> >> A CONVERSATION WITH / Anne Foerst
>> Do Androids Dream? M.I.T. Is Working on It
>> <By CLAUDIA DREIFUS<>>
>> Rick Friedman for The New York Times
>> Dr. Anne Foerst outside her office at the Artificial Intelligence
>> Related Articles
>> <• <A HREF="http://www.nytimes.com/library/tech/00/08/biztech/articles/31robot.html"><Aping Biology, Computer Guides Automated Evolution of a Robot </A>(Aug. 31,
>> <• <A HREF="http://www.nytimes.com/library/tech/00/06/circuits/articles/22next.html"><Trying to Give Bomb Squad Robots Brains to Match Their Brawn </A>(June 22,
>> <• <A HREF="http://www.nytimes.com/library/tech/00/06/circuits/articles/15fish.html"><Team Links Brain Cells With a Robot</A> (June 15, 2000)
>> <• <A HREF="http://www.nytimes.com/library/tech/00/03/circuits/articles/09next.html"><Shape-Shifting Robots From Xerox</A> (March 9, 2000)
>> Rick Friedman for The New York Times
>> "Cute" robots with social skills, like Kismet, are in favor at the M.I.T.
>> <>AMBRIDGE, Mass. — Dr. Anne Foerst, 34, a researcher at the Artificial
>> Intelligence Laboratory at the Massachusetts Institute of Technology and
>> the director of M.I.T.'s God and Computers project, apologized on a recent
>> afternoon that a certain robot named Kismet wouldn't be joining our
>> "Cynthia Breazeal, who built Kismet, is away in Japan right now and
>> there's no getting her going," Dr. Foerst said in her German accent, "but
>> you'd love her. She's oh so cute." A cute robot? Well, yes. At the
>> Artificial Intelligence Laboratory, engineers are trying to build robots
>> with social skills and humanlike experiences, and so, as an experiment,
>> they've created creatures that they think humans will relate to.
>> Dr. Foerst, a Lutheran minister who supported herself by repairing
>> computers during eight years of higher education in Germany, serves as
>> theological adviser to the scientists building Kismet and the robot's
>> brother, Cog.
>> Q. What exactly do people do here at this laboratory?
>> A. We are trying to build robots that are social and embodied.
>> We have four projects. I am the theological adviser for two of them: the
>> building of the humanoid machines, Cog and Kismet.
>> Cog is a robot built in analogy to a human infant. He has a torso, two
>> arms, a head, ears and eyes. He, it, learns to coordinate those limbs to
>> explore its environment, just as newborn babies do. Kismet is a robot who
>> interacts with humans through her body posture and facial expressions. The
>> aim of this project is to explore social interactions between humans and
>> robots and also between the humans themselves.
>> Q. Why a theologian here in this particular laboratory?
>> A. Two reasons. The first is when you build machines in analogy to humans,
>> you make assumptions about humans. Theologians explore the cultural and
>> spiritual dimensions of that very question, What does it mean to be human?
>> The idea is that as these robots are built, we can use the wisdom of
>> religious studies to enlarge our understanding of humans, and thus what
>> you build into the humanoid machines.
>> The other reason is that when we build social interactive robots that
>> force people to treat them as if they were persons, tricky moral questions
>> come up. For instance, Who are we, really? Are all our reactions actually
>> developed in a very mechanistic, functionalist way? Or is there a
>> dimension to social interaction that goes beyond that? What are ethics
>> here? Why should I treat someone else like a human, with dignity, when it
>> is just a mechanistic thing?
>> For instance, one question we discuss quite frequently is, What would be
>> the threshold when the robots are developed to a certain point that you
>> couldn't switch them off anymore? The question really is, When does a
>> creature deserve to be treated as intrinsically valuable?
>> Q. When do you think a robot should be treated as intrinsically valuable?
>> A. Well, that moment is 50 years down the road. At least. But it's pretty
>> clear that when it comes, those who built the robot will have to make that
>> decision because they won't be blinded by their fears of the seemingly
>> human qualities of the machines. They'll know what's inside. And if it
>> ever got to the point where the builders felt, Oops, now that has become
>> something, the builders could become the creature's strongest advocates.
>> Q. What make the robots Cog and Kismet different from previous ones?
>> A. Previous attempts put very abstract features of human intelligence into
>> a machine: chess playing, mathematical theorem-proving and natural
>> language processing. The idea now is, In order for a machine to really be
>> intelligent, it has to be embodied. We say intelligence cannot be
>> abstracted from the body. We feel that the body — the way it moves, grows,
>> digests food, gets older, all have an influence on how a person thinks.
>> That's why we've built Cog and Kismet to have humanoid features.
>> Cog moves and experiences the world the way someone who can walk upright
>> might. He experiences balance problems, friction problems, weight,
>> gravity, all the stuff that we do, so that he can have a body feeling that
>> is similar to ours. The humanoid features are also crafted into the
>> machines in order to trigger social responses from the people interacting
>> with them.
>> The other thing we believe is that humans are human because we are social.
>> Thus, we try to treat Cog and Kismet something like the way most of us
>> treat babies, as if they have intentionality, emotion, desires and
>> intelligence. We give them as much social interaction as we can.
>> Cog is a whole body and Kismet is mostly a head and facial expression. Our
>> work with Cog concentrates more on the embodiment stuff and Kismet more on
>> emotional-social learning.
>> Q. Is the robot Kismet a she?
>> A. Robots are its. But I can't help but think of her as a she. If you were
>> to see Kismet, you would be taken by her enormously expressive face: long
>> eyelashes, big blue eyes, movable brow, cute, kissy mouth. When Kismet
>> puts her eyes on you and looks sad, you want to make her happy. Of course,
>> part of you thinks, It's just a stupid machine. But you do react and you
>> can't help it.
>> The point of reacting to Kismet is the same as reacting to a baby. We
>> believe that only when you treat the machines as if they have all these
>> social characteristics, will they ever get them. If you want to have an
>> intelligent being, you need to create that circle. So we react here to
>> Kismet's emotional displays. When she's bored, you want to make her happy.
>> When she seems scared, you back off.
>> Q. Has the very social robot Kismet done anything yet that has astonished
>> A. Kismet has not yet learned. Cog is the one who learns. A former
>> graduate student, Matt Williamson, the guy who built Cog's arms, taught
>> the robot how to control his arm.
>> To coordinate the arms, Matt had to touch a part of Cog's body and then,
>> the arm would touch that part, too. After he did that for the first time,
>> Matt ran into my office and said, "You've got to come to look at this." It
>> looked so eerily human. It's not so much that Cog does something that's
>> unexpected, it's more the human reaction, like, it's alive!
>> Q. People often talk about humans having some indefinable extra above life
>> that makes for humanness — some call it "spirit." Can a robot have spirit?
>> A. Rod calls it "juice." He says, "Even though I get it all right, might
>> there not be some juice I'm missing?" I would say from a religious
>> perspective, the juice is that which comes from the outside world and
>> emerges in social interaction.
>> Q. Some people might complain that in building humanoid robots, you are
>> trying to supplant God.
>> A. Yes. I know. They say, "Do you want to be like God?" Actually, if you
>> use biology as your inspiration in your robot-building and focus on
>> embodiment and environment, you get much more humble instead of arrogant.
>> Suddenly, you realize that even the most brilliant robot that the most
>> brilliant engineers have worked on for years and years is still dumber
>> than an insect.
>> Q. So, in your view, God is, as the Latin Americans say, the "intellectual
>> author" of everything?
>> A. No. The creative author. When we are creative, the power of creation is
>> from God.
>> Q. In the many plays, novels and movies about robots, the dramatic climax
>> of the story always comes at moment when the machine achieves sentience.
>> Why do you think that is?
>> A. Well, I think it's the search to feel and to be treated like something
>> more than the sum of the parts that's inherently dramatic. This is the
>> moment when the robots start to participate in the all-too-human quest of
>> what does it mean to be me?
>> Q. In the movie "2001: A Space Odyssey," HAL becomes a danger to humans
>> once he's sentient.
>> A. In "Frankenstein," too. But in both cases, there is an explanation.
>> When you look at Frankenstein, he is never part of a community. His
>> creator left him right away. The people hated him, feared him, ran away
>> from him. The only person who ever loved him was a blind man who couldn't
>> see what he looked like. Frankenstein was never treated as a valuable
>> being, a person with dignity. He had to turn against the society that
>> shunned him. Where should the goodness come from when he never experienced
>> it himself?
>> HAL is the same thing. And he's disembodied. There is no body with which
>> to experience the world. I would even say that in such a setting a robot
>> couldn't even become sentient. In the movie, HAL becomes sentient at some
>> point and nobody notices. No one treats him properly and he's isolated and
>> what happens? He becomes psychotic.
>> Q. What's your favorite robot movie?
>> A. "Blade Runner." I teach it in my classes. The robots have this absolute
>> search for meaning, and when their quest is not taken seriously, it
>> becomes fatal. The movie raises this wonderful question: how do humanoid
>> creatures feel about having been created by us and how do they deal with
>> their human-made limitations?
This archive was generated by hypermail 2b30 : Mon May 28 2001 - 09:50:20 MDT