ABOUT THE SPEAKER
Rodney Brooks - Roboticist
Rodney Brooks builds robots based on biological principles of movement and reasoning. The goal: a robot who can figure things out.

Why you should listen

Former MIT professor Rodney Brooks studies and engineers robot intelligence, looking for the holy grail of robotics: the AGI, or artificial general intelligence. For decades, we've been building robots to do highly specific tasks -- welding, riveting, delivering interoffice mail -- but what we all want, really, is a robot that can figure things out on its own, the way we humans do.

Brooks realized that a top-down approach -- just building the biggest brain possible and teaching it everything we could think of -- would never work. What would work is a robot who learns like we do, by trial and error, and with many separate parts that learn separate jobs. The thesis of his work which was captured in Fast, Cheap and Out of Control,went on to become the title of the great Errol Morris documentary.

A founder of iRobot, makers of the Roomba vacuum, Brooks is now founder and CTO of Rethink Robotics, whose mission is to apply advanced robotic intelligence to manufacturing and physical labor. Its first robots: the versatile two-armed Baxter and one-armed Sawyer. Brooks is the former director of CSAIL, MIT's Computers Science and Artificial Intelligence Laboratory.

 
More profile about the speaker
Rodney Brooks | Speaker | TED.com
TED2003

Rodney Brooks: Robots will invade our lives

罗德尼·布鲁克斯预言:机器人将全面进入我们的生活

Filmed:
689,918 views

在这个2003年的预言式的演讲中,机器人学者罗德尼·布鲁克斯谈到了机器人将怎样在我们的生活中起作用——从成为玩具开始,到做家务...或者更多的事情。
- Roboticist
Rodney Brooks builds robots based on biological principles of movement and reasoning. The goal: a robot who can figure things out. Full bio

Double-click the English transcript below to play the video.

00:18
What I want to tell you about today今天 is how I see robots机器人 invading入侵 our lives生活
0
0
5000
今天我想跟大家讨论的话题是:我怎样看待机器人将多层面、跨时代地
00:23
at multiple levels水平, over multiple timescales时间表.
1
5000
3000
进入我们的生活。
00:26
And when I look out in the future未来, I can't imagine想像 a world世界, 500 years年份 from now,
2
8000
4000
当我展望未来的时候,我无法想象从现在起的五百年后,
00:30
where we don't have robots机器人 everywhere到处.
3
12000
2000
机器人竟然没有遍布世界。
00:32
Assuming假设 -- despite尽管 all the dire可怕的 predictions预测 from many许多 people about our future未来 --
4
14000
5000
假设——不管那些大多数人对未来所做的一切恐怖预言——
00:37
assuming假设 we're still around, I can't imagine想像 the world世界 not being存在 populated人口稠密 with robots机器人.
5
19000
4000
假设我们仍然存在,我无法想象这个世界不被机器人所占据。
00:41
And then the question is, well, if they're going to be here in 500 years年份,
6
23000
3000
那么问题就是,如果五百年后必然在这个世界上,
00:44
are they going to be everywhere到处 sooner than that?
7
26000
2000
那么他们会不会更早的时候就已无处不在了?
00:46
Are they going to be around in 50 years年份?
8
28000
2000
他们会不会在五十年之后就到处都有了?
00:48
Yeah, I think that's pretty漂亮 likely容易 -- there's going to be lots of robots机器人 everywhere到处.
9
30000
3000
是啊,我觉得那非常可能——现在已经有很多机器人在我们周围了。
00:51
And in fact事实 I think that's going to be a lot sooner than that.
10
33000
3000
实际上我认为这个进程还要快得多。
00:54
I think we're sort分类 of on the cusp风口浪尖 of robots机器人 becoming变得 common共同,
11
36000
4000
我认为我们可以说是处于日益普遍的机器人发展尖峰,
00:58
and I think we're sort分类 of around 1978 or 1980 in personal个人 computer电脑 years年份,
12
40000
6000
而且我认为我们可以说在1978或1980年左右的个人电脑时代,
01:04
where the first few少数 robots机器人 are starting开始 to appear出现.
13
46000
3000
新式的机器人才刚刚开始出现。
01:07
Computers电脑 sort分类 of came来了 around through通过 games游戏 and toys玩具.
14
49000
4000
电脑就像是通过游戏和玩具发展而来的。
01:11
And you know, the first computer电脑 most people had in the house
15
53000
3000
而且大家知道,大多数人的第一台家用电脑
01:14
may可能 have been a computer电脑 to play Pong,
16
56000
2000
是用来玩Pong(一种简单的视频游戏)的,
01:16
a little microprocessor微处理器 embedded嵌入式,
17
58000
2000
仅仅用一个小小的嵌入式微处理器就可以支持了,
01:18
and then other games游戏 that came来了 after that.
18
60000
3000
然后就涌现了更多其他游戏。
01:21
And we're starting开始 to see that same相同 sort分类 of thing with robots机器人:
19
63000
3000
我们现在开始发现机器人也发生了同样的事情:
01:24
LEGOLEGO Mindstorms头脑风暴, FurbiesFurbies -- who here -- did anyone任何人 here have a Furby菲比?
20
66000
4000
比如LEGO头脑风暴和Furbies(两者都是电子机器人玩具)——这里谁曾买过Furby?
01:28
Yeah, there's 38 million百万 of them sold出售 worldwide全世界.
21
70000
3000
对,这种玩具全球卖出了三千八百万套。
01:31
They are pretty漂亮 common共同. And they're a little tiny robot机器人,
22
73000
2000
它们很普通,只是小巧的机器人,
01:33
a simple简单 robot机器人 with some sensors传感器,
23
75000
2000
带着一些传感器的简单机器人。
01:35
a little bit of processing处理 actuation启动.
24
77000
2000
仅仅有一些驱动处理。
01:37
On the right there is another另一个 robot机器人 doll娃娃, who you could get a couple一对 of years年份 ago.
25
79000
3000
在右边有另外一个机器人娃娃,你可能几年前就可以买到。
01:40
And just as in the early days,
26
82000
2000
就像是早些年,
01:42
when there was a lot of sort分类 of amateur业余 interaction相互作用 over computers电脑,
27
84000
5000
有很多可以说是业余水平的在电脑上的互动,
01:47
you can now get various各个 hacking黑客 kits, how-to-hack如何对黑客 books图书.
28
89000
4000
而现在你可以买到多种多样的黑客工具包,以及如何做黑客的书籍。
01:51
And on the left there is a platform平台 from Evolution演化 Robotics机器人,
29
93000
4000
在左边有一个来自于“进化机器人”的平台,
01:55
where you put a PC个人计算机 on, and you program程序 this thing with a GUIGUI
30
97000
3000
这里你可以放一台电脑,然后使用图形用户界面来编程,
01:58
to wander漫步 around your house and do various各个 stuff东东.
31
100000
3000
让这个机器人在你的房间里到处巡视并且做很多事情。
02:01
And then there's a higher更高 price价钱 point sort分类 of robot机器人 toys玩具 --
32
103000
3000
然后是一个高价货,可以说是机器人玩具——
02:04
the Sony索尼 Aibo艾博. And on the right there, is one that the NECNEC developed发达,
33
106000
4000
索尼的Aibo。在右边,是NEC开发的一款机器人,
02:08
the PaPeRoPaPeRo, which哪一个 I don't think they're going to release发布.
34
110000
3000
PaPeRo,我认为他们很快就会发布它。
02:11
But nevertheless虽然, those sorts排序 of things are out there.
35
113000
3000
不过不管怎样,这所有的东西都过时了。
02:14
And we've我们已经 seen看到, over the last two or three years年份, lawn-mowing草坪修剪 robots机器人,
36
116000
4000
我们已经看到,在过去的两三年内,出现了锄草机器人,
02:18
HusqvarnaHusqvarna的 on the bottom底部, Friendly友善 Robotics机器人 on top最佳 there, an Israeli以色列 company公司.
37
120000
6000
是屏幕底部的Husqvarna公司生产的,还有顶部那个可爱的机器人,是一家以色列公司生产的。
02:24
And then in the last 12 months个月 or so
38
126000
2000
然后,在最近的一年左右,
02:26
we've我们已经 started开始 to see a bunch of home-cleaning家居清洁 robots机器人 appear出现.
39
128000
4000
我们开始看到一些家用清洁机器人的面世。
02:30
The top最佳 left one is a very nice不错 home-cleaning家居清洁 robot机器人
40
132000
3000
左上角是一个非常好的家用清洁机器人,
02:33
from a company公司 called Dyson戴森, in the U.K. Except it was so expensive昂贵 --
41
135000
4000
它是一家叫“Dyson”的英国公司生产的。不过它相当贵——
02:37
3,500 dollars美元 -- they didn't release发布 it.
42
139000
2000
3500美元——而且并没有上市。
02:39
But at the bottom底部 left, you see Electrolux伊莱克斯, which哪一个 is on sale拍卖.
43
141000
3000
在左下角你可以看到正在发售的Electrolux,
02:42
Another另一个 one from Karcher凯驰.
44
144000
2000
它是Karcher公司的另外一款产品。
02:44
At the bottom底部 right is one that I built内置 in my lab实验室
45
146000
2000
右下角是我10年前在自己的实验室里造出来的机器人,
02:46
about 10 years年份 ago, and we finally最后 turned转身 that into a product产品.
46
148000
3000
最终我们把他做成了一款产品。
02:49
And let me just show显示 you that.
47
151000
2000
让我给你们演示一下吧。
02:51
We're going to give this away I think, Chris克里斯 said, after the talk.
48
153000
4000
我们将会把他送出去,我觉得克里斯在我演讲之后会这样说。
02:55
This is a robot机器人 that you can go out and buy购买, and that will clean清洁 up your floor地板.
49
157000
6000
这是一个你能外出买到的机器人,它能够清洁你的地板。
03:05
And it starts启动 off sort分类 of just going around in ever-increasing不断增加 circles.
50
167000
5000
它就像不断来回绕圈子似的出发、前进。
03:10
If it hits点击 something -- you people see that?
51
172000
4000
如果它撞到什么东西的话——你们看到了么?
03:14
Now it's doing wall-following墙下, it's following以下 around my feet
52
176000
3000
现在它开始沿着墙根走,它在绕着我的脚前进
03:17
to clean清洁 up around me. Let's see, let's --
53
179000
4000
并在我周围做清洁。让我们看一看——
03:21
oh, who stole偷了 my Rice白饭 Krispies脆片? They stole偷了 my Rice白饭 Krispies脆片!
54
183000
5000
哦,谁偷了我的“米饼”?他们偷了我的“米饼”。
03:26
(Laughter笑声)
55
188000
6000
(笑声)
03:32
Don't worry担心, relax放松, no, relax放松, it's a robot机器人, it's smart聪明!
56
194000
3000
别担心,放轻松,对,放轻松,这是一个机器人,它很聪明。
03:35
(Laughter笑声)
57
197000
3000
(笑声)
03:38
See, the three-year-old三十岁 kids孩子, they don't worry担心 about it.
58
200000
4000
看,这些三岁的孩子,他们不担心这一点。
03:42
It's grown-ups大人 that get really upset烦乱.
59
204000
2000
大人们才会感到头疼。
03:44
(Laughter笑声)
60
206000
1000
(笑声)
03:45
We'll just put some crap掷骰子 here.
61
207000
2000
让我们在这儿放点乱七八糟的东西。
03:47
(Laughter笑声)
62
209000
4000
(笑声)
03:51
Okay.
63
213000
2000
很好。
03:53
(Laughter笑声)
64
215000
4000
(笑声)
03:57
I don't know if you see -- so, I put a bunch of Rice白饭 Krispies脆片 there,
65
219000
3000
我不知道你们是否看到——所以,我在这里放一些米饼,
04:00
I put some pennies便士, let's just shoot射击 it at that, see if it cleans清理 up.
66
222000
7000
我放一些硬币,让我们在这边拍摄,来看看它是否能够清理干净。
04:10
Yeah, OK. So --
67
232000
2000
对,很好。所以——
04:12
we'll leave离开 that for later后来.
68
234000
4000
我们把它留到后面。
04:16
(Applause掌声)
69
238000
5000
(鼓掌)
04:22
Part部分 of the trick was building建造 a better cleaning清洁的 mechanism机制, actually其实;
70
244000
4000
这个小把戏的一部分实际上就是建立一个更好的清扫机制;
04:26
the intelligence情报 on board was fairly相当 simple简单.
71
248000
4000
主板里的智能程序其实很简单。
04:30
And that's true真正 with a lot of robots机器人.
72
252000
2000
并且对于大多数机器人来讲这都是真理。
04:32
We've我们已经 all, I think, become成为, sort分类 of computational计算 chauvinists沙文主义者,
73
254000
4000
我认为,我们都变成了计算沙文主义者似的,
04:36
and think that computation计算 is everything,
74
258000
2000
都觉得计算就是一切,
04:38
but the mechanics机械学 still matter.
75
260000
2000
不过机械原理还是很重要。
04:40
Here's这里的 another另一个 robot机器人, the PackBotPackBot机器人,
76
262000
3000
这是另外一个机器人,PackBot,
04:43
that we've我们已经 been building建造 for a bunch of years年份.
77
265000
2000
我们已经几年之前就造出来了。
04:45
It's a military军事 surveillance监控 robot机器人, to go in ahead of troops军队 --
78
267000
6000
它是一种军事侦察机器人,诸如,在军队前面开路,
04:51
looking at caves洞穴, for instance.
79
273000
3000
观察洞穴之类的。
04:54
But we had to make it fairly相当 robust强大的,
80
276000
2000
但我们必须把它做得相当坚固,
04:56
much more robust强大的 than the robots机器人 we build建立 in our labs实验室.
81
278000
7000
要比我们在实验室里做出来的机器人要坚固的多。
05:03
(Laughter笑声)
82
285000
3000
(笑声)
05:12
On board that robot机器人 is a PC个人计算机 running赛跑 LinuxLinux的.
83
294000
4000
它的控制电脑系统是Linux。
05:16
It can withstand经受 a 400G shock休克. The robot机器人 has local本地 intelligence情报:
84
298000
6000
它可以承受400G的冲击力。这个机器人还有随环境而调节自身的智能:
05:22
it can flip翻动 itself本身 over, can get itself本身 into communication通讯 range范围,
85
304000
6000
它可以自己翻转,可以自己进入通讯范围,
05:28
can go upstairs楼上 by itself本身, et cetera等等.
86
310000
3000
可以自己爬楼梯等等。
05:38
Okay, so it's doing local本地 navigation导航 there.
87
320000
4000
好的,现在它在进行地区导航。
05:42
A soldier士兵 gives it a command命令 to go upstairs楼上, and it does.
88
324000
6000
一名士兵给它下达上楼的指令,它立即照做。
05:49
That was not a controlled受控 descent降落.
89
331000
3000
这不是一次预定的降落。
05:52
(Laughter笑声)
90
334000
2000
(笑声)
05:54
Now it's going to head off.
91
336000
2000
现在它将要开始拦截。
05:56
And the big breakthrough突破 for these robots机器人, really, was September九月 11th.
92
338000
5000
其实有关这些机器人最重大的技术突破发生于9月11日。
06:01
We had the robots机器人 down at the World世界 Trade贸易 Center中央 late晚了 that evening晚间.
93
343000
4000
傍晚的时候,我们把这些机器人放在世贸中心下面。
06:06
Couldn't不能 do a lot in the main主要 rubble瓦砾 pile,
94
348000
2000
它们并不能在主要的废墟下做很多事情,
06:08
things were just too -- there was nothing left to do.
95
350000
3000
事情仅仅是——那儿并没有什么可以做的了。
06:11
But we did go into all the surrounding周围 buildings房屋 that had been evacuated疏散,
96
353000
5000
不过我们进入周边那些已经疏散了人群的建筑中,
06:16
and searched搜索 for possible可能 survivors幸存者 in the buildings房屋
97
358000
3000
在那些由于危险而无法进入的建筑中
06:19
that were too dangerous危险 to go into.
98
361000
2000
寻找可能的生还者。
06:21
Let's run this video视频.
99
363000
2000
我们来看看这个视频。
06:23
Reporter记者: ...battlefield战场 companions同伴 are helping帮助 to reduce减少 the combat战斗 risks风险.
100
365000
3000
记者:……战场上的伙伴能够帮助我们降低战斗风险。
06:26
Nick缺口 Robertson罗伯逊 has that story故事.
101
368000
3000
来听听尼克·罗伯特森的故事。
06:31
Rodney罗德尼 Brooks布鲁克斯: Can we have another另一个 one of these?
102
373000
2000
罗德尼•布鲁克斯:我们能从里面拿一个吗?
06:38
Okay, good.
103
380000
2000
好的,很好。
06:43
So, this is a corporal下士 who had seen看到 a robot机器人 two weeks previously先前.
104
385000
3000
这是一个两周前刚刚见过机器人的下士。
06:48
He's sending发出 robots机器人 into caves洞穴, looking at what's going on.
105
390000
4000
他把机器人送入洞穴,去探索那里发生了什么。
06:52
The robot's机器人 being存在 totally完全 autonomous自主性.
106
394000
2000
机器人是全自动的。
06:54
The worst最差 thing that's happened发生 in the cave洞穴 so far
107
396000
4000
迄今为止那个洞穴里发生的最糟糕的事情
06:58
was one of the robots机器人 fell下跌 down ten meters.
108
400000
3000
就是这些机器人中的一个从十米高的地方摔了下去。
07:08
So one year ago, the US military军事 didn't have these robots机器人.
109
410000
3000
一年前,美国军队还没有这些机器人。
07:11
Now they're on active活性 duty义务 in Afghanistan阿富汗 every一切 day.
110
413000
2000
现在这些机器人每天都在阿富汗执行任务。
07:13
And that's one of the reasons原因 they say a robot机器人 invasion侵入 is happening事件.
111
415000
3000
这也是为什么人们说机器人正在入侵的原因之一了。
07:16
There's a sea change更改 happening事件 in how -- where technology's技术的 going.
112
418000
4000
技术以怎样的方式、迈向何方,都将带来巨变。
07:20
Thanks谢谢.
113
422000
2000
谢谢。
07:23
And over the next下一个 couple一对 of months个月,
114
425000
2000
在接下来的几个月里,
07:25
we're going to be sending发出 robots机器人 in production生产
115
427000
3000
我们将把机器人投入油井的生产
07:28
down producing生产 oil wells to get that last few少数 years年份 of oil out of the ground地面.
116
430000
4000
以便我们能够将近几年要用的石油开采出来。
07:32
Very hostile敌对 environments环境, 150˚ C, 10,000 PSIPSI.
117
434000
4000
非常恶劣的环境,150摄氏度的高温,10000PSI的压强。
07:36
Autonomous自主性 robots机器人 going down, doing this sort分类 of work.
118
438000
4000
自动控制的机器人将潜下去,开展工作。
07:40
But robots机器人 like this, they're a little hard to program程序.
119
442000
3000
但是像这样的机器人,其程序会有一点难写。
07:43
How, in the future未来, are we going to program程序 our robots机器人
120
445000
2000
未来,我们将如何给我们的机器人编程序
07:45
and make them easier更轻松 to use?
121
447000
2000
让他们更加易用呢?
07:47
And I want to actually其实 use a robot机器人 here --
122
449000
3000
而且实际上我想在这里使用一个机器人——
07:50
a robot机器人 named命名 Chris克里斯 -- stand up. Yeah. Okay.
123
452000
5000
一个叫做克里斯的机器人——站起来。对,很好。
07:57
Come over here. Now notice注意, he thinks robots机器人 have to be a bit stiff僵硬.
124
459000
4000
过来,到这里来。下面请注意,他认为机器人肯定有一点僵硬。
08:01
He sort分类 of does that. But I'm going to --
125
463000
3000
他就是那么想的。不过我要——
08:04
Chris克里斯 Anderson安德森: I'm just British英国的. RBRB: Oh.
126
466000
2000
克里斯安德森:我只是比较英国化。罗德尼布鲁克斯:哦。
08:06
(Laughter笑声)
127
468000
2000
(笑声)
08:08
(Applause掌声)
128
470000
2000
(掌声)
08:10
I'm going to show显示 this robot机器人 a task任务. It's a very complex复杂 task任务.
129
472000
3000
我将给这个机器人展示一件任务。一件很复杂的任务。
08:13
Now notice注意, he nodded点头 there, he was giving me some indication迹象
130
475000
3000
下面要注意,他点头了,他正在向我表示
08:16
he was understanding理解 the flow of communication通讯.
131
478000
3000
他理解了沟通的流程。
08:19
And if I'd said something completely全然 bizarre奇异的
132
481000
2000
而且如果我说了一些完全异乎寻常的事情,
08:21
he would have looked看着 askance怀疑地 at me, and regulated调控 the conversation会话.
133
483000
3000
他就会以询问的眼神看着我,控制这段对话。
08:24
So now I brought this up in front面前 of him.
134
486000
3000
所以,现在我把这个放在他的面前。
08:27
I'd looked看着 at his eyes眼睛, and I saw his eyes眼睛 looked看着 at this bottle瓶子 top最佳.
135
489000
4000
我看着他的眼睛,然后我发现他的眼睛盯着这个瓶子的上端。
08:31
And I'm doing this task任务 here, and he's checking检查 up.
136
493000
2000
我在做这件事情,而他在检查。
08:33
His eyes眼睛 are going back and forth向前 up to me, to see what I'm looking at --
137
495000
3000
他的眼睛前后移动,看着我正在看的东西,
08:36
so we've我们已经 got shared共享 attention注意.
138
498000
2000
所以我们有共同的关注点。
08:38
And so I do this task任务, and he looks容貌, and he looks容貌 to me
139
500000
3000
所以我在做事情,他在看,而且是看着我,
08:41
to see what's happening事件 next下一个. And now I'll give him the bottle瓶子,
140
503000
4000
想要知道下一步将发生什么。那么现在我将把瓶子给他,
08:45
and we'll see if he can do the task任务. Can you do that?
141
507000
2000
我们看看他是不是能做这件事情。你能做么?
08:47
(Laughter笑声)
142
509000
3000
(笑声)
08:50
Okay. He's pretty漂亮 good. Yeah. Good, good, good.
143
512000
4000
好的,他表现非常好。是的,很好,很好,很好。
08:54
I didn't show显示 you how to do that.
144
516000
2000
我并没有给你演示应该怎么做。
08:56
Now see if you can put it back together一起.
145
518000
2000
下面看看你是不是能够把他们补好。
08:58
(Laughter笑声)
146
520000
2000
(笑声)
09:00
And he thinks a robot机器人 has to be really slow.
147
522000
1000
他认为一个机器人肯定动作相当缓慢。
09:01
Good robot机器人, that's good.
148
523000
2000
好机器人,非常好。
09:03
So we saw a bunch of things there.
149
525000
2000
那么,我们看到了很多。
09:06
We saw when we're interacting互动,
150
528000
3000
大家看到了当我们在交流的时候,
09:09
we're trying to show显示 someone有人 how to do something, we direct直接 their visual视觉 attention注意.
151
531000
4000
我们试着给某些人演示怎么做事情的时候,我们引导着他们视觉的注意力。
09:13
The other thing communicates相通 their internal内部 state to us,
152
535000
4000
其他事情会告诉我们他们的内心状态,
09:17
whether是否 he's understanding理解 or not, regulates调整对象 a social社会 interaction相互作用.
153
539000
3000
他是否正在理解,规范着一次社交活动。
09:20
There was shared共享 attention注意 looking at the same相同 sort分类 of thing,
154
542000
2000
当看着一样的东西时,双方就有了共同的关注点,
09:22
and recognizing认识 socially社交上 communicated传达 reinforcement加强 at the end结束.
155
544000
4000
并且最终会认识到社交的力量。
09:26
And we've我们已经 been trying to put that into our lab实验室 robots机器人
156
548000
3000
我们正试图把这些理念加入我们的实验室机器人
09:29
because we think this is how you're going to want to interact相互作用 with robots机器人 in the future未来.
157
551000
4000
因为我们觉得这才是未来我们想与机器人沟通的方式。
09:33
I just want to show显示 you one technical技术 diagram here.
158
555000
2000
在这里我只想给大家展示一个技术图表。
09:35
The most important重要 thing for building建造 a robot机器人 that you can interact相互作用 with socially社交上
159
557000
4000
要制作一个能与人进行社交的机器人,最重要的就是
09:39
is its visual视觉 attention注意 system系统.
160
561000
2000
它的视觉焦点系统。
09:41
Because what it pays支付 attention注意 to is what it's seeing眼看
161
563000
3000
因为他所关注的焦点就是他所看到
09:44
and interacting互动 with, and what you're understanding理解 what it's doing.
162
566000
3000
和交流的对象,以及所理解的正在发生的事情。
09:47
So in the videos视频 I'm about to show显示 you,
163
569000
3000
因此,在视频里,我将向大家展示,
09:50
you're going to see a visual视觉 attention注意 system系统 on a robot机器人
164
572000
4000
你将看到一个机器人身上的视觉焦点系统——
09:54
which哪一个 has -- it looks容貌 for skin皮肤 tone in HSVHSV space空间,
165
576000
4000
这台机器人在HSV空间中辨别肤色,
09:58
so it works作品 across横过 all human人的 colorings色素.
166
580000
4000
所以他能够适应于所有的人种。
10:02
It looks容貌 for highly高度 saturated饱和的 colors颜色, from toys玩具.
167
584000
2000
它能够识别玩具的高度饱和色。
10:04
And it looks容貌 for things that move移动 around.
168
586000
2000
他也能识别运动的物体。
10:06
And it weights权重 those together一起 into an attention注意 window窗口,
169
588000
3000
他把所有这些放在一个关注窗口中衡量,
10:09
and it looks容貌 for the highest-scoring最高分 place地点 --
170
591000
2000
并找出得到最高分的地方——
10:11
the stuff东东 where the most interesting有趣 stuff东东 is happening事件 --
171
593000
2000
哪里发生最有趣的事情就关注哪里。
10:13
and that is what its eyes眼睛 then segueSEGUE to.
172
595000
4000
那就是他的眼睛所关注的地方。
10:17
And it looks容貌 right at that.
173
599000
2000
他盯着那里。
10:19
At the same相同 time, some top-down自顶向下 sort分类 of stuff东东:
174
601000
3000
与此同时,一些自上而下之类的东西
10:22
might威力 decide决定 that it's lonely孤独 and look for skin皮肤 tone,
175
604000
3000
可能会决定他的孤独,识别肤色,
10:25
or might威力 decide决定 that it's bored无聊 and look for a toy玩具 to play with.
176
607000
3000
或可能会决定他的无聊,找玩具来玩。
10:28
And so these weights权重 change更改.
177
610000
2000
因此这些权重会发生变化。
10:30
And over here on the right,
178
612000
2000
在右边的那个地方,
10:32
this is what we call the Steven史蒂芬 Spielberg斯皮尔伯格 memorial纪念馆 module.
179
614000
3000
是我们所说的史蒂芬斯皮尔伯格记忆模块。
10:35
Did people see the movie电影 "AIAI"? (Audience听众: Yes.)
180
617000
2000
大家看过电影“人工智能”么?听众:是的。
10:37
RBRB: Yeah, it was really bad, but --
181
619000
2000
罗德尼布鲁克斯:哦,那片子拍的真糟糕,不过——
10:39
remember记得, especially特别 when Haley海利 Joel乔尔 Osment奥斯蒙特, the little robot机器人,
182
621000
4000
记住,特别是哈利乔奥斯蒙特,那个小机器人,
10:43
looked看着 at the blue蓝色 fairy仙女 for 2,000 years年份 without taking服用 his eyes眼睛 off it?
183
625000
4000
整整2000年都在盯着那个蓝色精灵,眼睛眨都不眨?
10:47
Well, this gets得到 rid摆脱 of that,
184
629000
2000
那么,这个机器人不会像他那样,
10:49
because this is a habituation习惯 Gaussian高斯 that gets得到 negative,
185
631000
4000
因为这是一个为负的习惯性高斯曲线,
10:53
and more and more intense激烈 as it looks容貌 at one thing.
186
635000
3000
当他盯着一样东西时,曲线会变得越来越陡。
10:56
And it gets得到 bored无聊, so it will then look away at something else其他.
187
638000
3000
然后他会感觉到厌烦,这样他将转开视线去看别的东西。
10:59
So, once一旦 you've got that -- and here's这里的 a robot机器人, here's这里的 Kismet天命,
188
641000
4000
所以,当你知道这些以后——这里有个机器人,叫Kismet,
11:03
looking around for a toy玩具. You can tell what it's looking at.
189
645000
4000
正在四处寻找一个玩具。你能够知道他正在看什么。
11:07
You can estimate估计 its gaze凝视 direction方向 from those eyeballs眼球 covering覆盖 its camera相机,
190
649000
5000
你能从覆盖着下面相机的眼球转动估计到他将看向什么地方,
11:12
and you can tell when it's actually其实 seeing眼看 the toy玩具.
191
654000
3000
你也能知道他什么时候看到了那个玩具。
11:15
And it's got a little bit of an emotional情绪化 response响应 here.
192
657000
2000
这里他有一点点的情绪反应。
11:17
(Laughter笑声)
193
659000
1000
(笑声)
11:18
But it's still going to pay工资 attention注意
194
660000
2000
不过如果比较明显的东西进入他的视野
11:20
if something more significant重大 comes into its field领域 of view视图 --
195
662000
4000
他会一直关注的——
11:24
such这样 as Cynthia辛西娅 BreazealBreazeal, the builder建设者 of this robot机器人, from the right.
196
666000
4000
比如辛西娅布里泽尔,这个机器人的制造者。
11:28
It sees看到 her, pays支付 attention注意 to her.
197
670000
5000
他看到了她,对她很关注。
11:33
Kismet天命 has an underlying底层, three-dimensional三维 emotional情绪化 space空间,
198
675000
4000
Kismet具备底层、三维的情感空间,
11:37
a vector向量 space空间, of where it is emotionally感情上.
199
679000
3000
一个向量的空间,那是有情感的。
11:40
And at different不同 places地方 in that space空间, it expresses表达 --
200
682000
5000
在那个空间中不同的地方,他会表达——
11:46
can we have the volume on here?
201
688000
2000
我们能把音量调高点么?
11:48
Can you hear that now, out there? (Audience听众: Yeah.)
202
690000
2000
现在你们能听到那里面的声音么?听众:是的。
11:50
Kismet天命: Do you really think so? Do you really think so?
203
692000
5000
Kismet:你真这样认为么?你真这样认为么?
11:57
Do you really think so?
204
699000
2000
你真这样认为么?
12:00
RBRB: So it's expressing表达 its emotion情感 through通过 its face面对
205
702000
3000
罗德尼布鲁克斯:所以他在通过他的面部动作和声音的韵律
12:03
and the prosody韵律学 in its voice语音.
206
705000
2000
表达他的情感。
12:05
And when I was dealing交易 with my robot机器人 over here,
207
707000
4000
当我跟我的机器人在这里交流的时候
12:09
Chris克里斯, the robot机器人, was measuring测量 the prosody韵律学 in my voice语音,
208
711000
3000
Chris,这个机器人,正在分析我的声音韵律,
12:12
and so we have the robot机器人 measure测量 prosody韵律学 for four basic基本 messages消息
209
714000
5000
所以,我们已经有了能分析声音韵律来判断四种基本信息的机器人,
12:17
that mothers母亲 give their children孩子 pre-linguistically前语言.
210
719000
4000
这四种信息是妈妈给学说话之前的孩子们传达的。
12:21
Here we've我们已经 got naive幼稚 subjects主题 praising赞美 the robot机器人:
211
723000
3000
现在我们拿出一些幼稚的话题来奖励这个机器人,
12:26
Voice语音: Nice尼斯 robot机器人.
212
728000
2000
声音:乖,
12:29
You're such这样 a cute可爱 little robot机器人.
213
731000
2000
你就是一个可爱的小机器人。
12:31
(Laughter笑声)
214
733000
2000
(笑声)
12:33
RBRB: And the robot's机器人 reacting反应 appropriately适当.
215
735000
2000
这个机器人的反应非常好。
12:35
Voice语音: ...very good, Kismet天命.
216
737000
4000
声音:。。。很好,Kismet。
12:40
(Laughter笑声)
217
742000
2000
(笑声)
12:42
Voice语音: Look at my smile微笑.
218
744000
2000
声音:看着我笑。
12:46
RBRB: It smiles笑容. She imitates模仿 the smile微笑. This happens发生 a lot.
219
748000
3000
罗德尼布鲁克斯:他笑了。她模仿着这个微笑。这样的情形经常发生。
12:49
These are naive幼稚 subjects主题.
220
751000
2000
这些只是比较幼稚的话题。
12:51
Here we asked them to get the robot's机器人 attention注意
221
753000
3000
下面我们让他们吸引这个机器人的注意力
12:54
and indicate表明 when they have the robot's机器人 attention注意.
222
756000
3000
并且当成功的时候向我们示意。
12:57
Voice语音: Hey, Kismet天命, ah, there it is.
223
759000
4000
声音:嘿,Kismet,啊,就是那样。
13:01
RBRB: So she realizes实现 she has the robot's机器人 attention注意.
224
763000
4000
罗德尼布鲁克斯:所以她意识到她已经吸引到了这个机器人的注意力。
13:08
Voice语音: Kismet天命, do you like the toy玩具? Oh.
225
770000
4000
声音:Kismet,你喜欢这个玩具么?噢。
13:13
RBRB: Now, here they're asked to prohibit禁止 the robot机器人,
226
775000
2000
罗德尼布鲁克斯:现在,现在他们被要求教会机器人令行禁止,
13:15
and this first woman女人 really pushes the robot机器人 into an emotional情绪化 corner.
227
777000
4000
首先出来的这个女的确实将机器人推入一个情绪化的角落里了。
13:19
Voice语音: No. No. You're not to do that. No.
228
781000
5000
声音:不,不。你不能做那件事。不。
13:24
(Laughter笑声)
229
786000
3000
(笑声)
13:27
Not appropriate适当. No. No.
230
789000
6000
声音:这不合适。不,不。
13:33
(Laughter笑声)
231
795000
3000
(笑声)
13:36
RBRB: I'm going to leave离开 it at that.
232
798000
2000
罗德尼布鲁克斯:我会把他留在那里。
13:38
We put that together一起. Then we put in turn taking服用.
233
800000
2000
我们把那个放在一起。然后我们进行轮换。
13:40
When we talk to someone有人, we talk.
234
802000
3000
当我们跟某人谈话时,我们就谈话。
13:43
Then we sort分类 of raise提高 our eyebrows眉毛, move移动 our eyes眼睛,
235
805000
4000
然后我们就像是耸动眉毛,转动眼睛那样
13:47
give the other person the idea理念 it's their turn to talk.
236
809000
3000
告诉别人轮到他们讲话了。
13:50
And then they talk, and then we pass通过 the baton指挥棒 back and forth向前 between之间 each other.
237
812000
4000
接着他们开始讲话,这样我们就来来回回的接力。
13:54
So we put this in the robot机器人.
238
816000
2000
所以我们把这个因素加入机器人。
13:56
We got a bunch of naive幼稚 subjects主题 in,
239
818000
2000
我们把很多这样幼稚的话题加入,
13:58
we didn't tell them anything about the robot机器人,
240
820000
2000
我们没有告诉他们任何关于这个机器人的消息,
14:00
satSAT them down in front面前 of the robot机器人 and said, talk to the robot机器人.
241
822000
2000
只是让他们坐在机器人前面,让他们跟机器人讲话。
14:02
Now what they didn't know was,
242
824000
2000
现在他们所不明白的是,
14:04
the robot机器人 wasn't understanding理解 a word they said,
243
826000
2000
机器人完全不懂他们在讲什么,
14:06
and that the robot机器人 wasn't speaking请讲 English英语.
244
828000
3000
而且机器人也不说英语。
14:09
It was just saying random随机 English英语 phonemes音素.
245
831000
2000
他仅仅发出一些英语音节。
14:11
And I want you to watch carefully小心, at the beginning开始 of this,
246
833000
2000
我希望你们能仔细看,最初,
14:13
where this person, Ritchie里奇, who happened发生 to talk to the robot机器人 for 25 minutes分钟 --
247
835000
4000
这个叫Ritchie的人,跟这个机器人讲了25分钟的话——
14:17
(Laughter笑声)
248
839000
2000
(笑声)
14:19
-- says, "I want to show显示 you something.
249
841000
2000
——他说,“我想给你看一些东西。
14:21
I want to show显示 you my watch."
250
843000
2000
我想给你看看我的手表。”
14:23
And he brings带来 the watch center中央, into the robot's机器人 field领域 of vision视力,
251
845000
5000
然后他把手表放在机器人的视野范围内,
14:28
points to it, gives it a motion运动 cue球杆,
252
850000
2000
指着它,给了他一个情感上的暗示,
14:30
and the robot机器人 looks容貌 at the watch quite相当 successfully顺利.
253
852000
2000
这个机器人非常成功的注意到了手表。
14:32
We don't know whether是否 he understood了解 or not that the robot机器人 --
254
854000
3000
我们不知道他是否意识到,这个机器人——
14:36
Notice注意 the turn-taking转回吐.
255
858000
2000
意识到了角色的轮换。
14:38
Ritchie里奇: OK, I want to show显示 you something. OK, this is a watch
256
860000
3000
Ritchie:好,我想让你看一些东西。好,这是一块手表,
14:41
that my girlfriend女朋友 gave me.
257
863000
3000
是我女朋友给我的。
14:44
Robot机器人: Oh, cool.
258
866000
2000
机器人:哦,酷!
14:46
Ritchie里奇: Yeah, look, it's got a little blue蓝色 light in it too. I almost几乎 lost丢失 it this week.
259
868000
4000
Ritchie:是的,看,它还在发着蓝光呢。这周我差点丢了它。
14:51
(Laughter笑声)
260
873000
4000
(笑声)
14:55
RBRB: So it's making制造 eye contact联系 with him, following以下 his eyes眼睛.
261
877000
3000
罗德尼布鲁克斯:他正在与他进行眼神接触,跟随着他的眼睛。
14:58
Ritchie里奇: Can you do the same相同 thing? Robot机器人: Yeah, sure.
262
880000
2000
Ritchie:你能做到同样的事情么?机器人:是的,当然了。
15:00
RBRB: And they successfully顺利 have that sort分类 of communication通讯.
263
882000
2000
罗德尼布鲁克斯:他们成功的开始了交流。
15:02
And here's这里的 another另一个 aspect方面 of the sorts排序 of things that Chris克里斯 and I were doing.
264
884000
4000
另外,这里还有一个方面的事情是克里斯和我正在做的。
15:06
This is another另一个 robot机器人, CogCOG.
265
888000
2000
这是另外一个机器人,Cog。
15:08
They first make eye contact联系, and then, when Christie克里斯蒂 looks容貌 over at this toy玩具,
266
890000
6000
他们首次进行了眼神接触,然后,当克里斯蒂看着那个玩具,
15:14
the robot机器人 estimates估计 her gaze凝视 direction方向
267
896000
2000
这个机器人判断出她的关注方向
15:16
and looks容貌 at the same相同 thing that she's looking at.
268
898000
2000
并且跟她一起盯着同样的东西。
15:18
(Laughter笑声)
269
900000
1000
(笑声)
15:19
So we're going to see more and more of this sort分类 of robot机器人
270
901000
3000
未来的几年后,我们将会在实验室里
15:22
over the next下一个 few少数 years年份 in labs实验室.
271
904000
2000
看到越来越多像这样的机器人。
15:24
But then the big questions问题, two big questions问题 that people ask me are:
272
906000
5000
不过随之而来是大问题,人们问我的最大的两个问题是:
15:29
if we make these robots机器人 more and more human-like类人,
273
911000
2000
如果我们造出的机器人越来越像人类,
15:31
will we accept接受 them, will we -- will they need rights权利 eventually终于?
274
913000
5000
我们会承认他们,我们会——他们最终会需要权利么?
15:36
And the other question people ask me is, will they want to take over?
275
918000
3000
人们问我的另一个问题是,他们将会接管世界么?
15:39
(Laughter笑声)
276
921000
1000
(笑声)
15:40
And on the first -- you know, this has been a very Hollywood好莱坞 theme主题
277
922000
3000
第一个问题,这已经是一个非常好莱坞的主题了,
15:43
with lots of movies电影. You probably大概 recognize认识 these characters人物 here --
278
925000
3000
相关的电影也非常多。你可能会意识到这些角色——
15:46
where in each of these cases, the robots机器人 want more respect尊重.
279
928000
4000
不管在任何一个例子里,机器人都渴望得到更多的尊重。
15:50
Well, do you ever need to give robots机器人 respect尊重?
280
932000
3000
那么,你是否需要给予机器人尊重?
15:54
They're just machines, after all.
281
936000
2000
他们只是机器而已。
15:56
But I think, you know, we have to accept接受 that we are just machines.
282
938000
4000
不过我觉得,我们也不得不承认我们是机器。
16:00
After all, that's certainly当然 what modern现代 molecular分子 biology生物学 says about us.
283
942000
5000
毕竟,那确实是现代分子生物学对我们的描述。
16:05
You don't see a description描述 of how, you know,
284
947000
3000
你看不到这样的描述,也就是关于
16:08
Molecule分子 A, you know, comes up and docks码头 with this other molecule分子.
285
950000
4000
分子A,靠近并与另外一个分子对接。
16:12
And it's moving移动 forward前锋, you know, propelled推进的 by various各个 charges收费,
286
954000
3000
然后向前移动,被不同的电荷所推动,
16:15
and then the soul灵魂 steps脚步 in and tweaks调整 those molecules分子 so that they connect.
287
957000
4000
然后soul加入,调整这些分子使它们连接起来。
16:19
It's all mechanistic机械. We are mechanism机制.
288
961000
3000
这都是机械化的,我们就是机器体。
16:22
If we are machines, then in principle原理 at least最小,
289
964000
3000
如果我们是机器,那么至少在原则上,
16:25
we should be able能够 to build建立 machines out of other stuff东东,
290
967000
4000
我们应该能够用其他东西造出机器,
16:29
which哪一个 are just as alive as we are.
291
971000
4000
这些机器就像我们一样是活着的。
16:33
But I think for us to admit承认 that,
292
975000
2000
不过我觉得我们应该承认,
16:35
we have to give up on our special-ness特别的烦躁, in a certain某些 way.
293
977000
3000
我们不得不以某种方式在我们的特殊性上放弃。
16:38
And we've我们已经 had the retreat撤退 from special-ness特别的烦躁
294
980000
2000
至少在过去的几百年,我们已经很多次
16:40
under the barrage弹幕 of science科学 and technology技术 many许多 times
295
982000
3000
想证明我们的特殊性,但总是
16:43
over the last few少数 hundred years年份, at least最小.
296
985000
2000
在科学技术的攻势下收获失败。
16:45
500 years年份 ago we had to give up the idea理念
297
987000
2000
当地球开始围绕太阳转时,500年前
16:47
that we are the center中央 of the universe宇宙
298
989000
3000
我们就放弃了我们处于
16:50
when the earth地球 started开始 to go around the sun太阳;
299
992000
2000
宇宙中心的想法;
16:52
150 years年份 ago, with Darwin达尔文, we had to give up the idea理念 we were different不同 from animals动物.
300
994000
5000
150年前,因为达尔文,我们不得不摒弃我们与动物不同的想法。
16:57
And to imagine想像 -- you know, it's always hard for us.
301
999000
3000
大家知道,去想象这些——对于我们来说很困难。
17:00
Recently最近 we've我们已经 been battered受虐 with the idea理念 that maybe
302
1002000
3000
近来,我们又震惊的得知
17:03
we didn't even have our own拥有 creation创建 event事件, here on earth地球,
303
1005000
2000
我们在地球上甚至没有自己的创造活动,
17:05
which哪一个 people didn't like much. And then the human人的 genome基因组 said,
304
1007000
3000
这样的想法大家都不喜欢。然后人类基因告诉我们,
17:08
maybe we only have 35,000 genes基因. And that was really --
305
1010000
3000
可能我们只有35000个基因。那真是——
17:11
people didn't like that, we've我们已经 got more genes基因 than that.
306
1013000
3000
大家不喜欢这样说,我们拥有比那个数目更多的基因。
17:14
We don't like to give up our special-ness特别的烦躁, so, you know,
307
1016000
3000
我们不愿意放弃自己的特殊性,所以,大家知道,
17:17
having the idea理念 that robots机器人 could really have emotions情绪,
308
1019000
2000
了解到机器人能够真正拥有感情的想法,
17:19
or that robots机器人 could be living活的 creatures生物 --
309
1021000
2000
或者机器人可能是生物的想法——
17:21
I think is going to be hard for us to accept接受.
310
1023000
2000
我想对我们来说都很难接受。
17:23
But we're going to come to accept接受 it over the next下一个 50 years年份 or so.
311
1025000
4000
不过到下一个50年之后我们将会接受这样的想法。
17:27
And the second第二 question is, will the machines want to take over?
312
1029000
3000
第二个问题是,机器人想要接管世界么?
17:30
And here the standard标准 scenario脚本 is that we create创建 these things,
313
1032000
5000
这里标准的脚本就是我们创造了他们,
17:35
they grow增长, we nurture培育 them, they learn学习 a lot from us,
314
1037000
3000
他们成长着,我们哺育着他们,他们从我们这里学到很多,
17:38
and then they start开始 to decide决定 that we're pretty漂亮 boring无聊, slow.
315
1040000
4000
然后他们开始认为我们很无聊,行动缓慢。
17:42
They want to take over from us.
316
1044000
2000
他们希望从我们手中接管世界。
17:44
And for those of you that have teenagers青少年, you know what that's like.
317
1046000
3000
对于那些家里由青少年的人,你们知道那是什么光景。
17:47
(Laughter笑声)
318
1049000
1000
(笑声)
17:48
But Hollywood好莱坞 extends扩展 it to the robots机器人.
319
1050000
3000
只不过好莱坞把它拓展到了机器人身上。
17:51
And the question is, you know,
320
1053000
3000
问题是,
17:54
will someone有人 accidentally偶然 build建立 a robot机器人 that takes over from us?
321
1056000
4000
某些人会突然制造出一个接管世界的机器人么?
17:58
And that's sort分类 of like this lone孤单 guy in the backyard后院,
322
1060000
3000
这就像一个在后院生活的孤独的人说,
18:01
you know -- "I accidentally偶然 built内置 a 747."
323
1063000
3000
“我不小心造出一架波音747。”
18:04
I don't think that's going to happen发生.
324
1066000
2000
我认为这不会发生。
18:06
And I don't think --
325
1068000
2000
我并不认为——
18:08
(Laughter笑声)
326
1070000
1000
(笑声)
18:09
-- I don't think we're going to deliberately故意 build建立 robots机器人
327
1071000
3000
——我觉得我们不会故意制造一些
18:12
that we're uncomfortable不舒服 with.
328
1074000
2000
让我们不好受的机器人。
18:14
We'll -- you know, they're not going to have a super bad robot机器人.
329
1076000
2000
我们造出的——不会一下子是个超级机器人大坏蛋。
18:16
Before that has to come to be a mildly温和 bad robot机器人,
330
1078000
3000
在此之前,先得有个轻度的不良机器人,
18:19
and before that a not so bad robot机器人.
331
1081000
2000
而在此之前,会变成一个不那么坏的机器人。
18:21
(Laughter笑声)
332
1083000
1000
(笑声)
18:22
And we're just not going to let it go that way.
333
1084000
2000
我们不会就让事态向那么坏的方向发展。
18:24
(Laughter笑声)
334
1086000
1000
(笑声)
18:25
So, I think I'm going to leave离开 it at that: the robots机器人 are coming未来,
335
1087000
6000
所以,我觉得我会下这样的结论:机器人正在到来,
18:31
we don't have too much to worry担心 about, it's going to be a lot of fun开玩笑,
336
1093000
3000
我们无需太过担忧,这将很有趣,
18:34
and I hope希望 you all enjoy请享用 the journey旅程 over the next下一个 50 years年份.
337
1096000
4000
我希望大家能够享受着下一个50年的旅程。
18:38
(Applause掌声)
338
1100000
2000
(掌声)
Translated by Zhijun Wang
Reviewed by Zhu Jie

▲Back to top

ABOUT THE SPEAKER
Rodney Brooks - Roboticist
Rodney Brooks builds robots based on biological principles of movement and reasoning. The goal: a robot who can figure things out.

Why you should listen

Former MIT professor Rodney Brooks studies and engineers robot intelligence, looking for the holy grail of robotics: the AGI, or artificial general intelligence. For decades, we've been building robots to do highly specific tasks -- welding, riveting, delivering interoffice mail -- but what we all want, really, is a robot that can figure things out on its own, the way we humans do.

Brooks realized that a top-down approach -- just building the biggest brain possible and teaching it everything we could think of -- would never work. What would work is a robot who learns like we do, by trial and error, and with many separate parts that learn separate jobs. The thesis of his work which was captured in Fast, Cheap and Out of Control,went on to become the title of the great Errol Morris documentary.

A founder of iRobot, makers of the Roomba vacuum, Brooks is now founder and CTO of Rethink Robotics, whose mission is to apply advanced robotic intelligence to manufacturing and physical labor. Its first robots: the versatile two-armed Baxter and one-armed Sawyer. Brooks is the former director of CSAIL, MIT's Computers Science and Artificial Intelligence Laboratory.

 
More profile about the speaker
Rodney Brooks | Speaker | TED.com