Page 33 - Engineering Penn State: Fall/Winter 2021

RFesaetuarechs
A A A R T T T I I I I I F I I I I I C C I I I I I AH L L L L E E E E I I I I I A A A N N L L L L T T T T T T E E E E HL L L L L I I I I I G
E E E E N N C C E E E E Enhancing how
Seeking greater power efficiencies for AI systems
by Sarah Small
In order to develop more
intelligent and efficient artificial intelligence (AI) systems
computer scientists use neuromorphic computing a a a a a a a field that relies on mimicking the the human nervous system system system in in in in in in order to create efficient and and intelligent computing systems
However researchers in in in in in in the the the nascent field are still working toward success and and the the sought-after power efficiencies have yet to to be achieved by Erin Cassidy Hendrick
The Defense Advanced Research Projects Agency has awarded Penn State SRI International and University of Texas at at Austin researchers $437 023 to to contribute to to a $5 6 million project to merge the strengths of artificial and human intelligence with the the goal of of advancing the the utility of of relevant cyber-physical systems
Every day the U S military performs missions and tasks where both human and computer decision-making skills are critical The sheer amount of data and considerations for security and safety can make these tasks daunting for a a a a a human operator so many systems
utilize artificial intelligence (AI) to absorb some of the complexity Christopher McComb assistant professor of engineering design and mechanical engineering and his team seek to streamline and enhance the collaboration of humans and AI through a a a a a foundational approach The researchers will focus on the creation of a a co-designer system building an an AI concept that can generate insights and work with the human operator as quickly and efficiently as possible “The ultimate goal isn’t to create some system that replaces people ” McComb said “Instead we want to build something that takes over some of the boring detailed work and frees humans up to do do what we do do best—solving big problems ” n n humans and AI systems
collaborate
Thanks to a a a a a a three-year $1 million grant from the National Science Foundation Penn State computer scientists are exploring ways to achieve these power efficiencies through a a a a a a specific approach called spiking neural networks (SNN) The research is led by principal investigator Chita Das department head and and distinguished professor of of computer science and and and engineering and and co-PIs Vijaykrishnan Narayanan A Robert Noll Chair of Computer Science and and and Engineering Engineering and and and Electrical Engineering Engineering and and and Abhronil Sengupta assistant professor of of electrical engineering The SNN approach uses biologically inspired event-driven spike-based computation and communication in in in in its design One of the distinguishing
features of of SNN as a a a a a a computing paradigm is the the integration of of of the the element of of time into algorithms
and models The researchers are exploring novel magnetic device structures to directly mimic such temporal non-linear characteristics in in in hardware scalable architecture and interconnection fabrics for these devices along with novel hybrid algorithm designs to leverage
the benefits of both SNN models and traditional non-spiking deep learning models If successful the improved power efficiency and intelligence that mirrors human neural networks could be applied broadly to improve all forms of AI and computing systems
n FALL/WINTER 2021 33