A group of multidisciplinary scientists at Argonne National Laboratory, Sandia National Laboratory and the Rochester Institute of Technology have designed a neuromorphic computer chip that not only performs and adapts better than conventional chips, but does so on using a minuscule amount of power — around one watt.
Despite the many benefits of artificial intelligence, it has some limitations. AI relies on lots of data and hardware — to which it must always be connected — demands a great deal of power and has limited flexibility.
To address this problem, an Argonne-led team of scientists designed and simulated a new neuromorphic chip, which has two key components:
• Dynamic filters and weights that change the strength of various neural connections, depending on what the system finds important in real time.
• Tungsten‐aluminum oxide, an award-winning nanocomposite material created by Elam and Mane, which would allow the chip to operate at power levels far below one watt.
The new chip design is as accurate as the standard design, but it learns much more quickly and retains its accuracy — even under 60 percent error rates in its internal operation.
The design led the team to win the Best Paper Award at the 2019 Institute of Electrical and Electronics Engineers (IEEE) Computer Society’s Space Computing Conference.
Applications and Industries
Aerospace, Sensing, Urban monitoring & smart cities
Flexible and resilient computing power in extreme environments