ABOUT THE SPEAKER
Deborah Gordon - Ecologist
By studying how ant colonies work without any one leader, Deborah Gordon has identified striking similarities in how ant colonies, brains, cells and computer networks regulate themselves.

Why you should listen

Ecologist Deborah M. Gordon has learned that ant colonies can work without central control by using simple interactions like how often the insects touch antennae. Contrary to the notion that colonies are organized by efficient ants, she has instead discovered that evolution has produced “noisy” systems that tolerate accident and respond flexibly to the environment. When conditions are tough, natural selection favors colonies that conserve resources.

Her studies of ant colonies have led her and her Stanford colleagues to the discovery of the “Anternet,” which regulates foraging in ants in the same way the internet regulates data traffic. But as she said to Wired in 2013, "Insect behavior mimicking human networks ... is actually not what’s most interesting about ant networks. What’s far more interesting are the parallels in the other direction: What have the ants worked out that we humans haven’t thought of yet?" Her latest exploration: How do ants behave in space?

More profile about the speaker
Deborah Gordon | Speaker | TED.com
TED2014

Deborah Gordon: What ants teach us about the brain, cancer and the Internet

Filmed:
1,481,089 views

Ecologist Deborah Gordon studies ants wherever she can find them -- in the desert, in the tropics, in her kitchen ... In this fascinating talk, she explains her obsession with insects most of us would happily swat away without a second thought. She argues that ant life provides a useful model for learning about many other topics, including disease, technology and the human brain.
- Ecologist
By studying how ant colonies work without any one leader, Deborah Gordon has identified striking similarities in how ant colonies, brains, cells and computer networks regulate themselves. Full bio

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

00:13
I study ants
0
1099
1619
00:14
in the desert, in the tropical forest
1
2718
3297
00:18
and in my kitchen,
2
6015
1528
00:19
and in the hills around Silicon Valley where I live.
3
7543
4020
00:23
I've recently realized that ants
4
11563
1676
00:25
are using interactions differently
5
13239
2278
00:27
in different environments,
6
15517
1828
00:29
and that got me thinking
that we could learn from this
7
17345
2000
00:31
about other systems,
8
19345
1555
00:32
like brains and data networks that we engineer,
9
20900
5588
00:38
and even cancer.
10
26488
2846
00:41
So what all these systems have in common
11
29334
1752
00:43
is that there's no central control.
12
31086
2857
00:45
An ant colony consists of sterile female workers --
13
33943
3820
00:49
those are the ants you see walking around —
14
37763
2386
00:52
and then one or more reproductive females
15
40149
2156
00:54
who just lay the eggs.
16
42305
2103
00:56
They don't give any instructions.
17
44408
2009
00:58
Even though they're called queens,
18
46417
1984
01:00
they don't tell anybody what to do.
19
48401
2165
01:02
So in an ant colony, there's no one in charge,
20
50566
3276
01:05
and all systems like this without central control
21
53842
2778
01:08
are regulated using very simple interactions.
22
56620
3919
01:12
Ants interact using smell.
23
60549
2727
01:15
They smell with their antennae,
24
63276
2187
01:17
and they interact with their antennae,
25
65463
2761
01:20
so when one ant touches another with its antennae,
26
68224
3009
01:23
it can tell, for example, if the other ant
27
71233
1607
01:24
is a nestmate
28
72840
1634
01:26
and what task that other ant has been doing.
29
74474
4236
01:30
So here you see a lot of ants moving around
30
78710
3200
01:33
and interacting in a lab arena
31
81910
1933
01:35
that's connected by tubes to two other arenas.
32
83843
3438
01:39
So when one ant meets another,
33
87281
2504
01:41
it doesn't matter which ant it meets,
34
89785
2089
01:43
and they're actually not transmitting
35
91874
1885
01:45
any kind of complicated signal or message.
36
93759
3465
01:49
All that matters to the ant is the rate
37
97224
2111
01:51
at which it meets other ants.
38
99335
2821
01:54
And all of these interactions, taken together,
39
102156
2757
01:56
produce a network.
40
104913
2505
01:59
So this is the network of the ants
41
107418
2210
02:01
that you just saw moving around in the arena,
42
109628
2644
02:04
and it's this constantly shifting network
43
112272
3287
02:07
that produces the behavior of the colony,
44
115559
2520
02:10
like whether all the ants are hiding inside the nest,
45
118079
2881
02:12
or how many are going out to forage.
46
120960
2694
02:15
A brain actually works in the same way,
47
123654
1759
02:17
but what's great about ants is
48
125413
1500
02:18
that you can see the whole network as it happens.
49
126913
4867
02:23
There are more than 12,000 species of ants,
50
131780
3100
02:26
in every conceivable environment,
51
134880
2125
02:29
and they're using interactions differently
52
137005
2649
02:31
to meet different environmental challenges.
53
139654
2768
02:34
So one important environmental challenge
54
142422
2305
02:36
that every system has to deal with
55
144727
1870
02:38
is operating costs, just what it takes
56
146597
2083
02:40
to run the system.
57
148680
2265
02:42
And another environmental challenge is resources,
58
150945
2548
02:45
finding them and collecting them.
59
153493
2336
02:47
In the desert, operating costs are high
60
155829
2783
02:50
because water is scarce,
61
158612
1773
02:52
and the seed-eating ants that I study in the desert
62
160385
2499
02:54
have to spend water to get water.
63
162884
2900
02:57
So an ant outside foraging,
64
165784
2139
02:59
searching for seeds in the hot sun,
65
167923
2037
03:01
just loses water into the air.
66
169960
2007
03:03
But the colony gets its water
67
171967
1664
03:05
by metabolizing the fats out of the seeds
68
173631
1766
03:07
that they eat.
69
175397
1934
03:09
So in this environment, interactions are used
70
177331
3159
03:12
to activate foraging.
71
180490
1549
03:14
An outgoing forager doesn't go out unless
72
182039
2264
03:16
it gets enough interactions with returning foragers,
73
184303
2573
03:18
and what you see are the returning foragers
74
186876
1990
03:20
going into the tunnel, into the nest,
75
188866
1824
03:22
and meeting outgoing foragers on their way out.
76
190690
2636
03:25
This makes sense for the ant colony,
77
193326
1623
03:26
because the more food there is out there,
78
194949
2323
03:29
the more quickly the foragers find it,
79
197272
1851
03:31
the faster they come back,
80
199123
1236
03:32
and the more foragers they send out.
81
200359
2801
03:35
The system works to stay stopped,
82
203160
2507
03:37
unless something positive happens.
83
205667
2015
03:39
So interactions function to activate foragers.
84
207682
3881
03:43
And we've been studying
the evolution of this system.
85
211563
2907
03:46
First of all, there's variation.
86
214470
1091
03:47
It turns out that colonies are different.
87
215561
2479
03:50
On dry days, some colonies forage less,
88
218040
2696
03:52
so colonies are different in how
89
220736
1505
03:54
they manage this trade-off
90
222241
1354
03:55
between spending water to search for seeds
91
223595
3168
03:58
and getting water back in the form of seeds.
92
226763
3320
04:02
And we're trying to understand why
93
230083
1719
04:03
some colonies forage less than others
94
231802
2140
04:05
by thinking about ants as neurons,
95
233942
2325
04:08
using models from neuroscience.
96
236267
2338
04:10
So just as a neuron adds up its stimulation
97
238605
2674
04:13
from other neurons to decide whether to fire,
98
241279
2006
04:15
an ant adds up its stimulation from other ants
99
243285
2885
04:18
to decide whether to forage.
100
246170
2023
04:20
And what we're looking for is whether there might be
101
248193
1637
04:21
small differences among colonies
102
249830
1992
04:23
in how many interactions each ant needs
103
251822
3314
04:27
before it's willing to go out and forage,
104
255136
1926
04:29
because a colony like that would forage less.
105
257062
3767
04:32
And this raises an analogous question about brains.
106
260829
3139
04:35
We talk about the brain,
107
263968
1412
04:37
but of course every brain is slightly different,
108
265380
2890
04:40
and maybe there are some individuals
109
268270
1379
04:41
or some conditions
110
269649
1319
04:42
in which the electrical properties of neurons are such
111
270968
3222
04:46
that they require more stimulus to fire,
112
274190
3870
04:50
and that would lead to differences in brain function.
113
278060
3726
04:53
So in order to ask evolutionary questions,
114
281786
2334
04:56
we need to know about reproductive success.
115
284120
2679
04:58
This is a map of the study site
116
286799
2325
05:01
where I have been tracking this population
117
289124
2553
05:03
of harvester ant colonies for 28 years,
118
291677
3293
05:06
which is about as long as a colony lives.
119
294970
2262
05:09
Each symbol is a colony,
120
297232
2099
05:11
and the size of the symbol is
how many offspring it had,
121
299331
3266
05:14
because we were able to use genetic variation
122
302597
1913
05:16
to match up parent and offspring colonies,
123
304510
2711
05:19
that is, to figure out which colonies
124
307221
3297
05:22
were founded by a daughter queen
125
310518
1925
05:24
produced by which parent colony.
126
312443
2068
05:26
And this was amazing for me, after all these years,
127
314511
2002
05:28
to find out, for example, that colony 154,
128
316513
3128
05:31
whom I've known well for many years,
129
319641
2176
05:33
is a great-grandmother.
130
321817
1819
05:35
Here's her daughter colony,
131
323636
1740
05:37
here's her granddaughter colony,
132
325376
2684
05:40
and these are her great-granddaughter colonies.
133
328060
2392
05:42
And by doing this, I was able to learn
134
330452
2025
05:44
that offspring colonies resemble parent colonies
135
332477
3283
05:47
in their decisions about which days are so hot
136
335760
2217
05:49
that they don't forage,
137
337977
1778
05:51
and the offspring of parent colonies
138
339755
1464
05:53
live so far from each other that the ants never meet,
139
341219
2906
05:56
so the ants of the offspring colony
140
344125
2289
05:58
can't be learning this from the parent colony.
141
346414
2244
06:00
And so our next step is to look
142
348658
1381
06:02
for the genetic variation
underlying this resemblance.
143
350039
5276
06:07
So then I was able to ask, okay, who's doing better?
144
355315
4125
06:11
Over the time of the study,
145
359440
1460
06:12
and especially in the past 10 years,
146
360900
1465
06:14
there's been a very severe and deepening drought
147
362365
3308
06:17
in the Southwestern U.S.,
148
365673
2127
06:19
and it turns out that the
colonies that conserve water,
149
367800
3053
06:22
that stay in when it's really hot outside,
150
370853
4429
06:27
and thus sacrifice getting as much food as possible,
151
375282
2632
06:29
are the ones more likely to have offspring colonies.
152
377914
2945
06:32
So all this time, I thought that colony 154
153
380859
2269
06:35
was a loser, because on really dry days,
154
383128
2670
06:37
there'd be just this trickle of foraging,
155
385798
1868
06:39
while the other colonies were out
156
387666
1591
06:41
foraging, getting lots of food,
157
389257
2117
06:43
but in fact, colony 154 is a huge success.
158
391374
3020
06:46
She's a matriarch.
159
394394
1332
06:47
She's one of the rare great-grandmothers on the site.
160
395726
3052
06:50
To my knowledge, this is the first time
161
398778
2785
06:53
that we've been able to track
162
401563
1639
06:55
the ongoing evolution of collective behavior
163
403202
3001
06:58
in a natural population of animals
164
406203
2117
07:00
and find out what's actually working best.
165
408320
4657
07:04
Now, the Internet uses an algorithm
166
412977
2421
07:07
to regulate the flow of data
167
415398
2853
07:10
that's very similar to the one
168
418251
2227
07:12
that the harvester ants are using to regulate
169
420478
2376
07:14
the flow of foragers.
170
422854
1541
07:16
And guess what we call this analogy?
171
424395
3366
07:19
The anternet is coming.
172
427761
1518
07:21
(Applause)
173
429279
1721
07:23
So data doesn't leave the source computer
174
431000
3454
07:26
unless it gets a signal that there's enough bandwidth
175
434454
2861
07:29
for it to travel on.
176
437315
2729
07:32
In the early days of the Internet,
177
440044
1441
07:33
when operating costs were really high
178
441485
2274
07:35
and it was really important not to lose any data,
179
443759
3227
07:38
then the system was set up for interactions
180
446986
2157
07:41
to activate the flow of data.
181
449143
3062
07:44
It's interesting that the ants are using an algorithm
182
452205
2385
07:46
that's so similar to the one that we recently invented,
183
454590
3896
07:50
but this is only one of a handful of ant algorithms
184
458486
2953
07:53
that we know about,
185
461439
1419
07:54
and ants have had 130 million years
186
462858
3193
07:58
to evolve a lot of good ones,
187
466051
2026
08:00
and I think it's very likely
188
468077
1506
08:01
that some of the other 12,000 species
189
469583
2557
08:04
are going to have interesting algorithms
190
472140
2697
08:06
for data networks
191
474837
1024
08:07
that we haven't even thought of yet.
192
475861
2697
08:10
So what happens when operating costs are low?
193
478558
3085
08:13
Operating costs are low in the tropics,
194
481643
1787
08:15
because it's very humid, and it's easy for the ants
195
483430
2096
08:17
to be outside walking around.
196
485526
2824
08:20
But the ants are so abundant
197
488350
1653
08:22
and diverse in the tropics
198
490003
1818
08:23
that there's a lot of competition.
199
491821
2598
08:26
Whatever resource one species is using,
200
494419
1952
08:28
another species is likely to be using that
201
496371
3172
08:31
at the same time.
202
499543
2379
08:33
So in this environment, interactions are used
203
501922
2608
08:36
in the opposite way.
204
504530
1945
08:38
The system keeps going
205
506475
1395
08:39
unless something negative happens,
206
507870
1554
08:41
and one species that I study makes circuits
207
509424
2167
08:43
in the trees of foraging ants
208
511591
2159
08:45
going from the nest to a food source and back,
209
513750
2981
08:48
just round and round,
210
516731
1329
08:50
unless something negative happens,
211
518060
1442
08:51
like an interaction
212
519502
1609
08:53
with ants of another species.
213
521111
2739
08:55
So here's an example of ant security.
214
523850
2937
08:58
In the middle, there's an ant
215
526787
1858
09:00
plugging the nest entrance with its head
216
528645
2463
09:03
in response to interactions with another species.
217
531108
2993
09:06
Those are the little ones running around
218
534101
1659
09:07
with their abdomens up in the air.
219
535760
2751
09:10
But as soon as the threat is passed,
220
538511
2046
09:12
the entrance is open again,
221
540557
2755
09:15
and maybe there are situations
222
543312
1790
09:17
in computer security
223
545102
1089
09:18
where operating costs are low enough
224
546191
2206
09:20
that we could just block access temporarily
225
548397
3375
09:23
in response to an immediate threat,
226
551772
2193
09:25
and then open it again,
227
553965
2026
09:27
instead of trying to build
228
555991
1269
09:29
a permanent firewall or fortress.
229
557260
3980
09:33
So another environmental challenge
230
561240
1940
09:35
that all systems have to deal with
231
563180
1695
09:36
is resources, finding and collecting them.
232
564875
5452
09:42
And to do this, ants solve the problem
233
570327
1673
09:44
of collective search,
234
572000
1258
09:45
and this is a problem that's of great interest
235
573258
1576
09:46
right now in robotics,
236
574834
1484
09:48
because we've understood that,
237
576318
1732
09:50
rather than sending a single,
238
578050
1614
09:51
sophisticated, expensive robot out
239
579664
3495
09:55
to explore another planet
240
583159
1417
09:56
or to search a burning building,
241
584576
2567
09:59
that instead, it may be more effective
242
587143
2600
10:01
to get a group of cheaper robots
243
589743
4377
10:06
exchanging only minimal information,
244
594120
2545
10:08
and that's the way that ants do it.
245
596665
2799
10:11
So the invasive Argentine ant
246
599464
1765
10:13
makes expandable search networks.
247
601229
2330
10:15
They're good at dealing with the main problem
248
603559
2273
10:17
of collective search,
249
605832
1331
10:19
which is the trade-off between
250
607163
2566
10:21
searching very thoroughly
251
609729
1336
10:23
and covering a lot of ground.
252
611065
1997
10:25
And what they do is,
253
613062
895
10:25
when there are many ants in a small space,
254
613957
2387
10:28
then each one can search very thoroughly
255
616344
2213
10:30
because there will be another ant nearby
256
618557
1624
10:32
searching over there,
257
620181
1357
10:33
but when there are a few ants
258
621538
1647
10:35
in a large space,
259
623185
2055
10:37
then they need to stretch out their paths
260
625240
2414
10:39
to cover more ground.
261
627654
1803
10:41
I think they use interactions to assess density,
262
629457
2954
10:44
so when they're really crowded,
263
632411
1229
10:45
they meet more often,
264
633640
1102
10:46
and they search more thoroughly.
265
634742
2465
10:49
Different ant species must use different algorithms,
266
637207
3400
10:52
because they've evolved to deal with
267
640607
2522
10:55
different resources,
268
643129
1671
10:56
and it could be really useful to know about this,
269
644800
2559
10:59
and so we recently asked ants
270
647359
1642
11:01
to solve the collective search problem
271
649001
2450
11:03
in the extreme environment
272
651451
1368
11:04
of microgravity
273
652819
1558
11:06
in the International Space Station.
274
654377
1976
11:08
When I first saw this picture, I thought,
275
656353
1545
11:09
Oh no, they've mounted the habitat vertically,
276
657898
2857
11:12
but then I realized that, of course, it doesn't matter.
277
660755
2618
11:15
So the idea here is that the ants
278
663373
2637
11:18
are working so hard to hang on
279
666010
1970
11:19
to the wall or the floor or whatever you call it
280
667980
3057
11:23
that they're less likely to interact,
281
671037
3009
11:26
and so the relationship between
282
674046
1244
11:27
how crowded they are and how often they meet
283
675290
2120
11:29
would be messed up.
284
677410
1725
11:31
We're still analyzing the data.
285
679135
1395
11:32
I don't have the results yet.
286
680530
1964
11:34
But it would be interesting to know
287
682494
1694
11:36
how other species solve this problem
288
684188
2459
11:38
in different environments on Earth,
289
686647
2564
11:41
and so we're setting up a program
290
689211
1266
11:42
to encourage kids around the world
291
690477
2160
11:44
to try this experiment with different species.
292
692637
2536
11:47
It's very simple.
293
695173
1767
11:48
It can be done with cheap materials.
294
696940
2090
11:51
And that way, we could make a global map
295
699030
2833
11:53
of ant collective search algorithms.
296
701863
3347
11:57
And I think it's pretty likely that the invasive species,
297
705210
2483
11:59
the ones that come into our buildings,
298
707693
2149
12:01
are going to be really good at this,
299
709842
1742
12:03
because they're in your kitchen
300
711584
1848
12:05
because they're really good
at finding food and water.
301
713432
3907
12:09
So the most familiar resource for ants
302
717339
3265
12:12
is a picnic,
303
720604
1315
12:13
and this is a clustered resource.
304
721919
2145
12:16
When there's one piece of fruit,
305
724064
999
12:17
there's likely to be another piece of fruit nearby,
306
725063
2515
12:19
and the ants that specialize on clustered resources
307
727578
3432
12:23
use interactions for recruitment.
308
731010
1942
12:24
So when one ant meets another,
309
732952
1277
12:26
or when it meets a chemical deposited
310
734229
1625
12:27
on the ground by another,
311
735854
1736
12:29
then it changes direction to follow
312
737590
1833
12:31
in the direction of the interaction,
313
739423
1573
12:32
and that's how you get the trail of ants
314
740996
1977
12:34
sharing your picnic.
315
742973
1413
12:36
Now this is a place where I think we might be able
316
744386
1695
12:38
to learn something from ants about cancer.
317
746081
3741
12:41
I mean, first, it's obvious that we could do a lot
318
749822
1981
12:43
to prevent cancer
319
751803
1610
12:45
by not allowing people to spread around
320
753413
2577
12:47
or sell the toxins that promote
321
755990
1944
12:49
the evolution of cancer in our bodies,
322
757934
2780
12:52
but I don't think the ants can help us much with this
323
760714
2346
12:55
because ants never poison their own colonies.
324
763060
3358
12:58
But we might be able to learn something from ants
325
766418
1332
12:59
about treating cancer.
326
767750
1763
13:01
There are many different kinds of cancer.
327
769513
2225
13:03
Each one originates in a particular part of the body,
328
771738
2978
13:06
and then some kinds of cancer will spread
329
774716
2966
13:09
or metastasize to particular other tissues
330
777682
2830
13:12
where they must be getting
resources that they need.
331
780512
2880
13:15
So if you think from the perspective
332
783392
1808
13:17
of early metastatic cancer cells
333
785200
1950
13:19
as they're out searching around
334
787150
1623
13:20
for the resources that they need,
335
788773
2317
13:23
if those resources are clustered,
336
791090
1983
13:25
they're likely to use interactions for recruitment,
337
793073
3013
13:28
and if we can figure out how
cancer cells are recruiting,
338
796086
3093
13:31
then maybe we could set traps
339
799179
2347
13:33
to catch them before they become established.
340
801526
4049
13:37
So ants are using interactions in different ways
341
805575
3235
13:40
in a huge variety of environments,
342
808810
2602
13:43
and we could learn from this
343
811412
1821
13:45
about other systems that operate
344
813233
1777
13:47
without central control.
345
815010
2337
13:49
Using only simple interactions,
346
817347
1979
13:51
ant colonies have been performing
347
819326
1795
13:53
amazing feats for more than 130 million years.
348
821121
3633
13:56
We have a lot to learn from them.
349
824754
2140
13:58
Thank you.
350
826894
2738
14:01
(Applause)
351
829632
2733

▲Back to top

ABOUT THE SPEAKER
Deborah Gordon - Ecologist
By studying how ant colonies work without any one leader, Deborah Gordon has identified striking similarities in how ant colonies, brains, cells and computer networks regulate themselves.

Why you should listen

Ecologist Deborah M. Gordon has learned that ant colonies can work without central control by using simple interactions like how often the insects touch antennae. Contrary to the notion that colonies are organized by efficient ants, she has instead discovered that evolution has produced “noisy” systems that tolerate accident and respond flexibly to the environment. When conditions are tough, natural selection favors colonies that conserve resources.

Her studies of ant colonies have led her and her Stanford colleagues to the discovery of the “Anternet,” which regulates foraging in ants in the same way the internet regulates data traffic. But as she said to Wired in 2013, "Insect behavior mimicking human networks ... is actually not what’s most interesting about ant networks. What’s far more interesting are the parallels in the other direction: What have the ants worked out that we humans haven’t thought of yet?" Her latest exploration: How do ants behave in space?

More profile about the speaker
Deborah Gordon | Speaker | TED.com