TED2017
Radhika Nagpal: What intelligent machines can learn from a school of fish
Filmed:
Readability: 4.3
1,266,261 views
Science fiction visions of the future show us AI built to replicate our way of thinking -- but what if we modeled it instead on the other kinds of intelligence found in nature? Robotics engineer Radhika Nagpal studies the collective intelligence displayed by insects and fish schools, seeking to understand their rules of engagement. In a visionary talk, she presents her work creating artificial collective power and previews a future where swarms of robots work together to build flood barriers, pollinate crops, monitor coral reefs and form constellations of satellites.
Radhika Nagpal - Robotics engineer
Taking cues from bottom-up biological networks like those of social insects, Radhika Nagpal helped design an unprecedented “swarm” of ant-like robots. Full bio
Taking cues from bottom-up biological networks like those of social insects, Radhika Nagpal helped design an unprecedented “swarm” of ant-like robots. Full bio
Double-click the English transcript below to play the video.
00:12
In my early days as a graduate student,
0
755
2015
00:14
I went on a snorkeling trip
off the coast of the Bahamas.
off the coast of the Bahamas.
1
2794
3555
00:18
I'd actually never swum
in the ocean before,
in the ocean before,
2
6789
2949
00:21
so it was a bit terrifying.
3
9762
1844
00:24
What I remember the most is,
as I put my head in the water
as I put my head in the water
4
12016
3000
00:27
and I was trying really hard
to breathe through the snorkel,
to breathe through the snorkel,
5
15040
4250
00:31
this huge group
of striped yellow and black fish
of striped yellow and black fish
6
19314
5366
00:36
came straight at me ...
7
24704
1483
00:38
and I just froze.
8
26817
1397
00:40
And then, as if it had
suddenly changed its mind,
suddenly changed its mind,
9
28975
3613
00:44
came towards me
and then swerved to the right
and then swerved to the right
10
32612
2437
00:47
and went right around me.
11
35073
1515
00:48
It was absolutely mesmerizing.
12
36920
1526
00:50
Maybe many of you
have had this experience.
have had this experience.
13
38470
2182
00:53
Of course, there's the color
and the beauty of it,
and the beauty of it,
14
41239
3422
00:56
but there was also
just the sheer oneness of it,
just the sheer oneness of it,
15
44685
2928
00:59
as if it wasn't hundreds of fish
16
47637
2343
01:02
but a single entity
with a single collective mind
with a single collective mind
17
50004
3135
01:05
that was making decisions.
18
53163
1507
01:07
When I look back, I think that experience
really ended up determining
really ended up determining
19
55666
3682
01:11
what I've worked on for most of my career.
20
59372
2222
01:16
I'm a computer scientist,
21
64157
1280
01:17
and the field that I work in
is artificial intelligence.
is artificial intelligence.
22
65461
2747
01:20
And a key theme in AI
23
68639
1517
01:22
is being able to understand intelligence
by creating our own computational systems
by creating our own computational systems
24
70180
4443
01:26
that display intelligence
the way we see it in nature.
the way we see it in nature.
25
74647
3253
01:30
Now, most popular views of AI, of course,
come from science fiction and the movies,
come from science fiction and the movies,
26
78467
4438
01:34
and I'm personally a big Star Wars fan.
27
82929
2577
01:38
But that tends to be a very human-centric
view of intelligence.
view of intelligence.
28
86501
3889
01:43
When you think of a fish school,
29
91144
2207
01:45
or when I think of a flock of starlings,
30
93375
2953
01:48
that feels like a really different
kind of intelligence.
kind of intelligence.
31
96352
3340
01:52
For starters, any one fish is just so tiny
32
100945
3913
01:56
compared to the sheer size
of the collective,
of the collective,
33
104882
2887
01:59
so it seems that any one individual
34
107793
3110
02:02
would have a really limited
and myopic view of what's going on,
and myopic view of what's going on,
35
110927
2993
02:05
and intelligence
isn't really about the individual
isn't really about the individual
36
113944
2334
02:08
but somehow a property
of the group itself.
of the group itself.
37
116302
2677
02:12
Secondly, and the thing
that I still find most remarkable,
that I still find most remarkable,
38
120118
3231
02:15
is that we know that there are no leaders
supervising this fish school.
supervising this fish school.
39
123373
5032
02:21
Instead, this incredible
collective mind behavior
collective mind behavior
40
129163
3501
02:24
is emerging purely from the interactions
of one fish and another.
of one fish and another.
41
132688
4532
02:29
Somehow, there are these interactions
or rules of engagement
or rules of engagement
42
137244
3968
02:33
between neighboring fish
43
141236
1755
02:35
that make it all work out.
44
143015
1467
02:37
So the question for AI then becomes,
45
145736
2651
02:40
what are those rules of engagement
that lead to this kind of intelligence,
that lead to this kind of intelligence,
46
148411
4158
02:44
and of course, can we create our own?
47
152593
1907
02:46
And that's the primary thing
that I work on with my team in my lab.
that I work on with my team in my lab.
48
154999
3587
02:50
We work on it through theory,
49
158943
1637
02:52
looking at abstract rule systems
50
160604
2348
02:54
and thinking about
the mathematics behind it.
the mathematics behind it.
51
162976
2349
02:57
We also do it through biology,
working closely with experimentalists.
working closely with experimentalists.
52
165897
4285
03:02
But mostly, we do it through robotics,
53
170579
1953
03:04
where we try to create
our own collective systems
our own collective systems
54
172556
3904
03:08
that can do the kinds of things
that we see in nature,
that we see in nature,
55
176484
2707
03:11
or at least try to.
56
179215
1237
03:13
One of our first robotic quests
along this line
along this line
57
181907
2804
03:16
was to create our very own colony
of a thousand robots.
of a thousand robots.
58
184735
4045
03:21
So very simple robots,
59
189140
1334
03:22
but they could be programmed
to exhibit collective intelligence,
to exhibit collective intelligence,
60
190498
3603
03:26
and that's what we were able to do.
61
194125
1729
03:28
So this is what a single robot looks like.
62
196194
2032
03:30
It's quite small,
about the size of a quarter,
about the size of a quarter,
63
198250
2523
03:32
and you can program how it moves,
64
200797
2310
03:35
but it can also wirelessly
communicate with other robots,
communicate with other robots,
65
203131
3416
03:38
and it can measure distances from them.
66
206571
2167
03:40
And so now we can start to program
exactly an interaction,
exactly an interaction,
67
208762
3476
03:44
a rule of engagement between neighbors.
68
212262
2094
03:46
And once we have this system,
69
214713
1894
03:48
we can start to program many
different kinds of rules of engagement
different kinds of rules of engagement
70
216631
3416
03:52
that you would see in nature.
71
220071
1506
03:53
So for example,
spontaneous synchronization,
spontaneous synchronization,
72
221601
2976
03:56
how audiences are clapping
and suddenly start all clapping together,
and suddenly start all clapping together,
73
224601
5238
04:01
the fireflies flashing together.
74
229863
2068
04:06
We can program rules
for pattern formation,
for pattern formation,
75
234919
2691
04:09
how cells in a tissue
76
237634
1786
04:11
determine what role
they're going to take on
they're going to take on
77
239444
2102
04:13
and set the patterns of our bodies.
78
241570
1706
04:17
We can program rules for migration,
79
245045
2089
04:19
and in this way, we're really learning
from nature's rules.
from nature's rules.
80
247158
2977
04:22
But we can also take it a step further.
81
250595
2647
04:25
We can actually take these rules
that we've learned from nature
that we've learned from nature
82
253266
2992
04:28
and combine them and create
entirely new collective behaviors
entirely new collective behaviors
83
256282
3794
04:32
of our very own.
84
260100
1198
04:33
So for example,
85
261960
1478
04:35
imagine that you had
two different kinds of rules.
two different kinds of rules.
86
263462
2352
04:38
So your first rule is a motion rule
87
266374
2119
04:40
where a moving robot can move
around other stationary robots.
around other stationary robots.
88
268517
4341
04:44
And your second rule is a pattern rule
89
272882
1811
04:46
where a robot takes on a color
based on its two nearest neighbors.
based on its two nearest neighbors.
90
274717
3157
04:50
So if I start with a blob of robots
in a little pattern seed,
in a little pattern seed,
91
278679
3445
04:54
it turns out that these two rules
are sufficient for the group
are sufficient for the group
92
282148
2906
04:57
to be able to self-assemble
a simple line pattern.
a simple line pattern.
93
285078
2752
05:01
And if I have more
complicated pattern rules,
complicated pattern rules,
94
289114
2544
05:03
and I design error correction rules,
95
291682
2317
05:06
we can actually create really,
really complicated self assemblies,
really complicated self assemblies,
96
294023
3097
05:09
and here's what that looks like.
97
297144
1644
05:11
So here, you're going to see
a thousand robots
a thousand robots
98
299874
2985
05:14
that are working together
to self-assemble the letter K.
to self-assemble the letter K.
99
302883
3462
05:18
The K is on its side.
100
306369
1306
05:20
And the important thing
is that no one is in charge.
is that no one is in charge.
101
308223
2731
05:22
So any single robot is only talking
to a small number of robots nearby it,
to a small number of robots nearby it,
102
310978
4825
05:27
and it's using its motion rule
to move around the half-built structure
to move around the half-built structure
103
315827
3937
05:31
just looking for a place to fit in
based on its pattern rules.
based on its pattern rules.
104
319788
3007
05:35
And even though no robot
is doing anything perfectly,
is doing anything perfectly,
105
323794
4398
05:40
the rules are such that we can get
the collective to do its goal
the collective to do its goal
106
328216
3660
05:43
robustly together.
107
331900
1473
05:46
And the illusion becomes
almost so perfect, you know --
almost so perfect, you know --
108
334033
2982
05:49
you just start to not even notice
that they're individual robots at all,
that they're individual robots at all,
109
337039
3416
05:52
and it becomes a single entity,
110
340479
1683
05:54
kind of like the school of fish.
111
342186
1721
06:00
So these are robots and rules
in two dimensions,
in two dimensions,
112
348013
2739
06:02
but we can also think about robots
and rules in three dimensions.
and rules in three dimensions.
113
350776
3311
06:06
So what if we could create robots
that could build together?
that could build together?
114
354111
3603
06:10
And here, we can take inspiration
from social insects.
from social insects.
115
358576
3255
06:14
So if you think about
mound-building termites
mound-building termites
116
362189
2660
06:16
or you think about army ants,
117
364873
2052
06:18
they create incredible,
complex nest structures out of mud
complex nest structures out of mud
118
366949
4253
06:23
and even out of their own bodies.
119
371226
2144
06:26
And like the system I showed you before,
120
374602
2220
06:28
these insects actually
also have pattern rules
also have pattern rules
121
376846
2970
06:31
that help them determine what to build,
122
379840
2038
06:33
but the pattern can be made
out of other insects,
out of other insects,
123
381902
2302
06:36
or it could be made out of mud.
124
384228
1787
06:39
And we can use that same idea
to create rules for robots.
to create rules for robots.
125
387178
4361
06:44
So here, you're going to see
some simulated robots.
some simulated robots.
126
392221
3161
06:47
So the simulated robot has a motion rule,
127
395406
2483
06:49
which is how it traverses
through the structure,
through the structure,
128
397913
2333
06:52
looking for a place to fit in,
129
400270
1997
06:54
and it has pattern rules
where it looks at groups of blocks
where it looks at groups of blocks
130
402291
3000
06:57
to decide whether to place a block.
131
405315
2205
07:00
And with the right motion rules
and the right pattern rules,
and the right pattern rules,
132
408644
3063
07:03
we can actually get the robots
to build whatever we want.
to build whatever we want.
133
411731
3635
07:08
And of course, everybody wants
their own tower.
their own tower.
134
416197
2691
07:11
(Laughter)
135
419350
1982
07:14
So once we have these rules,
136
422000
1684
07:15
we can start to create the robot bodies
that go with these rules.
that go with these rules.
137
423708
3166
07:18
So here, you see a robot
that can climb over blocks,
that can climb over blocks,
138
426898
3309
07:22
but it can also lift and move these blocks
139
430231
2681
07:24
and it can start to edit
the very structure that it's on.
the very structure that it's on.
140
432936
2697
07:28
But with these rules,
141
436617
1148
07:29
this is really only one kind of robot body
that you could imagine.
that you could imagine.
142
437789
3479
07:33
You could imagine
many different kinds of robot bodies.
many different kinds of robot bodies.
143
441292
2579
07:35
So if you think about robots
that maybe could move sandbags
that maybe could move sandbags
144
443895
4610
07:40
and could help build levees,
145
448529
2549
07:43
or we could think of robots
that built out of soft materials
that built out of soft materials
146
451102
4301
07:47
and worked together
to shore up a collapsed building --
to shore up a collapsed building --
147
455427
3644
07:51
so just the same kind of rules
in different kinds of bodies.
in different kinds of bodies.
148
459095
2998
07:56
Or if, like my group, you are completely
obsessed with army ants,
obsessed with army ants,
149
464210
4223
08:00
then maybe one day we can make robots
that can climb over literally anything
that can climb over literally anything
150
468457
4374
08:04
including other members of their tribe,
151
472855
2174
08:07
and self-assemble things
out of their own bodies.
out of their own bodies.
152
475053
2349
08:10
Once you understand the rules,
153
478137
1681
08:11
just many different kinds
of robot visions become possible.
of robot visions become possible.
154
479842
3379
08:18
And coming back to the snorkeling trip,
155
486792
2234
08:21
we actually understand a great deal
about the rules that fish schools use.
about the rules that fish schools use.
156
489050
5345
08:26
So if we can invent
the bodies to go with that,
the bodies to go with that,
157
494769
2836
08:29
then maybe there is a future
158
497629
1428
08:31
where I and my group will get to snorkel
with a fish school of our own creation.
with a fish school of our own creation.
159
499081
4522
08:40
Each of these systems that I showed you
160
508850
2129
08:43
brings us closer to having
the mathematical and the conceptual tools
the mathematical and the conceptual tools
161
511003
4277
08:47
to create our own versions
of collective power,
of collective power,
162
515304
3381
08:50
and this can enable many different kinds
of future applications,
of future applications,
163
518709
3001
08:53
whether you think about robots
that build flood barriers
that build flood barriers
164
521734
3164
08:56
or you think about robotic bee colonies
that could pollinate crops
that could pollinate crops
165
524922
4297
09:01
or underwater schools of robots
that monitor coral reefs,
that monitor coral reefs,
166
529243
3524
09:04
or if we reach for the stars
and we thinking about programming
and we thinking about programming
167
532791
3103
09:07
constellations of satellites.
168
535918
1619
09:10
In each of these systems,
169
538148
1612
09:11
being able to understand
how to design the rules of engagement
how to design the rules of engagement
170
539784
3547
09:15
and being able to create
good collective behavior
good collective behavior
171
543355
2514
09:17
becomes a key to realizing these visions.
172
545893
2374
09:22
So, so far I've talked about
rules for insects and for fish
rules for insects and for fish
173
550742
4107
09:26
and for robots,
174
554873
2369
09:29
but what about the rules that apply
to our own human collective?
to our own human collective?
175
557266
3103
09:32
And the last thought
that I'd like to leave you with
that I'd like to leave you with
176
560866
2430
09:35
is that science is of course itself
177
563320
1681
09:37
an incredible manifestation
of collective intelligence,
of collective intelligence,
178
565025
3484
09:40
but unlike the beautiful
fish schools that I study,
fish schools that I study,
179
568533
3318
09:43
I feel we still have a much longer
evolutionary path to walk.
evolutionary path to walk.
180
571875
3943
09:48
So in addition to working on improving
the science of robot collectives,
the science of robot collectives,
181
576746
4604
09:53
I also work on creating robots
and thinking about rules
and thinking about rules
182
581374
3277
09:56
that will improve
our own scientific collective.
our own scientific collective.
183
584675
2460
10:00
There's this saying that I love:
184
588198
1668
10:01
who does science
determines what science gets done.
determines what science gets done.
185
589890
3404
10:06
Imagine a society
186
594239
2941
10:09
where we had rules of engagement
187
597204
1651
10:10
where every child grew up believing
that they could stand here
that they could stand here
188
598879
3303
10:14
and be a technologist of the future,
189
602206
2422
10:16
or where every adult
190
604652
1501
10:18
believed that they had the ability
not just to understand but to change
not just to understand but to change
191
606177
4119
10:22
how science and technology
impacts their everyday lives.
impacts their everyday lives.
192
610320
3555
10:26
What would that society look like?
193
614820
1899
10:30
I believe that we can do that.
194
618386
1508
10:31
I believe that we can choose our rules,
195
619918
2291
10:34
and we engineer not just robots
196
622233
1757
10:36
but we can engineer
our own human collective,
our own human collective,
197
624014
2596
10:38
and if we do and when we do,
it will be beautiful.
it will be beautiful.
198
626634
3834
10:42
Thank you.
199
630492
1151
10:43
(Applause)
200
631667
6547
ABOUT THE SPEAKER
Radhika Nagpal - Robotics engineerTaking cues from bottom-up biological networks like those of social insects, Radhika Nagpal helped design an unprecedented “swarm” of ant-like robots.
Why you should listen
With a swarm of 1,024 robots inspired by the design of ant colonies, Radhika Nagpal and her colleagues at Harvard’s SSR research group have redefined expectations for self-organizing robotic systems. Guided by algorithms, Nagpal’s shockingly simple robots guide themselves into a variety of shapes -- an ability that, brought to scale, might lead to applications like disaster rescue, space exploration and beyond.
In addition to her work with biologically inspired robots, Nagpal helped create ROOT, a simple robot to teach coding to would-be programmers through a simple user interface suitable for students of all ages.
Radhika Nagpal | Speaker | TED.com