DARPA tech chief envisions the future
- By Frank Tiboni
- Apr 19, 2004
Ron Brachman's curiosity about robots programmed to think on their own dates back to his childhood in New Jersey. It was the 1960s, "Star Trek" first appeared on television and putting a man on the moon became a remarkable reality.
When he was in high school, Brachman and his father, an electrical engineer, built a robot equipped with a light and photo sensor that allowed it to follow a taped line on the floor of his basement, even in darkness. He became a science-fiction aficionado, watching Mr. Spock and "The Twilight Zone."
As an undergraduate at Princeton University, Brachman developed more academic hobbies as he studied logic, mathematics and computer science. In doctoral work at Harvard University, he added his love of linguistics to his research on building a computer capable of reasoning — his lifelong pursuit.
"There's a lot to try to understand about the kind of reasoning and learning that people do that's very different than sort of classical, [mathematician John] von Neumann, step-by-step computer algorithms," Brachman said. "So I just got extremely interested in this."
Brachman later earned a reputation as a world-renowned expert in artificial intelligence while working as a researcher and executive at AT&T Bell Laboratories. During his 15 years there, he helped build the CLASSIC and PROSE systems that use artificial intelligence to speed up the processing and delivery of equipment for companies.
Now Brachman works at the Defense Advanced Research Projects Agency as director of its Information Processing Technology Office, where he oversees programs that study and develop cognitive computing.
He wants to solve the same problem he pondered as a teenager watching "Star Trek" — how to get people and computers to collaborate. Military officials think robots, with their superior memory, can aid generals in command and control centers, Brachman said.
"My sense of what it takes to put together a cognitive agent that is successful, like a really good executive assistant, is that you just don't put all these [technologies] in a pot and stir and hope that it all adds up," he said.
He will rely on his boyhood aspirations, research knowledge and industry management experience to build the Defense Department's version of Lt. Cmdr. Data, the intuitive but emotionless robotic officer in "Star Trek: The Next Generation." Brachman's team will take an eclectic approach to building a robot similar to Data.
"The challenge we have asked people to look at is how do we put all of these pieces together," Brachman said. "Maybe we don't need the world's best computer vision or speech-understanding technology. But what would happen if they both work together?"
DARPA spends $29 million per year on its Perceptive Assistant That Learns program, which develops robots programmed to think. The agency and its contractors will first develop an architecture that considers reasoning, learning, perception, language and action for robots, he said.
"One of the things that's amazing about humans is our ability to sort of focus on problems at hand — to use experiences when necessary, to do an analytic kind of reasoning and deduction when necessary, to draw sketches," Brachman said. "We bring these things to bear at the right time somehow — almost magically, if you will."
If anyone can make this kind of project a reality, Brachman can, said Eric Sumner, a colleague and friend for 14 years.
"He's energetic, positive and obviously very bright," Sumner said. "He's good at judging talent, bringing [people] together and making them better with his technical judgment — and his energy." Sumner worked with Brachman at AT&T Bell Laboratories. He is now a partner at Open Data Partners LLC, a firm that helps companies use data.
Sumner believes DARPA can build a thinking robot by 2030, but he thinks the agency will likely field elements of the robot's artificial intelligence, such as reasoning and speech recognition, sooner.
Nurtured by computer and science-fiction fantasies, Brachman's imagination will help turn artificial intelligence possibilities into reality, Sumner said.
"He can see into the future," Sumner said. "Then he sets a vision that people can go toward."