Hod Lipson: Robots that are "self-aware"

翻译人员: Qing Zhang 校对人员: Yongming Luo

So, where are the robots?

啊，机器人们都在哪儿呢？

We've been told for 40 years already that they're coming soon.

四十年前人们就说它们很快就要来了

Very soon they'll be doing everything for us.

很快，它们就能为我们做一切事情

They'll be cooking, cleaning, buying things, shopping, building. But they aren't here.

它们会做饭，打扫，买东西，购物，甚至是建房子。但直到今天，它们也没能进入我们的生活。

Meanwhile, we have illegal immigrants doing all the work,

这段时间里，非法移民承担着这些工作，

but we don't have any robots.

但我们什么机器人都没有。

So what can we do about that? What can we say?

我们又能做些或说些什么呢？

So I want to give a little bit of a different perspective

我想给大家带来点不同的启发

of how we can perhaps look at these things in a little bit of a different way.

看看今天我们能不能换一个角度看待这些事情

And this is an x-ray picture

这是一张X光片

of a real beetle, and a Swiss watch, back from '88. You look at that --

上面有一只活甲虫，和一只88年的瑞士手表，你看——

what was true then is certainly true today.

无论是当时还是现在

We can still make the pieces. We can make the right pieces.

我们都可以做出这些零件，一模一样的零件，

We can make the circuitry of the right computational power,

我们能做出有同样计算能力的电路，

but we can't actually put them together to make something

但我们并不能把它们放在一起，再做出个什么东西

that will actually work and be as adaptive as these systems.

能和这些系统（甲虫）有一样的适应力。

So let's try to look at it from a different perspective.

那么就让我们试着换个角度再看看这个问题。

Let's summon the best designer, the mother of all designers.

让我们召集最好的设计师，所有设计师的鼻祖，

Let's see what evolution can do for us.

看看进化论能给我们做些什么。

So we threw in -- we created a primordial soup

所以我们就找来了--我们创造了一锅“原汤”，

with lots of pieces of robots -- with bars, with motors, with neurons.

里面有组装机器人需要的各种零件：条状的，带马达的，还有些带神经元的。

Put them all together, and put all this under kind of natural selection,

把它们都放在一起，然后把所有这些放到一种自然选择，

under mutation, and rewarded things for how well they can move forward.

可能产生突变的环境中，并奖励那些发生“进化”的零件组。

A very simple task, and it's interesting to see what kind of things came out of that.

这是一个非常简单的任务，并且得出的结果非常有趣。

So if you look, you can see a lot of different machines

看一下，你能看到各式各样的机器

come out of this. They all move around.

在这种模式下制造出来，它们到处移动，

They all crawl in different ways, and you can see on the right,

以不同的方式爬行，在右边你可以看到

that we actually made a couple of these things,

我们真的做出了几个这样的玩意儿，

and they work in reality. These are not very fantastic robots,

它们在现实中真的能工作。这些还算不上非常先进的机器人，

but they evolved to do exactly what we reward them for:

但它们的确按照我们奖励的方向进化了：

for moving forward. So that was all done in simulation,

那就是向前进化。这些都是模拟的，

but we can also do that on a real machine.

但我们在真的机器上也做成功了，

Here's a physical robot that we actually

这是一个真实的机器人，

have a population of brains,

我们动用了大量的人力脑力

competing, or evolving on the machine.

让机器们相互竞争，共同进化

It's like a rodeo show. They all get a ride on the machine,

这有点像一场套马表演：他们都骑在机器上

and they get rewarded for how fast or how far

根据他们能够驾驶机器前进的速度和距离

they can make the machine move forward.

得到奖励

And you can see these robots are not ready

你可以看到这些机器人都还没完全准备好

to take over the world yet, but

占领这个世界，但是

they gradually learn how to move forward,

它们渐渐地学会如何前进

and they do this autonomously.

并且是自发地学习。

So in these two examples, we had basically

所以从这两个例子中，我们已经基本上

machines that learned how to walk in simulation,

得到了能在虚拟中学习走路

and also machines that learned how to walk in reality.

和在现实中学习走路的机器人。

But I want to show you a different approach,

现在我还要再给你们展示另一项进展，

and this is this robot over here, which has four legs.

就是这里的这个机器人，在这儿，它有四条腿

It has eight motors, four on the knees and four on the hip.

它身上安了八个马达，膝盖上四个，腿上四个

It has also two tilt sensors that tell the machine

它还配备了两个倾斜度传感器，可以告诉自己

which way it's tilting.

正在向哪个方向倾斜。

But this machine doesn't know what it looks like.

但这个机器并不知道自己长啥样

You look at it and you see it has four legs,

你可以看到它长了四条腿，

the machine doesn't know if it's a snake, if it's a tree,

但它自己并不知道自己是一条蛇还是一棵树

it doesn't have any idea what it looks like,

它完全给蒙在鼓里，不晓得自己的相貌

but it's going to try to find that out.

但它马上就要试着找到自己

Initially, it does some random motion,

首先，它会做一些随机的动作

and then it tries to figure out what it might look like.

然后试着弄清楚那些动作都看起来是什么样子的——

And you're seeing a lot of things passing through its minds,

你可以看到它的脑海里闪过许许多多的东西，

a lot of self-models that try to explain the relationship

大量的自我尝试的动作模型，试着理清

between actuation and sensing. It then tries to do

行动和感官之间的关系——然后它将再做第二个动作

a second action that creates the most disagreement

在所有可能的动作模型中

among predictions of these alternative models,

最诡异的一个动作，

like a scientist in a lab. Then it does that

就像科学家在实验室里的试验。接着，它重复那个动作，

and tries to explain that, and prune out its self-models.

并且试着解释那个动作，然后梳理出自己的动作模型。

This is the last cycle, and you can see it's pretty much

这是最后一个环节，你可以看到它已经基本上

figured out what its self looks like. And once it has a self-model,

清楚自己的样子了。一旦它理清自己的动作模型，

it can use that to derive a pattern of locomotion.

就可以从模型中得出一种运动模式。

So what you're seeing here are a couple of machines --

好了，你现在看到的是几个机器——

a pattern of locomotion.

嗯，一种运动模式，

We were hoping that it wass going to have a kind of evil, spidery walk,

我们期待它能产生一种邪恶的，蜘蛛式的运动，

but instead it created this pretty lame way of moving forward.

结果它却自创出这种相当脑残的前进方式。

But when you look at that, you have to remember

但当你看着它前进的时候，你必须记得

that this machine did not do any physical trials on how to move forward,

这个机器并没有接受任何物理指令，控制着它们向前进，

nor did it have a model of itself.

它也没有任何已有的自我模型。

It kind of figured out what it looks like, and how to move forward,

它相当于是自己推理出了自己的样子，以及应该如何向前进

and then actually tried that out.

并且进行了亲身的尝试。

(Applause)

鼓掌~~~♫

So, we'll move forward to a different idea.

那么现在，我们再来看看另一个想法。

So that was what happened when we had a couple of --

那是我们将几个...

that's what happened when you had a couple of -- OK, OK, OK --

把几个...放在一块儿就会...好啦好啦好啦——

(Laughter)

（笑）

-- they don't like each other. So

——它们不大喜欢对方，所以啦~

there's a different robot.

这是另外一个机器人。

That's what happened when the robots actually

刚才的那些都是在机器人做对了动作，

are rewarded for doing something.

获得奖励的情况下发生的。

What happens if you don't reward them for anything, you just throw them in?

那么如果我们不给它们奖励，只是把它们扔到一块，又会怎么样呢？

So we have these cubes, like the diagram showed here.

所以我们拿来了这些立方体，就像这里的这些图，

The cube can swivel, or flip on its side,

它们能旋转，或者翻筋斗

and we just throw 1,000 of these cubes into a soup --

我们把1000个这样的立方体放入“原汤”——

this is in simulation --and don't reward them for anything,

这是模拟效果——我们没给它们任何奖励，

we just let them flip. We pump energy into this

我们就让它们自己活动。我们给它们注入了些能量，

and see what happens in a couple of mutations.

看看经过几次突变，会发生点什么。

So, initially nothing happens, they're just flipping around there.

刚开始的时候，什么也没发生，它们光在那儿跳来跳去。

But after a very short while, you can see these blue things

但过了一小会儿，你可以看到这些蓝色的小东西，

on the right there begin to take over.

它们在右边逐渐地开始占取主动。

They begin to self-replicate. So in absence of any reward,

它们开始自我复制。由此可见即使没有任何奖励

the intrinsic reward is self-replication.

它们也会用自我复制的方式来奖励自己。

And we've actually built a couple of these,

事实上我们已经造了好几个这类的玩意儿，

and this is part of a larger robot made out of these cubes.

这是一部分以这些立方体为单位造出来的大机器人，

It's an accelerated view, where you can see the robot actually

这是快进的效果，可以让你看到机器人

carrying out some of its replication process.

进行自我复制的过程。

So you're feeding it with more material -- cubes in this case --

如果你给它多喂点儿——就是这些立方体——

and more energy, and it can make another robot.

再多给它点能量，它就能自己造出另一个机器人。

So of course, this is a very crude machine,

当然，这还是一个非常粗糙，不成熟的机器，

but we're working on a micro-scale version of these,

但我们正在研究微缩版的这类机器人，

and hopefully the cubes will be like a powder that you pour in.

希望这些立方体能小到像倒出的面粉一般。

OK, so what can we learn? These robots are of course

好的，那么我们都了解到了什么？这些机器人当然

not very useful in themselves, but they might teach us something

自己并不是有很大用处，但它们能教会我们一些东西，

about how we can build better robots,

让我们知道如何造出更好的机器人，

and perhaps how humans, animals, create self-models and learn.

甚至是人类，动物创造自我模型和学习机制的原理。

And one of the things that I think is important

我觉得这其中最重要的，

is that we have to get away from this idea

就是我们必须摒弃之前的观念，

of designing the machines manually,

手动地设计这些机器

but actually let them evolve and learn, like children,

而是让它们自己进化，学习，像孩子一样，

and perhaps that's the way we'll get there. Thank you.

这大概才是我们成功的必经之路。谢谢！

(Applause)

（鼓掌♫）