How we can teach computers to make sense of our emotions | Raphael Arar

Translator: Ivana Korom Reviewer: Joanna Pietrulewicz

譯者: Lilian Chiu 審譯者: Yanyan Hong

I consider myself one part artist and one part designer.

我自認一部分是藝術家， 一部分是設計師。

And I work at an artificial intelligence research lab.

我在一間人工智能研究實驗室工作。

We're trying to create technology

我們在試圖創造的技術，

that you'll want to interact with in the far future.

是在遙遠的未來 你會想要和它互動的技術。

Not just six months from now, but try years and decades from now.

不只是現在算起六個月之後， 而是數年、數十年之後。

And we're taking a moonshot

我們做了個很大膽的猜測，

that we'll want to be interacting with computers

猜想我們將來會想要和電腦

in deeply emotional ways.

用帶有深度感情的方式來互動。

So in order to do that,

為了做到這一點，

the technology has to be just as much human as it is artificial.

這項技術不能只是很人工， 也要很人性。

It has to get you.

它必須要能懂你。

You know, like that inside joke that'll have you and your best friend

就像只有你和你最好的朋友之間 才懂的「圈內笑話」

on the floor, cracking up.

能讓你們笑到在地上打滾。

Or that look of disappointment that you can just smell from miles away.

或是你遠遠就能嗅到的 那種失望表情。

I view art as the gateway to help us bridge this gap between human and machine:

我把藝術視為是一種途徑， 能協助我們在人、機之間搭起橋樑：

to figure out what it means to get each other

找出「了解彼此」是什麼意思，

so that we can train AI to get us.

這樣才能訓練人工智能來了解我們。

See, to me, art is a way to put tangible experiences

對我而言，藝術是一種 可以把有形的經驗

to intangible ideas, feelings and emotions.

放到無形的想法、感受， 及情緒中的方式。

And I think it's one of the most human things about us.

我認為它是我們 最有人性的事物之一。

See, we're a complicated and complex bunch.

我們是很複雜又難懂的物種。

We have what feels like an infinite range of emotions,

我們的情緒範圍感覺 起來似乎是無限的，

and to top it off, we're all different.

更麻煩的是，我們每個人都不同。

We have different family backgrounds,

我們有不同的家庭背景、

different experiences and different psychologies.

不同的經驗，以及不同的心理。

And this is what makes life really interesting.

正是因為如此，人生才有意思。

But this is also what makes working on intelligent technology

但也是因為如此，

發展人工智能才極度困難。

extremely difficult.

And right now, AI research, well,

而現在，人工智能研究，嗯，

it's a bit lopsided on the tech side.

它有點不平衡，傾向技術面。

And that makes a lot of sense.

那是非常合理的。

See, for every qualitative thing about us --

對於人類的質性成分──

you know, those parts of us that are emotional, dynamic and subjective --

我們有很多情緒、 動態、主觀的部分──

we have to convert it to a quantitative metric:

我們得將之轉換為量性的度量：

something that can be represented with facts, figures and computer code.

可以用事實、數字， 和電腦程式來呈現的東西。

The issue is, there are many qualitative things

問題在於，有許多質性的東西

that we just can't put our finger on.

是我們實在無法認出來的。

So, think about hearing your favorite song for the first time.

試想看看，當你第一次 聽見你最喜歡的歌曲時，

What were you doing?

你在做什麼？

How did you feel?

你的感受是什麼？

Did you get goosebumps?

你有起雞皮疙瘩嗎？

Or did you get fired up?

你有充滿激情嗎？

Hard to describe, right?

很難形容，對吧？

See, parts of us feel so simple,

我們的一些部分感覺起來很簡單，

but under the surface, there's really a ton of complexity.

但在表面底下， 是相當可觀的複雜度。

And translating that complexity to machines

要把那種複雜翻譯給機器了解，

is what makes them modern-day moonshots.

讓這個目標變得非常高難度。

And I'm not convinced that we can answer these deeper questions

我不相信我們能夠只用零和一來回答

with just ones and zeros alone.

這些較深的問題，

So, in the lab, I've been creating art

所以，在實驗室中， 我一直在創造藝術，

as a way to help me design better experiences

藝術協助我為最先進的技術

for bleeding-edge technology.

設計出更佳的體驗。

And it's been serving as a catalyst

它一直是一種催化劑，

to beef up the more human ways that computers can relate to us.

用來讓電腦用更人性的 方式和我們相處。

Through art, we're tacking some of the hardest questions,

我們透過藝術來處理 一些最困難的問題，

like what does it really mean to feel?

比如「去感覺」究竟是什麼意思？

Or how do we engage and know how to be present with each other?

或者我們要如何與人互動， 如何在彼此相處時不是心不在焉？

And how does intuition affect the way that we interact?

直覺又如何影響我們互動的方式？

So, take for example human emotion.

以人類情緒為例。

Right now, computers can make sense of our most basic ones,

現在，電腦能夠理解 我們最基本的情緒，

like joy, sadness, anger, fear and disgust,

比如喜悅、悲傷、 憤怒、恐懼，及作噁，

by converting those characteristics to math.

做法是將那些特徵轉換為數學。

But what about the more complex emotions?

但遇到更複雜的情緒怎麼辦？

You know, those emotions

你們知道的，

that we have a hard time describing to each other?

有些情緒我們都很難對彼此形容？

Like nostalgia.

像是鄉愁。

So, to explore this, I created a piece of art, an experience,

為了探究這一點，我創作出了 一件藝術作品，一項體驗，

that asked people to share a memory,

我請大家分享一段記憶，

and I teamed up with some data scientists

我和一些資料科學家合作，

to figure out how to take an emotion that's so highly subjective

想辦法把一種高度主觀性的情緒

and convert it into something mathematically precise.

轉換成很精確的數學。

So, we created what we call a nostalgia score

於是我們創造出了鄉愁分數，

and it's the heart of this installation.

它是這項裝置的核心。

To do that, the installation asks you to share a story,

做法是這樣的：這項裝置 會先請你分享一個故事，

the computer then analyzes it for its simpler emotions,

接著，電腦會分析 這個故事中比較簡單的情緒，

it checks for your tendency to use past-tense wording

它會檢查你是否傾向 用比較多過去式修辭，

and also looks for words that we tend to associate with nostalgia,

也會尋找我們談鄉愁時 比較會用到的字眼，

like "home," "childhood" and "the past."

比如「家」、「童年」，和「過去」。

It then creates a nostalgia score

接著，它就會算出鄉愁分數，

to indicate how nostalgic your story is.

表示你的故事有多麼具有鄉愁。

And that score is the driving force behind these light-based sculptures

那個分數，就是這些以光線為 基礎之雕塑背後的驅動力，

that serve as physical embodiments of your contribution.

這些雕塑就是將你的貢獻 給具體化的結果。

And the higher the score, the rosier the hue.

分數越高，色調就會越偏玫瑰色。

You know, like looking at the world through rose-colored glasses.

就像是戴上玫瑰色的 眼鏡來看世界。

So, when you see your score

所以，當你看到你的分數

and the physical representation of it,

以及它的實體代表呈現之後，

sometimes you'd agree and sometimes you wouldn't.

有時你可能可以認同，有時不能。

It's as if it really understood how that experience made you feel.

它就像是真正去了解 那體驗帶給你什麼感覺。

But other times it gets tripped up

但其他時候，這裝置會犯錯，

and has you thinking it doesn't understand you at all.

讓你認為它完全不了解你。

But the piece really serves to show

但這件作品實際上呈現出

that if we have a hard time explaining the emotions that we have to each other,

如果我們都很難向彼此 解釋我們的某些情緒，

how can we teach a computer to make sense of them?

我們要如何教電腦理解那些情緒？

So, even the more objective parts about being human are hard to describe.

即使是人類比較客觀的 部分，也很難形容。

Like, conversation.

比如，對談。

Have you ever really tried to break down the steps?

你可曾真正試過把它拆解成步驟？

So think about sitting with your friend at a coffee shop

試想一下，你和朋友 坐在咖啡廳裡，

and just having small talk.

只是隨意閒聊。

How do you know when to take a turn?

怎麼知道何時要換人說話？

How do you know when to shift topics?

怎麼知道何時要換主題？

And how do you even know what topics to discuss?

甚至，怎麼知道要聊什麼主題？

See, most of us don't really think about it,

我們大部分人並不會去思考這些，

because it's almost second nature.

因為我們早就習慣成自然。

And when we get to know someone, we learn more about what makes them tick,

當我們漸漸了解一個人， 就會更清楚什麼會打動他，

and then we learn what topics we can discuss.

然後就會知道我們能討論什麼話題。

But when it comes to teaching AI systems how to interact with people,

但若要教導人工智能系統 怎麼和人互動，

we have to teach them step by step what to do.

我們得要教它們每一個步驟。

And right now, it feels clunky.

現在，感覺還很笨拙。

If you've ever tried to talk with Alexa, Siri or Google Assistant,

如果你曾經試過和 Alexa、 Siri，或 Google 助理說話，

you can tell that it or they can still sound cold.

你就知道它們聽起來還是很冰冷。

And have you ever gotten annoyed

你是否曾被它們惹惱，

when they didn't understand what you were saying

因為它們聽不懂你在說什麼，

and you had to rephrase what you wanted 20 times just to play a song?

你得要換二十種說法， 只為播放一首歌？

Alright, to the credit of the designers, realistic communication is really hard.

好吧，設計師是很辛苦的， 真實的溝通真的很難。

And there's a whole branch of sociology,

而社會學還有一整個分支，

called conversation analysis,

就叫做談話分析，

that tries to make blueprints for different types of conversation.

試圖為不同類型的對談繪製出藍圖。

Types like customer service or counseling, teaching and others.

像客服、諮詢、教學和其他類型。

I've been collaborating with a conversation analyst at the lab

我在實驗室和一位對談分析師合作，

to try to help our AI systems hold more human-sounding conversations.

試圖協助我們的人工智慧系統 在進行對談時聽起來更像人類。

This way, when you have an interaction with a chatbot on your phone

這麼一來，當你用手機 和聊天機器人互動，

or a voice-based system in the car,

或和車上的語音系統互動時，

it sounds a little more human and less cold and disjointed.

它聽起來會更像人一點， 不那麼冰冷，不那麼沒條理。

So I created a piece of art

我創作了一件藝術作品，

that tries to highlight the robotic, clunky interaction

試圖強調出機器式、笨拙的互動，

to help us understand, as designers,

來協助我們設計師了解

why it doesn't sound human yet and, well, what we can do about it.

為什麼它聽起來還不像人， 以及對此我們能怎麼辦。

The piece is called Bot to Bot

這件作品叫機器人對機器人，

and it puts one conversational system against another

它讓一個交談系統 和另一個交談系統互動，

and then exposes it to the general public.

接著讓系統接觸一般民眾。

And what ends up happening is that you get something

最後發生的狀況就是， 你得到一種產物，

that tries to mimic human conversation,

它會試圖模仿人類交談，

but falls short.

但達不到標準。

Sometimes it works and sometimes it gets into these, well,

有時它行得通，

有時它會陷入誤解的迴圈當中。

loops of misunderstanding.

So even though the machine-to-machine conversation can make sense,

雖然機器對機器的交談是有意義的，

grammatically and colloquially,

在文法上和口語上皆是如此，

it can still end up feeling cold and robotic.

但它最後的感覺 可能還是很冰冷、很機械化。

And despite checking all the boxes, the dialogue lacks soul

儘管所有的條件都做到了， 對話還是沒有靈魂，

and those one-off quirks that make each of us who we are.

缺乏讓我們每個人 獨一無二的個人特色。

So while it might be grammatically correct

所以，即使文法是正確的，

and uses all the right hashtags and emojis,

所有「＃」和表情符號的 應用也都沒錯，

it can end up sounding mechanical and, well, a little creepy.

聽起來還是很機械， 且還有一點詭異。

And we call this the uncanny valley.

我們稱之為恐怖谷。

You know, that creepiness factor of tech

技術中的怪異因子，

where it's close to human but just slightly off.

很接近人類，但又還差那麼一點。

And the piece will start being

這件作品將開始成為一種方式，

one way that we test for the humanness of a conversation

讓我們能夠測試對話的人性

and the parts that get lost in translation.

和翻譯中迷失的部分。

So there are other things that get lost in translation, too,

在翻譯中還有其他的東西會遺失，

like human intuition.

比如人類直覺。

Right now, computers are gaining more autonomy.

目前，電腦越來越自動化。

They can take care of things for us,

它們能為我們處理事情，

like change the temperature of our houses based on our preferences

比如根據我們的偏好 幫我們改變房子裡的室溫，

and even help us drive on the freeway.

甚至協助我們在高速公路上開車。

But there are things that you and I do in person

但有些我們個人會做的事情，

that are really difficult to translate to AI.

很難翻譯讓人工智能理解。

So think about the last time that you saw an old classmate or coworker.

想想看你上回碰見一位 老同學或同事的情況。

Did you give them a hug or go in for a handshake?

你是給他一個擁抱，還是握個手？

You probably didn't think twice

你可能根本不用思考，

because you've had so many built up experiences

因為你有許多既有經驗，

that had you do one or the other.

會影響你決定做前者或後者。

And as an artist, I feel that access to one's intuition,

身為藝術家，我覺得 正是因為會使用直覺，

your unconscious knowing,

潛意識中所知道的事，

is what helps us create amazing things.

才讓我們能創造出了不起的事物。

Big ideas, from that abstract, nonlinear place in our consciousness

大點子是來自於我們意識中 那塊抽象、非線性的地方，

that is the culmination of all of our experiences.

我們所有經驗的最高點。

And if we want computers to relate to us and help amplify our creative abilities,

若我們想讓電腦能和我們相處， 並協助強化我們的創意能力，

I feel that we'll need to start thinking about how to make computers be intuitive.

我覺得我們得要開始思考 如何讓電腦能有直覺。

So I wanted to explore how something like human intuition

所以，我想要探究的是， 像人類直覺這類東西

could be directly translated to artificial intelligence.

如何能被直接翻譯成 人工智慧能懂的語言。

And I created a piece that explores computer-based intuition

我創作了一件作品， 來探究實體空間中

in a physical space.

以電腦為基礎的直覺。

The piece is called Wayfinding,

這件作品叫做「找路」，

and it's set up as a symbolic compass that has four kinetic sculptures.

它的設計是個象徵性的羅盤， 具有四個動態雕塑。

Each one represents a direction,

每一個都代表一個方向，

north, east, south and west.

北、東、南、西。

And there are sensors set up on the top of each sculpture

每個雕塑上都裝有感測器，

that capture how far away you are from them.

能知道你離它們有多遠。

And the data that gets collected

資料會被收集起來，

ends up changing the way that sculptures move

最後會改變雕塑移動的方式，

and the direction of the compass.

以及羅盤的方向。

The thing is, the piece doesn't work like the automatic door sensor

重點是，這件作品並不是像 自動門的感測器那樣運作，

that just opens when you walk in front of it.

當你走到自動門前時它就會打。

See, your contribution is only a part of its collection of lived experiences.

它在收集活體的經驗， 而你的貢獻只是其中一部分。

And all of those experiences affect the way that it moves.

所有經驗都會影響它移動的方式。

So when you walk in front of it,

所以，當你走到它前方時，

it starts to use all of the data

它會開始用所有資料，

that it's captured throughout its exhibition history --

所有它在展示期間 所捕捉到的資料──

or its intuition --

可以說是它的直覺──

to mechanically respond to you based on what it's learned from others.

根據它從其他人身上學到的， 來對你做出機械式的反應。

And what ends up happening is that as participants

最後的結果是，我們參與者開始

we start to learn the level of detail that we need

學到許多細節， 我們需要這些細節，

in order to manage expectations

才能管理來自人類

from both humans and machines.

以及機器的期望。

We can almost see our intuition being played out on the computer,

我們幾乎可以看見 我們的直覺在電腦上呈現出來，

picturing all of that data being processed in our mind's eye.

描繪出我們心靈之眼所看見的 所有被處理的資料。

My hope is that this type of art

我希望這類藝術

will help us think differently about intuition

能夠協助我們對直覺 有不同的想法，

and how to apply that to AI in the future.

及未來要如何 把它用在人工智能上。

So these are just a few examples of how I'm using art to feed into my work

這些只是幾個例子，

用來說明我這個 人工智能設計師兼研究者，

as a designer and researcher of artificial intelligence.

如何把藝術注入工作中。

And I see it as a crucial way to move innovation forward.

我認為，要讓創新再向前邁進， 這種方式十分關鍵。

Because right now, there are a lot of extremes when it comes to AI.

因為，目前談到人工智能時， 會有許多極端想法。

Popular movies show it as this destructive force

熱門電影把人工智能 呈現成毀滅性的力量，

while commercials are showing it as a savior

廣告又把它呈現成救星，

to solve some of the world's most complex problems.

用來解決世界上最困難的一些問題。

But regardless of where you stand,

但，不論你的立場為何，

it's hard to deny that we're living in a world

很難否認，我們所居住的世界

that's becoming more and more digital by the second.

每一秒鐘都在變得越來越數位。

Our lives revolve around our devices, smart appliances and more.

我們的生活圍繞著我們的裝置、 智慧設備等等在轉動。

And I don't think this will let up any time soon.

我不認為近期內這會減緩下來。

So, I'm trying to embed more humanness from the start.

所以，我試著打從一開始 就嵌入更多的人性。

And I have a hunch that bringing art into an AI research process

我有預感，將藝術帶入

人工智能的研究過程是一種方法。

is a way to do just that.

Thank you.

謝謝。

(Applause)

（掌聲）