Leaders at the Defense Advanced Research Projects Agency announced Sept. 7 a $2 billion initiative, known as AI Next, that aims to substantially improve artificial intelligence research.
“With AI Next, we are making multiple research investments aimed at transforming computers from specialized tools to partners in problem-solving,” Steven Walker, the agency’s director, said in a press release. “Today, machines lack contextual reasoning capabilities, and their training must cover every eventuality, which is not only costly, but ultimately impossible. We want to explore how machines can acquire human-like communication and reasoning capabilities, with the ability to recognize new situations and environments and adapt to them.”
As a result of the effort, DARPA leaders envision improving day-to-day processes, such as vetting for security clearances in a week or accrediting software systems for operational deployment in a day or less. DARPA said it plans to announce multiple broad agency announcements in the next year.
The announcement marks the second major artificial intelligence-related initiative in the last three months. In late June, the Department of Defense announced its Joint AI Center, which is expected to help provide a road map for the department’s nearly 600 AI-related programs, including the controversial Project Maven program.
The funding is a parallel effort to the Department of Defense’s Joint AI Center, known as JAIC, which is expected to cost about $1.7 billion in the coming years. That program is run through the DOD CIO’s office and is expected to focus more on applications for artificial intelligence while the DARPA programs will concentrate on answering questions related to the “foundational science” related to AI.
DARPA has a history of working on AI project. It is currently funding about 20 AI related programs and said its initial funding on the topic dates back to the 1960s.
Leaders describe the new investment as working toward what they call the third wave of AI. The first wave focused on rule-based systems for narrow tasks and a second wave has been used to find statistical patterns in large data sets. But DARPA leaders are describing a third wave as one of “contextual adaptation.”
In a press conference at the conclusion of DARPA’s D60 symposium, agency leaders said the new initiative would include a focus on explainable AI, in other words, how a machine arrived at its answer. They also said they expect a broader discussion of ethics related to artificial intelligence.