Running AI models without matrix math means far less power consumption—and fewer GPUs?

  • Pennomi
    link
    English
    25 months ago

    Only for maximum efficiency. LLMs already run tolerably well on normal CPUs and this technique would make it much more efficient there as well.