By Chris Adams
Through the Defense Advanced Research Projects Agency, researchers are seeking to bring a third wave to artificial intelligence.
“We’re looking to take computers as tools to computers as partners,” said Peter Highnam, DARPA’s deputy director.
In a session with National Press Foundation fellows in the Virginia suburbs of Washington, DARPA officials and experts laid out the agency’s vision and priorities.
Right now, DARPA is at front end of a $2 billion artificial intelligence program that seeks to fund projects that could make it possible for machines to adapt to changing situations. Called the “AI Next” campaign, the multi-year program will fund both existing programs and new ones.
DARPA doesn’t have its own laboratories but instead funds work at universities and private-sector labs. Seven DARPA leaders and program managers detailed the history of AI and what may come next.
The first wave of AI, according to John Everett, deputy director of DARPA’s Information Innovation Office, was symbolic AI – essentially a long series of if/then questions that led a system to a conclusion. Think of a tax program such as TurboTax. “After a while, these systems become brittle and they don’t handle surprises too well,” Everett said.
The second wave moved beyond that into the development of neural networks and what is called “deep learning” – using far more and computational power to teach computers to learn by example. An example would be facial-recognition technology.
The third wave seeks to build on that, addressing the limitations of first and second wave technologies by making it possible for machines to contextually adapt to changing situations.
Consider trying to teach a computer common sense. Think about the issues that would have to be overcome if you told a domestic robot to “pick up the living room.” What does that even mean? Humans understand the nuance involved in such a question. But right now, a computer doesn’t.
“We’re looking at developing a program that can understand common sense,” Everett said.