SMARTER WAY TO LEARN MULTIPLE THINGS ONE AFTER ANOTHER

Summary from the TLDR AI newsletter

This paper proposes a system that helps models learn new tasks without forgetting the old ones.