Some challenges that are relatively easy for children, including sensorimotor tasks, turn out to be among the most difficult for AI entities.

1988

MORAVEC’S PARADOX

“If you wanted to beat Magnus Carlsen, the world chess champion, you would choose a computer,” writes journalist Larry Elliott. “If you wanted to clean the chess pieces after the game, you would choose a human being.” This is the essence of Moravec’s paradox, emphasized by several AI researchers in the 1980s, who ironically noted that the seemingly difficult tasks involving high-level reasoning were becoming easier and easier for computers to perform. At the same time, seemingly easy tasks that involve human sensorimotor skills (e.g., walking around and picking up a piece of lint from a shoe) can be quite difficult for computerized systems. The paradox is named after roboticist Hans Moravec, who wrote in his 1988 book Mind Children that “it is comparatively easy to make computers exhibit adult-level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility.”

Moravec has noted that millions of years of evolution contributed to our ability to perform some tasks nearly unconsciously, like walking and recognizing faces and voices, which are so very necessary for survival. However, abstract thought—for example, reasoning that involves mathematics and logic, like chess—is newer and more difficult for humans. However, this type of cognition is actually less difficult to engineer in AI systems. For many tasks, AI systems still need to evolve even more sensitive touch and motion controls to assist humans in roles like patient care, food service, and plumbing. As elegantly summarized by cognitive scientist Steven Pinker (b. 1954), Moravec’s paradox means that future human workers may be left with lower-paying jobs that have been around for centuries, even millennia: “The main lesson of thirty-five years of AI research is that the hard problems are easy and the easy problems are hard. The mental abilities of a four-year-old that we take for granted—recognizing a face, lifting a pencil, walking across a room, answering a question—in fact solve some of the hardest engineering problems ever conceived. . . . As the new generation of intelligent devices appears, it will be the stock analysts and petrochemical engineers and parole board members who are in danger of being replaced by machines. The gardeners, receptionists, and cooks are secure in their jobs for decades to come.”