Enabling autonomous systems - like driverless cars - to learn and adapt without human supervision is the aim of a new multi-site research endeavor drawing on the expertise of UCI cognitive scientists Jeffrey Krichmar and Emre Neftci. As a part of HRL Laboratories’ Super Turing Evolving Lifelong Learning Architecture (STELLAR) project, the UCI researchers will explore biologically inspired machine learning methods to make transformative improvements to Artificial Intelligence.  If successful, the breakthrough in machine-learning architectures will be able to rapidly adapt to unforeseen situations, remember and learn from each experience, and consolidate new tasks with previous ones. This would represent significant improvement over current machine learning systems, which forget old tasks when learning new ones, and are not capable of learning online and responding to new situations not presented during offline training.

“Unlike current AI systems which are brittle, animals are able to learn rapidly, learn over a lifetime, and not forget what has been learned before. We plan to incorporate how biological organisms evolve and learn into the next generation of AI,” says Krichmar who directs the Cognitive Anteater Robotics Laboratory at UCI.

“Brain inspiration in AI can have a profound impact in making them robust to the dynamical nature of real-world problems. We strive to bridge machine learning and neuroscience to reproduce the underpinnings of dynamical learning in AI,” says Neftci, director of the Neuromorphic Machine Intelligence Lab at UCI.

In addition to UCI, the HRL team led by Dr. Praveen Pilly includes collaborators at Stanford University (Prof. James McClelland), University of Texas Austin (Prof. Risto Miikkulainen), Loughborough University (Prof. Andrea Soltoggio), IT University of Copenhagen (Prof. Sebastian Risi), and The French Institute for Research in Computer Science and Automation (INRIA, Prof. Jean-Baptiste Mouret) with world-leading expertise in neuroevolution, brain-inspired learning and memory models, neuromodulation, and autonomous systems.

“This is truly an all-star team and we are thrilled to be a part of this effort,” says Krichmar.

UCI’s portion of the research is funded for two years with $250,069 by the Lifelong Learning Machines (L2M) program, which is led by Hava Siegelmann from the Defense Advanced Research Project Agency (DARPA). If performance metrics are met, phase two funding for the UCI team comes with a potential $379,495 in grant funding.

This material is based upon work supported by the United States Air Force and DARPA under contract number FA8750-18-C-0103.  Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the United States Air Force and DARPA.