Well, this is a kick in the teeth to all the AI hardware companies. This technique reduces memory size of ai models to 1/10th the size and removes the need for matmult. So no need for GPU or NPU to do ai.
https://arstechnica.com/information-technology/2024/06/researchers-upend-ai-status-quo-by-eliminating-matrix-multiplication-in-llms/
No comments:
Post a Comment