Program Overview

Sponsored by the Air Force Office of Sponsored Research

Program Monitor Dr. Ali Sayir

History and Mission Statement

CEREBRAL began in November 2017 and addresses the fundamental limitations of neuromorphic computing so as to expand machine learning (sometimes called artificial intelligence) using inherent behaviors naturally encoded inside adaptive oxide materials. CEREBRAL’s mission statement is to “enable neuromorphic technologies by understanding and manipulating the atomistic world”.

Background

Presently, the majority of machine learning implementations are performed in silicon circuitry or in software. With remarkable advances in these fields, engineers have only managed to achieve the neural density equal to about half a fruit fly and at a tremendous power and size requirements. CEREBRAL seeks more neural density than is presently imaginable with traditional means, Therefore CEREBRAL is pursuing a new approach and a new way of miniaturizing the fundamental building blocks for neural computing. With this in mind, CERBRAL is examining the fundamental science of adaptive oxides, materials that can naturally reconfigure themselves based on the history of voltage, current, heat, and light that has acted upon the material. This unique property establishes a record of past events, a memory that allows computation is a new form that emulates the way a human brain functions, i.e. neuromorphic computing. Unfortunately, much is still unknown about these remarkable adaptive oxide materials. This discovery is at the heart of CERBRAL’s goals. The question addressed is bold. The science—including the materials used, the exploratory tools developed and the architectures and “intelligence” algorithms applied—are new. CEREBRAL is ambitious but hopes to lay scientific foundations for the way your children's computers will be constructed.

Program Overview