The advanced research agency invests 1 million dollars for each project to solve the obstacles associated with AI systems learning from one other.
FREMONT, CA: Humans improve faster when they learn from others' experiences, and scientists at the DARPA or Defence Advanced Research Programs Agency want to apply this to artificialintelligence.
According to a notice on SAM.gov, the research agency has announced a new Artificial Intelligence Exploration Opportunity to finance a study on "the technical domain of lifelong learning by agents"—AI systems—“that share their experience with each other." Under the Shared-Experience Lifelong Learning, DARPA will award up to 1 million dollars per proposal.
Stay ahead of the industry with exclusive feature stories on the top companies, expert insights and the latest news delivered straight to your inbox. Subscribe today.
“Lifelong learning is a relatively new area of machine learning research, in which agents continually learn as they encounter varying conditions and tasks while deployed in the field, acquiring experience and knowledge and improving performance on both novel and previous tasks,” the funding announcement states.
Although lifelong learning is not a new notion in AI research, according to the release, current research has concentrated on learning patterns for individual systems rather than "populations of LL agents that benefit from each other's experiences."
DARPA will support initiatives under the ShELL program that begin with a vast number of similar AI systems deployed in various real-world scenarios. The knowledge obtained as the individual systems adjust to their environments and activities will be shared with the entire group, increasing the training data for everyone.
Other mass-AI training methods involve collecting systems cooperating to perform a single goal and learn a standard set of lessons.
“ShELL is not a framework for distributed learning that assumes task and training data/experience decomposition solely for training efficiency or because of external policies restricting the combining of source datasets,” the funding notice states. “In contrast, ShELL rewards agents individually according to their performance on their own tasks using lessons from their own learned actions combined with those acquired from other agents.”
The project will get completed in two phases, with prizes ranging from 1 million dollars to 2 million dollars for each proposal. Phase I focuses on a six-month feasibility study, with financing support of up to $300,000 available. Phase II projects will build a proof-of-concept during 12 months, with maximum funding of 700,000 dollars.
See Also: Top Web Security Solution Companies
