Brain-Like Computing Objective of the new initiative
As society adapts to become information-centric, conventional computing is approaching its fundamental limits. The need for computing to become faster and more energy efficient is growing exponentially.
Texas A&M University is set to lead its first Energy Frontiers Research Center (EFRC) funded by the Department of Energy (DOE). The center will focus on Reconfigurable Electronic Materials Inspired by Nonlinear Neural Dynamics (REMIND), an initiative that strives to transform computing to behave more like a human brain for fast and efficient processing.
R. Stanley Williams, Professor in the Department of Electrical and Computer Engineering, will be the director of the EFRC, and Sarbajit Banerjee, a professor in the Department of Materials Science and Engineering and the Department of Chemistry, will be the associate director.
“We are at a crossroads for the future of computing,” Banerjee said. “Self-driving cars, networked networks and personalized medicine are on the rise, which require huge amounts of energy. A whole new approach to brain-like computing is essential to meet the needs of society. »
Modern computers excel in various essential functions, such as high-precision arithmetic and solving known equations. However, they perform poorly when it comes to natural human abilities such as real-time learning, concept identification, and decision making.
This ability to process information is possible because the human brain has nerve cells (neurons) that constantly compare incoming stimuli with previously learned data. Neurons communicate with each other via electrical and chemical signals through connections called synapses that store memories. Although the individual biological stages are slow compared to those of transistors, an enormous number of them operate simultaneously to perform sophisticated calculations with energy-efficient orders of magnitude greater than the most advanced electronic computers.
“Let’s say we’re looking at a picture of a dog,” Banerjee said. “A human brain can almost immediately recognize the dog itself, its type and its relative age. A computer will struggle with basic recognition and can make a big mistake while using a lot more energy to do it.
Researchers involved in the REMIND initiative are discovering ways to emulate human neurons and synapses in electrical circuits by designing, creating, and assembling materials that exhibit tunable nonlinear responses to incoming electrical signals, such as thresholding, l amplification, integration and built-in memory. In other words, they mimic the processing system of the human brain and attempt to assemble it into a highly efficient and capable computer.
“Our center seeks to uncover the basic science of artificial neurons and synapses,” Williams said. “We look forward to solving a generational challenge that will transform the future of computing.”
If successful in implementing these techniques, their findings could significantly reduce the energy consumption used by computers, helping to tackle the energy crisis.
“We are focused on moving computers from computing mathematical functions to learning and making decisions in uncertain and changing environments,” Banerjee said. “We are discovering the fundamental chemistries and materials to fabricate the next generation of brain-like computing.”
The DOE recently announced the EFRC awards for developing technologies capable of transforming energy production and reducing harmful emissions. Research efforts will have input from 54 universities and 11 national laboratories in 34 states.
The EFRC is a collaboration between the College of Engineering, Department of Chemistry, Texas A&M Engineering Experiment Station, National Renewable Energy Laboratory, Lawrence Berkeley National Laboratory, and Sandia National Laboratories. In addition to Banerjee and Williams, REMIND researchers include Raymundo Arroyave, Matt Pharr, Xiaofeng Qian and Patrick Shamberger from the Department of Materials Science and Engineering; Perla Balbuena of the Artie McFerrin Department of Chemical Engineering; and Marcetta Darensbourg and Kim Dunbar from the Department of Chemistry. The team represents the breadth and depth of expertise needed to tackle this complex challenge spanning multiple disciplines.