1
00:00:00,564 --> 00:00:03,086
My job is to design, build and study

2
00:00:03,086 --> 00:00:05,086
robots that communicate with people.

3
00:00:05,086 --> 00:00:06,639
But this story doesn't start with robotics at all,

4
00:00:06,639 --> 00:00:08,531
it starts with animation.

5
00:00:08,531 --> 00:00:11,117
When I first saw Pixar's "Luxo Jr.,"

6
00:00:11,117 --> 00:00:13,103
I was amazed by how much emotion

7
00:00:13,103 --> 00:00:14,861
they could put into something

8
00:00:14,861 --> 00:00:17,452
as trivial as a desk lamp.

9
00:00:17,452 --> 00:00:19,285
I mean, look at them -- at the end of this movie,

10
00:00:19,285 --> 00:00:21,789
you actually feel something for two pieces of furniture.

11
00:00:21,789 --> 00:00:23,617
(Laughter)

12
00:00:23,617 --> 00:00:25,737
And I said, I have to learn how to do this.

13
00:00:25,737 --> 00:00:29,271
So I made a really bad career decision.

14
00:00:29,271 --> 00:00:31,675
And that's what my mom was like when I did it.

15
00:00:31,675 --> 00:00:33,686
(Laughter)

16
00:00:33,686 --> 00:00:35,866
I left a very cozy tech job in Israel

17
00:00:35,866 --> 00:00:38,471
at a nice software company and I moved to New York

18
00:00:38,471 --> 00:00:39,482
to study animation.

19
00:00:39,482 --> 00:00:40,639
And there I lived

20
00:00:40,639 --> 00:00:43,635
in a collapsing apartment building
in Harlem with roommates.

21
00:00:43,635 --> 00:00:45,003
I'm not using this phrase metaphorically,

22
00:00:45,003 --> 00:00:46,963
the ceiling actually collapsed one day

23
00:00:46,963 --> 00:00:48,273
in our living room.

24
00:00:48,273 --> 00:00:50,939
Whenever they did those news stories 
about building violations in New York,


25
00:00:50,939 --> 00:00:53,185
they would put the report in front of our building.

26
00:00:53,185 --> 00:00:56,783
As kind of like a backdrop 
to show how bad things are.

27
00:00:56,783 --> 00:00:58,729
Anyway, during the day I went to school and at night

28
00:00:58,729 --> 00:01:02,251
I would sit and draw frame by frame 
of pencil animation.

29
00:01:02,251 --> 00:01:04,525
And I learned two surprising lessons --

30
00:01:04,525 --> 00:01:07,078
one of them was that

31
00:01:07,078 --> 00:01:09,038
when you want to arouse emotions,

32
00:01:09,038 --> 00:01:10,764
it doesn't matter so much how something looks, 


33
00:01:10,764 --> 00:01:13,110
it's all in the motion -- it's in the timing

34
00:01:13,110 --> 00:01:14,717
of how the thing moves.

35
00:01:14,717 --> 00:01:17,944
And the second, was something 
one of our teachers told us.

36
00:01:17,944 --> 00:01:20,347
He actually did the weasel in Ice Age.

37
00:01:20,347 --> 00:01:21,738
And he said:

38
00:01:21,738 --> 00:01:24,895
"As an animator you are not 
a director, you're an actor."

39
00:01:24,895 --> 00:01:27,876
So, if you want to find the 
right motion for a character,

40
00:01:27,876 --> 00:01:30,227
don't think about it, go use your body to find it --

41
00:01:30,227 --> 00:01:31,863
stand in front of a mirror, act it out

42
00:01:31,863 --> 00:01:33,565
in front of a camera -- whatever you need.

43
00:01:33,565 --> 00:01:36,459
And then put it back in your character.

44
00:01:36,459 --> 00:01:38,619
A year later I found myself at MIT

45
00:01:38,619 --> 00:01:40,769
in the robotic life group, it was one of the first groups

46
00:01:40,769 --> 00:01:43,414
researching the relationships 
between humans and robots.

47
00:01:43,414 --> 00:01:45,347
And I still had this dream to make

48
00:01:45,347 --> 00:01:48,064
an actual, physical Luxo Jr. lamp.

49
00:01:48,064 --> 00:01:49,631
But I found that robots didn't move at all

50
00:01:49,631 --> 00:01:50,909
in this engaging way that I was used to

51
00:01:50,909 --> 00:01:52,544
for my animation studies.

52
00:01:52,544 --> 00:01:54,761
Instead, they were all --

53
00:01:54,761 --> 00:01:57,255
how should I put it, they were all kind of robotic.

54
00:01:57,255 --> 00:01:59,152
(Laughter)

55
00:01:59,152 --> 00:02:02,787
And I thought, what if I took whatever 
I learned in animation school,

56
00:02:02,787 --> 00:02:05,455
and used that to design my robotic desk lamp.

57
00:02:05,455 --> 00:02:07,539
So I went and designed frame by frame

58
00:02:07,539 --> 00:02:09,393
to try to make this robot

59
00:02:09,393 --> 00:02:12,050
as graceful and engaging as possible.

60
00:02:12,050 --> 00:02:14,092
And here when you see the robot interacting with me

61
00:02:14,092 --> 00:02:15,723
on a desktop.

62
00:02:15,723 --> 00:02:17,934
And I'm actually redesigning the robot so,

63
00:02:17,934 --> 00:02:19,731
unbeknownst to itself,

64
00:02:19,731 --> 00:02:22,419
it's kind of digging its own grave by helping me.

65
00:02:22,419 --> 00:02:24,449
(Laughter)

66
00:02:24,449 --> 00:02:26,418
I wanted it to be less of a mechanical structure

67
00:02:26,418 --> 00:02:27,712
giving me light,

68
00:02:27,712 --> 00:02:30,754
and more of a helpful, kind of quiet apprentice

69
00:02:30,754 --> 00:02:33,948
that's always there when you need 
it and doesn't really interfere.

70
00:02:33,948 --> 00:02:35,625
And when, for example, I'm looking for a battery

71
00:02:35,625 --> 00:02:37,158
that I can't find,

72
00:02:37,158 --> 00:02:41,753
in a subtle way, it will show me where the battery is.

73
00:02:41,753 --> 00:02:44,319
So you can see my confusion here.

74
00:02:44,319 --> 00:02:48,578
I'm not an actor.

75
00:02:48,578 --> 00:02:50,026
And I want you to notice how the same

76
00:02:50,026 --> 00:02:52,000
mechanical structure can at one point,

77
00:02:52,000 --> 00:02:55,004
just by the way it moves seem gentle and caring --

78
00:02:55,004 --> 00:02:58,237
and in the other case, seem 
violent and confrontational.

79
00:02:58,237 --> 00:03:01,548
And it's the same structure, 
just the motion is different.

80
00:03:07,026 --> 00:03:12,900
Actor: "You want to know something? 
Well, you want to know something?

81
00:03:12,900 --> 00:03:13,837
He was already dead!

82
00:03:13,837 --> 00:03:17,854
Just laying there, eyes glazed over!"

83
00:03:17,854 --> 00:03:18,857
(Laughter)

84
00:03:18,857 --> 00:03:22,755
But, moving in graceful ways is just one 
building block of this whole structure

85
00:03:22,755 --> 00:03:24,339
called human-robot interaction.

86
00:03:24,339 --> 00:03:25,828
I was at the time doing my Ph.D.,

87
00:03:25,828 --> 00:03:27,998
I was working on human robot teamwork;

88
00:03:27,998 --> 00:03:30,000
teams of humans and robots working together.

89
00:03:30,000 --> 00:03:31,334
I was studying the engineering,

90
00:03:31,334 --> 00:03:34,116
the psychology, the philosophy of teamwork.

91
00:03:34,116 --> 00:03:35,583
And at the same time I found myself

92
00:03:35,583 --> 00:03:37,438
in my own kind of teamwork situation

93
00:03:37,438 --> 00:03:39,659
with a good friend of mine who is actually here.

94
00:03:39,659 --> 00:03:42,273
And in that situation we can easily imagine robots

95
00:03:42,273 --> 00:03:44,366
in the near future being there with us.

96
00:03:44,366 --> 00:03:46,213
It was after a Passover seder.

97
00:03:46,213 --> 00:03:48,053
We were folding up a lot of folding chairs,

98
00:03:48,053 --> 00:03:50,986
and I was amazed at how quickly 
we found our own rhythm.

99
00:03:50,986 --> 00:03:52,553
Everybody did their own part.

100
00:03:52,553 --> 00:03:54,182
We didn't have to divide our tasks.

101
00:03:54,182 --> 00:03:56,278
We didn't have to communicate verbally about this.

102
00:03:56,278 --> 00:03:57,889
It all just happened.

103
00:03:57,889 --> 00:03:58,783
And I thought,

104
00:03:58,783 --> 00:04:00,632
humans and robots don't look at all like this.

105
00:04:00,632 --> 00:04:02,297
When humans and robots interact,

106
00:04:02,297 --> 00:04:03,495
it's much more like a chess game.

107
00:04:03,495 --> 00:04:04,758
The human does a thing,

108
00:04:04,758 --> 00:04:06,638
the robot analyzes whatever the human did,

109
00:04:06,638 --> 00:04:08,168
then the robot decides what to do next,

110
00:04:08,168 --> 00:04:09,344
plans it and does it.

111
00:04:09,344 --> 00:04:11,475
And then the human waits, until it's their turn again.

112
00:04:11,475 --> 00:04:12,665
So, it's much more like a chess game

113
00:04:12,665 --> 00:04:14,875
and that makes sense because chess is great

114
00:04:14,875 --> 00:04:16,396
for mathematicians and computer scientists.

115
00:04:16,396 --> 00:04:18,654
It's all about information analysis,

116
00:04:18,654 --> 00:04:21,503
decision making and planning.

117
00:04:21,503 --> 00:04:25,191
But I wanted my robot to be less of a chess player,

118
00:04:25,191 --> 00:04:27,152
and more like a doer

119
00:04:27,152 --> 00:04:29,177
that just clicks and works together.

120
00:04:29,177 --> 00:04:32,570
So I made my second horrible career choice:

121
00:04:32,570 --> 00:04:35,230
I decided to study acting for a semester.

122
00:04:35,230 --> 00:04:38,090
I took off from a Ph.D. I went to acting classes.

123
00:04:38,090 --> 00:04:40,550
I actually participated in a play,

124
00:04:40,550 --> 00:04:43,342
I hope theres no video of that around still.

125
00:04:43,342 --> 00:04:45,969
And I got every book I could find about acting,

126
00:04:45,969 --> 00:04:47,945
including one from the 19th century

127
00:04:47,945 --> 00:04:49,012
that I got from the library.

128
00:04:49,012 --> 00:04:52,470
And I was really amazed because my 
name was the second name on the list --

129
00:04:52,470 --> 00:04:55,286
the previous name was in 1889. (Laughter)

130
00:04:55,286 --> 00:04:57,000
And this book was kind of waiting for 100 years

131
00:04:57,000 --> 00:05:00,191
to be rediscovered for robotics.

132
00:05:00,191 --> 00:05:01,669
And this book shows actors

133
00:05:01,669 --> 00:05:04,034
how to move every muscle in the body

134
00:05:04,034 --> 00:05:06,742
to match every kind of emotion 
that they want to express.

135
00:05:06,742 --> 00:05:08,574
But the real relevation was

136
00:05:08,574 --> 00:05:09,910
when I learned about method acting.

137
00:05:09,910 --> 00:05:12,186
It became very popular in the 20th century.

138
00:05:12,186 --> 00:05:14,830
And method acting said, you don't have 
to plan every muscle in your body.

139
00:05:14,830 --> 00:05:17,503
Instead you have to use your body 
to find the right movement.

140
00:05:17,503 --> 00:05:19,683
You have to use your sense memory

141
00:05:19,683 --> 00:05:21,539
to reconstruct the emotions and kind of

142
00:05:21,539 --> 00:05:24,440
think with your body to find the right expression.

143
00:05:24,440 --> 00:05:26,460
Improvise, play off yor scene partner.

144
00:05:26,460 --> 00:05:29,708
And this came at the same time 
as I was reading about this trend

145
00:05:29,708 --> 00:05:32,804
in cognitive psychology called embodied cognition.

146
00:05:32,804 --> 00:05:34,478
Which also talks about the same ideas --

147
00:05:34,478 --> 00:05:35,987
We use our bodies to think,

148
00:05:35,987 --> 00:05:38,212
we don't just think with our brains 
and use our bodies to move.

149
00:05:38,212 --> 00:05:40,621
but our bodies feed back into our brain

150
00:05:40,621 --> 00:05:42,838
to generate the way that we behave.

151
00:05:42,838 --> 00:05:44,209
And it was like a lightning bolt.

152
00:05:44,209 --> 00:05:45,571
I went back to my office.

153
00:05:45,571 --> 00:05:48,470
I wrote this paper -- which I never really published

154
00:05:48,470 --> 00:05:50,836
called "Acting Lessons for Artificial Intelligence."

155
00:05:50,836 --> 00:05:52,233
And I even took another month

156
00:05:52,233 --> 00:05:54,844
to do what was then the first theater play

157
00:05:54,844 --> 00:05:56,696
with a human and a robot acting together.

158
00:05:56,696 --> 00:06:00,263
That's what you saw before with the actors.

159
00:06:00,263 --> 00:06:01,692
And I thought:

160
00:06:01,692 --> 00:06:04,599
How can we make an artificial intelligence model --

161
00:06:04,599 --> 00:06:06,460
computer, computational model --

162
00:06:06,460 --> 00:06:08,784
that will model some of these ideas of improvisation,

163
00:06:08,784 --> 00:06:10,941
of taking risks, of taking chances,

164
00:06:10,941 --> 00:06:12,508
even of making mistakes.

165
00:06:12,508 --> 00:06:15,280
Maybe it can make for better robotic teammates.

166
00:06:15,280 --> 00:06:17,908
So I worked for quite a long time on these models

167
00:06:17,908 --> 00:06:20,257
and I implemented them on a number of robots.

168
00:06:20,257 --> 00:06:22,272
Here you can see a very early example

169
00:06:22,272 --> 00:06:25,984
with the robots trying to use this 
embodied artificial intelligence,

170
00:06:25,984 --> 00:06:28,742
to try to match my movements 
as closely as possible,

171
00:06:28,742 --> 00:06:30,229
sort of like a game.

172
00:06:30,229 --> 00:06:32,086
Let's look at it.

173
00:06:35,594 --> 00:06:39,698
You can see when I psych it out, it gets fooled.

174
00:06:39,698 --> 00:06:41,899
And it's a little bit like what you might see actors do

175
00:06:41,899 --> 00:06:43,786
when they try to mirror each other

176
00:06:43,786 --> 00:06:46,304
to find the right synchrony between them.

177
00:06:46,304 --> 00:06:48,220
And then, I did another experiment,

178
00:06:48,220 --> 00:06:52,303
and I got people off the street 
to use the robotic desk lamp,

179
00:06:52,303 --> 00:06:56,257
and try out this idea of embodied 
artificial intelligence.

180
00:06:56,257 --> 00:07:00,543
So, I actually used two kinds 
of brains for the same robot.

181
00:07:00,543 --> 00:07:01,863
The robot is the same lamp that you saw,

182
00:07:01,863 --> 00:07:03,793
and I put in it two brains.

183
00:07:03,793 --> 00:07:05,633
For one half of the people,

184
00:07:05,633 --> 00:07:08,457
I put in a brain that's kind of the traditional,

185
00:07:08,457 --> 00:07:09,711
calculated robotic brain.

186
00:07:09,711 --> 00:07:12,265
It waits for its turn, it analyzes everything, it plans.

187
00:07:12,265 --> 00:07:13,921
Let's call it the calculated brain.

188
00:07:13,921 --> 00:07:17,553
The other got more the stage actor, risk taker brain.

189
00:07:17,553 --> 00:07:19,633
Let's call it the adventurous brain.

190
00:07:19,633 --> 00:07:22,594
It sometimes acts without knowing 
everything it has to know.

191
00:07:22,594 --> 00:07:24,771
It sometimes makes mistakes and corrects them.

192
00:07:24,771 --> 00:07:27,354
And I had them do this very tedious task

193
00:07:27,354 --> 00:07:28,852
that took almost 20 minutes

194
00:07:28,852 --> 00:07:30,353
and they had to work together.

195
00:07:30,353 --> 00:07:32,634
Somehow simulating like a factory job

196
00:07:32,634 --> 00:07:35,299
of repetitively doing the same thing.

197
00:07:35,299 --> 00:07:37,073
And what I found was that people actually loved

198
00:07:37,073 --> 00:07:38,721
the adventurous robot.

199
00:07:38,721 --> 00:07:40,103
And they thought it was more intelligent,

200
00:07:40,103 --> 00:07:42,171
more committed, a better member of the team,

201
00:07:42,171 --> 00:07:44,268
contributed to the success of the team more.

202
00:07:44,268 --> 00:07:45,983
They even called it 'he' and 'she,'

203
00:07:45,983 --> 00:07:48,899
whereas people with the calculated brain called it 'it.'

204
00:07:48,899 --> 00:07:51,646
And nobody ever called it 'he' or 'she'.

205
00:07:51,646 --> 00:07:53,495
When they talked about it after the task

206
00:07:53,495 --> 00:07:55,193
with the adventurous brain,

207
00:07:55,193 --> 00:07:59,275
they said, "By the end, we were good 
friends and high-fived mentally."

208
00:07:59,275 --> 00:08:00,714
Whatever that means.

209
00:08:00,714 --> 00:08:03,961
(Laughter) Sounds painful.

210
00:08:03,961 --> 00:08:06,676
Whereas the people with the calculated brain

211
00:08:06,676 --> 00:08:09,049
said it was just like a lazy apprentice.

212
00:08:09,049 --> 00:08:11,750
It only did what it was supposed 
to do and nothing more.

213
00:08:11,750 --> 00:08:14,307
Which is almost what people expect robots to do,

214
00:08:14,307 --> 00:08:16,742
so I was surprised that people 
had higher expectations

215
00:08:16,742 --> 00:08:21,886
of robots, than what anybody in robotics 
thought robots should be doing.

216
00:08:21,886 --> 00:08:23,856
And in a way, I thought, maybe it's time --

217
00:08:23,856 --> 00:08:26,717
just like method acting changed the way

218
00:08:26,717 --> 00:08:28,149
people thought about acting in the 19th century,

219
00:08:28,149 --> 00:08:29,992
from going from the very calculated,

220
00:08:29,992 --> 00:08:31,774
planned way of behaving,

221
00:08:31,774 --> 00:08:35,295
to a more intuitive, risk-taking, 
embodied way of behaving.

222
00:08:35,295 --> 00:08:36,699
Maybe it's time for robots

223
00:08:36,699 --> 00:08:39,845
to have the same kind of revolution.

224
00:08:39,845 --> 00:08:40,966
A few years later,

225
00:08:40,966 --> 00:08:43,474
I was at my next research job 
at Georgia Tech in Atlanta,

226
00:08:43,474 --> 00:08:44,539
and I was working in a group

227
00:08:44,539 --> 00:08:45,821
dealing with robotic musicians.

228
00:08:45,821 --> 00:08:48,662
And I thought, music, that's the perfect place

229
00:08:48,662 --> 00:08:51,059
to look at teamwork, coordination,

230
00:08:51,059 --> 00:08:52,948
timing, improvisation --

231
00:08:52,948 --> 00:08:55,484
and we just got this robot playing marimba.

232
00:08:55,484 --> 00:08:57,499
Marimba, for everybody who was like me,

233
00:08:57,499 --> 00:09:00,344
it was this huge, wooden xylophone.

234
00:09:00,344 --> 00:09:03,039
And, when I was looking at this,

235
00:09:03,039 --> 00:09:05,633
I looked at other works in 
human-robot improvisation --

236
00:09:05,633 --> 00:09:08,012
yes, there are other works in 
human-robot improvisation --

237
00:09:08,012 --> 00:09:10,105
and they were also a little bit like a chess game.

238
00:09:10,105 --> 00:09:11,409
The human would play,

239
00:09:11,409 --> 00:09:13,989
the robot would analyze what was played,

240
00:09:13,989 --> 00:09:16,201
would improvise their own part.

241
00:09:16,201 --> 00:09:17,868
So, this is what musicians called

242
00:09:17,868 --> 00:09:19,351
a call and response interaction,

243
00:09:19,351 --> 00:09:23,188
and it also fits very well, robots 
and artificial intelligence.

244
00:09:23,188 --> 00:09:25,095
But I thought, if I use the same ideas I used

245
00:09:25,095 --> 00:09:28,232
in the theater play and in the teamwork studies,

246
00:09:28,232 --> 00:09:30,684
maybe I can make the robots jam together

247
00:09:30,684 --> 00:09:32,305
like a band.

248
00:09:32,305 --> 00:09:36,118
Everybody's riffing off each other, 
nobody is stopping it for a moment.

249
00:09:36,118 --> 00:09:38,910
And so, I tried to do the same 
things, this time with music,

250
00:09:38,910 --> 00:09:40,297
where the robot doesn't really know

251
00:09:40,297 --> 00:09:41,272
what it's about to play.

252
00:09:41,272 --> 00:09:42,545
It just sort of moves its body

253
00:09:42,545 --> 00:09:44,773
and uses opportunities to play,

254
00:09:44,773 --> 00:09:47,494
And does what my jazz teacher
when I was 17 taught me.

255
00:09:47,494 --> 00:09:48,758
She said, when you improvise,

256
00:09:48,758 --> 00:09:50,419
sometimes you don't know what you're doing

257
00:09:50,419 --> 00:09:51,434
and you're still doing it.

258
00:09:51,434 --> 00:09:52,864
And so I tried to make a robot that doesn't actually

259
00:09:52,864 --> 00:09:54,809
know what it's doing, but it's still doing it.

260
00:09:54,809 --> 00:09:57,753
So let's look at a few seconds 
from this performance.

261
00:09:57,753 --> 00:10:00,670
Where the robot listens to the human musician

262
00:10:00,670 --> 00:10:02,473
and improvises.

263
00:10:02,473 --> 00:10:04,846
And then, look at how the human musician also

264
00:10:04,846 --> 00:10:06,633
responds to what the robot is doing, and picking up

265
00:10:06,633 --> 00:10:09,453
from its behavior.

266
00:10:09,453 --> 00:10:13,981
And at some point can even be surprised 
by what the robot came up with.

267
00:10:13,981 --> 00:11:00,056
(Music)

268
00:11:00,056 --> 00:11:05,087
(Applause)

269
00:11:05,087 --> 00:11:07,073
Being a musician is not just about making notes,

270
00:11:07,073 --> 00:11:09,459
otherwise nobody would every go see a live show.

271
00:11:09,459 --> 00:11:11,382
Musicians also communicate with their bodies,

272
00:11:11,382 --> 00:11:13,329
with other band members, with the audience,

273
00:11:13,329 --> 00:11:15,492
they use their bodies to express the music.

274
00:11:15,492 --> 00:11:17,872
And I thought, we already have 
a robot musician on stage,

275
00:11:17,872 --> 00:11:20,820
why not make it be a full-fledged musician.

276
00:11:20,820 --> 00:11:23,336
And I started designing a socially expressive head

277
00:11:23,336 --> 00:11:24,885
for the robot.

278
00:11:24,885 --> 00:11:26,712
The head does't actually touch the marimba,

279
00:11:26,712 --> 00:11:28,348
it just expresses what the music is like.

280
00:11:28,348 --> 00:11:31,428
These are some napkin sketches 
from a bar in Atlanta,

281
00:11:31,428 --> 00:11:34,050
that was dangerously located exactly halfway

282
00:11:34,050 --> 00:11:35,502
between my lab and my home. (Laughter)

283
00:11:35,502 --> 00:11:36,918
So I spent, I would say on average,

284
00:11:36,918 --> 00:11:39,846
three to four hours a day there.

285
00:11:39,846 --> 00:11:42,744
I think. (Laughter)

286
00:11:42,744 --> 00:11:45,723
And I went back to my animation 
tools and tried to figure out

287
00:11:45,723 --> 00:11:47,905
not just what a robotic musician would look like,

288
00:11:47,905 --> 00:11:50,713
but especially what a robotic 
musician would move like.

289
00:11:50,713 --> 00:11:54,013
To sort of show that it doesn't like 
what the other person is playing --

290
00:11:54,013 --> 00:11:56,492
and maybe show whatever beat it's feeling

291
00:11:56,492 --> 00:11:58,141
at the moment.

292
00:11:58,141 --> 00:12:02,668
So we ended up actually getting the money 
to build this robot, which was nice.

293
00:12:02,668 --> 00:12:04,878
I'm going to show you now the 
same kind of performance,

294
00:12:04,878 --> 00:12:07,220
this time with a socially expressive head.

295
00:12:07,220 --> 00:12:09,184
And notice one thing --

296
00:12:09,184 --> 00:12:10,880
how the robot is really showing us

297
00:12:10,880 --> 00:12:12,744
the beat it's picking up from the human.

298
00:12:12,744 --> 00:12:16,760
We're also giving the human a sense 
that the robot knows what it's doing.

299
00:12:16,760 --> 00:12:18,140
And also how it changes the way it moves


300
00:12:18,140 --> 00:12:20,680
as soon as it starts its own solo.

301
00:12:20,680 --> 00:12:24,600
(Music)

302
00:12:24,600 --> 00:12:27,501
Now it's looking at me to make sure I'm listening.

303
00:12:27,501 --> 00:12:48,601
(Music)

304
00:12:48,601 --> 00:12:52,228
And now look at the final chord of the piece again,

305
00:12:52,228 --> 00:12:55,067
and this time the robot communicates with its body

306
00:12:55,067 --> 00:12:57,204
when it's busy doing its own thing.

307
00:12:57,204 --> 00:12:58,816
And when it's ready

308
00:12:58,816 --> 00:13:02,168
to coordinate the final chord with me.

309
00:13:02,168 --> 00:13:15,097
(Music)

310
00:13:15,128 --> 00:13:20,906
(Applause)

311
00:13:20,906 --> 00:13:25,263
Thanks. I hope you see how much this totally not --

312
00:13:25,263 --> 00:13:27,944
how much this part of the body 
that doesn't touch the instrument

313
00:13:27,944 --> 00:13:31,310
actually helps with the musical performance.

314
00:13:31,310 --> 00:13:34,530
And at some point, we are in Atlanta, 
so obviously some rapper

315
00:13:34,530 --> 00:13:36,399
will come into our lab at some point.

316
00:13:36,399 --> 00:13:38,639
And we had this rapper come in

317
00:13:38,639 --> 00:13:41,212
and do a little jam with the robot.

318
00:13:41,212 --> 00:13:43,914
And here you can see the robot

319
00:13:43,914 --> 00:13:45,143
basically responding to the beat and --

320
00:13:45,143 --> 00:13:47,997
notice two things. One, how irresistible it is

321
00:13:47,997 --> 00:13:50,510
to join the robot while it's moving its head.

322
00:13:50,051 --> 00:13:52,412
and you kind of want to move 
your own head when it does it.

323
00:13:52,412 --> 00:13:56,207
And second, even though the rapper 
is really focused on his iPhone,

324
00:13:56,207 --> 00:13:59,121
as soon as the robot turns to him, he turns back.

325
00:13:59,121 --> 00:14:01,200
So even though it's just in 
the periphery of his vision --

326
00:14:01,200 --> 00:14:03,770
it's just in the corner of his eye -- 
it's very powerful.

327
00:14:03,770 --> 00:14:05,821
And the reason is that we can't ignore

328
00:14:05,821 --> 00:14:07,522
physical things moving in our environment.

329
00:14:07,522 --> 00:14:09,286
We are wired for that.

330
00:14:09,286 --> 00:14:12,618
So, if you have a problem with maybe your partners

331
00:14:12,618 --> 00:14:15,489
looking at the iPhone too much 
or their smartphone too much,

332
00:14:15,489 --> 00:14:17,082
you might want to have a robot there

333
00:14:17,082 --> 00:14:18,636
to get their attention. (Laughter)

334
00:14:18,636 --> 00:14:38,211
(Music)

335
00:14:38,211 --> 00:14:44,872
(Applause)

336
00:14:44,872 --> 00:14:47,382
Just to introduce the last robot

337
00:14:47,382 --> 00:14:49,526
that we've worked on,

338
00:14:49,526 --> 00:14:51,599
that came out of something kind 
of surprising that we found:

339
00:14:51,599 --> 00:14:54,634
At some point people didn't care anymore 
about the robot being so intelligent,

340
00:14:54,634 --> 00:14:56,342
and can improvise and listen,

341
00:14:56,342 --> 00:15:00,750
and do all these embodied intelligence 
things that I spent years on developing.

342
00:15:00,750 --> 00:15:04,263
They really liked that the robot was 
enjoying the music. (Laughter)

343
00:15:04,263 --> 00:15:06,668
And they didn't say that the 
robot was moving to the music,

344
00:15:06,668 --> 00:15:08,340
they said that the robot was enjoying the music.

345
00:15:08,340 --> 00:15:10,744
And we thought, why don't we take this idea,

346
00:15:10,744 --> 00:15:13,629
and I designed a new piece of furniture.

347
00:15:13,629 --> 00:15:15,718
This time it wasn't a desk 
lamp; it was a speaker dock.

348
00:15:15,718 --> 00:15:18,826
It was one of those things you 
plug your smartphone in.

349
00:15:18,826 --> 00:15:20,545
And I thought, what would happen

350
00:15:20,545 --> 00:15:23,146
if your speaker dock didn't 
just play the music for you,

351
00:15:23,146 --> 00:15:25,657
but it would actually enjoy it too. (Laughter)

352
00:15:25,657 --> 00:15:27,434
And so again, here are some animation tests

353
00:15:27,434 --> 00:15:31,835
from an early stage. (Laughter)

354
00:15:31,835 --> 00:15:36,005
And this is what the final product looked like.

355
00:15:47,223 --> 00:16:09,461
("Drop It Like It's Hot")

356
00:16:09,461 --> 00:16:12,105
So, a lot of bobbing head.

357
00:16:12,105 --> 00:16:15,327
(Applause)

358
00:16:15,327 --> 00:16:17,228
A lot of bobbing heads in the audience,

359
00:16:17,228 --> 00:16:20,386
so we can still see robots influence people.

360
00:16:20,386 --> 00:16:23,028
And it's not just fun and games.

361
00:16:23,028 --> 00:16:25,420
I think one of the reasons I care so much

362
00:16:25,420 --> 00:16:27,214
about robots that use their body to communicate

363
00:16:27,214 --> 00:16:29,407
and use their body to move --

364
00:16:29,407 --> 00:16:32,640
and I'm going to let you in on a little 
secret we roboticists are hiding --

365
00:16:32,640 --> 00:16:35,001
is that every one of you is 
going to be living with a robot

366
00:16:35,001 --> 00:16:37,254
at some point in their life.

367
00:16:37,254 --> 00:16:39,550
Somewhere in your future there's 
going to be a robot in your life.

368
00:16:39,550 --> 00:16:41,678
And if not in yours, then in your children's lives.

369
00:16:41,678 --> 00:16:43,467
And I want these robots to be --

370
00:16:43,467 --> 00:16:46,961
to be more fluent, more engaging, more graceful

371
00:16:46,961 --> 00:16:48,749
than currently they seem to be.

372
00:16:48,749 --> 00:16:50,668
And for that I think that maybe robots

373
00:16:50,668 --> 00:16:52,034
need to be less like chess players

374
00:16:52,034 --> 00:16:54,744
and more like stage actors and more like musicians.

375
00:16:54,744 --> 00:16:57,683
Maybe they should be able to 
take chances and improvise.

376
00:16:57,683 --> 00:17:00,180
And maybe they should be able to 
anticipate what you're about to do.

377
00:17:00,180 --> 00:17:02,756
And maybe they need to be able to make mistakes

378
00:17:02,756 --> 00:17:03,979
and correct them,

379
00:17:03,979 --> 00:17:06,015
because in the end we are human.

380
00:17:06,015 --> 00:17:09,386
And maybe as humans, robots 
that are a little less than perfect

381
00:17:09,386 --> 00:17:11,252
are just perfect for us.

382
00:17:11,252 --> 00:17:12,766
Thank you.

383
00:17:12,766 --> 00:17:16,188
(Applause)