Implementing brain-like systems using nano functional devices

Abstract

Despite their enormous computational powers, digital computers today are inferior to humans in such tasks, like seeing events happening in front, perceiving and recognizing them by intuition and association, and making a decision to take an immediate action. We are aiming to develop a new-paradigm computing system most suited to such human-like intelligent information processing by best utilizing the stateof-the-art silicon technology. For this end, we have developed a series of VLSI chips dedicated to specific brain-mimicking processing using digital, analog as well as mixed-signal circuit technologies. In this paper, we will discuss how nano devices could be utilized in this context. Such a brain-like system could be build by a vast-scale integration of nono functional devices having a simple correlation function arising from the resonance-type I-V characteristics along with a primitive memory function. Thanks to the majority-voting decision making principle, such a system could be made tolerable to device-level errors, thus presenting a promising opportunity for system applications of nano devices in the beyond CMOS era.