Little Progress Toward Complex Autonomous Robots

(p. A8) [In June 2015] . . . , the Defense Advanced Research Projects Agency, a Pentagon research arm, . . . [held] the final competition in its Robotics Challenge in Pomona, Calif. With $2 million in prize money for the robot that performs best in a series of rescue-oriented tasks in under an hour, the event . . . offer[ed] what engineers refer to as the “ground truth” — a reality check on the state of the art in the field of mobile robotics.

A preview of their work suggests that nobody needs to worry about a Terminator creating havoc anytime soon. Given a year and a half to improve their machines, the roboticists, who shared details about their work in interviews before the contest in June, appear to have made limited progress.
. . .
“The extraordinary thing that has happened in the last five years is that we have seemed to make extraordininary progress in machine perception,” said Gill Pratt, the Darpa program manager in charge of the Robotics Challenge.
Pattern recognition hardware and software has made it possible for computers to make dramatic progress in computer vision and speech understanding. In contrast, Dr. Pratt said, little headway has been made in “cognition,” the higher-level humanlike processes required for robot planning and true autonomy. As a result, both in the Darpa contest and in the field of robotics more broadly, there has been a re-emphasis on the idea of human-machine partnerships.
“It is extremely important to remember that the Darpa Robotics Challenge is about a team of humans and machines working together,” he said. “Without the person, these machines could hardly do anything at all.”
In fact, the steep challenge in making progress toward mobile robots that can mimic human capabilities is causing robotics researchers worldwide to rethink their goals. Now, instead of trying to build completely autonomous robots, many researchers have begun to think instead of creating ensembles of humans and robots, an approach they describe as co-robots or “cloud robotics.”
Ken Goldberg, a University of California, Berkeley, roboticist, has called on the computing world to drop its obsession with singularity, the much-ballyhooed time when computers are predicted to surpass their human designers. Rather, he has proposed a concept he calls “multiplicity,” with diverse groups of humans and machines solving problems through collaboration.
For decades, artificial-intelligence researchers have noted that the simplest tasks for humans, such as reaching into a pocket to retrieve a quarter, are the most challenging for machines.
“The intuitive idea is that the more money you spend on a robot, the more autonomy you will be able to design into it,” said Rodney Brooks, an M.I.T. roboticist and co-founder two early companies, iRobot and Rethink Robotics. “The fact is actually the opposite is true: The cheaper the robot, the more autonomy it has.”
For example, iRobot’s Roomba robot is autonomous, but the vacuuming task it performs by wandering around rooms is extremely simple. By contrast, the company’s Packbot is more expensive, designed for defusing bombs, and must be teleoperated or controlled wirelessly by people.

For the full story, see:
JOHN MARKOFF. “A Reality Check for A.I.” The New York Times (Tues., MAY 26, 2015): D2.
(Note: ellipses, and bracketed expressions, added. I corrected a misspelling of “extraordinary.”)
(Note: the date of the online version of the story is MAY 25, 2015, and has the title “Relax, the Terminator Is Far Away.”)

Leave a Reply

Your email address will not be published.