Johns Hopkins electrical and computer engineers are pioneering a new approach to creating neural network chips—neuromorphic accelerators that could power energy-efficient, real-time machine intelligence for next-generation embodied systems like autonomous vehicles and robots.