Officials of the U.S. Defense Advanced research Projects Agency (DARPA) have issued an artificial intelligence exploration opportunity for the Shared-Experience Lifelong Learning (ShELL) project. ShELL will be a new kind of artificial intelligence computer programming which will enable computers to learn from their experiences, and also share their experiences with other computers.
Typical machine learning systems use the train-then-deploy process, which can result in unpredictable outcomes; catastrophic forgetting of previously learned knowledge; and the inability to execute new tasks effectively, if at all. Lifelong learning seeks to have computers continually learning as they encounter various conditions and tasks while deployed in the field. Current lifelong learning research assumes one independent computer that learns from its own actions and surroundings, but has not yet approached using multiple lifelong learning computers that can benefit from each other’s experiences.
The ShELL program will advance current lifelong learning approaches to large numbers of originally identical computers. When these computers are deployed they may encounter different input and environmental conditions and execute variants of a task, therefore learning different lessons. They will then be expected to share this new knowledge with each other.
ShELL has three core challenges:
- what knowledge should be shared and incorporated;
- when and how should computers share their knowledge; and,
- developing lifelong learning algorithms that account for the size, weight, computing, and communications constraints of the platforms supporting each learning computer.
The total award value for the combined phase-one base and phase-two option is limited to $1 million US per proposal. DARPA intends to award a ShELL contract by late September of this year. Companies interested should upload proposals no later than 27 July 2021 to the DARPA BAA portal at https://baa.darpa.mil.