‘pacman,’ Other Arcade Games Inspire Energy Efficiency In Nanosystems For Machine Learning

In a study published by Berkeley Lab, Stephen Whitelam from Lawrence Berkeley National Laboratory of the Department of Energy used neural networks in computer simulations to train nanosystems to work with greater energy efficiency. The research was conducted at the Molecular Foundry, the DOE Office of Science user facility at Berkeley Lab. The simulations reveal that learned protocols can draw heat from the systems by constant measurement to identify the most efficient operations. This discovery applies to operating large systems such as computer data centers and is based on the idea of the retro arcade game Pac-Man. The simulations developed by Whitelam also seem to answer the thought experiment in physics with a hypothetical intelligent being called Maxwell’s Demon.