ARLINGTON, Virginia – US military researchers are calling on industry to design a new type of artificial intelligence (AI) computer programming that allows computers not only to learn from their experiences, but also to share their experiences with other computers.
Officials at the US Defense Advanced Research Projects Agency (DARPA) in Arlington, Va., On Monday released an artificial intelligence exploration opportunity (DARPA-PA-20-02-11) for the Shared-Experience Lifelong Learning project ( ShELL).
ShELL seeks to advance computer science in lifelong learning through computers that share their experiences with each other. Lifelong learning is a relatively new area of ââmachine learning research, where computers continually learn as they encounter varying conditions and tasks when deployed in the field.
This differs from the process of training and then deploying typical machine learning systems, which often leads to unpredictable results; catastrophic forgetting of previously acquired knowledge; and the inability to perform new tasks effectively, if at all.
Related: IBM Combines Cloud Computing and Artificial Intelligence to Fuse Data into Decision Making
Current research on lifelong learning assumes an independent computer that learns from its own actions and from its environment; it did not take into account the populations of lifelong learning computers that benefit from each other’s experiences.
The total award value for the combined basic and phase two option is limited to $ 1 million per proposal.
The algorithms used for lifelong learning typically require large amounts of computing resources, including server farms, graphics processing units (GPUs), and other resource-intensive hardware, and typically do not need not to meet the limitations of communication resources.
The Shared Experiential Lifelong Learning (ShELL) program extends current lifelong learning approaches to a large number of originally identical computers. When these computers are deployed, they may encounter different input and environmental conditions, perform variations of a task, and therefore learn different lessons.
Related: Military Cybersecurity: Threats and Solutions
Other computers could benefit if one computer could share what it learned with other computers. Such sharing of experiences could reduce the amount of training required by a personal computer.
ShELL differs from approaches that reward a federation of computers for collaborating or competing on a common overall task, either breaking the task into chunks, assembling alternative approaches for the same task, or evolving specialist roles.
ShELL rewards individual computers based on their performance in their own tasks using lessons learned from their own actions combined with those learned from other computers.
Related: Deployable Simulation and Training
ShELL has three main challenges: what knowledge needs to be shared and integrated; when and how should computers share knowledge; and develop lifelong learning algorithms that take into account the size, weight, computing and communication constraints of the platforms supporting each learning computer.
DARPA researchers say they would like to award a ShELL contract by the end of September. Interested companies should upload the proposals no later than July 27, 2021 on the DARPA BAA portal at https://baa.darpa.mil.