1-bit LLMs could solve AI’s energy demands: « “Imprecise” language models are smaller, speedier—and nearly as accurate. »

submitted by /u/fchung
[link] [comments]