A recent research from MIT Lincoln Laboratory and Northeastern University has investigated the savings that can be made by power capping GPUs used in model training and inference and several different methods to reduce AI energy use in light of growing concern over huge machine learning models’ energy demands. Power