Why I draw with robots
-
0:01 - 0:04Many of us here use technology
in our day-to-day. -
0:04 - 0:07And some of us rely
on technology to do our jobs. -
0:07 - 0:11For a while, I thought of machines
and the technologies that drive them -
0:11 - 0:16as perfect tools that could make my work
more efficient and more productive. -
0:16 - 0:20But with the rise of automation
across so many different industries, -
0:20 - 0:21it led me to wonder:
-
0:21 - 0:23If machines are starting
to be able to do the work -
0:23 - 0:25traditionally done by humans,
-
0:25 - 0:27what will become of the human hand?
-
0:28 - 0:32How does our desire for perfection,
precision and automation -
0:32 - 0:34affect our ability to be creative?
-
0:35 - 0:39In my work as an artist and researcher,
I explore AI and robotics -
0:39 - 0:42to develop new processes
for human creativity. -
0:42 - 0:43For the past few years,
-
0:43 - 0:48I've made work alongside machines,
data and emerging technologies. -
0:48 - 0:50It's part of a lifelong fascination
-
0:50 - 0:53about the dynamics
of individuals and systems -
0:53 - 0:55and all the messiness that that entails.
-
0:55 - 1:00It's how I'm exploring questions about
where AI ends and we begin -
1:00 - 1:02and where I'm developing processes
-
1:02 - 1:05that investigate potential
sensory mixes of the future. -
1:06 - 1:09I think it's where philosophy
and technology intersect. -
1:09 - 1:11Doing this work
has taught me a few things. -
1:12 - 1:14It's taught me how embracing imperfection
-
1:14 - 1:17can actually teach us
something about ourselves. -
1:17 - 1:20It's taught me that exploring art
-
1:20 - 1:23can actually help shape
the technology that shapes us. -
1:23 - 1:26And it's taught me
that combining AI and robotics -
1:26 - 1:30with traditional forms of creativity --
visual arts in my case -- -
1:30 - 1:32can help us think a little bit more deeply
-
1:32 - 1:35about what is human
and what is the machine. -
1:36 - 1:38And it's led me to the realization
-
1:38 - 1:41that collaboration is the key
to creating the space for both -
1:41 - 1:42as we move forward.
-
1:42 - 1:45It all started with a simple
experiment with machines, -
1:45 - 1:48called "Drawing Operations
Unit: Generation 1." -
1:48 - 1:51I call the machine "D.O.U.G." for short.
-
1:51 - 1:52Before I built D.O.U.G,
-
1:52 - 1:55I didn't know anything
about building robots. -
1:55 - 1:58I took some open-source
robotic arm designs, -
1:58 - 2:01I hacked together a system
where the robot would match my gestures -
2:02 - 2:03and follow [them] in real time.
-
2:03 - 2:05The premise was simple:
-
2:05 - 2:07I would lead, and it would follow.
-
2:07 - 2:10I would draw a line,
and it would mimic my line. -
2:10 - 2:14So back in 2015, there we were,
drawing for the first time, -
2:14 - 2:17in front of a small audience
in New York City. -
2:17 - 2:19The process was pretty sparse --
-
2:19 - 2:23no lights, no sounds,
nothing to hide behind. -
2:23 - 2:27Just my palms sweating
and the robot's new servos heating up. -
2:27 - 2:29(Laughs) Clearly, we were
not built for this. -
2:30 - 2:33But something interesting happened,
something I didn't anticipate. -
2:33 - 2:38See, D.O.U.G., in its primitive form,
wasn't tracking my line perfectly. -
2:38 - 2:40While in the simulation
that happened onscreen -
2:40 - 2:42it was pixel-perfect,
-
2:42 - 2:44in physical reality,
it was a different story. -
2:44 - 2:47It would slip and slide
and punctuate and falter, -
2:47 - 2:49and I would be forced to respond.
-
2:50 - 2:51There was nothing pristine about it.
-
2:51 - 2:55And yet, somehow, the mistakes
made the work more interesting. -
2:55 - 2:57The machine was interpreting
my line but not perfectly. -
2:57 - 2:59And I was forced to respond.
-
2:59 - 3:01We were adapting
to each other in real time. -
3:01 - 3:03And seeing this taught me a few things.
-
3:03 - 3:08It showed me that our mistakes
actually made the work more interesting. -
3:09 - 3:13And I realized that, you know,
through the imperfection of the machine, -
3:13 - 3:17our imperfections became
what was beautiful about the interaction. -
3:18 - 3:21And I was excited,
because it led me to the realization -
3:21 - 3:24that maybe part of the beauty
of human and machine systems -
3:24 - 3:27is their shared inherent fallibility.
-
3:27 - 3:29For the second generation of D.O.U.G.,
-
3:29 - 3:31I knew I wanted to explore this idea.
-
3:31 - 3:36But instead of an accident produced
by pushing a robotic arm to its limits, -
3:36 - 3:39I wanted to design a system
that would respond to my drawings -
3:39 - 3:41in ways that I didn't expect.
-
3:41 - 3:44So, I used a visual algorithm
to extract visual information -
3:44 - 3:47from decades of my digital
and analog drawings. -
3:47 - 3:50I trained a neural net on these drawings
-
3:50 - 3:52in order to generate
recurring patterns in the work -
3:52 - 3:56that were then fed through custom software
back into the machine. -
3:56 - 4:00I painstakingly collected
as many of my drawings as I could find -- -
4:00 - 4:05finished works, unfinished experiments
and random sketches -- -
4:05 - 4:07and tagged them for the AI system.
-
4:07 - 4:10And since I'm an artist,
I've been making work for over 20 years. -
4:10 - 4:12Collecting that many drawings took months,
-
4:12 - 4:14it was a whole thing.
-
4:14 - 4:16And here's the thing
about training AI systems: -
4:16 - 4:19it's actually a lot of hard work.
-
4:19 - 4:21A lot of work goes on behind the scenes.
-
4:21 - 4:24But in doing the work,
I realized a little bit more -
4:24 - 4:27about how the architecture
of an AI is constructed. -
4:27 - 4:30And I realized it's not just made
of models and classifiers -
4:30 - 4:32for the neural network.
-
4:32 - 4:35But it's a fundamentally
malleable and shapable system, -
4:35 - 4:38one in which the human hand
is always present. -
4:38 - 4:42It's far from the omnipotent AI
we've been told to believe in. -
4:42 - 4:45So I collected these drawings
for the neural net. -
4:45 - 4:49And we realized something
that wasn't previously possible. -
4:49 - 4:53My robot D.O.U.G. became
a real-time interactive reflection -
4:53 - 4:56of the work I'd done
through the course of my life. -
4:56 - 5:00The data was personal,
but the results were powerful. -
5:00 - 5:01And I got really excited,
-
5:01 - 5:06because I started thinking maybe
machines don't need to be just tools, -
5:06 - 5:09but they can function
as nonhuman collaborators. -
5:10 - 5:11And even more than that,
-
5:11 - 5:14I thought maybe
the future of human creativity -
5:14 - 5:15isn't in what it makes
-
5:15 - 5:19but how it comes together
to explore new ways of making. -
5:19 - 5:21So if D.O.U.G._1 was the muscle,
-
5:21 - 5:23and D.O.U.G._2 was the brain,
-
5:23 - 5:26then I like to think
of D.O.U.G._3 as the family. -
5:26 - 5:31I knew I wanted to explore this idea
of human-nonhuman collaboration at scale. -
5:31 - 5:33So over the past few months,
-
5:33 - 5:36I worked with my team
to develop 20 custom robots -
5:36 - 5:38that could work with me as a collective.
-
5:38 - 5:39They would work as a group,
-
5:39 - 5:42and together, we would collaborate
with all of New York City. -
5:42 - 5:45I was really inspired
by Stanford researcher Fei-Fei Li, -
5:45 - 5:48who said, "if we want to teach
machines how to think, -
5:48 - 5:50we need to first teach them how to see."
-
5:50 - 5:52It made me think of the past decade
of my life in New York, -
5:52 - 5:56and how I'd been all watched over by these
surveillance cameras around the city. -
5:56 - 5:58And I thought it would be
really interesting -
5:58 - 6:01if I could use them
to teach my robots to see. -
6:01 - 6:03So with this project,
-
6:03 - 6:05I thought about the gaze of the machine,
-
6:05 - 6:08and I began to think about vision
as multidimensional, -
6:08 - 6:10as views from somewhere.
-
6:10 - 6:12We collected video
-
6:12 - 6:15from publicly available
camera feeds on the internet -
6:15 - 6:17of people walking on the sidewalks,
-
6:17 - 6:19cars and taxis on the road,
-
6:19 - 6:20all kinds of urban movement.
-
6:21 - 6:24We trained a vision algorithm
on those feeds -
6:24 - 6:26based on a technique
called "optical flow," -
6:26 - 6:28to analyze the collective density,
-
6:28 - 6:32direction, dwell and velocity states
of urban movement. -
6:32 - 6:36Our system extracted those states
from the feeds as positional data -
6:36 - 6:40and became pads for my
robotic units to draw on. -
6:40 - 6:42Instead of a collaboration of one-to-one,
-
6:42 - 6:45we made a collaboration of many-to-many.
-
6:45 - 6:49By combining the vision of human
and machine in the city, -
6:49 - 6:52we reimagined what
a landscape painting could be. -
6:52 - 6:54Throughout all of my
experiments with D.O.U.G., -
6:54 - 6:57no two performances
have ever been the same. -
6:57 - 6:58And through collaboration,
-
6:58 - 7:01we create something that neither of us
could have done alone: -
7:01 - 7:04we explore the boundaries
of our creativity, -
7:04 - 7:07human and nonhuman working in parallel.
-
7:08 - 7:10I think this is just the beginning.
-
7:11 - 7:13This year, I've launched Scilicet,
-
7:13 - 7:17my new lab exploring human
and interhuman collaboration. -
7:17 - 7:19We're really interested
in the feedback loop -
7:19 - 7:24between individual, artificial
and ecological systems. -
7:24 - 7:27We're connecting human and machine output
-
7:27 - 7:30to biometrics and other kinds
of environmental data. -
7:30 - 7:34We're inviting anyone who's interested
in the future of work, systems -
7:34 - 7:35and interhuman collaboration
-
7:35 - 7:37to explore with us.
-
7:37 - 7:40We know it's not just technologists
that have to do this work -
7:40 - 7:42and that we all have a role to play.
-
7:42 - 7:45We believe that by teaching machines
-
7:45 - 7:47how to do the work
traditionally done by humans, -
7:47 - 7:50we can explore and evolve our criteria
-
7:50 - 7:53of what's made possible by the human hand.
-
7:53 - 7:56And part of that journey
is embracing the imperfections -
7:56 - 8:00and recognizing the fallibility
of both human and machine, -
8:00 - 8:03in order to expand the potential of both.
-
8:03 - 8:05Today, I'm still in pursuit
of finding the beauty -
8:05 - 8:08in human and nonhuman creativity.
-
8:08 - 8:11In the future, I have no idea
what that will look like, -
8:12 - 8:14but I'm pretty curious to find out.
-
8:14 - 8:15Thank you.
-
8:15 - 8:17(Applause)
- Title:
- Why I draw with robots
- Speaker:
- Sougwen Chung
- Description:
-
What happens when humans and robots make art together? In this awe-inspiring talk, artist Sougwen Chung shows how she "taught" her artistic style to a machine -- and shares the results of their collaboration after making an unexpected discovery: robots make mistakes, too. "Part of the beauty of human and machine systems is their inherent, shared fallibility," she says.
- Video Language:
- English
- Team:
- closed TED
- Project:
- TEDTalks
- Duration:
- 08:30
marialadias edited English subtitles for Why I draw with robots | ||
Erin Gregory approved English subtitles for Why I draw with robots | ||
Erin Gregory edited English subtitles for Why I draw with robots | ||
Camille Martínez accepted English subtitles for Why I draw with robots | ||
Camille Martínez edited English subtitles for Why I draw with robots | ||
Camille Martínez edited English subtitles for Why I draw with robots | ||
Ivana Korom edited English subtitles for Why I draw with robots | ||
Ivana Korom edited English subtitles for Why I draw with robots |