What intelligent machines can learn from a school of fish | Radhika Nagpal

譯者: Lilian Chiu 審譯者: Wilde Luo

In my early days as a graduate student,

在我剛開始成為研究生的時候，

I went on a snorkeling trip off the coast of the Bahamas.

我到巴哈馬海岸去浮潛。

I'd actually never swum in the ocean before,

我其實從未在海洋中游泳過，

so it was a bit terrifying.

所以我有點害怕。

What I remember the most is, as I put my head in the water

我最難忘的是，當我把頭沉入水中，

and I was trying really hard to breathe through the snorkel,

並竭力透過呼吸管呼吸，

this huge group of striped yellow and black fish

有一大群黃黑條紋的魚

came straight at me ...

筆直朝我遊來……

and I just froze.

我呆住了。

And then, as if it had suddenly changed its mind,

然後，牠們好像突然轉念了一樣，

came towards me and then swerved to the right

朝我過來之後就向右急轉彎，

and went right around me.

從我身邊繞過。

It was absolutely mesmerizing.

那實在非常迷人。

Maybe many of you have had this experience.

也許在座有許多人有過這種體驗。

Of course, there's the color and the beauty of it,

當然，魚群的顏色及美麗都很難忘，

but there was also just the sheer oneness of it,

但牠們還有著一種純粹的一體感，

as if it wasn't hundreds of fish

彷彿牠們並不是數百條魚，

but a single entity with a single collective mind

而是一個整體，包含著 一個做出決策的集體思維。

that was making decisions.

When I look back, I think that experience really ended up determining

回想起來，我認為那段經歷 使我最終下定決心

what I've worked on for most of my career.

去做這份佔據我大半生涯的工作。

I'm a computer scientist,

我是個計算機科學家，

and the field that I work in is artificial intelligence.

我研究的領域是人工智慧。

And a key theme in AI

人工智慧的關鍵主題 是要能理解「智慧」的本質，

is being able to understand intelligence by creating our own computational systems

做法是創建自己的計算系統 （computational system)

that display intelligence the way we see it in nature.

來展現類似於自然生物的智慧。

Now, most popular views of AI, of course, come from science fiction and the movies,

當然，目前最熱門的人工智慧觀點 來自科幻小說和電影，

and I'm personally a big Star Wars fan.

我個人是《星際大戰》的忠實粉絲。

But that tends to be a very human-centric view of intelligence.

但那往往是個非常 以人為中心的智慧觀。

When you think of a fish school,

當你思考魚群

or when I think of a flock of starlings,

或想像一群椋鳥，

that feels like a really different kind of intelligence.

那感覺是一種完全 不同的智慧形式。

For starters, any one fish is just so tiny

首先，和整體魚群的大小相比較，

compared to the sheer size of the collective,

一條魚真的是太小了，

so it seems that any one individual

所以，似乎其中任何一個個體

would have a really limited and myopic view of what's going on,

對正在發生的事應該 眼光短淺、缺乏遠見。

and intelligence isn't really about the individual

而且「智慧」並不體現在個體身上，

but somehow a property of the group itself.

而是團體本身的一種特性。

Secondly, and the thing that I still find most remarkable,

第二，我仍然認為是最了不起的事，

is that we know that there are no leaders supervising this fish school.

就是我們知道在這魚群中 並不存在管理著群體的領導者。

Instead, this incredible collective mind behavior

反而，這個集體思維 所做出的非凡行為

is emerging purely from the interactions of one fish and another.

單純來自魚與魚間的互動。

Somehow, there are these interactions or rules of engagement

不知何故，相鄰近的魚之間 會存在著這些互動，

between neighboring fish

或者說是約定好的行為規則，

that make it all work out.

從而產生這集體行為。

So the question for AI then becomes,

所以，對人工智慧的問題變成是：

what are those rules of engagement that lead to this kind of intelligence,

是什麼約定規則產生這種智慧的？

and of course, can we create our own?

當然還有，我們能否自己創造一個？

And that's the primary thing that I work on with my team in my lab.

這是我與團隊的實驗研究主題。

We work on it through theory,

我們透過理論來研究，

looking at abstract rule systems

探究抽象的規則系統，

and thinking about the mathematics behind it.

思考其背後的數學原理。

We also do it through biology, working closely with experimentalists.

我們也透過生物學來研究，

與實驗者密切合作。

But mostly, we do it through robotics,

但最主要是通過機器人研究，

where we try to create our own collective systems

嘗試創造我們自己的集體系統，

that can do the kinds of things that we see in nature,

讓系統能做出，或至少試著做出 自然界中的智慧行為。

or at least try to.

One of our first robotic quests along this line

我們最初以這種方式 在機器人方面的探索之一，

was to create our very own colony of a thousand robots.

是創造我們自己的千人機器人群體。

So very simple robots,

機器人非常簡單，

but they could be programmed to exhibit collective intelligence,

但能通過程式設計讓它們 展現出集體智慧，

and that's what we were able to do.

這是我們能夠做到的。

So this is what a single robot looks like.

單個的機器人看起來是這樣的。

It's quite small, about the size of a quarter,

它很小，約 25 分硬幣的大小，

and you can program how it moves,

你可以設計程式來規範它如何移動，

but it can also wirelessly communicate with other robots,

它也能以無線的方式 和其他機器人溝通，

and it can measure distances from them.

能測量與其他機器人的距離。

And so now we can start to program exactly an interaction,

我們就可以開始 針對一套互動規則來設計程式，

a rule of engagement between neighbors.

指定鄰近機器人之間的行為規則。

And once we have this system,

一旦有了這個系統，

we can start to program many different kinds of rules of engagement

我們就可針對自然界中的 各類約定規則來編寫程式。

that you would see in nature.

So for example, spontaneous synchronization,

比如「自發性同步」，

how audiences are clapping and suddenly start all clapping together,

一旦有觀眾開始拍手， 全部都驟然跟著拍手，

the fireflies flashing together.

螢火蟲也會一起發光。

We can program rules for pattern formation,

我們可以編寫圖案形成的規則， （pattern formation）

how cells in a tissue

組織中的細胞

determine what role they're going to take on

如何決定它們將扮演什麼角色

and set the patterns of our bodies.

並設定我們身體的模式。

We can program rules for migration,

我們可編寫遷移的規則，

and in this way, we're really learning from nature's rules.

以這種方式，我們能真正地 向自然界的規則學習。

But we can also take it a step further.

但，我們也可以再進一步。

We can actually take these rules that we've learned from nature

我們可以組合這些 向自然界學來的規則，

and combine them and create entirely new collective behaviors

創造出我們自己的、 全新的集體行為。

of our very own.

So for example,

比如，

imagine that you had two different kinds of rules.

想像你有兩種不同的規則。

So your first rule is a motion rule

第一種是動作規則，

where a moving robot can move around other stationary robots.

讓移動中的機器人 可以繞著靜止的機器人轉動。

And your second rule is a pattern rule

第二種是模式規則，

where a robot takes on a color based on its two nearest neighbors.

機器人會根據旁邊 兩名同伴的顔色來呈現顏色。

So if I start with a blob of robots in a little pattern seed,

所以，最開始我只需一小群機器人，

就能埋下一顆「模式種子」，

it turns out that these two rules are sufficient for the group

結果，對這個群體而言，

有這兩種規則就足以自我組裝出

to be able to self-assemble a simple line pattern.

一個簡單的線條樣式。

And if I have more complicated pattern rules,

如果我有更複雜的模式規則

and I design error correction rules,

且設計出修正錯誤的規則，

we can actually create really, really complicated self assemblies,

我們就能實際造出 非常複雜的自我組裝樣式，

and here's what that looks like.

看起來就會像是這樣。

So here, you're going to see a thousand robots

所以，各位將會在這裡 看到一千個機器人，

that are working together to self-assemble the letter K.

它們正在合作並自我組裝出 英文字母「K」。

The K is on its side.

這是一個側過來的 K 。

And the important thing is that no one is in charge.

重要的是，沒有人在主導。

So any single robot is only talking to a small number of robots nearby it,

所以任何一個機器人都只是在 和它附近的少數幾個機器人交談，

and it's using its motion rule to move around the half-built structure

它會用它的動作規則， 在這個半成品周圍移動，

just looking for a place to fit in based on its pattern rules.

根據它的模式規則， 找個適合的位置插進去。

And even though no robot is doing anything perfectly,

雖然沒有任一機器人 完美地做好一件事，

the rules are such that we can get the collective to do its goal

規則是這樣的，

我們可以讓集體一起 穩健地完成目標。

robustly together.

And the illusion becomes almost so perfect, you know --

這個幻覺幾乎完美，

you just start to not even notice that they're individual robots at all,

你甚至會忘了它們各自是個機器人，

and it becomes a single entity,

合起來成了單一的實體，

kind of like the school of fish.

就像一群魚。

So these are robots and rules in two dimensions,

上面這些是二維世界中的 機器人及規則，

but we can also think about robots and rules in three dimensions.

但我們也可以思考 三維世界中的機器人及規則。

So what if we could create robots that could build together?

如果我們造出能 共同建設的機器人會如何呢？

And here, we can take inspiration from social insects.

這裡，我們的靈感來自於群居昆蟲。

So if you think about mound-building termites

如果你想到建立土墩的白蟻

or you think about army ants,

或是行軍蟻，

they create incredible, complex nest structures out of mud

牠們造出很了不起、 很複雜的巢穴結構，

用泥巴，甚至用自己的身體。

and even out of their own bodies.

And like the system I showed you before,

就像我先前給各位看的系統，

these insects actually also have pattern rules

這些昆蟲其實也有模式規則

that help them determine what to build,

來協助牠們決定要建造什麼，

but the pattern can be made out of other insects,

做模型的材料可以是其他昆蟲

or it could be made out of mud.

甚至是泥巴。

And we can use that same idea to create rules for robots.

我們可以把同樣的想法 用來為機器人創造規則。

So here, you're going to see some simulated robots.

在這裡你將看到的 是一些模擬的機器人。

So the simulated robot has a motion rule,

這模擬機器人有一條動作規則：

which is how it traverses through the structure,

以何種方式在結構中來回移動，

looking for a place to fit in,

並尋找一個適合插入的地方。

and it has pattern rules where it looks at groups of blocks

同樣它也有一套模式規則，

使它在看到一堆積木時 決定是否放下手中的積木。

to decide whether to place a block.

And with the right motion rules and the right pattern rules,

有正確的動作規則 和正確的模式規則，

we can actually get the robots to build whatever we want.

我們就能夠讓機器人建造出 任何我們想要的東西。

And of course, everybody wants their own tower.

當然，每個人都想擁有 屬於自己的一座塔。

(Laughter)

（笑聲）

So once we have these rules,

一旦我們有了這些規則，

we can start to create the robot bodies that go with these rules.

我們就可以配合這些規則 開始打造機器人的身體。

So here, you see a robot that can climb over blocks,

在這裡，各位可以看到， 機器人能爬過積木，

but it can also lift and move these blocks

它也可以舉起和搬動這些積木，

and it can start to edit the very structure that it's on.

它可以自己開始修建這個結構。

But with these rules,

但是配合這些規則，

this is really only one kind of robot body that you could imagine.

這其實只是所有你能想到的 機器人身體構造情況中的一種。

You could imagine many different kinds of robot bodies.

你還可想像出多種 不同的機器人身體構造。

So if you think about robots that maybe could move sandbags

所以，你也許可以想像出 會搬移沙袋的機器人，

and could help build levees,

它們能協助築堤，

or we could think of robots that built out of soft materials

我們或許也可用軟材料做機器人，

and worked together to shore up a collapsed building --

共同撐起倒塌的建築物。

so just the same kind of rules in different kinds of bodies.

這只是把同樣的規則 放到不同類的身體中。

Or if, like my group, you are completely obsessed with army ants,

或者，和我的團隊一樣， 你可能對行軍蟻很著迷，

then maybe one day we can make robots that can climb over literally anything

那麼也許有一天

我們做出能爬過任何東西的機器人，

including other members of their tribe,

包括爬過它們自己的夥伴成員，

and self-assemble things out of their own bodies.

用它們自己的身體組裝出東西。

Once you understand the rules,

一旦你瞭解了規則，

just many different kinds of robot visions become possible.

多種不同類型的 機器人遠景都變為可能。

And coming back to the snorkeling trip,

回到我的浮潛之旅，

we actually understand a great deal about the rules that fish schools use.

其實我們瞭解很多魚群的規則。

So if we can invent the bodies to go with that,

所以，若我們能發明出 配合這些規則的身體，

then maybe there is a future

那麼也許在未來，

where I and my group will get to snorkel with a fish school of our own creation.

我和團隊會和我們創造出的 魚群一起浮潛。

Each of these systems that I showed you

每一個我展現給你們的系統

brings us closer to having the mathematical and the conceptual tools

讓我們更進一步邁向 這些數學和概念性工具

to create our own versions of collective power,

來創造我們自己的集體力量，

and this can enable many different kinds of future applications,

這就能讓許多種 未來技術都成為可能，

whether you think about robots that build flood barriers

你可考慮用機器人來建立防洪設施，

or you think about robotic bee colonies that could pollinate crops

用機器蜜蜂群來授粉，

or underwater schools of robots that monitor coral reefs,

或用水底機器人群體來監看珊瑚礁；

or if we reach for the stars and we thinking about programming

或是我們雄心萬丈，

可以考慮為一群衛星設計程式。

constellations of satellites.

In each of these systems,

在所有這些系統中，

being able to understand how to design the rules of engagement

能夠瞭解如何設計出約定規則，

and being able to create good collective behavior

以及能夠創造出好的集體行為，

becomes a key to realizing these visions.

是實現這些遠景的關鍵。

So, so far I've talked about rules for insects and for fish

目前，我已經談過了昆蟲、魚

and for robots,

和機器人之間的規則，

but what about the rules that apply to our own human collective?

那麼用在我們自己 人類群體上的規則呢？

And the last thought that I'd like to leave you with

最後我想留給各位 去思考的一件事是

is that science is of course itself

當然科學本身是

an incredible manifestation of collective intelligence,

集體智慧的一種偉大表現形式，

but unlike the beautiful fish schools that I study,

但不像我研究的美麗魚群，

I feel we still have a much longer evolutionary path to walk.

我覺得我們還有 非常長的演化之路要走。

So in addition to working on improving the science of robot collectives,

所以除了致力於發展機器人 群體的科學研究之外，

I also work on creating robots and thinking about rules

我也從事創造機器人的工作， 並且思考一些規則，

that will improve our own scientific collective.

它將對我們自己的 科學研究群體大有裨益。

There's this saying that I love:

分享一句我喜歡的話：

who does science determines what science gets done.

做科學的人，決定了科學能做什麽。

Imagine a society

想像一個這樣的社會：

where we had rules of engagement

我們有個約定規則：

where every child grew up believing that they could stand here

每個孩子在成長的過程中都相信

他們能站在這個講臺上

and be a technologist of the future,

成為未來的科技專家；

or where every adult

或每個成年人都相信他們有能力

believed that they had the ability not just to understand but to change

不僅理解而且改變 科技對日常生活的影響。

how science and technology impacts their everyday lives.

What would that society look like?

那樣的社會會是怎樣的？

I believe that we can do that.

我相信我們能讓它成真。

I believe that we can choose our rules,

我相信我們能選擇我們的規則，

and we engineer not just robots

除了機器人之外，

but we can engineer our own human collective,

我們也能設計我們自己的人類群體，

and if we do and when we do, it will be beautiful.

如果我們做到了， 世界會變得無比美好。

Thank you.

謝謝。

(Applause)

（掌聲）