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

翻译人员: 校对人员: Yolanda Zhang

In my early days as a graduate student,

当我还在读研究生的时候

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

我在巴哈马海岸参加了一次潜泳之旅

I'd actually never swum in the ocean before,

我之前从没在海里游过泳

so it was a bit terrifying.

心里难免有些忐忑

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

让我印象最深的是 当我把头潜入水中时

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

很努力地试图通过通气管呼吸

this huge group of striped yellow and black fish

这时一大群拥有黄黑相间条纹的鱼

came straight at me ...

径直向我游过来

and I just froze.

我瞬间变得不知所措

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

之后鱼群好像突然改变了主意

came towards me and then swerved to the right

游到我面前 然后突然右转

and went right around me.

与我擦肩而过

It was absolutely mesmerizing.

这鱼群彻底惊艳到我了

Maybe many of you have had this experience.

也许你们中的许多人也有过这样的经历

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

当然 令人着迷的不但 有它的艳丽和壮美

but there was also just the sheer oneness of it,

同时也有它纯粹的同一性

as if it wasn't hundreds of fish

好像它不是成百上千条鱼

but a single entity with a single collective mind

而是单一 有唯一集体思维的个体

that was making decisions.

在做决定

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

回想起来 这段经历 对我职业生涯的大部分工作

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

起到了决定性的作用

I'm a computer scientist,

我是一个计算机科学家

and the field that I work in is artificial intelligence.

从事的领域是人工智能（AI）

And a key theme in AI

人工智能中一个关键的主题

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

是能通过创造我们自己的 计算机系统来了解智能

that display intelligence the way we see it in nature.

这些系统展现了我们在 自然中所观察到的智能

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

如今AI最受欢迎的想法是 源于科幻小说和电影

and I'm personally a big Star Wars fan.

我本人就是星球大战的超级粉丝

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

但那些关于AI的想法 往往以人类为中心

When you think of a fish school,

当你想到鱼群

or when I think of a flock of starlings,

或当我想到一群八哥时

that feels like a really different kind of intelligence.

那是一种全新的关于智能的体会

For starters, any one fish is just so tiny

首先 每一条鱼在庞大鱼群的映衬下

compared to the sheer size of the collective,

都显得非常微不足道

so it seems that any one individual

所以似乎任何一个个体

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

对于所发生事物的视角都是 非常有限和片面的

and intelligence isn't really about the individual

智能似乎与个体无关

but somehow a property of the group itself.

而是群体的某种属性

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

其次 我觉得最引人注目的是

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

鱼群中没有指挥者监督它们

Instead, this incredible collective mind behavior

这种简直不可思议的集体行为

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

仅仅源于个体之间的互动

Somehow, there are these interactions or rules of engagement

在相邻的鱼之间以某种方式

between neighboring fish

存在着合作的规律

that make it all work out.

得以呈现这样的景象

So the question for AI then becomes,

所以AI的问题演变成了

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

引导这种智能背后的规则是什么呢

and of course, can we create our own?

当然还包括 我们可以把它创造出来吗

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

这就是我和我的团队 在实验室中的主要工作内容

We work on it through theory,

我们通过理论研究

looking at abstract rule systems

观察其中抽象的规则系统

and thinking about the mathematics behind it.

然后思考蕴藏在背后的运算规则

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

我们也会和生物学方面的 研究人员密切合作

But mostly, we do it through robotics,

但最主要的是 我们通过机器人进行研究

where we try to create our own collective systems

尝试着去创造 属于我们自己的群体系统

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

使其可以完成我们在 自然中所观察到的现象

or at least try to.

至少可以尝试一下

One of our first robotic quests along this line

由此 我们第一批的机器人探索活动

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

创造了属于我们的 由一千个机器人组成的群体

So very simple robots,

非常简单的机器人

but they could be programmed to exhibit collective intelligence,

但可以通过编程让它们展现集体智慧

and that's what we were able to do.

这也是我们能够实现的

So this is what a single robot looks like.

这是单个的机器人

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

它非常小 只有一枚硬币大小

and you can program how it moves,

你可以通过编程控制它的移动

but it can also wirelessly communicate with other robots,

但它也可以和其他的机器人无线连接

and it can measure distances from them.

同时也可以测量它们之间的距离

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

现在我们可以开始 精确地编码一个交互运动

a rule of engagement between neighbors.

两个相邻机器人之间得运动规则

And once we have this system,

一旦我们有了这个系统

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

我们就能开始编码你可以 在自然界中观察到的

that you would see in nature.

许多不同种类的规则

So for example, spontaneous synchronization,

例如自发同步现象

how audiences are clapping and suddenly start all clapping together,

观众是怎样鼓掌 怎样突然一起开始鼓掌的

the fireflies flashing together.

萤火虫如何同步闪烁

We can program rules for pattern formation,

我们可以编码图案形成的规则

how cells in a tissue

组织中的细胞

determine what role they're going to take on

如何决定它们将会起什么样的作用

and set the patterns of our bodies.

进而决定我们的人体结构的

We can program rules for migration,

我们可以编码迁移的规则

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

这样一来我们真的是在借鉴自然法则

But we can also take it a step further.

但我们也可以更进一步

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

我们可以将这些从自然中学到的规律

and combine them and create entirely new collective behaviors

结合在一起 创造全新的 属于我们自己的

of our very own.

集体行为

So for example,

例如

imagine that you had two different kinds of rules.

想象一下你有两种不同的规则

So your first rule is a motion rule

第一种是运动规则

where a moving robot can move around other stationary robots.

运动的机器人可以在 其他固定的机器人周围移动

And your second rule is a pattern rule

第二种是样式规则

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

机器人参考相邻的两个机器人 呈现与其不同的颜色

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

如果从一小组机器人开始

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

结果显示这两种规则已经 足够实现让群体自我组装成

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

一个简单的线性图案这一目标

And if I have more complicated pattern rules,

如果我用更复杂的样式规则

and I design error correction rules,

并且设计错误更正规则

we can actually create really, really complicated self assemblies,

我们就可以创造 非常复杂的自组装结构了

and here's what that looks like.

给大家展示一下初步成果

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

这里你们可以看到一千个机器人

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

合作自发组成字母K

The K is on its side.

完成后的字母位于上方

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

重要的是没有一个 机器人起指导作用

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

每个机器人 只和周围的一些机器人交流

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

它利用自身的运动规则 在建了一半的结构周围运动

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

只是根据其样式规律 寻找一个合适的位置

And even though no robot is doing anything perfectly,

即使没有一个机器人的行为是完美的

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

这些规则却足以使我们通过集体协作

顺利达成目标

robustly together.

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

这种幻想近乎完美

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

你们甚至不需要注意到 它们是个体机器人

and it becomes a single entity,

它们变成了一个单一的实体

kind of like the school of fish.

就像刚刚提到的鱼群一样

So these are robots and rules in two dimensions,

这些是从二维层面看 机器人及其运动规则

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

但我们也可以从三维的角度 考虑机器人和规则

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

如果我们能创造 有建筑能力的机器人呢

And here, we can take inspiration from social insects.

看看这里 我们从 社会性昆虫身上获得了启发

So if you think about mound-building termites

如果你想到了白蚁丘

or you think about army ants,

或者是行军蚁

they create incredible, complex nest structures out of mud

它们用泥土 甚至用它们自己的身体

and even out of their own bodies.

惊人 复杂的巢穴结构

And like the system I showed you before,

就像我之前给你们展示的系统一样

these insects actually also have pattern rules

这些昆虫实际上也有排布规则

that help them determine what to build,

可以帮助它们决定要建造什么

but the pattern can be made out of other insects,

但是这种样式结构 也可以由其他昆虫造出来

or it could be made out of mud.

或者是用泥土

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

我们也可以将同样的想法 用于创建机器人规则上

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

这里你们可以看到一些模拟机器人

So the simulated robot has a motion rule,

模拟机器人有自己的运动规则

which is how it traverses through the structure,

指导它如何穿越结构体

looking for a place to fit in,

找到一个合适的地方

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

同时它也有排布规则 它会查看组块

to decide whether to place a block.

来决定是否放下一个方块

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

有了正确的运动和样式规则之后

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

我们就能让机器人建造出 我们想要的任何东西

And of course, everybody wants their own tower.

当然每个人都想要属于自己的塔楼

(Laughter)

（笑）

So once we have these rules,

一旦我们有了这些规则

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

我们就可以开始创造 能遵守这些规则的机器人

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

在这儿你们能看到 可以爬上方块的机器人

but it can also lift and move these blocks

但它也可以举起和移动这些方块

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

可以编辑它所在的结构体

But with these rules,

但即便有了这些规则

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

这也仅仅只是你们能想到的 机器人中的一种

You could imagine many different kinds of robot bodies.

你们可以想象许多不同种类的机器人

So if you think about robots that maybe could move sandbags

如果你觉得机器人也许可以移动沙袋

and could help build levees,

并帮助建筑堤坝

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

或者我们想象机器人 可以用柔软的材料造成

and worked together to shore up a collapsed building --

共同支撑倒塌的建筑

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

只是把同样的规则 用于不同种类的机器人

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

或是像我的团队一样 你完全着迷于行军蚁

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

那么也许有一天我们能 让机器人攀爬所有的东西

including other members of their tribe,

包括它们部落中的其它成员

and self-assemble things out of their own bodies.

或者用它们自身自组装一些东西

Once you understand the rules,

一旦你明白了这些规律

just many different kinds of robot visions become possible.

许多不同的机器人愿景都将成为可能

And coming back to the snorkeling trip,

再回到那个潜泳之旅

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

事实上我们非常了解鱼群的运动规则

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

如果我们能创造一个那样的群体

then maybe there is a future

那么也许在未来

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

我的团队将和我们创造的鱼群 展开一次潜泳之旅

Each of these systems that I showed you

我向你们展示的每一个系统

brings us closer to having the mathematical and the conceptual tools

都使我们更接近利用数学和概念工具

to create our own versions of collective power,

来创造属于我们的集体力量这一目标

and this can enable many different kinds of future applications,

这也使许多不同的未来应用成为可能

whether you think about robots that build flood barriers

机器人可以建筑防洪堤

or you think about robotic bee colonies that could pollinate crops

可以给农作物授粉

or underwater schools of robots that monitor coral reefs,

或是让水下机器人模拟珊瑚礁

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

又或者假设我们抵达某个星球

constellations of satellites.

并设计了卫星的行星

In each of these systems,

在每一个系统中

being able to understand how to design the rules of engagement

能明白如何设计交互规则

and being able to create good collective behavior

以及创造良好的集体行为

becomes a key to realizing these visions.

成为了实现这些愿景的关键

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

那么到目前为 止 我谈到了昆虫 鱼群

and for robots,

还有机器人的各种行为规则

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

但是那些适用于人类集体的规则呢

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

我想留给你们的最后一点思考是

is that science is of course itself

科学本身 毫无疑问就是

an incredible manifestation of collective intelligence,

集体智慧的伟大体现

but unlike the beautiful fish schools that I study,

但不同于我所研究的美丽鱼群

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

我觉得我们还有很长的一段路要走

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

除了要致力于改进机器人集体的科学

I also work on creating robots and thinking about rules

我还致力于创造机器人和思考

that will improve our own scientific collective.

能够改善我们的科学团队的规则

There's this saying that I love:

有一句谚语我很喜欢

who does science determines what science gets done.

科学的成果掌握在从事科学的人手中

Imagine a society

想象这样一个社会

我们有交互规则

where we had rules of engagement

where every child grew up believing that they could stand here

使得每个孩子从小就 相信自己长大后能站在这里

and be a technologist of the future,

并成为未来的技术专家

or where every adult

或者每一个成年人

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

都相信他们不仅有能力去理解 更有能力去改变

how science and technology impacts their everyday lives.

科技如何影响他们的日常生活

What would that society look like?

那样的社会该有多么美好啊

I believe that we can do that.

我相信我们能做得到

I believe that we can choose our rules,

我相信我们可以选择自己的规则

and we engineer not just robots

我们不仅能设计机器人

but we can engineer our own human collective,

更可以设计我们的人类群体

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

在我们开始行动时 一切都将变得非常美好

Thank you.

谢谢

(Applause)

（掌声）