The Defense Advanced Research Programs Agency (DARPA) wants artificial intelligence models with lifelong learning capabilities from other AI systems’ experience. DARPA is putting up $1 million per project under Shared-Experience Lifelong Learning.
Humans advance quicker and better when learned from other’s experiences, researchers at DARPA want to impose this into artificial intelligence models. DARPA inaugurated a new artificial intelligence exploration opportunity to pursue work on “the technical domain of lifelong learning by agents” in AI, according to the SAM.gov website.
The concept of lifelong learning is not new in AI research. According to the announcement, the ongoing research is focused on learning from the same system rather than populations of LL agents that know each other’s experiences. Shared-Experience Lifelong Learning (ShELL) program extends LL approaches to large numbers of originally identical agents for shared learning. Sharing of experiences could significantly reduce the amount of training required by individual agents.
Read more: Google Announces New Cloud TPU VMs For AI Workloads
The selected proposals should be completed over two phases, Phase I, which is of 6-months, the first 2 months for a detailed research plan, month 4 for preliminary design, and month 6 on the feasibility study of the model with a funding of $300,000. The shortlisted projects from Phase II will develop a proof-of-concept over 12 months with a $700,000 budget.
The proposers have to address three major sections in their bids; Content, a brief description of shareable and non-sharable knowledge and knowledge that needs to be incorporated, and the one that needs to be ignored by SHell agents. Communications: When and how the knowledge should be shared. Computation: Ensure that the LL algorithms have enough computing power through a mix of edge and cloud resources.
The proposals must contain the mentioned sections and should be uploaded in DARPA’s BAA Portal. DARPA will acknowledge receipt of complete submissions via email and assign identifying numbers that should be used in all further correspondence regarding those submissions. The awarding will be done on 24th September.
