Samsung Advanced Institute of Technology

Reverse-engineering the brain by mimicking the structure and function of its neuronal network on a silicon integrated circuit was the original goal of neuromorphic engineering and is still the ultimate goal of the field. Unfortunately, the architecture of the biological neuronal networks responsible for the brain’s higher function still remains one of the greatest scientific mysteries, making the original pursuit difficult. As a result, the focus of neuromorphic engineering has been expanded from the rigorous brain mimicry to brain-inspired design that embodies some qualitative features of the brain, such as asynchronous, event-driven signaling and in-memory information processing. In this Perspective, we will articulate our vision of bringing the neuromorphic electronics paradigm back to its original goal of brain reverse-engineering. The essence of our vision is to copy the functional synaptic connectivity map of a mammalian neuronal network by leveraging cutting-edge advances in neuroscience tools and to paste this map onto a state-of-the-art ultrahigh density three-dimensional (3D) network of solid-state memories. This “copying and pasting” of the biological synaptic organization may bring us a step closer to creating silicon integrated circuits that approximate the brain’s unique computing abilities.