Hod Lipson: Robots that are "self-aware"

譯者: Sofia Lee 審譯者: Marie Wu

So, where are the robots?

嗯? 所以呢? 那些機器人咧?

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

40年來一直有人告訴我們, 機器人很快就會出現在這個世界上了。

Very soon they'll be doing everything for us.

很快地, 他們會替我們做每一件事,

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

他們會煮飯, 打掃, 買東西, 購物血拼, 蓋房子, 但是, 他們並沒有出現。

Meanwhile, we have illegal immigrants doing all the work,

現在這當兒, 我們雇用非法移民來替我們完成所有的工作,

but we don't have any robots.

但是, 我們還是沒有機器人呀!

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

所以, 對於這件事我們可以做些什麼? 或者說些什麼呢?

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

所以, 我想跟你們分享一些不同的觀點,

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

看看我們能怎樣從不同的角度看待這些事。

And this is an x-ray picture

這是一張大甲蟲和瑞士名錶的X光圖,

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

是在1988年拍攝的, 你們看看這裡----

what was true then is certainly true today.

當年確實存在的, 現在還是存在。

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

我們還是能做出零件, 而且是對的零件,

We can make the circuitry of the right computational power,

我們可以畫出具有運算功能的電路圖,

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

但是我們卻沒有辦法把他們組合在一起然後創造出一個東西,

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

而那個東西又要能夠跟這些系統一樣運作良好又具備適應能力。

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

那麼讓我來試著從不一樣的角度看看,

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

我們來召喚一個設計師--他是所有設計師的老師:

Let's see what evolution can do for us.

我們來看看演化為我們做了些什麼。

So we threw in -- we created a primordial soup

我們創造一種最原始的湯汁,

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

我們丟入很多機器人的碎片, 裡面包含了拉桿, 引擎和神經,

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

把他們全部放在一起, 然後讓他們面對物競天擇、

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

突變,並依據他們發展的情況給予獎賞。

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

這是很簡單的工作,而且觀察這個演化過程也十分有趣。

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

仔細一瞧,你就會發現很多不一樣的機器被創造出來了

come out of this. They all move around.

他們到處走來走去,

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

他們都往不同的方向爬,你們可以在右邊

that we actually made a couple of these things,

看到我們做出來的成果,

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

他們都可以在現實生活中執行任務, 你看到的這些都不是多高檔的機器人,

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

但是他們卻完全依照我們所給的獎賞而演化。

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

這些雖然都是在電腦上模擬出來的,

but we can also do that on a real machine.

但我們也可以讓真的機器做出相同的事。

Here's a physical robot that we actually

這就是一個我們實際上可以看到的機器人,

have a population of brains,

他有好幾個大腦,

competing, or evolving on the machine.

這幾個大腦在這機器上彼此競爭並且演化,

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

就像賽馬一樣: 他們會騎到機器上,

and they get rewarded for how fast or how far

他們之中讓機器跑得越快越遠的,

they can make the machine move forward.

就可以得到越多獎賞。

And you can see these robots are not ready

現在你可以看到這些機器人

to take over the world yet, but

還沒準備好取代人類統治這個世界,

they gradually learn how to move forward,

不過他們慢慢學會要怎麼往前走了,

and they do this autonomously.

而且他們是完全自主地前進著。

So in these two examples, we had basically

在剛剛提到的兩個例子裡, 我們基本上擁有兩種機器,

machines that learned how to walk in simulation,

第一種是以電腦模擬的方式學習走路,

and also machines that learned how to walk in reality.

第二種則是在現實生活中學習前進。

But I want to show you a different approach,

但是我要讓你們看的是另一種更不一樣的方式,

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

請看, 這隻機器人有四隻腳,

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

八個引擎, 四個在膝蓋的地方, 另外四個在臀部。

It has also two tilt sensors that tell the machine

他還有兩個傾斜感應器,

which way it's tilting.

用來感應自己向哪裡傾斜了。

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

但是這機器並不知道他自己長什麼樣子,

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

你看得到他, 所以知道他有四隻腳。

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

但是這機器卻沒辦法知道他自己是一條蛇還是一顆樹,

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

他完全不清楚自己長什麼樣子,

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

但是他會想辦法知道。

Initially, it does some random motion,

一開始, 他會隨機做一些動作,

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

接著他試圖看出自己大概長什麼樣子--

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

你會看到他腦海中浮現非常多東西,

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

有很多他自己創造的動作模式, 他試圖去釐清

between actuation and sensing. It then tries to do

動作和感知之間的關係─然後他再試著做第二個動作,

a second action that creates the most disagreement

那個動作不在既有的動作模式內,

among predictions of these alternative models,

完全出乎我們的意料,

like a scientist in a lab. Then it does that

就像在實驗室裡的科學家一樣。接著, 他重複那個動作,

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

並且試著解釋那個動作, 然後創造出自己的動作模式。

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

這是他最後一次重覆這整個循環, 你可以看到他

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

已經弄清楚自己的樣子了, 一旦他整理出自己的動作模式,

it can use that to derive a pattern of locomotion.

就可以從中發展出一種運動模式。

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

你現在看到的是幾個機器─

a pattern of locomotion.

一種運動模式。

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

我們期待他做出一種如惡魔或者蜘蛛般的行走模式,

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

但是他卻創造出這種看似殘障的前進方法。

But when you look at that, you have to remember

但是當你看著他前進的時候, 你必須記得,

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

這機器並沒有做過任何往前行進的物理試驗,

nor did it have a model of itself.

他也沒有任何屬於自己的模式,

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

他等於是自己發現了自己的樣子, 然後找出前進的方法,

and then actually tried that out.

並實際驗證成功。

(Applause)

(掌聲響起)

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

那麼現在, 我們再來看看另一種想法,

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

那是我們將幾個─

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

把幾個放在一塊兒就會......好啦好啦好啦!!!

(Laughter)

(笑聲)

-- they don't like each other. So

─他們不太喜歡對方, 所以

there's a different robot.

這是另一個機器人。

That's what happened when the robots actually

剛剛的事情都是因為機器人做對動作,

are rewarded for doing something.

並且得到獎勵才發生的。

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

那如果們不給他們獎勵, 直接把他們丟在一塊會怎麼樣呢?

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

所以我們拿來了這些立方體, 就像你在圖上看到的,

The cube can swivel, or flip on its side,

他們會旋轉或者翻轉。

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

我們把一千個這樣的立方體放入"原始湯汁"裡─

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

這是電腦模擬效果─我們沒有給他們任何獎勵,

we just let them flip. We pump energy into this

就讓他們翻轉而已。我們給他們一些能量,

and see what happens in a couple of mutations.

看看經過幾次突變以後會怎樣。

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

剛開始什麼都沒發生, 他們就只是跳來跳去,

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

但又過了一下, 你就會看到右邊那些

on the right there begin to take over.

藍色的東西開始掌控全局。

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

他們開始自我複製, 由此可見就算沒有獎勵,

the intrinsic reward is self-replication.

他們也會用自我複製的方式獎勵自己。

And we've actually built a couple of these,

實際上我們已經製造了好幾個像這樣的玩意兒,

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

這是用這樣的立方體做出來的機器人其中的一部分,

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

我們用快轉的方式,讓你看看這機器人

carrying out some of its replication process.

進行自我複製的過程。

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

如果你多餵給這個機器人一些原料─那些立方體─

and more energy, and it can make another robot.

還有很多能量, 他可以製造出另一個機器人。

So of course, this is a very crude machine,

當然, 這是一個很粗糙的機器,

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

但是我們正努力做出這種機器人的縮小版,

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

希望這些立方體可以小到跟粉末一樣。

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

好吧!那麼我們學到了些什麼呢?這些機器人

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

本身不一定多有用, 但他們卻可以教會我們一些事情,

about how we can build better robots,

關於我們如何做出更好的機器人,

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

甚至是人類或動物創造自我模式跟學習的機制原理。

And one of the things that I think is important

還有一樣我覺得最重要的，

is that we have to get away from this idea

就是我們要放棄

of designing the machines manually,

以人工設計機器的想法，

but actually let them evolve and learn, like children,

放手讓機器自己演化與學習, 像孩子一樣，

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

這或許才是讓我們成功的辦法, 謝謝！

(Applause)

(掌聲)