WEBVTT
1
00:00:00.000 --> 00:00:01.030
[otherworldly music]
2
00:00:01.030 --> 00:00:02.100
It's easy to get lost
3
00:00:02.100 --> 00:00:04.440
in Refik Anadol's surreal installations.
4
00:00:04.440 --> 00:00:06.000
But dig a little deeper,
5
00:00:06.000 --> 00:00:07.030
and you realize his work
6
00:00:07.030 --> 00:00:09.530
is made up of millions of tiny pieces,
7
00:00:09.530 --> 00:00:11.110
and that every single little point
8
00:00:11.110 --> 00:00:12.690
represents a piece of data
9
00:00:12.690 --> 00:00:14.440
that he's fed through a neural network
10
00:00:14.440 --> 00:00:17.123
to show us his vision of our future.
11
00:00:17.123 --> 00:00:19.201
I was always trying to speculate
12
00:00:19.201 --> 00:00:22.280
this idea of, "Can data become a pigment?"
13
00:00:22.280 --> 00:00:24.810 line:15%
At the end, the data is truly numbers.
14
00:00:24.810 --> 00:00:27.490 line:15%
It has no kind of inner skin or skeleton.
15
00:00:27.490 --> 00:00:30.330
But what I'm trying to do as an artist is find algorithm
16
00:00:30.330 --> 00:00:32.740
that can narrate the moment of data,
17
00:00:32.740 --> 00:00:35.490
kind of make that invisible moment visible.
18
00:00:35.490 --> 00:00:38.226
Enormous data sets inspire most of Anadol's art
19
00:00:38.226 --> 00:00:41.047
and he uses machine intelligence and algorithms
20
00:00:41.047 --> 00:00:44.950
to create visualizations, what he calls data sculptures
21
00:00:44.950 --> 00:00:47.730
like this piece, called "Machine Hallucinations."
22
00:00:47.730 --> 00:00:52.160
He started by finding 113 million images of New York online.
23
00:00:52.160 --> 00:00:53.545
I let the machine learn
24
00:00:53.545 --> 00:00:54.969
from this entire corpus of data,
25
00:00:54.969 --> 00:00:56.969
and then we also find a way
26
00:00:56.969 --> 00:00:59.668
to erase the humans from this data
27
00:00:59.668 --> 00:01:03.470
and only focus on the buildings, nature, and environments
28
00:01:03.470 --> 00:01:05.650
that is a collective memory of New York.
29
00:01:05.650 --> 00:01:08.020
Once all of the images of people were removed,
30
00:01:08.020 --> 00:01:10.960
Anadol was left with 10 million pictures of New York.
31
00:01:10.960 --> 00:01:13.230
He fed them through a machine learning algorithm
32
00:01:13.230 --> 00:01:16.228
that generates visual associations as it learns.
33
00:01:16.228 --> 00:01:18.820
For example, when it sees multiple photos
34
00:01:18.820 --> 00:01:19.850
of the Statue of Liberty,
35
00:01:19.850 --> 00:01:21.680
taken from slightly different angles,
36
00:01:21.680 --> 00:01:23.650
the algorithm interpolates information
37
00:01:23.650 --> 00:01:26.080
to help it create a moving, shifting image
38
00:01:26.080 --> 00:01:28.990
that represents the entire life cycle of the structure.
39
00:01:28.990 --> 00:01:32.960
Machine looks at this information like a human being,
40
00:01:32.960 --> 00:01:35.273
but it's kind of more like collective memories
41
00:01:35.273 --> 00:01:36.620
than personal memories,
42
00:01:36.620 --> 00:01:38.200
because a building in New York
43
00:01:38.200 --> 00:01:40.530
can be explored by thousands of perspectives,
44
00:01:40.530 --> 00:01:41.760
from multiple angles,
45
00:01:41.760 --> 00:01:43.290
from a different time of the year.
46
00:01:43.290 --> 00:01:46.690
It's more like an honest memory for a machine
47
00:01:46.690 --> 00:01:48.489
'cause it feels more totalitarian,
48
00:01:48.489 --> 00:01:50.910
and feels everything and everyone
49
00:01:50.910 --> 00:01:52.793
than just one person.
50
00:01:52.793 --> 00:01:54.692
Anadol covered the walls and the floor
51
00:01:54.692 --> 00:01:57.570
of a boiler room in a building in Lower Manhattan
52
00:01:57.570 --> 00:02:00.291
with this machine-interpreted dynamic landscape.
53
00:02:00.291 --> 00:02:01.790
If you ask him,
54
00:02:01.790 --> 00:02:03.769
he'll tell you that the machine is dreaming.
55
00:02:03.769 --> 00:02:05.932
When a machine learns
56
00:02:05.932 --> 00:02:08.353
from outputs and memories like this,
57
00:02:08.353 --> 00:02:11.350
it can create an alternative reality.
58
00:02:11.350 --> 00:02:13.780
It look at the patterns of the trees, the buildings,
59
00:02:13.780 --> 00:02:15.310
the nature, the people,
60
00:02:15.310 --> 00:02:18.880
every single thing hidden inside these image corpus.
61
00:02:18.880 --> 00:02:21.793
Seeing a machine giving a context of data
62
00:02:21.793 --> 00:02:24.270
and giving an hallucinative output
63
00:02:24.270 --> 00:02:26.951
was something really inspiring.
64
00:02:26.951 --> 00:02:28.970
Anadol created this piece
65
00:02:28.970 --> 00:02:31.470
to celebrate the centennial of the LA Philharmonic
66
00:02:31.470 --> 00:02:33.165
using half a million images,
67
00:02:33.165 --> 00:02:34.990
thousands of audio recordings,
68
00:02:34.990 --> 00:02:37.890
and hundreds of videos from the orchestra's archives.
69
00:02:37.890 --> 00:02:40.270
He fed all that into a series of algorithms
70
00:02:40.270 --> 00:02:43.370
that turned it into these extraordinary projected images.
71
00:02:43.370 --> 00:02:45.030
By using machine learning algorithms,
72
00:02:45.030 --> 00:02:47.600
the entire archive of LA Philharmonic
73
00:02:47.600 --> 00:02:51.130
becomes a shape of three-dimensional outputs.
74
00:02:51.130 --> 00:02:52.870
We were able to see all these data points
75
00:02:52.870 --> 00:02:54.390
becoming together or disappearing,
76
00:02:54.390 --> 00:02:56.370
and making kind of new sculptures.
77
00:02:56.370 --> 00:02:57.500
And the building itself,
78
00:02:57.500 --> 00:02:59.450
Frank Gehry's Walt Disney Concert Hall,
79
00:02:59.450 --> 00:03:01.806
became part of the sculpture too.
80
00:03:01.806 --> 00:03:05.250
I was always looking for inspiring buildings,
81
00:03:05.250 --> 00:03:07.040
and Frank Gehry became my hero.
82
00:03:07.040 --> 00:03:11.120
And I rent a car, went to downtown, it was 2:00 a.m.
83
00:03:11.120 --> 00:03:12.750
The building's light was off,
84
00:03:12.750 --> 00:03:14.250
like there was nothing around it.
85
00:03:14.250 --> 00:03:16.610
And that night, I was really struck by an idea,
86
00:03:16.610 --> 00:03:19.760
like, "What will happen if this building can remember?"
87
00:03:19.760 --> 00:03:21.230
Like, "Can it dream?"
88
00:03:21.230 --> 00:03:22.130
So he projected
89
00:03:22.130 --> 00:03:23.680
the machine interpreted archives
90
00:03:23.680 --> 00:03:26.526
onto a smaller model of the Walt Disney Concert Hall.
91
00:03:26.526 --> 00:03:28.530
Once he settled on something he liked,
92
00:03:28.530 --> 00:03:31.980
he projected it onto the bigger, real life building.
93
00:03:31.980 --> 00:03:34.610
What we are doing not a typical painting,
94
00:03:34.610 --> 00:03:35.806
not a typical sculpture,
95
00:03:35.806 --> 00:03:38.386
but more like an experience from a near future.
96
00:03:38.386 --> 00:03:39.950
And when they come together
97
00:03:39.950 --> 00:03:42.130
with sound, data, machine intelligence,
98
00:03:42.130 --> 00:03:43.960
and light and architecture,
99
00:03:43.960 --> 00:03:44.920
they have a new meaning,
100
00:03:44.920 --> 00:03:47.604
a new kind of symbiotic connections.
101
00:03:47.604 --> 00:03:49.530
I think I was really hooked up
102
00:03:49.530 --> 00:03:53.120
by the idea of, "A building can become an interface."
103
00:03:53.120 --> 00:03:55.500
It can be a boring advertising,
104
00:03:55.500 --> 00:03:57.465
but it can be a beautiful interface.
105
00:03:57.465 --> 00:04:00.250
The ideation of a human instinct,
106
00:04:00.250 --> 00:04:01.998
becoming part of a machine,
107
00:04:01.998 --> 00:04:05.270
is really striking me emotionally.
108
00:04:05.270 --> 00:04:06.170
Anadol and his team
109
00:04:06.170 --> 00:04:08.070
used 42 large scale projectors
110
00:04:08.070 --> 00:04:10.584
with an astonishing 50K video resolution
111
00:04:10.584 --> 00:04:13.610
to create a dramatic display in Downtown LA
112
00:04:13.610 --> 00:04:15.330
every night for 10 days,
113
00:04:15.330 --> 00:04:17.444
a nod to one of his favorite movies.
114
00:04:17.444 --> 00:04:20.025
[whimsical music]
115
00:04:20.025 --> 00:04:21.310
I was eight years old
116
00:04:21.310 --> 00:04:22.890
when I watched the movie "Blade Runner."
117
00:04:22.890 --> 00:04:24.650
I was extremely inspired by the movie.
118
00:04:24.650 --> 00:04:27.785
In Downtown Los Angeles, a building suddenly becomes alive,
119
00:04:27.785 --> 00:04:31.023
and it has a cognitive capacity of remembering,
120
00:04:31.023 --> 00:04:33.770
and it has a capacity of dreaming.
121
00:04:33.770 --> 00:04:36.745
So this was all very high-level science fiction narratives.
122
00:04:36.745 --> 00:04:38.526
In the futuristic dystopian films,
123
00:04:38.526 --> 00:04:40.788
exploration of more complex themes
124
00:04:40.788 --> 00:04:43.700
echo Anadol's interest in the relationship
125
00:04:43.700 --> 00:04:45.950
between humans and technology.
126
00:04:45.950 --> 00:04:48.170
You think I'm a replicant, don't you?
127
00:04:48.170 --> 00:04:49.460
When Deckard and Rachael
128
00:04:49.460 --> 00:04:51.700
contemplates each other and has a dialogue,
129
00:04:51.700 --> 00:04:54.120
where Deckard defines Rachael as a replicant,
130
00:04:54.120 --> 00:04:56.120
it was a very interesting moment.
131
00:04:56.120 --> 00:04:58.206
Where a human defines another machine,
132
00:04:58.206 --> 00:05:01.030
that is kind of this,
133
00:05:01.030 --> 00:05:03.390
defining what is real or what is not.
134
00:05:03.390 --> 00:05:07.508
That kind of human consciousness and machine consciousness
135
00:05:07.508 --> 00:05:09.785
were having its dialogue moment,
136
00:05:09.785 --> 00:05:11.908
that was truly inspiring.
137
00:05:11.908 --> 00:05:13.340
You remember the spider
138
00:05:13.340 --> 00:05:15.140
that lived in a bush outside your window?
139
00:05:15.140 --> 00:05:16.440
The egg hatched.
140
00:05:16.440 --> 00:05:17.280
Those aren't your memories.
141
00:05:17.280 --> 00:05:18.840
Those are somebody else's.
142
00:05:18.840 --> 00:05:20.040
I think, in humanity,
143
00:05:20.040 --> 00:05:22.302
we have very certain findings
144
00:05:22.302 --> 00:05:24.780
that transform who we are.
145
00:05:24.780 --> 00:05:26.170
When we found the fire,
146
00:05:26.170 --> 00:05:28.206
we cook with it, we create communities.
147
00:05:28.206 --> 00:05:29.990
With the same technology,
148
00:05:29.990 --> 00:05:32.200
we kill each other and destroy.
149
00:05:32.200 --> 00:05:35.968
And clearly, AI is one of the discovery in humanity
150
00:05:35.968 --> 00:05:37.947
that has a potential to make communities
151
00:05:37.947 --> 00:05:39.966
or destroy each other.
152
00:05:39.966 --> 00:05:41.580
Anadol finds data
153
00:05:41.580 --> 00:05:44.066
for his machine intelligence collaborations everywhere,
154
00:05:44.066 --> 00:05:46.230
and when he learned that when data is
155
00:05:46.230 --> 00:05:48.990
continuously being collected at the Boston Airport,
156
00:05:48.990 --> 00:05:50.805
he knew he'd found a gold mine.
157
00:05:50.805 --> 00:05:52.900
He turned it all into this project
158
00:05:52.900 --> 00:05:54.667
called "Winds of Boston."
159
00:05:54.667 --> 00:05:58.060
I was always inspired by nature, as an inspiration,
160
00:05:58.060 --> 00:05:59.765
from the wind itself, the water,
161
00:05:59.765 --> 00:06:01.980
the nature, like a core nature,
162
00:06:01.980 --> 00:06:05.400
like aspect of motion, theory, and life in general,
163
00:06:05.400 --> 00:06:07.050
and I thought this can be an incredible opportunity
164
00:06:07.050 --> 00:06:10.521
to use wind as a data and as a pigment.
165
00:06:10.521 --> 00:06:13.730
So we first located our data source
166
00:06:13.730 --> 00:06:16.090
from the airport, Logan Airport.
167
00:06:16.090 --> 00:06:18.365
The wind is extremely important for the flight networks.
168
00:06:18.365 --> 00:06:22.465
So we took a one-year-long wind data of Boston,
169
00:06:22.465 --> 00:06:26.880
and this data consists of the gust, speed, and direction,
170
00:06:26.880 --> 00:06:28.920
and the temperature of the weather.
171
00:06:28.920 --> 00:06:30.590
Anadol fed all this wind data
172
00:06:30.590 --> 00:06:32.344
into a series of algorithms,
173
00:06:32.344 --> 00:06:35.325
then he built custom 13-foot-tall LED screens
174
00:06:35.325 --> 00:06:37.550
to display the visualized data.
175
00:06:37.550 --> 00:06:40.010
And I thought that algorithms can be an incredible way
176
00:06:40.010 --> 00:06:42.210
of visualizing this invisible pattern of things,
177
00:06:42.210 --> 00:06:44.910
and transform them into a poetic-like motion.
178
00:06:44.910 --> 00:06:46.130
Because it's very inspiring
179
00:06:46.130 --> 00:06:48.264
that a machine can find something interesting,
180
00:06:48.264 --> 00:06:50.164
that I didn't even think about it.
181
00:06:50.164 --> 00:06:52.340
And I found that the true collaboration
182
00:06:52.340 --> 00:06:53.730
starts, happen there.
183
00:06:53.730 --> 00:06:56.240
And sometimes machine gives opportunities
184
00:06:56.240 --> 00:06:57.385
that I didn't think about it,
185
00:06:57.385 --> 00:06:59.790
and I do believe that is really something
186
00:06:59.790 --> 00:07:01.366
that is inspiring for humanity,
187
00:07:01.366 --> 00:07:02.990
that will most likely bring
188
00:07:02.990 --> 00:07:04.885
some new kind of imagination methods
189
00:07:04.885 --> 00:07:06.853
that is not discovered yet.
190
00:07:08.490 --> 00:07:09.970
The inspiration for this project,
191
00:07:09.970 --> 00:07:11.310
called "Melting Memories,"
192
00:07:11.310 --> 00:07:12.870
came from a personal place.
193
00:07:12.870 --> 00:07:14.908
I went to Istanbul, my home town,
194
00:07:14.908 --> 00:07:19.020
and my uncle just couldn't remember
195
00:07:19.020 --> 00:07:20.890
where I am coming from.
196
00:07:20.890 --> 00:07:24.270
And apparently it was his first stage of Alzheimer's.
197
00:07:24.270 --> 00:07:26.470
When this disease happens,
198
00:07:26.470 --> 00:07:29.308
we literally melt our memories.
199
00:07:29.308 --> 00:07:31.809
Our brain tissue disappears.
200
00:07:31.809 --> 00:07:33.560
But I also was very curious about,
201
00:07:33.560 --> 00:07:36.470
scientifically, honestly, what memory means.
202
00:07:36.470 --> 00:07:37.610
Like where do they come from?
203
00:07:37.610 --> 00:07:39.747
What is the cognitive representation of a memory/
204
00:07:39.747 --> 00:07:41.825
And the images
205
00:07:41.825 --> 00:07:42.740
on these 20-foot-tall LED screens
206
00:07:42.740 --> 00:07:45.185
represent the real data behind that process.
207
00:07:45.185 --> 00:07:48.246
And these moments are really trying to give a sense
208
00:07:48.246 --> 00:07:51.160
of a tangible feeling of a memory.
209
00:07:51.160 --> 00:07:52.750
I know that we are not there,
210
00:07:52.750 --> 00:07:55.526
as technology that truly transcribes memory,
211
00:07:55.526 --> 00:07:58.520
but it at least gives a glimpse
212
00:07:58.520 --> 00:08:00.550
of an abstract language of a memory.
213
00:08:00.550 --> 00:08:02.966
That's one of those feelings that is hidden
214
00:08:02.966 --> 00:08:05.808
in those algorithmic exploration of data.
215
00:08:05.808 --> 00:08:07.500
Anadol partnered with scientists
216
00:08:07.500 --> 00:08:09.180
at a neuroscience lab.
217
00:08:09.180 --> 00:08:11.810
They asked subjects to concentrate on childhood memories,
218
00:08:11.810 --> 00:08:14.980
and recorded all their brain pulses using an EEG.
219
00:08:14.980 --> 00:08:16.566
The data itself is literally pulsing
220
00:08:16.566 --> 00:08:19.430
in four milliseconds of human brain activity,
221
00:08:19.430 --> 00:08:21.520
in the two location of brain,
222
00:08:21.520 --> 00:08:23.945
from hippocampus to frontal left lobe,
223
00:08:23.945 --> 00:08:27.600
and the algorithm symbolizes the moment of remembering,
224
00:08:27.600 --> 00:08:29.201
and it is how poetic,
225
00:08:29.201 --> 00:08:31.443
moment of this firing location
226
00:08:31.443 --> 00:08:33.480
in two locations of the brain.
227
00:08:33.480 --> 00:08:35.720
Anadol and his team created custom software
228
00:08:35.720 --> 00:08:37.670
that transformed all of that brain data
229
00:08:37.670 --> 00:08:40.326
into an artistic interpretation of the neurons firing.
230
00:08:40.326 --> 00:08:43.260
Another example of how we use this technology
231
00:08:43.260 --> 00:08:44.920
to describe life.
232
00:08:44.920 --> 00:08:46.450
I never stop thinking about data
233
00:08:46.450 --> 00:08:49.140
as a mean of like material.
234
00:08:49.140 --> 00:08:50.561
Sometimes it can be wind data.
235
00:08:50.561 --> 00:08:53.770
It can be wifi, be a little signal.
236
00:08:53.770 --> 00:08:56.639
It can be machine decisions.
237
00:08:56.639 --> 00:08:59.570
It can be CPU data, GPU data.
238
00:08:59.570 --> 00:09:01.718
It's honestly anything that machine needs
239
00:09:01.718 --> 00:09:04.170
to understand life,
240
00:09:04.170 --> 00:09:06.980
can become a material for this imagination.
241
00:09:06.980 --> 00:09:08.850
One problem we have to solve today [laughs]
242
00:09:08.850 --> 00:09:10.539
is the visual discontinuity problem.
243
00:09:10.539 --> 00:09:12.200
Anadol has built a team
244
00:09:12.200 --> 00:09:13.830
with expertise in different fields
245
00:09:13.830 --> 00:09:15.830
to help him fulfill his vision.
246
00:09:15.830 --> 00:09:18.801
AI experts, computer scientists, architects, and designers.
247
00:09:18.801 --> 00:09:21.830
Produce them, we want to make them continuous.
248
00:09:21.830 --> 00:09:23.250
And they're working at a new work flow
249
00:09:23.250 --> 00:09:26.118
for a data-driven public art project set for Portland.
250
00:09:26.118 --> 00:09:29.100
For this one, Anadol's letting hundreds of thousands
251
00:09:29.100 --> 00:09:31.720
of images of Portland inform both the projections
252
00:09:31.720 --> 00:09:33.718
and the structure they'll be projected on.
253
00:09:33.718 --> 00:09:36.030
The team is using a 3D printer
254
00:09:36.030 --> 00:09:38.520
to build the structure one panel at a time.
255
00:09:38.520 --> 00:09:41.740
It will ultimately be a 21-foot-tall sculpture.
256
00:09:41.740 --> 00:09:42.910
I found that we were stuck
257
00:09:42.910 --> 00:09:44.290
in this just virtual world.
258
00:09:44.290 --> 00:09:45.937
I found that we are in the screens,
259
00:09:45.937 --> 00:09:48.518
2D flat world of imagination.
260
00:09:48.518 --> 00:09:49.950
I was really looking forward,
261
00:09:49.950 --> 00:09:52.881
how we can take this machine consciousness
262
00:09:52.881 --> 00:09:56.377
out of the screen and bring it to the 3D world.
263
00:09:56.377 --> 00:09:57.810
Whatever's in store
264
00:09:57.810 --> 00:09:59.650
for the future of Anadol's work,
265
00:09:59.650 --> 00:10:02.070
data machine learning will be the foundation.
266
00:10:02.070 --> 00:10:06.350
It's very clear that machines can really capture our data,
267
00:10:06.350 --> 00:10:09.372
machines can capture our decisions, our memories,
268
00:10:09.372 --> 00:10:12.214
but it's not clear what will happen to them.
269
00:10:12.214 --> 00:10:15.880
And I think, also, clearly,
270
00:10:15.880 --> 00:10:17.302
the data we leave behind
271
00:10:17.302 --> 00:10:19.860
is the memories of humanity.
272
00:10:19.860 --> 00:10:23.030
My obsession with this relationship
273
00:10:23.030 --> 00:10:25.366
is what else we can do with that.
274
00:10:25.366 --> 00:10:28.449
[otherworldly music]