is not a scientific fact,

人類不是一個科學事實，

that it's an ever-shifting, concatenating idea

人類是一個會不斷地變化、串聯想法、

and one that changes over time.

隨時間改變的物種。

So that when we begin to grapple with the ideas

所以，當我們開始要努力克服

of artificial intelligence in the future,

未來人工智慧的這個想法時，

we shouldn't only be asking ourselves,

我們不應該只問我們自己，

"Can we build it?"

「我們可以建造它嗎？」

But we should also be asking ourselves,

我們還得問我們自己，

"What idea of the human do we want to have reflected back to us?"

「我們希望可以得到甚麼樣的人性回應？」

This is an essentially philosophical idea,

這絕對是個哲學想法，

and it's one that can't be answered with software alone,

而且不是單靠軟體就可以回答出來的，

but I think requires a moment of species-wide, existential reflection.

但我認為，這需要一個各類物種 共存的反應時刻，

謝謝各位。

Thank you.

(Applause)

（掌聲）