""It's notoriously difficult to consistently measure the energy usage of AI models, but DARPA wants to put an end to that uncertainty with new "energy-aware" machine learning systems. The Mapping Machine Learning to Physics program, which opened solicitations on Tuesday, aims to do something that's simple, at least on paper. It wants to map the efficiency of various forms of machine learning directly to, as the name suggests, physics.""
""Today when we build machine learning models, we only optimize for performance, and we miss other characteristics. A very important characteristic is how much energy it's using," ML2P founding program manager Bernard McShea said in a video accompanying a press release about the program DARPA published on Wednesday. "What we want to do is consider mapping machine learning model performance to physical characteristics," McShea added. "We want to do this so we can balance model performance with the amount of resources it's taking up.""
The ML2P program will quantify machine learning energy use with precise granular measurements in joules to directly map model performance to physical energy consumption. The program aims to balance model accuracy with energy cost so systems can be optimized for battery-powered, battlefield, and edge deployments. The initiative seeks to move beyond accuracy-only optimization by measuring performance per joule and using those metrics to guide design choices. Selected performers must release documentation, algorithms, code, and tutorials under a permissive open-source license to promote transparency and reuse. The effort targets smarter, leaner, and more useful AI for constrained environments.
Read at Theregister
Unable to calculate read time
Collection
[
|
...
]