[10p-W810-1~9] Science of the Material Intelligence: Bringing out the Intrinsic Learning and Optimization Capabilities of Materials

[10p-W810-2] Neuromorphic Engineering for Materials

〇Tetsuya Asai¹ (1.Hokkaido Univ.)

Keywords:neuromorphic engineering, nonvolatile memory, reservoir computing

Artificial neural networks have stricken back against conventional algorithm-based computing in recent years, and hence R&D of neural device/hardware is massively driven by practical demands for realizing compact, low power, and brain-morphic (brain-like) intelligent artifacts. Trends in building AI systems are of course based on silicon CMOS and memory technologies, which results in architectural competition of power efficiency and memory-logic bandwidth, as in present Neumann-based computer systems engineering. Although present fundamental CMOS devices are definitely logic and memory devices, I here introduce a “virtual” unit, i.e., brain-morphic 3-D AI devices and materials, optimized for physical reconstruction of fundamental brain structures. The device aims at not only implementing conventional AI systems, especially in the cloud edge, but encouraging both emergence and growth of advanced neuromorphic computing/AI systems.