DARPA program: Progress seen on lifelong learning for machines
ARLINGTON, Va. The Defense Advanced Research Projects Agency (DARPA) Lifelong Learning Machines (L2M) program reports some progress toward its goal of developing computer systems that can learn continuously and become increasingly expert while performing tasks.
Researchers at DARPA partner University of Southern California (USC) have published results regarding exploration into bio-inspired artificial intelligence (AI) algorithms: In an article outlined in the March cover of Nature Machine Intelligence, L2M researcher Francisco J. Valero-Cuevas, professor of biomedical engineering and biokinesiology at USC Viterbi School of Engineering, along with USC Viterbi School of Engineering doctoral students Ali Marjaninejad, Dario Urbina-Melendez, and Brian Cohn, details the successful creation of an AI-controlled robotic limb driven by animal-like tendons capable of teaching itself a walking task, even automatically recovering from a disruption to its balance.
Propelling the USC researchers’ robotic limb is a bio-inspired algorithm that can learn a walking task on its own after only five minutes of “unstructured play”; that is, conducting random movements that enable the robot to learn its own structure and its surrounding environment.
Current machine-learning approaches rely on preprogramming a system to handle all potential scenarios, which is complex, labor-intensive, and inefficient. In contrast, the USC researchers reveal that it is possible for AI systems to learn from relevant experience, as they work at finding and adapting solutions to challenges over time.
The L2M program, initially announced in 2017, is delving into research and development of next-generation AI systems and their components, together with learning mechanisms in biological organisms capable of translation into computational processes. The L2M effort currently encompasses a large base of 30 performer groups via grants and contracts of different duration and size.