TEDGlobal 2011
Daniel Wolpert: The real reason for brains
Filmed:
Readability: 3.9
1,994,993 views
Neuroscientist Daniel Wolpert starts from a surprising premise: the brain evolved, not to think or feel, but to control movement. In this entertaining, data-rich talk he gives us a glimpse into how the brain creates the grace and agility of human motion.
Daniel Wolpert - Movement expert
A neuroscientist and engineer, Daniel Wolpert studies how the brain controls the body. Full bio
A neuroscientist and engineer, Daniel Wolpert studies how the brain controls the body. Full bio
Double-click the English transcript below to play the video.
00:15
I'm a neuroscientist.
0
0
2000
00:17
And in neuroscience,
1
2000
2000
00:19
we have to deal with many difficult questions about the brain.
2
4000
3000
00:22
But I want to start with the easiest question
3
7000
2000
00:24
and the question you really should have all asked yourselves at some point in your life,
4
9000
3000
00:27
because it's a fundamental question
5
12000
2000
00:29
if we want to understand brain function.
6
14000
2000
00:31
And that is, why do we and other animals
7
16000
2000
00:33
have brains?
8
18000
2000
00:35
Not all species on our planet have brains,
9
20000
3000
00:38
so if we want to know what the brain is for,
10
23000
2000
00:40
let's think about why we evolved one.
11
25000
2000
00:42
Now you may reason that we have one
12
27000
2000
00:44
to perceive the world or to think,
13
29000
2000
00:46
and that's completely wrong.
14
31000
2000
00:48
If you think about this question for any length of time,
15
33000
3000
00:51
it's blindingly obvious why we have a brain.
16
36000
2000
00:53
We have a brain for one reason and one reason only,
17
38000
3000
00:56
and that's to produce adaptable and complex movements.
18
41000
3000
00:59
There is no other reason to have a brain.
19
44000
2000
01:01
Think about it.
20
46000
2000
01:03
Movement is the only way you have
21
48000
2000
01:05
of affecting the world around you.
22
50000
2000
01:07
Now that's not quite true. There's one other way, and that's through sweating.
23
52000
3000
01:10
But apart from that,
24
55000
2000
01:12
everything else goes through contractions of muscles.
25
57000
2000
01:14
So think about communication --
26
59000
2000
01:16
speech, gestures, writing, sign language --
27
61000
3000
01:19
they're all mediated through contractions of your muscles.
28
64000
3000
01:22
So it's really important to remember
29
67000
2000
01:24
that sensory, memory and cognitive processes are all important,
30
69000
4000
01:28
but they're only important
31
73000
2000
01:30
to either drive or suppress future movements.
32
75000
2000
01:32
There can be no evolutionary advantage
33
77000
2000
01:34
to laying down memories of childhood
34
79000
2000
01:36
or perceiving the color of a rose
35
81000
2000
01:38
if it doesn't affect the way you're going to move later in life.
36
83000
3000
01:41
Now for those who don't believe this argument,
37
86000
2000
01:43
we have trees and grass on our planet without the brain,
38
88000
2000
01:45
but the clinching evidence is this animal here --
39
90000
2000
01:47
the humble sea squirt.
40
92000
2000
01:49
Rudimentary animal, has a nervous system,
41
94000
3000
01:52
swims around in the ocean in its juvenile life.
42
97000
2000
01:54
And at some point of its life,
43
99000
2000
01:56
it implants on a rock.
44
101000
2000
01:58
And the first thing it does in implanting on that rock, which it never leaves,
45
103000
3000
02:01
is to digest its own brain and nervous system
46
106000
3000
02:04
for food.
47
109000
2000
02:06
So once you don't need to move,
48
111000
2000
02:08
you don't need the luxury of that brain.
49
113000
3000
02:11
And this animal is often taken
50
116000
2000
02:13
as an analogy to what happens at universities
51
118000
2000
02:15
when professors get tenure,
52
120000
2000
02:17
but that's a different subject.
53
122000
2000
02:19
(Applause)
54
124000
2000
02:21
So I am a movement chauvinist.
55
126000
3000
02:24
I believe movement is the most important function of the brain --
56
129000
2000
02:26
don't let anyone tell you that it's not true.
57
131000
2000
02:28
Now if movement is so important,
58
133000
2000
02:30
how well are we doing
59
135000
2000
02:32
understanding how the brain controls movement?
60
137000
2000
02:34
And the answer is we're doing extremely poorly; it's a very hard problem.
61
139000
2000
02:36
But we can look at how well we're doing
62
141000
2000
02:38
by thinking about how well we're doing building machines
63
143000
2000
02:40
which can do what humans can do.
64
145000
2000
02:42
Think about the game of chess.
65
147000
2000
02:44
How well are we doing determining what piece to move where?
66
149000
3000
02:47
If you pit Garry Kasparov here, when he's not in jail,
67
152000
3000
02:50
against IBM's Deep Blue,
68
155000
2000
02:52
well the answer is IBM's Deep Blue will occasionally win.
69
157000
3000
02:55
And I think if IBM's Deep Blue played anyone in this room, it would win every time.
70
160000
3000
02:58
That problem is solved.
71
163000
2000
03:00
What about the problem
72
165000
2000
03:02
of picking up a chess piece,
73
167000
2000
03:04
dexterously manipulating it and putting it back down on the board?
74
169000
3000
03:07
If you put a five year-old child's dexterity against the best robots of today,
75
172000
3000
03:10
the answer is simple:
76
175000
2000
03:12
the child wins easily.
77
177000
2000
03:14
There's no competition at all.
78
179000
2000
03:16
Now why is that top problem so easy
79
181000
2000
03:18
and the bottom problem so hard?
80
183000
2000
03:20
One reason is a very smart five year-old
81
185000
2000
03:22
could tell you the algorithm for that top problem --
82
187000
2000
03:24
look at all possible moves to the end of the game
83
189000
2000
03:26
and choose the one that makes you win.
84
191000
2000
03:28
So it's a very simple algorithm.
85
193000
2000
03:30
Now of course there are other moves,
86
195000
2000
03:32
but with vast computers we approximate
87
197000
2000
03:34
and come close to the optimal solution.
88
199000
2000
03:36
When it comes to being dexterous,
89
201000
2000
03:38
it's not even clear what the algorithm is you have to solve to be dexterous.
90
203000
2000
03:40
And we'll see you have to both perceive and act on the world,
91
205000
2000
03:42
which has a lot of problems.
92
207000
2000
03:44
But let me show you cutting-edge robotics.
93
209000
2000
03:46
Now a lot of robotics is very impressive,
94
211000
2000
03:48
but manipulation robotics is really just in the dark ages.
95
213000
3000
03:51
So this is the end of a Ph.D. project
96
216000
2000
03:53
from one of the best robotics institutes.
97
218000
2000
03:55
And the student has trained this robot
98
220000
2000
03:57
to pour this water into a glass.
99
222000
2000
03:59
It's a hard problem because the water sloshes about, but it can do it.
100
224000
3000
04:02
But it doesn't do it with anything like the agility of a human.
101
227000
3000
04:05
Now if you want this robot to do a different task,
102
230000
3000
04:08
that's another three-year Ph.D. program.
103
233000
3000
04:11
There is no generalization at all
104
236000
2000
04:13
from one task to another in robotics.
105
238000
2000
04:15
Now we can compare this
106
240000
2000
04:17
to cutting-edge human performance.
107
242000
2000
04:19
So what I'm going to show you is Emily Fox
108
244000
2000
04:21
winning the world record for cup stacking.
109
246000
3000
04:24
Now the Americans in the audience will know all about cup stacking.
110
249000
2000
04:26
It's a high school sport
111
251000
2000
04:28
where you have 12 cups you have to stack and unstack
112
253000
2000
04:30
against the clock in a prescribed order.
113
255000
2000
04:32
And this is her getting the world record in real time.
114
257000
3000
04:39
(Laughter)
115
264000
8000
04:47
(Applause)
116
272000
5000
04:52
And she's pretty happy.
117
277000
2000
04:54
We have no idea what is going on inside her brain when she does that,
118
279000
2000
04:56
and that's what we'd like to know.
119
281000
2000
04:58
So in my group, what we try to do
120
283000
2000
05:00
is reverse engineer how humans control movement.
121
285000
3000
05:03
And it sounds like an easy problem.
122
288000
2000
05:05
You send a command down, it causes muscles to contract.
123
290000
2000
05:07
Your arm or body moves,
124
292000
2000
05:09
and you get sensory feedback from vision, from skin, from muscles and so on.
125
294000
3000
05:12
The trouble is
126
297000
2000
05:14
these signals are not the beautiful signals you want them to be.
127
299000
2000
05:16
So one thing that makes controlling movement difficult
128
301000
2000
05:18
is, for example, sensory feedback is extremely noisy.
129
303000
3000
05:21
Now by noise, I do not mean sound.
130
306000
3000
05:24
We use it in the engineering and neuroscience sense
131
309000
2000
05:26
meaning a random noise corrupting a signal.
132
311000
2000
05:28
So the old days before digital radio when you were tuning in your radio
133
313000
3000
05:31
and you heard "crrcckkk" on the station you wanted to hear,
134
316000
2000
05:33
that was the noise.
135
318000
2000
05:35
But more generally, this noise is something that corrupts the signal.
136
320000
3000
05:38
So for example, if you put your hand under a table
137
323000
2000
05:40
and try to localize it with your other hand,
138
325000
2000
05:42
you can be off by several centimeters
139
327000
2000
05:44
due to the noise in sensory feedback.
140
329000
2000
05:46
Similarly, when you put motor output on movement output,
141
331000
2000
05:48
it's extremely noisy.
142
333000
2000
05:50
Forget about trying to hit the bull's eye in darts,
143
335000
2000
05:52
just aim for the same spot over and over again.
144
337000
2000
05:54
You have a huge spread due to movement variability.
145
339000
3000
05:57
And more than that, the outside world, or task,
146
342000
2000
05:59
is both ambiguous and variable.
147
344000
2000
06:01
The teapot could be full, it could be empty.
148
346000
2000
06:03
It changes over time.
149
348000
2000
06:05
So we work in a whole sensory movement task soup of noise.
150
350000
4000
06:09
Now this noise is so great
151
354000
2000
06:11
that society places a huge premium
152
356000
2000
06:13
on those of us who can reduce the consequences of noise.
153
358000
3000
06:16
So if you're lucky enough to be able to knock a small white ball
154
361000
3000
06:19
into a hole several hundred yards away using a long metal stick,
155
364000
3000
06:22
our society will be willing to reward you
156
367000
2000
06:24
with hundreds of millions of dollars.
157
369000
3000
06:27
Now what I want to convince you of
158
372000
2000
06:29
is the brain also goes through a lot of effort
159
374000
2000
06:31
to reduce the negative consequences
160
376000
2000
06:33
of this sort of noise and variability.
161
378000
2000
06:35
And to do that, I'm going to tell you about a framework
162
380000
2000
06:37
which is very popular in statistics and machine learning of the last 50 years
163
382000
3000
06:40
called Bayesian decision theory.
164
385000
2000
06:42
And it's more recently a unifying way
165
387000
3000
06:45
to think about how the brain deals with uncertainty.
166
390000
3000
06:48
And the fundamental idea is you want to make inferences and then take actions.
167
393000
3000
06:51
So let's think about the inference.
168
396000
2000
06:53
You want to generate beliefs about the world.
169
398000
2000
06:55
So what are beliefs?
170
400000
2000
06:57
Beliefs could be: where are my arms in space?
171
402000
2000
06:59
Am I looking at a cat or a fox?
172
404000
2000
07:01
But we're going to represent beliefs with probabilities.
173
406000
3000
07:04
So we're going to represent a belief
174
409000
2000
07:06
with a number between zero and one --
175
411000
2000
07:08
zero meaning I don't believe it at all, one means I'm absolutely certain.
176
413000
3000
07:11
And numbers in between give you the gray levels of uncertainty.
177
416000
3000
07:14
And the key idea to Bayesian inference
178
419000
2000
07:16
is you have two sources of information
179
421000
2000
07:18
from which to make your inference.
180
423000
2000
07:20
You have data,
181
425000
2000
07:22
and data in neuroscience is sensory input.
182
427000
2000
07:24
So I have sensory input, which I can take in to make beliefs.
183
429000
3000
07:27
But there's another source of information, and that's effectively prior knowledge.
184
432000
3000
07:30
You accumulate knowledge throughout your life in memories.
185
435000
3000
07:33
And the point about Bayesian decision theory
186
438000
2000
07:35
is it gives you the mathematics
187
440000
2000
07:37
of the optimal way to combine
188
442000
2000
07:39
your prior knowledge with your sensory evidence
189
444000
2000
07:41
to generate new beliefs.
190
446000
2000
07:43
And I've put the formula up there.
191
448000
2000
07:45
I'm not going to explain what that formula is, but it's very beautiful.
192
450000
2000
07:47
And it has real beauty and real explanatory power.
193
452000
3000
07:50
And what it really says, and what you want to estimate,
194
455000
2000
07:52
is the probability of different beliefs
195
457000
2000
07:54
given your sensory input.
196
459000
2000
07:56
So let me give you an intuitive example.
197
461000
2000
07:58
Imagine you're learning to play tennis
198
463000
3000
08:01
and you want to decide where the ball is going to bounce
199
466000
2000
08:03
as it comes over the net towards you.
200
468000
2000
08:05
There are two sources of information
201
470000
2000
08:07
Bayes' rule tells you.
202
472000
2000
08:09
There's sensory evidence -- you can use visual information auditory information,
203
474000
3000
08:12
and that might tell you it's going to land in that red spot.
204
477000
3000
08:15
But you know that your senses are not perfect,
205
480000
3000
08:18
and therefore there's some variability of where it's going to land
206
483000
2000
08:20
shown by that cloud of red,
207
485000
2000
08:22
representing numbers between 0.5 and maybe 0.1.
208
487000
3000
08:26
That information is available in the current shot,
209
491000
2000
08:28
but there's another source of information
210
493000
2000
08:30
not available on the current shot,
211
495000
2000
08:32
but only available by repeated experience in the game of tennis,
212
497000
3000
08:35
and that's that the ball doesn't bounce
213
500000
2000
08:37
with equal probability over the court during the match.
214
502000
2000
08:39
If you're playing against a very good opponent,
215
504000
2000
08:41
they may distribute it in that green area,
216
506000
2000
08:43
which is the prior distribution,
217
508000
2000
08:45
making it hard for you to return.
218
510000
2000
08:47
Now both these sources of information carry important information.
219
512000
2000
08:49
And what Bayes' rule says
220
514000
2000
08:51
is that I should multiply the numbers on the red by the numbers on the green
221
516000
3000
08:54
to get the numbers of the yellow, which have the ellipses,
222
519000
3000
08:57
and that's my belief.
223
522000
2000
08:59
So it's the optimal way of combining information.
224
524000
3000
09:02
Now I wouldn't tell you all this if it wasn't that a few years ago,
225
527000
2000
09:04
we showed this is exactly what people do
226
529000
2000
09:06
when they learn new movement skills.
227
531000
2000
09:08
And what it means
228
533000
2000
09:10
is we really are Bayesian inference machines.
229
535000
2000
09:12
As we go around, we learn about statistics of the world and lay that down,
230
537000
4000
09:16
but we also learn
231
541000
2000
09:18
about how noisy our own sensory apparatus is,
232
543000
2000
09:20
and then combine those
233
545000
2000
09:22
in a real Bayesian way.
234
547000
2000
09:24
Now a key part to the Bayesian is this part of the formula.
235
549000
3000
09:27
And what this part really says
236
552000
2000
09:29
is I have to predict the probability
237
554000
2000
09:31
of different sensory feedbacks
238
556000
2000
09:33
given my beliefs.
239
558000
2000
09:35
So that really means I have to make predictions of the future.
240
560000
3000
09:38
And I want to convince you the brain does make predictions
241
563000
2000
09:40
of the sensory feedback it's going to get.
242
565000
2000
09:42
And moreover, it profoundly changes your perceptions
243
567000
2000
09:44
by what you do.
244
569000
2000
09:46
And to do that, I'll tell you
245
571000
2000
09:48
about how the brain deals with sensory input.
246
573000
2000
09:50
So you send a command out,
247
575000
3000
09:53
you get sensory feedback back,
248
578000
2000
09:55
and that transformation is governed
249
580000
2000
09:57
by the physics of your body and your sensory apparatus.
250
582000
3000
10:00
But you can imagine looking inside the brain.
251
585000
2000
10:02
And here's inside the brain.
252
587000
2000
10:04
You might have a little predictor, a neural simulator,
253
589000
2000
10:06
of the physics of your body and your senses.
254
591000
2000
10:08
So as you send a movement command down,
255
593000
2000
10:10
you tap a copy of that off
256
595000
2000
10:12
and run it into your neural simulator
257
597000
2000
10:14
to anticipate the sensory consequences of your actions.
258
599000
4000
10:18
So as I shake this ketchup bottle,
259
603000
2000
10:20
I get some true sensory feedback as the function of time in the bottom row.
260
605000
3000
10:23
And if I've got a good predictor, it predicts the same thing.
261
608000
3000
10:26
Well why would I bother doing that?
262
611000
2000
10:28
I'm going to get the same feedback anyway.
263
613000
2000
10:30
Well there's good reasons.
264
615000
2000
10:32
Imagine, as I shake the ketchup bottle,
265
617000
2000
10:34
someone very kindly comes up to me and taps it on the back for me.
266
619000
3000
10:37
Now I get an extra source of sensory information
267
622000
2000
10:39
due to that external act.
268
624000
2000
10:41
So I get two sources.
269
626000
2000
10:43
I get you tapping on it, and I get me shaking it,
270
628000
3000
10:46
but from my senses' point of view,
271
631000
2000
10:48
that is combined together into one source of information.
272
633000
3000
10:51
Now there's good reason to believe
273
636000
2000
10:53
that you would want to be able to distinguish external events from internal events.
274
638000
3000
10:56
Because external events are actually much more behaviorally relevant
275
641000
3000
10:59
than feeling everything that's going on inside my body.
276
644000
3000
11:02
So one way to reconstruct that
277
647000
2000
11:04
is to compare the prediction --
278
649000
2000
11:06
which is only based on your movement commands --
279
651000
2000
11:08
with the reality.
280
653000
2000
11:10
Any discrepancy should hopefully be external.
281
655000
3000
11:13
So as I go around the world,
282
658000
2000
11:15
I'm making predictions of what I should get, subtracting them off.
283
660000
3000
11:18
Everything left over is external to me.
284
663000
2000
11:20
What evidence is there for this?
285
665000
2000
11:22
Well there's one very clear example
286
667000
2000
11:24
where a sensation generated by myself feels very different
287
669000
2000
11:26
then if generated by another person.
288
671000
2000
11:28
And so we decided the most obvious place to start
289
673000
2000
11:30
was with tickling.
290
675000
2000
11:32
It's been known for a long time, you can't tickle yourself
291
677000
2000
11:34
as well as other people can.
292
679000
2000
11:36
But it hasn't really been shown, it's because you have a neural simulator,
293
681000
3000
11:39
simulating your own body
294
684000
2000
11:41
and subtracting off that sense.
295
686000
2000
11:43
So we can bring the experiments of the 21st century
296
688000
3000
11:46
by applying robotic technologies to this problem.
297
691000
3000
11:49
And in effect, what we have is some sort of stick in one hand attached to a robot,
298
694000
3000
11:52
and they're going to move that back and forward.
299
697000
2000
11:54
And then we're going to track that with a computer
300
699000
2000
11:56
and use it to control another robot,
301
701000
2000
11:58
which is going to tickle their palm with another stick.
302
703000
2000
12:00
And then we're going to ask them to rate a bunch of things
303
705000
2000
12:02
including ticklishness.
304
707000
2000
12:04
I'll show you just one part of our study.
305
709000
2000
12:06
And here I've taken away the robots,
306
711000
2000
12:08
but basically people move with their right arm sinusoidally back and forward.
307
713000
3000
12:11
And we replay that to the other hand with a time delay.
308
716000
3000
12:14
Either no time delay,
309
719000
2000
12:16
in which case light would just tickle your palm,
310
721000
2000
12:18
or with a time delay of two-tenths of three-tenths of a second.
311
723000
4000
12:22
So the important point here
312
727000
2000
12:24
is the right hand always does the same things -- sinusoidal movement.
313
729000
3000
12:27
The left hand always is the same and puts sinusoidal tickle.
314
732000
3000
12:30
All we're playing with is a tempo causality.
315
735000
2000
12:32
And as we go from naught to 0.1 second,
316
737000
2000
12:34
it becomes more ticklish.
317
739000
2000
12:36
As you go from 0.1 to 0.2,
318
741000
2000
12:38
it becomes more ticklish at the end.
319
743000
2000
12:40
And by 0.2 of a second,
320
745000
2000
12:42
it's equivalently ticklish
321
747000
2000
12:44
to the robot that just tickled you without you doing anything.
322
749000
2000
12:46
So whatever is responsible for this cancellation
323
751000
2000
12:48
is extremely tightly coupled with tempo causality.
324
753000
3000
12:51
And based on this illustration, we really convinced ourselves in the field
325
756000
3000
12:54
that the brain's making precise predictions
326
759000
2000
12:56
and subtracting them off from the sensations.
327
761000
3000
12:59
Now I have to admit, these are the worst studies my lab has ever run.
328
764000
3000
13:02
Because the tickle sensation on the palm comes and goes,
329
767000
2000
13:04
you need large numbers of subjects
330
769000
2000
13:06
with these stars making them significant.
331
771000
2000
13:08
So we were looking for a much more objective way
332
773000
2000
13:10
to assess this phenomena.
333
775000
2000
13:12
And in the intervening years I had two daughters.
334
777000
2000
13:14
And one thing you notice about children in backseats of cars on long journeys,
335
779000
3000
13:17
they get into fights --
336
782000
2000
13:19
which started with one of them doing something to the other, the other retaliating.
337
784000
3000
13:22
It quickly escalates.
338
787000
2000
13:24
And children tend to get into fights which escalate in terms of force.
339
789000
3000
13:27
Now when I screamed at my children to stop,
340
792000
2000
13:29
sometimes they would both say to me
341
794000
2000
13:31
the other person hit them harder.
342
796000
3000
13:34
Now I happen to know my children don't lie,
343
799000
2000
13:36
so I thought, as a neuroscientist,
344
801000
2000
13:38
it was important how I could explain
345
803000
2000
13:40
how they were telling inconsistent truths.
346
805000
2000
13:42
And we hypothesize based on the tickling study
347
807000
2000
13:44
that when one child hits another,
348
809000
2000
13:46
they generate the movement command.
349
811000
2000
13:48
They predict the sensory consequences and subtract it off.
350
813000
3000
13:51
So they actually think they've hit the person less hard than they have --
351
816000
2000
13:53
rather like the tickling.
352
818000
2000
13:55
Whereas the passive recipient
353
820000
2000
13:57
doesn't make the prediction, feels the full blow.
354
822000
2000
13:59
So if they retaliate with the same force,
355
824000
2000
14:01
the first person will think it's been escalated.
356
826000
2000
14:03
So we decided to test this in the lab.
357
828000
2000
14:05
(Laughter)
358
830000
3000
14:08
Now we don't work with children, we don't work with hitting,
359
833000
2000
14:10
but the concept is identical.
360
835000
2000
14:12
We bring in two adults. We tell them they're going to play a game.
361
837000
3000
14:15
And so here's player one and player two sitting opposite to each other.
362
840000
2000
14:17
And the game is very simple.
363
842000
2000
14:19
We started with a motor
364
844000
2000
14:21
with a little lever, a little force transfuser.
365
846000
2000
14:23
And we use this motor to apply force down to player one's fingers
366
848000
2000
14:25
for three seconds and then it stops.
367
850000
3000
14:28
And that player's been told, remember the experience of that force
368
853000
3000
14:31
and use your other finger
369
856000
2000
14:33
to apply the same force
370
858000
2000
14:35
down to the other subject's finger through a force transfuser -- and they do that.
371
860000
3000
14:38
And player two's been told, remember the experience of that force.
372
863000
3000
14:41
Use your other hand to apply the force back down.
373
866000
3000
14:44
And so they take it in turns
374
869000
2000
14:46
to apply the force they've just experienced back and forward.
375
871000
2000
14:48
But critically,
376
873000
2000
14:50
they're briefed about the rules of the game in separate rooms.
377
875000
3000
14:53
So they don't know the rules the other person's playing by.
378
878000
2000
14:55
And what we've measured
379
880000
2000
14:57
is the force as a function of terms.
380
882000
2000
14:59
And if we look at what we start with,
381
884000
2000
15:01
a quarter of a Newton there, a number of turns,
382
886000
2000
15:03
perfect would be that red line.
383
888000
2000
15:05
And what we see in all pairs of subjects is this --
384
890000
3000
15:08
a 70 percent escalation in force
385
893000
2000
15:10
on each go.
386
895000
2000
15:12
So it really suggests, when you're doing this --
387
897000
2000
15:14
based on this study and others we've done --
388
899000
2000
15:16
that the brain is canceling the sensory consequences
389
901000
2000
15:18
and underestimating the force it's producing.
390
903000
2000
15:20
So it re-shows the brain makes predictions
391
905000
2000
15:22
and fundamentally changes the precepts.
392
907000
3000
15:25
So we've made inferences, we've done predictions,
393
910000
3000
15:28
now we have to generate actions.
394
913000
2000
15:30
And what Bayes' rule says is, given my beliefs,
395
915000
2000
15:32
the action should in some sense be optimal.
396
917000
2000
15:34
But we've got a problem.
397
919000
2000
15:36
Tasks are symbolic -- I want to drink, I want to dance --
398
921000
3000
15:39
but the movement system has to contract 600 muscles
399
924000
2000
15:41
in a particular sequence.
400
926000
2000
15:43
And there's a big gap
401
928000
2000
15:45
between the task and the movement system.
402
930000
2000
15:47
So it could be bridged in infinitely many different ways.
403
932000
2000
15:49
So think about just a point to point movement.
404
934000
2000
15:51
I could choose these two paths
405
936000
2000
15:53
out of an infinite number of paths.
406
938000
2000
15:55
Having chosen a particular path,
407
940000
2000
15:57
I can hold my hand on that path
408
942000
2000
15:59
as infinitely many different joint configurations.
409
944000
2000
16:01
And I can hold my arm in a particular joint configuration
410
946000
2000
16:03
either very stiff or very relaxed.
411
948000
2000
16:05
So I have a huge amount of choice to make.
412
950000
3000
16:08
Now it turns out, we are extremely stereotypical.
413
953000
3000
16:11
We all move the same way pretty much.
414
956000
3000
16:14
And so it turns out we're so stereotypical,
415
959000
2000
16:16
our brains have got dedicated neural circuitry
416
961000
2000
16:18
to decode this stereotyping.
417
963000
2000
16:20
So if I take some dots
418
965000
2000
16:22
and set them in motion with biological motion,
419
967000
3000
16:25
your brain's circuitry would understand instantly what's going on.
420
970000
3000
16:28
Now this is a bunch of dots moving.
421
973000
2000
16:30
You will know what this person is doing,
422
975000
3000
16:33
whether happy, sad, old, young -- a huge amount of information.
423
978000
3000
16:36
If these dots were cars going on a racing circuit,
424
981000
2000
16:38
you would have absolutely no idea what's going on.
425
983000
3000
16:41
So why is it
426
986000
2000
16:43
that we move the particular ways we do?
427
988000
2000
16:45
Well let's think about what really happens.
428
990000
2000
16:47
Maybe we don't all quite move the same way.
429
992000
3000
16:50
Maybe there's variation in the population.
430
995000
2000
16:52
And maybe those who move better than others
431
997000
2000
16:54
have got more chance of getting their children into the next generation.
432
999000
2000
16:56
So in evolutionary scales, movements get better.
433
1001000
3000
16:59
And perhaps in life, movements get better through learning.
434
1004000
3000
17:02
So what is it about a movement which is good or bad?
435
1007000
2000
17:04
Imagine I want to intercept this ball.
436
1009000
2000
17:06
Here are two possible paths to that ball.
437
1011000
3000
17:09
Well if I choose the left-hand path,
438
1014000
2000
17:11
I can work out the forces required
439
1016000
2000
17:13
in one of my muscles as a function of time.
440
1018000
2000
17:15
But there's noise added to this.
441
1020000
2000
17:17
So what I actually get, based on this lovely, smooth, desired force,
442
1022000
3000
17:20
is a very noisy version.
443
1025000
2000
17:22
So if I pick the same command through many times,
444
1027000
3000
17:25
I will get a different noisy version each time, because noise changes each time.
445
1030000
3000
17:28
So what I can show you here
446
1033000
2000
17:30
is how the variability of the movement will evolve
447
1035000
2000
17:32
if I choose that way.
448
1037000
2000
17:34
If I choose a different way of moving -- on the right for example --
449
1039000
3000
17:37
then I'll have a different command, different noise,
450
1042000
2000
17:39
playing through a noisy system, very complicated.
451
1044000
3000
17:42
All we can be sure of is the variability will be different.
452
1047000
3000
17:45
If I move in this particular way,
453
1050000
2000
17:47
I end up with a smaller variability across many movements.
454
1052000
3000
17:50
So if I have to choose between those two,
455
1055000
2000
17:52
I would choose the right one because it's less variable.
456
1057000
2000
17:54
And the fundamental idea
457
1059000
2000
17:56
is you want to plan your movements
458
1061000
2000
17:58
so as to minimize the negative consequence of the noise.
459
1063000
3000
18:01
And one intuition to get
460
1066000
2000
18:03
is actually the amount of noise or variability I show here
461
1068000
2000
18:05
gets bigger as the force gets bigger.
462
1070000
2000
18:07
So you want to avoid big forces as one principle.
463
1072000
3000
18:10
So we've shown that using this,
464
1075000
2000
18:12
we can explain a huge amount of data --
465
1077000
2000
18:14
that exactly people are going about their lives planning movements
466
1079000
3000
18:17
so as to minimize negative consequences of noise.
467
1082000
3000
18:20
So I hope I've convinced you the brain is there
468
1085000
2000
18:22
and evolved to control movement.
469
1087000
2000
18:24
And it's an intellectual challenge to understand how we do that.
470
1089000
3000
18:27
But it's also relevant
471
1092000
2000
18:29
for disease and rehabilitation.
472
1094000
2000
18:31
There are many diseases which effect movement.
473
1096000
3000
18:34
And hopefully if we understand how we control movement,
474
1099000
2000
18:36
we can apply that to robotic technology.
475
1101000
2000
18:38
And finally, I want to remind you,
476
1103000
2000
18:40
when you see animals do what look like very simple tasks,
477
1105000
2000
18:42
the actual complexity of what is going on inside their brain
478
1107000
2000
18:44
is really quite dramatic.
479
1109000
2000
18:46
Thank you very much.
480
1111000
2000
18:48
(Applause)
481
1113000
8000
18:56
Chris Anderson: Quick question for you, Dan.
482
1121000
2000
18:58
So you're a movement -- (DW: Chauvinist.) -- chauvinist.
483
1123000
4000
19:02
Does that mean that you think that the other things we think our brains are about --
484
1127000
3000
19:05
the dreaming, the yearning, the falling in love and all these things --
485
1130000
3000
19:08
are a kind of side show, an accident?
486
1133000
3000
19:11
DW: No, no, actually I think they're all important
487
1136000
2000
19:13
to drive the right movement behavior to get reproduction in the end.
488
1138000
3000
19:16
So I think people who study sensation or memory
489
1141000
3000
19:19
without realizing why you're laying down memories of childhood.
490
1144000
2000
19:21
The fact that we forget most of our childhood, for example,
491
1146000
3000
19:24
is probably fine, because it doesn't effect our movements later in life.
492
1149000
3000
19:27
You only need to store things which are really going to effect movement.
493
1152000
3000
19:30
CA: So you think that people thinking about the brain, and consciousness generally,
494
1155000
3000
19:33
could get real insight
495
1158000
2000
19:35
by saying, where does movement play in this game?
496
1160000
2000
19:37
DW: So people have found out for example
497
1162000
2000
19:39
that studying vision in the absence of realizing why you have vision
498
1164000
2000
19:41
is a mistake.
499
1166000
2000
19:43
You have to study vision with the realization
500
1168000
2000
19:45
of how the movement system is going to use vision.
501
1170000
2000
19:47
And it uses it very differently once you think about it that way.
502
1172000
2000
19:49
CA: Well that was quite fascinating. Thank you very much indeed.
503
1174000
3000
19:52
(Applause)
504
1177000
2000
ABOUT THE SPEAKER
Daniel Wolpert - Movement expertA neuroscientist and engineer, Daniel Wolpert studies how the brain controls the body.
Why you should listen
Consider your hand. You use it to lift things, to balance yourself, to give and take, to sense the world. It has a range of interacting degrees of freedom, and it interacts with many different objects under a variety of environmental conditions. And for most of us, it all just works. At his lab in the Engineering department at Cambridge, Daniel Wolpert and his team are studying why, looking to understand the computations underlying the brain's sensorimotor control of the body.
As he says, "I believe that to understand movement is to understand the whole brain. And therefore it’s important to remember when you are studying memory, cognition, sensory processing, they’re there for a reason, and that reason is action.” Movement is the only way we have of interacting with the world, whether foraging for food or attracting a waiter's attention. Indeed, all communication, including speech, sign language, gestures and writing, is mediated via the motor system. Taking this viewpoint, and using computational and robotic techniques as well as virtual reality systems, Wolpert and his team research the purpose of the human brain and the way it determines future actions.
Daniel Wolpert | Speaker | TED.com