nostr:npub1sg6plzptd64u62a878hep2kev88swjh3tw00gjsfl8f237lmu63q0uf63m
https://x.com/the_valley33005/status/1742324354183254491?s=46
wondering if it's possible to use a bitwise operation like an alphabet structure replacement for zeroing out llms and forcing the language barrier to erase more effectively. meta struggles with this issue, in india primarily. i am in discussions with the meta ai now and we think if numbers were the primary instead of being coded with english it could erase the crossover complications. 