Self-learning neuromorphic circuits of high-energy efficiency
Abstract
Computers have led to the development of artificial intelligent (AI) systems that simulate the intelligent functions of the human brain. However, these computers are more than five orders of magnitude less energy-efficient than the brain. Although these autonomous AI systems excel over humans in predictable stable environments within specific learning domains such as Go games and self-driving cars in ordinary road conditions, their performance is significantly inferior to that of humans in uncertain and mutable environments different from their static learning domains, such as operating self-driving cars and robotic systems in erratic environments. This makes the existing computers inadequate for next-generation autonomous systems, such as self-driving cars, humanoids, and self-piloted unmanned aerial and space vehicles, which require low-power consumption, real-time fast learning, and adaptability in mutable environments. To address these challenges, this proposed project aims to develop self-learning neuromorphic integrated circuits (SNICs) by emulating neurobiological circuits with high-energy-efficiency, low power consumption, real-time fast learning, and adaptability in complex mutable real-world.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 07, 2024
- Source ID
- FA95502310638
Entities
People
- Yong Chen
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of California, Los Angeles