The Defense Advanced Research Projects Agency wants to start a new program that will take machine learning to a new level: to help computers develop common sense.
The agency that takes on some of the Department of Defense’s hardest problems asked for $6.2 million for the program in its fiscal 2019 budget request.
“Recent advances in machine learning have resulted in exciting new artificial intelligence (AI) capabilities in areas such as image recognition, natural language processing, and two-person strategy games (Chess, Go),” according to the document. “But in all of these application domains, the machine reasoning is narrow and highly specialized; broad, commonsense reasoning by machines remains elusive.”
The request details how humans and machines will cultivate a symbiotic relationship in which humans teach machines to learn and reason.
“The program will create more human-like knowledge representations, for example, perceptually-grounded representations, to enable commonsense reasoning by machines about the physical world and spatio-temporal phenomena,” the document reads.
The problem of learning common sense has long been one of the most difficult hurdles facing AI researchers. In order for a machine to be considered truly intelligent, it must be able to reason using the broad scope of information that humans are expected to possess.
For example, imagine an autonomous robot programmed to be a waiter at a cocktail party. When the robot sees it has picked up a broken glass, common sense reasoning tells it not to pour liquid into the glass, but instead pick up a different one.
The potential list of applications for a system equipped with with general common sense reasoning is endless. This prospect of general AI, equipped with superhuman level computing power and learning capabilities, prompted tech mogul Elon Musk to warn that humanity risks “summoning a demon” if we’re not careful. Others, such as Stephen Hawking and Bill Gates, have echoed Musk’s sentiment.
Many Defense Department leaders, however, have taken a different view and embraced the prospect of machine learning technology as not only inevitable, but a force for good. Former Deputy Defense Secretary, Bob Work, touted machine learning technology as a necessary innovation that will allow machines to help humans make decisions faster.
Work’s Third Offset strategy called for the Pentagon to embrace advances in artificial intelligence technology and encouraged human-machine teaming on the battlefield.