Self-Assembling Robots and the Potential of Artificial Evolution | Emma Hart | TED

77,092 views

2022-04-01 ・ TED


New videos

Self-Assembling Robots and the Potential of Artificial Evolution | Emma Hart | TED

77,092 views ・ 2022-04-01

TED


μ•„λž˜ μ˜λ¬Έμžλ§‰μ„ λ”λΈ”ν΄λ¦­ν•˜μ‹œλ©΄ μ˜μƒμ΄ μž¬μƒλ©λ‹ˆλ‹€.

λ²ˆμ—­: Tae Lee κ²€ν† : Jihyeon J. Kim
00:04
Imagine a scientist
0
4876
1251
ν•œ κ³Όν•™μžλ₯Ό 상상해 λ³΄μ„Έμš”.
00:06
who wants to send a robot to explore in a faraway place,
1
6127
3587
탐사λ₯Ό μœ„ν•΄ μ•„μ£Ό λ¨Ό 곳으둜 λ‘œλ΄‡μ„ 보내고 μ‹Άμ–΄ ν•©λ‹ˆλ‹€.
00:09
a place whose geography might be completely unknown
2
9756
2628
κ·Έ 곳의 μ§€ν˜•μ„ μ „ν˜€ μ•Œμ§€λ„ λͺ»ν•˜κ³ 
00:12
and perhaps inhospitable.
3
12425
2044
λ¨Έλ¬΄λŠ” 것 μžμ²΄λ„ μ–΄λ ΅κ² μ£ .
00:15
Now imagine that instead of first designing that robot
4
15387
3670
이제 λ‘œλ΄‡μ΄ 잘 견디기λ₯Ό κΈ°λŒ€ν•˜λ©°
00:19
and sending it off in the hope that it might be suitable,
5
19099
3295
κ·Έ λ‘œλ΄‡μ„ μ„€κ³„ν•˜κ³  λ³΄λ‚΄λŠ” λŒ€μ‹ μ—
00:22
instead, she sends a robot-producing technology
6
22435
3796
λ‘œλ΄‡μ„ μƒμ‚°ν•˜λŠ” κΈ°μˆ μ„ λ³΄λ‚΄μ„œ
00:26
that figures out what kind of robot is needed once it arrives,
7
26273
3503
일단 κ·Έ 곳에 λ„μ°©ν•œ λ’€ μ–΄λ–€ μ’…λ₯˜μ˜ λ‘œλ΄‡μ΄ ν•„μš”ν•œμ§€ λΆ„μ„ν•˜κ³ 
00:29
builds it and then enables it to continue to evolve
8
29818
3295
그것을 λ§Œλ“€κ³  그것이 계속 진화할 수 있게 ν•΄μ„œ
00:33
to adapt to its new surroundings.
9
33154
2002
μƒˆλ‘œμš΄ ν™˜κ²½μ— μ μ‘ν•˜κ²Œ ν•˜λŠ” κ²λ‹ˆλ‹€.
00:36
It’s exactly what my collaborators and I are working on:
10
36241
3170
저와 제 ν˜‘λ ₯μžλ“€μ΄ λ°”λ‘œ 이것을 μ—°κ΅¬ν•˜κ³  μžˆμŠ΅λ‹ˆλ‹€.
00:39
a radical new technology which enables robots to be created,
11
39452
4296
λ‘œλ΄‡μ΄ λ§Œλ“€μ–΄μ§€κ³ , μž¬μƒμ‚°ν•˜κ³ , κΈ΄ μ‹œκ°„μ— 걸쳐 μ§„ν™”ν•˜λŠ” 것을
00:43
reproduce and evolve over long periods of time,
12
43748
3337
κ°€λŠ₯ν•˜κ²Œ ν•˜λŠ” 근본적으둜 μƒˆλ‘œμš΄ κΈ°μˆ μž…λ‹ˆλ‹€.
00:47
a technology where robot design and fabrication becomes a task
13
47961
3920
이 κΈ°μˆ μ€ λ‘œλ΄‡ 섀계와 μ‘°ν•©μ˜ μž„λ¬΄λ₯Ό
00:51
for machines rather than humans.
14
51923
2461
μ‚¬λžŒμ΄ μ•„λ‹Œ κΈ°κ³„μ—κ²Œ λΆ€μ—¬ν•©λ‹ˆλ‹€.
00:55
Robots are already all around us, in factories, in hospitals, in our home.
15
55885
4588
λ‘œλ΄‡μ€ 이미 곡μž₯에, 병원에, 가정에, 우리 μ£Όλ³€ μ–΄λŠ κ³³μ—λ‚˜ μžˆμŠ΅λ‹ˆλ‹€.
01:01
But from an engineer's perspective,
16
61391
1877
기술자의 κ΄€μ μ—μ„œ λ³Ό λ•Œ
01:03
designing a shelf-stacking robot or a Roomba to clean our home
17
63310
3503
μ„ λ°˜μ„ μŒ“λŠ” λ‘œλ΄‡μ΄λ‚˜ μ§‘μ•ˆμ„ μ²­μ†Œν•˜λŠ” λ‘œλ΄‡μ„ μ„€κ³„ν•˜λŠ” 것은
01:06
is relatively straightforward.
18
66813
2211
μƒλŒ€μ μœΌλ‘œ λ‹¨μˆœν•©λ‹ˆλ‹€.
01:09
We know exactly what they need to do,
19
69065
2253
μš°λ¦¬λŠ” λ‘œλ΄‡μ΄ ν•  일을 μ •ν™•νžˆ μ•Œκ³ 
01:11
and we can imagine the kind of situations they might find themselves in.
20
71318
3461
λ‘œλ΄‡μ΄ μ²˜ν•  μ˜¨κ°– 상황을 μ˜ˆμƒν•  수 μžˆμŠ΅λ‹ˆλ‹€.
01:14
So we design with this in mind.
21
74821
1794
μš°λ¦¬λŠ” μ΄λŸ¬ν•œ 점을 κ³ λ €ν•΄ μ„€κ³„ν•©λ‹ˆλ‹€.
01:18
But what if we want to send that robot to operate
22
78033
2794
ν•˜μ§€λ§Œ μš°λ¦¬κ°€ 잘 μ•Œμ§€λ„ λͺ»ν•˜λŠ” 곳에
01:20
in a place that we have little or even no knowledge about?
23
80827
3003
λ‘œλ΄‡μ„ 보내 μž‘λ™ν•˜κ²Œ ν•˜λ €λ©΄μš”?
01:24
For example, cleaning up legacy waste inside a nuclear reactor
24
84456
3295
예λ₯Ό λ“€μžλ©΄, μ‚¬λžŒμ„ λ³΄λ‚΄κΈ°μ—λŠ” μœ„ν—˜ν•œ
01:27
where it's unsafe to send humans,
25
87751
3003
μ›μžλ‘œ μ•ˆμ˜ 방사λŠ₯ 페기물 μ²­μ†Œ,
01:30
mining for minerals deep in a trench at the bottom of the ocean,
26
90754
3879
κΉŠμ€ λ°”λ‹€ λ°‘λ°”λ‹₯ ν•΄κ΅¬μ—μ„œμ˜ 채꡴,
01:34
or exploring a faraway asteroid.
27
94674
2169
λ˜λŠ” λ¨Ό 곳의 μ†Œν–‰μ„± νƒν—˜μ΄μ£ .
01:38
How frustrating would it be if the human-designed robot,
28
98345
3670
이런 경우 μ–Όλ§ˆλ‚˜ μ’Œμ ˆμŠ€λŸ¬μšΈκΉŒμš”, λ§Œμ•½ 인간이 μ„€κ³„ν•œ λ‘œλ΄‡μ΄
01:42
that had taken years to get to the asteroid
29
102057
2794
수 년이 κ±Έλ € μ†Œν–‰μ„±μ— λ„μ°©ν•œ 뒀에
01:44
suddenly found it needed to drill a hole
30
104851
2336
ν‘œλ³Έμ˜ μ±„μ§‘μ΄λ‚˜ 절벽 λ“±λ°˜μ„ μœ„ν•΄
01:47
to collect a sample or clamber up a cliff
31
107187
3044
ꡬ멍을 λš«μ„ λ“œλ¦΄μ˜ ν•„μš”λ₯Ό μ•Œκ²Œ λ˜μ§€λ§Œ
01:50
but it didn't have the right tools
32
110273
1627
μ λ‹Ήν•œ 도ꡬ가 μ—†κ±°λ‚˜
01:51
or the right means of locomotion to do so?
33
111941
2169
μ ν•©ν•œ 이동 μˆ˜λ‹¨μ΄ μ—†λ‹€λ©΄μš”.
01:55
If instead we had a technology
34
115236
2586
μ•„λ‹ˆλ©΄, μš°λ¦¬μ—κ²Œ 이런 기술이 μžˆλ‹€λ©΄μš”?
01:57
that enabled the robots to be designed and optimized in situ,
35
117864
4254
λ‘œλ΄‡μ΄ 슀슀둜 섀계λ₯Ό ν•˜κ³  상황에 맞게 κ°œμ„ ν•˜κ³ 
02:02
in the environment in which they need to live and work,
36
122160
3253
λ¨Έλ¬Όκ³  μΌν•˜λŠ” ν™˜κ²½μ— μ μ‘ν•œλ‹€λ©΄
02:05
then we could potentially save years of wasted effort
37
125455
2878
μˆ˜λ…„κ°„μ˜ μˆ˜κ³ κ°€ ν—ˆλΉ„λ˜λŠ” 것을 방지할 수 있고
02:08
and produce robots that are uniquely adapted
38
128375
2252
μŠ€μŠ€λ‘œκ°€ μ²˜ν•œ ν™˜κ²½μ—
02:10
to the environments that they find themselves in.
39
130669
2335
λ…νŠΉν•˜κ²Œ μ μ‘ν•˜λŠ” λ‘œλ΄‡μ„ λ§Œλ“œλŠ” κ²λ‹ˆλ‹€.
02:15
So to realize this technology, we've been turning to nature for help.
40
135173
4046
μ΄λŸ¬ν•œ κΈ°μˆ μ„ μ•ŒκΈ° μœ„ν•΄ μš°λ¦¬λŠ” μžμ—°μ—κ²Œμ„œ 도움을 κ΅¬ν–ˆμŠ΅λ‹ˆλ‹€.
02:20
All around us,
41
140345
1209
우리 μ£Όλ³€μ—μ„œ
02:21
we see examples of biological species
42
141596
3003
μ—¬λŸ¬ 생물적 쒅듀을 보게 λ˜λŠ”λ°
02:24
that have evolved smart adaptations
43
144599
2211
ν˜„λͺ…ν•˜κ²Œ μ μ‘ν•˜λ„λ‘ μ§„ν™”ν•˜μ—¬
02:26
that enable them to thrive in a given environment.
44
146810
2752
주어진 ν™˜κ²½μ—μ„œ λ²ˆμ°½ν•  수 μžˆμ—ˆμŠ΅λ‹ˆλ‹€.
02:31
For example, in the Cuban rainforest,
45
151064
2669
예λ₯Ό λ“€λ©΄, μΏ λ°”μ˜ λ°€λ¦Όμ—μ„œ
02:33
we find vines that have evolved leaves
46
153775
2669
λ°œκ²¬ν•œ 덩꡴은 잎이 μ§„ν™”ν•΄μ„œ
02:36
that are shaped like human-designed satellite dishes.
47
156486
3086
μΈκ³΅μœ„μ„± μ ‘μ‹œ 같은 λͺ¨μ–‘μ΄μ—ˆμŠ΅λ‹ˆλ‹€.
02:39
These leaves direct bats to their flowers
48
159614
2544
이 잎이 λ°•μ₯κ°€ λ³΄λ‚΄λŠ” μ‹ ν˜Έλ₯Ό μ¦ν­μ‹œμΌœ
02:42
by amplifying the signals that the bats send out,
49
162158
2795
λ°•μ₯λ₯Ό μžμ‹ μ˜ κ½ƒμœΌλ‘œ μΈλ„ν•˜λŠ”λ°
02:44
therefore, improving pollination.
50
164953
1960
κ½ƒμ˜ μˆ˜μ •μ΄ ν–₯μƒλ©λ‹ˆλ‹€.
02:48
What if we could create an artificial version of evolution
51
168665
4046
μš°λ¦¬κ°€ 인곡적인 진화λ₯Ό λ§Œλ“€μ–΄ λ‚Ό 수 μžˆμ–΄μ„œ
02:52
that would enable robots to evolve in a similar manner
52
172752
3253
생λͺ…μ²΄μ˜ 진화와 λΉ„μŠ·ν•œ λ°©μ‹μœΌλ‘œ
02:56
as biological organisms?
53
176047
2211
λ‘œλ΄‡μ„ μ§„ν™”ν•˜κ²Œ ν•œλ‹€λ©΄ μ–΄λ–¨κΉŒμš”?
03:00
I'm not talking about biomimicry,
54
180051
2169
μ €λŠ” μžμ—°μ—μ„œ κ΄€μ°°λ˜λŠ” 것을
03:02
a technology which simply copies what's observed in nature.
55
182262
3545
λ‹¨μˆœνžˆ λ³΅μ‚¬ν•˜λŠ” 생체λͺ¨λ°©μ„ λ§ν•˜λŠ” 게 μ•„λ‹™λ‹ˆλ‹€.
03:06
What we're hoping to harness is the creativity of evolution,
56
186808
4212
μ €ν¬λŠ” μ§„ν™”μ˜ 창쑰성을 배우고 싢은데,
03:11
to discover designs that are not observed here on Earth,
57
191020
3587
인간 κ³΅ν•™μžκ°€ μ—¬νƒœ 생각지 λͺ»ν–ˆκ±°λ‚˜ κΏˆλ„ 꾸지 λͺ»ν•œ,
03:14
the human engineer might not have thought of
58
194649
2419
이 μ§€κ΅¬μ—μ„œ κ΄€μ°°λ˜μ§€ μ•Šμ€
03:17
or even be capable of conceiving.
59
197068
2211
λ””μžμΈμ„ λ°œκ²¬ν•˜λ €λŠ” κ²λ‹ˆλ‹€.
03:20
In theory,
60
200947
1168
μ΄λ‘ μƒμœΌλ‘œλŠ”
03:22
this evolutionary design technology could operate completely autonomously
61
202157
3920
이런 진화적 섀계 기술이 μ•„μ£Ό λ¨Ό κ³³μ—μ„œ
03:26
in a faraway place.
62
206119
2127
μ™„μ „ μžλ™μœΌλ‘œ μž‘λ™ν•  수 μžˆμŠ΅λ‹ˆλ‹€.
03:28
But equally it could be guided by humans.
63
208288
2377
ν•˜μ§€λ§Œ 그와 λ™μ‹œμ— μ‚¬λžŒμ— μ˜ν•΄μ„œ μœ λ„λ  μˆ˜λ„ μžˆμŠ΅λ‹ˆλ‹€.
03:31
Just as we breed plants for qualities such as drought resistance or taste,
64
211291
4588
가뭄에 κ²¬λ””λŠ” μ„±μ§ˆμ΄λ‚˜ νŠΉμ •ν•œ 맛을 μ–»κΈ° μœ„ν•΄ 식물을 κ΅λ°°ν•˜λŠ” κ²ƒμ²˜λŸΌ
03:35
the human robot breeder could guide artificial evolution to producing robots
65
215879
5171
인간이 인곡 진화λ₯Ό μœ λ„ν•΄ νŠΉμ • μ„±λŠ₯을 가진 λ‘œλ΄‡μ„ λ§Œλ“€λ„λ‘
03:41
with specific qualities.
66
221050
1711
ν•  수 μžˆμ„ κ²λ‹ˆλ‹€.
03:42
For example,
67
222802
1126
예λ₯Ό λ“€λ©΄,
03:43
the ability to squeeze through a narrow gap
68
223970
2377
쒁은 사이λ₯Ό 비집고 ν†΅κ³Όν•˜κ±°λ‚˜
03:46
or perhaps operate at low energy.
69
226389
2711
μž‘μ€ μ—λ„ˆμ§€λ‘œ μž‘λ™ν•˜λŠ” λŠ₯λ ₯μž…λ‹ˆλ‹€.
03:51
This idea of artificial evolution imitating biological evolution
70
231686
3962
컴퓨터 ν”„λ‘œκ·Έλž¨μ„ μ΄μš©ν•΄ κΈ΄ μ‹œκ°„μ— 걸친
03:55
using a computer program
71
235648
1669
λ¬Έμ œμ— λŒ€ν•œ 점점 λ‚˜μ€ 해결책을 μ°ΎλŠ”
03:57
to breed better and better solutions to problems over time
72
237317
3587
생물적 진화λ₯Ό λͺ¨λ°©ν•˜λŠ” 인곡 μ§„ν™”μ˜ μ•„μ΄λ””μ–΄λŠ”
04:00
isn't actually new.
73
240904
1543
사싀 μƒˆλ‘­μ§€ μ•ŠμŠ΅λ‹ˆλ‹€
04:03
In fact, artificial evolution,
74
243323
2544
μ‹œμ‹€, 인곡 μ§„ν™”λŠ”
04:05
algorithms operating inside a computer,
75
245909
2711
컴퓨터 λ‚΄λΆ€μ—μ„œ μž‘λ™ν•˜λŠ” 연산방식인데
04:08
have been used to design everything from tables to turbine blades.
76
248620
4254
νƒμžμ—μ„œ 원동기 λ‚ κ°œκΉŒμ§€ λͺ¨λ“ κ²ƒμ„ μ„€κ³„ν•˜λŠ” 데 μ‚¬μš©ν•΄ μ™”μŠ΅λ‹ˆλ‹€.
04:13
Back in 2006,
77
253291
1668
μ§€λ‚œ 2006λ…„
04:15
NASA even sent a satellite into space with a communication antenna
78
255001
4004
λ‚˜μ‚¬μ—μ„œ 인곡 μ§„ν™”λ‘œ μ„€κ³„ν•œ 톡신 μ•ˆν…Œλ‚˜λ₯Ό μž₯μ°©ν•œ
04:19
that had been designed by artificial evolution.
79
259047
2711
μΈκ³΅μœ„μ„±μ„ 우주둜 보내기도 ν–ˆμŠ΅λ‹ˆλ‹€.
04:23
But evolving robots is actually much harder
80
263843
2544
ν•˜μ§€λ§Œ νƒμžμ²˜λŸΌ μ§„ν™”ν•˜λŠ” μˆ˜λ™μ  κ°œμ²΄λ³΄λ‹€λŠ”
04:26
than evolving passive objects such as tables,
81
266429
2878
μ§„ν™”ν•˜λŠ” λ‘œλ΄‡μ΄ 사싀상 훨씬 더 μ–΄λ ΅μŠ΅λ‹ˆλ‹€.
04:29
because robots need brains as well as bodies
82
269307
3420
λ‘œλ΄‡μ€ λͺΈμ²΄ 뿐만 μ•„λ‹ˆλΌ λ‘λ‡Œλ„ ν•„μš”ν•˜κΈ° λ•Œλ¬ΈμΈλ°,
04:32
in order to make sense of the information in the world around them
83
272769
5047
μžμ‹ μ„ λ‘˜λŸ¬μ‹Ό μ„Έκ³„μ˜ 정보λ₯Ό 이해해야 ν•˜κ³ 
04:37
and translate that into appropriate behaviors.
84
277816
2836
κ·Έ 정보λ₯Ό μ μ •ν•œ ν–‰λ™μœΌλ‘œ μ „ν™˜ν•΄μ•Ό ν•˜κΈ° λ•Œλ¬Έμž…λ‹ˆλ‹€.
04:41
So how do we do it?
85
281903
1543
그럼 μ–΄λ–»κ²Œ ν•΄μ•Ό ν• κΉŒμš”?
04:44
Surprisingly, evolution only needs three ingredients:
86
284489
4046
λ†€λžκ²Œλ„, μ§„ν™”λŠ” 단 μ„Έ 개의 μš”μ†Œλ§Œ ν•„μš”ν•©λ‹ˆλ‹€.
04:49
a population of individuals which exhibit some physical variations;
87
289369
5589
ν•˜λ‚˜, μ•½κ°„μ˜ 물리적 차이λ₯Ό λ³΄μ΄λŠ” κ°œμ²΄λ“€μ˜ 집합,
04:54
a method of reproduction
88
294999
1877
λ‘˜, λΆ€λͺ¨μ˜ νŠΉμ„± 일뢀와
04:56
in which offspring inherit some traits from their parents
89
296876
3045
가끔씩 변이λ₯Ό 톡해 μ–»λŠ” νŠΉμ„±μ΄
04:59
and occasionally acquire new ones via mutation;
90
299963
3545
ν›„μ„ΈλŒ€λ‘œ μœ μ „λ˜λŠ” μž¬μƒμ‚°μ˜ 방법,
05:03
and finally, a means of natural selection.
91
303550
2460
λ§ˆμ§€λ§‰ μ…‹, μžμ—° μ„ νƒμ˜ μˆ˜λ‹¨μž…λ‹ˆλ‹€.
05:07
So we can replicate these three ingredients to evolve robots
92
307178
3295
μ΄λŸ¬ν•œ μ„Έ 개의 μš”μ†Œλ₯Ό λͺ¨ν˜•μœΌλ‘œ ν•˜λ“œμ›¨μ–΄μ™€ μ†Œν”„νŠΈμ›¨μ–΄λ₯Ό μ‚¬μš©ν•΄
05:10
using a mixture of hardware and software.
93
310473
2836
λ‘œλ΄‡μ΄ μ§„ν™”ν•˜κ²Œ ν•  수 μžˆμŠ΅λ‹ˆλ‹€.
05:14
The first task is to design a digital version of DNA.
94
314769
4296
첫 번째 μž„λ¬΄λŠ” μœ μ „μžλ₯Ό λ””μ§€ν„Έλ‘œ μ„€κ³„ν•˜λŠ” κ²ƒμž…λ‹ˆλ‹€.
05:19
That is a digital blueprint that describes the robot's brain, its body,
95
319107
5464
이것은 λ‘œλ΄‡μ˜ λ‘λ‡Œ, λͺΈμ²΄, 감각 κΈ°κ΄€, μš΄λ™ μˆ˜λ‹¨μ„ λ¬˜μ‚¬ν•˜λŠ”
05:24
its sensory mechanisms and its means of locomotion.
96
324612
3212
디지털 청사진 μž…λ‹ˆλ‹€.
05:29
Using a randomly generated set of these blueprints,
97
329492
2795
이 진화 과정을 μ‹œμž‘μ‹œν‚€κΈ° μœ„ν•΄
05:32
we can create an initial population of 10 or more robots
98
332328
3420
λ¬΄μž‘μœ„λ‘œ μƒμ„±ν•œ μ΄λŸ¬ν•œ μ²­μ‚¬μ§„μ˜ μ‘°ν•©μœΌλ‘œ
05:35
to kick-start this evolutionary process.
99
335790
3212
10개 μ΄μƒμ˜ 첫 λ‘œλ΄‡ 집단을 μ œμž‘ν•  수 μžˆμŠ΅λ‹ˆλ‹€.
05:40
We've designed a technology that can take the digital blueprint
100
340587
3586
μ €ν¬λŠ” μΈκ°„μ˜ 도움 없이 디지털 청사진을 가지고
05:44
and turn it into a physical robot without any need for human assistance.
101
344215
4088
물리적인 λ‘œλ΄‡μœΌλ‘œ λ§Œλ“œλŠ” κΈ°μˆ μ„ μ„€κ³„ν–ˆμŠ΅λ‹ˆλ‹€.
05:49
For example, it uses a 3D printer to print the skeleton of the robot
102
349220
4463
예둜, 이것은 3차원 ν”„λ¦°ν„°λ‘œ λ‘œλ΄‡μ˜ 골격을 λ§Œλ“€κ³ 
05:53
and then an automated assembly arm like you might find in a factory
103
353725
3962
μ—¬λŸ¬λΆ„μ΄ 곡μž₯μ—μ„œ λ³Ό 수 μžˆλŠ” μžλ™ 쑰립 νŒ”λ‘œ
05:57
to add any electronics and moving parts,
104
357687
2878
λ‘λ‡Œ 역할을 ν•  μ†Œν˜• 컴퓨터λ₯Ό 포함해
06:00
including a small computer that acts as a brain.
105
360607
2794
μ „μž λΆ€ν’ˆκ³Ό λ™μž‘ λΆ€ν’ˆμ„ μΆ”κ°€ν•©λ‹ˆλ‹€.
06:04
And to enable this brain to adapt to the new body of the robot,
106
364611
3795
이 λ‘λ‡Œλ₯Ό λ‘œλ΄‡μ˜ μƒˆ λͺΈμ²΄μ— μ μ‘μ‹œν‚€κΈ° μœ„ν•΄
06:08
we send every robot produced to an equivalent of a kindergarten,
107
368448
5255
저희가 μƒμ‚°ν•œ λͺ¨λ“  λ‘œλ΄‡μ„ μœ μΉ˜μ›μ— κ²¬μ€„λ§Œν•œ 곳으둜 보내어
06:13
a place where the newborn robot can refine its motor skills
108
373745
3753
μƒˆλ‘œ λ§Œλ“€μ–΄μ§„ λ‘œλ΄‡λ“€μ΄ 어린이듀이 λ°°μš°λŠ” 것 처럼
06:17
almost like a small child would.
109
377540
1877
μš΄λ™ 감각을 μ„Έλ°€ν•˜κ²Œ μ‘°μ •ν•©λ‹ˆλ‹€.
06:21
To mimic natural selection,
110
381961
1835
μžμ—° 선택을 λͺ¨λ°©ν•˜λŠ” λ°©μ‹μœΌλ‘œ
06:23
we score these robots on the ability to conduct a task.
111
383796
3254
μ €ν¬λŠ” λ‘œλ΄‡λ“€μ˜ μž„λ¬΄ μˆ˜ν–‰ λŠ₯λ ₯에 점수λ₯Ό λΆ€μ—¬ν–ˆμŠ΅λ‹ˆλ‹€.
06:27
And then we use these scores
112
387926
1710
κ·Έλ¦¬κ³ μ„œ κ·Έ 점수λ₯Ό μ΄μš©ν•΄
06:29
to selectively decide which robots get to reproduce.
113
389677
3879
μ–΄λŠ λ‘œλ΄‡μ΄ μž¬μƒμ‚°λ  지 μ„ λ³„ν–ˆμŠ΅λ‹ˆλ‹€.
06:34
The reproduction mechanism
114
394891
1585
μž¬μƒμ‚° 방식은
06:36
mixes the digital DNA of the chosen parent robots
115
396476
4004
μ„ λ³„λœ λ‘œλ΄‡λ“€μ˜ 디지털 μœ μ „μžλ₯Ό μ„žμ–΄μ„œ
06:40
to create a new blueprint for a child robot
116
400521
3462
μƒˆλ‘œμš΄ ν›„μ„ΈλŒ€ λ‘œλ΄‡μ˜ 청사진을 λ§Œλ“€μ—ˆμŠ΅λ‹ˆλ‹€.
06:44
that inherits some of the characteristics from its parents
117
404025
3378
λΆ€λͺ¨ μ„ΈλŒ€ λ‘œλ΄‡μ˜ νŠΉμ„±λ„ 일뢀 μ „ν•΄μ‘Œμ§€λ§Œ
06:47
but occasionally also exhibits some new ones.
118
407445
2920
가끔 μ•½κ°„μ˜ μƒˆλ‘œμš΄ νŠΉμ„±λ„ λ‚˜νƒ€λ‚¬μŠ΅λ‹ˆλ‹€.
06:51
And by repeating the cycle of selection and reproduction over and over again,
119
411366
5297
선별과 μž¬μƒμ‚°μ˜ μ£ΌκΈ°λ₯Ό μ—¬λŸ¬λ²ˆ λ°˜λ³΅ν•¨μœΌλ‘œμ¨
06:56
we hope that we can breed successive generations of robots
120
416704
3128
μ €ν¬λŠ” μ΄μ–΄μ§ˆ μ„ΈλŒ€μ˜ λ‘œλ΄‡μ΄ 계속 λ§Œλ“€μ–΄μ§€κΈΈ κΈ°λŒ€ν•©λ‹ˆλ‹€.
06:59
where, just like is often observed in biological evolution,
121
419874
3921
생물학적 μ§„ν™”μ—μ„œ μ’…μ’… κ΄€μ°°λ˜λŠ” κ²ƒμ²˜λŸΌ
07:03
each generation gets better than the last,
122
423836
2586
각 μ„ΈλŒ€κ°€ κ·Έ 이전 μ„ΈλŒ€λ³΄λ‹€ 점점 λ‚˜μ•„μ§‘λ‹ˆλ‹€.
07:06
with the robots gradually optimizing their form and their behavior
123
426422
3712
λ‘œλ΄‡λ„ 그듀이 μ²˜ν•œ ν™˜κ²½κ³Ό μž„λ¬΄μ— 맞게
07:10
to the task and the environment that they find themselves in.
124
430176
3337
슀슀둜의 ν˜•νƒœμ™€ 행동을 점차적으둜 μ΅œμ ν™”ν•©λ‹ˆλ‹€.
07:15
Now, although this can all take place
125
435139
2211
λ•Œλ‘œ μˆ˜μ²œλ…„μ— 걸쳐 μ§„ν–‰λ˜λŠ”
07:17
in a time frame that's much faster than biological evolution,
126
437392
3253
생물학적 진화에 λΉ„ν•΄ 인곡 진화가
07:20
which sometimes takes thousands of years,
127
440687
2544
훨씬 λΉ λ₯΄κ²Œ μ§„ν–‰λ˜μ§€λ§Œ
07:23
it's still relatively slow in terms of the time frames we might expect
128
443272
3546
μš°λ¦¬κ°€ ν˜„λŒ€ μ‚¬νšŒμ—μ„œ κΈ°λŒ€ν•˜λŠ” μ‹œκ°„ν‹€λ‘œ 보자면
07:26
in our modern world
129
446818
1251
κ°€κ³΅ν’ˆμ„ μ„€κ³„ν•˜κ³  μƒμ‚°ν•˜κΈ°μ—λŠ”
07:28
to design and produce an artifact.
130
448111
2502
μ—¬μ „νžˆ μƒλŒ€μ μœΌλ‘œ λŠλ¦½λ‹ˆλ‹€.
07:30
It's mainly due to the 3D printing process,
131
450947
2252
κ·Έ μ£Όμš” 원인이 3차원 ν”„λ¦°ν„°μ˜ μ²˜λ¦¬μž…λ‹ˆλ‹€.
07:33
which can take more than four hours per robot,
132
453241
2502
λ‚œμ΄λ„μ™€ λ‘œλ΄‡μ˜ ν˜•νƒœμ— 따라
07:35
depending on the complexity and the shape of the robot.
133
455785
2961
λ‘œλ΄‡ ν•˜λ‚˜μ— 4μ‹œκ°„μ΄μƒ 걸릴 수 μžˆμŠ΅λ‹ˆλ‹€.
07:40
But we can give our artificial evolutionary process a helping hand
134
460456
3504
ν•˜μ§€λ§Œ 인곡 μ§„ν™”μ˜ 진행에 λ„μ›€μ˜ 손길을 쀄 μˆ˜λ„ μžˆμŠ΅λ‹ˆλ‹€.
07:44
to reduce the number of physical robots that we actually need to make.
135
464002
3586
λ§Œλ“€μ–΄μ•Ό ν•  물리적 λ‘œλ΄‡μ˜ 숫자λ₯Ό μ€„μ΄λŠ” κ²λ‹ˆλ‹€.
07:49
We create a digital copy of every robot produced
136
469298
2837
μ»΄ν“¨ν„°μ˜ 가상 μ‹€ν—˜κΈ°μ—
07:52
inside a simulation in a computer,
137
472176
2837
μƒμ‚°λ˜μ—ˆλ˜ λͺ¨λ“  λ‘œλ΄‡μ„ 디지털 λ³΅μ‚¬ν•΄μ„œ
07:55
and we allow this virtual population of robots to evolve.
138
475013
4170
κ°€μƒμœΌλ‘œ λ‘œλ΄‡λ“€μ΄ μ§„ν™”ν•˜κ²Œ ν•˜λŠ” κ²λ‹ˆλ‹€.
07:59
Now it's quite likely that the simulation isn't a very accurate representation
139
479642
5047
가상 μ‹€ν—˜μ΄ ν˜„μ‹€ 세계λ₯Ό μ•„μ£Ό μ •ν™•νžˆ λ°˜μ˜ν•˜λŠ” 것은
08:04
of the real world.
140
484689
1668
μ–΄λ ΅κ² μ£ .
08:06
But it has an advantage that it enables models of robots to be created
141
486983
4004
ν•˜μ§€λ§Œ μ΄κ²ƒμ˜ μž₯점으둜 λ‘œλ΄‡μ˜ λͺ¨ν˜•μ„ λ§Œλ“€κ³  μ‹œν—˜ν•˜λŠ”λ°
08:11
and tested in seconds rather than hours.
142
491029
2377
λͺ‡μ‹œκ°„이 μ•„λ‹Œ 단 λͺ‡μ΄ˆκ°€ κ±Έλ¦½λ‹ˆλ‹€.
08:14
So using the simulator technology,
143
494032
2460
κ·ΈλŸ¬λ―€λ‘œ 가상 μ‹€ν—˜ κΈ°μˆ μ„ μ΄μš©ν•˜λ©΄
08:16
we can quickly explore the potential of a wide range of robot types
144
496534
3795
λ‹€λ₯Έ λͺ¨μ–‘κ³Ό 크기, λ‹€λ₯Έ 감각 κΈ°λŠ₯ 섀정을 가진
08:20
of different shapes and sizes, of different sensory configurations,
145
500371
3921
폭넓은 λ²”μœ„μ˜ λ‹€λ₯Έ μ’…λ₯˜μ˜ λ‘œλ΄‡μ΄ 가진 잠재λ ₯을 λΉ λ₯΄κ²Œ νŒŒμ•…ν•  수 있고
08:24
and quickly get a rough estimate of how useful each robot may be
146
504292
3837
물리적으둜 λ§Œλ“€κΈ° 전에 각 λ‘œλ΄‡μ΄ μ–Όλ§ˆλ‚˜ μœ μš©ν•  지
08:28
before we physically make it.
147
508171
2127
λΉ λ₯΄κ²Œ λŒ€λž΅ μ˜ˆμƒν•  수 μžˆμŠ΅λ‹ˆλ‹€.
08:32
And we predict that by allowing a novel form of breeding
148
512258
3545
물리적 λ‘œλ΄‡κ³Ό κ°€μƒμ˜ λ‘œλ΄‡μ„ κ΅λ°°μ‹œν‚€λŠ”
08:35
in which a physical robot can breed with one of its virtual cousins,
149
515803
5172
ν›Œλ₯­ν•œ ν˜•νƒœμ˜ ꡐ배λ₯Ό 톡해 λ‘œλ΄‡μ˜ 잠재λ ₯을 μ˜ˆμΈ‘ν•˜κ³ 
08:41
then the useful traits that have been discovered in simulation
150
521017
3211
가상 μ‹€ν—˜μ—μ„œ λ°œκ²¬ν•œ μœ μš©ν•œ νŠΉμ„±μ„
08:44
will quickly spread into the physical robot population,
151
524270
2961
λΉ λ₯΄κ²Œ 물리적 λ‘œλ΄‡λ“€μ—κ²Œ μ μš©μ‹œν‚€λ©΄
08:47
where they can be further refined in situ.
152
527273
2836
그듀이 상황에 맞게 더 μ„Έλ°€νžˆ μ‘°μ •ν•  κ²λ‹ˆλ‹€.
08:52
It might sound like science fiction,
153
532487
2127
κ³΅μƒκ³Όν•™μ²˜λŸΌ λ“€λ¦¬κ² μ§€λ§Œ,
08:54
but actually there's a serious point.
154
534655
2253
사싀 μ§„μ§€ν•œ μš”μ μ΄ ν•œκ°€μ§€ μžˆμŠ΅λ‹ˆλ‹€.
09:00
While we expect the technology that I've just described
155
540286
3504
저희가 κΈ°λŒ€ν•˜λŠ” 것은 μ œκ°€ λ§μ”€λ“œλ¦° 기술이
09:03
to be useful in designing robots,
156
543831
2336
λ‘œλ΄‡ 섀계에 μœ μš©ν•˜λ¦¬λΌλŠ” κ²λ‹ˆλ‹€
09:06
for example, to work in situations where it's unsafe to send humans
157
546167
4087
예둜, μ‚¬λžŒμ„ λ³΄λ‚΄κΈ°μ—λŠ” μœ„ν—˜ν•œ κ³³μ—μ„œμ˜ μž‘μ—…μ΄λ‚˜
09:10
or to help us pursue our scientific quest for exoplanetary exploration,
158
550296
5589
외계행성 νƒν—˜μ„ μœ„ν•œ 과학적 탐ꡬλ₯Ό λ•λŠ” κ²ƒμ²˜λŸΌμš”.
09:15
there are some more pragmatic reasons
159
555927
1793
ν•˜μ§€λ§Œ 인곡 진화λ₯Ό κ³ λ €ν•΄μ•Ό ν• 
09:17
why we should consider artificial evolution.
160
557720
2795
또 λ‹€λ₯Έ μ‹€μ œμ μΈ μ΄μœ κ°€ μžˆμŠ΅λ‹ˆλ‹€.
09:22
As climate change gathers pace,
161
562475
2002
κΈ°ν›„ λ³€ν™”κ°€ 점점 빨라짐에 따라
09:24
it is clear that we need a radical rethink
162
564477
2169
μƒνƒœμ  발자취λ₯Ό 쀄이기 μœ„ν•΄μ„œλŠ”
09:26
to our approach to robotic design here on Earth
163
566687
2420
μ§€κ΅¬μƒμ—μ„œ λ‘œλ΄‡ 섀계에 λŒ€ν•œ 접근법에 λŒ€ν•œ
09:29
in order to reduce that ecological footprint.
164
569148
2670
근본적인 μƒκ°μ˜ μ „ν™˜μ΄ ν•„μš”ν•¨μ΄ λΆ„λͺ…ν•©λ‹ˆλ‹€.
09:32
For example,
165
572485
1168
예λ₯Ό λ“€λ©΄,
09:33
creating new designs of robot built from sustainable materials
166
573653
4337
μƒˆλ‘œμš΄ λ‘œλ΄‡μ˜ 섀계에 μΉœν™˜κ²½ 재료λ₯Ό μ‚¬μš©ν•˜κ³ 
09:38
that operate at low energy,
167
578032
1919
μž‘μ€ μ—λ„ˆμ§€λ‘œ μžλ™ν•˜κ²Œ ν•˜κ³ 
09:39
that are repairable and recyclable.
168
579992
2837
μˆ˜λ¦¬μ™€ μž¬ν™œμš©μ΄ κ°€λŠ₯ν•˜κ²Œ ν•˜λŠ” κ²λ‹ˆλ‹€.
09:44
It's quite likely that this new generation of robots
169
584330
2544
μ΄λŸ¬ν•œ μƒˆλ‘œμš΄ μ„ΈλŒ€μ˜ λ‘œλ΄‡μ˜ λͺ¨μŠ΅μ€
09:46
won't look anything like the robots that we see around us today,
170
586874
3629
μ˜€λŠ˜λ‚  우리 μ£Όλ³€μ—μ„œ λ³΄λŠ” λ‘œλ΄‡κ³Ό μ „ν˜€ λ‹€λ₯Έ 양상일 수 μžˆμŠ΅λ‹ˆλ‹€.
09:50
but that's exactly why artificial evolution might help.
171
590503
3670
우리의 κ³Όν•™ κΈ°μˆ μ— λŒ€ν•œ 이해가 섀계 곡정에 μ œμ•½μ„ κ°€ν•˜λŠ”λ°
09:55
Discovering novel designs by processes that are unfettered by the constraints
172
595007
4713
κ·ΈλŸ¬ν•œ μ œμ•½μ„ 받지 μ•ŠλŠ” ν›Œλ₯­ν•œ 섀계λ₯Ό ν•˜λŠ”λ°
09:59
that our own understanding of engineering science
173
599762
2753
λ°”λ‘œ 인곡 진화가
10:02
imposes on the design process.
174
602557
2043
도움이 될 수 μžˆμŠ΅λ‹ˆλ‹€.
10:05
Thank you.
175
605560
1167
κ°μ‚¬ν•©λ‹ˆλ‹€.
10:06
(Applause)
176
606769
3337
(λ°•μˆ˜)
이 μ›Ήμ‚¬μ΄νŠΈ 정보

이 μ‚¬μ΄νŠΈλŠ” μ˜μ–΄ ν•™μŠ΅μ— μœ μš©ν•œ YouTube λ™μ˜μƒμ„ μ†Œκ°œν•©λ‹ˆλ‹€. μ „ 세계 졜고의 μ„ μƒλ‹˜λ“€μ΄ κ°€λ₯΄μΉ˜λŠ” μ˜μ–΄ μˆ˜μ—…μ„ 보게 될 κ²ƒμž…λ‹ˆλ‹€. 각 λ™μ˜μƒ νŽ˜μ΄μ§€μ— ν‘œμ‹œλ˜λŠ” μ˜μ–΄ μžλ§‰μ„ 더블 ν΄λ¦­ν•˜λ©΄ κ·Έκ³³μ—μ„œ λ™μ˜μƒμ΄ μž¬μƒλ©λ‹ˆλ‹€. λΉ„λ””μ˜€ μž¬μƒμ— 맞좰 μžλ§‰μ΄ μŠ€ν¬λ‘€λ©λ‹ˆλ‹€. μ˜κ²¬μ΄λ‚˜ μš”μ²­μ΄ μžˆλŠ” 경우 이 문의 양식을 μ‚¬μš©ν•˜μ—¬ λ¬Έμ˜ν•˜μ‹­μ‹œμ˜€.

https://forms.gle/WvT1wiN1qDtmnspy7