The authors combine biologically-inspired learning techniques with neuromorphic hardware to implement an energy-efficient system that demonstrates rapid learning of unseen tasks across various domains.