Page 145 - Modern Robotics Building Versatile Macines
P. 145

SOCIABLE ROBOTS   125


              When Kismet did not have human contact for some time, it became
            lonely. If a visitor arrived, the robot began an attention-getting dis-
            play. It tilted its head forward. Its ears swiveled a bit like those of an
            excited terrier, while its vocal babbling conveyed excitement.
              The need for human company was one of Kismet’s three major
            motivational drives. Another was stimulation from seeing “interest-
            ing” objects. Because Kismet had no arms, it conveyed its interest
            in an object to a person, who usually reacted by bringing the object
            closer to the robot. Like an infant, Kismet also got “tired” after
            prolonged interaction because of its fatigue drive.
              In an interview with Douglas Whynott and Fenella Saunders of
            Discover magazine, Breazeal stressed that


              The behavior [of Kismet] is not canned. It is being computed and is
              not a random thing. The interaction is rich enough so that you can’t
              tell what’s going to happen next. The overarching behavior is that the
              robot is seeking someone out, but the internal factors are changing
              all the time.


              Perhaps the most remarkable thing was how fluidly Kismet’s
            behavior arose out of a system that has 15 separate computers run-
            ning several different operating systems.



            Leonardo

            Seeing how much can be elicited in both robots and humans even
            by the relatively simple Kismet, Breazeal was eager to build on
            that experience. One important challenge she faced was to link the
            cognitive and learning processes to the emotional drives and social
            interactions. Thus, as she explained to Time reporter Adam Cohen
            in 2000, a future “sociable robot” would learn language much in the
            way an infant does. The words it would learn most quickly would
            therefore be those that are connected with emotional needs—being
            able to ask for a favorite toy, for example. In the article, Breazeal
            said she hopes that eventually a robot will be able to make the kinds
            of links that seem to be almost instinctive in toddlers: “When I’m in
   140   141   142   143   144   145   146   147   148   149   150