If LLMs predict the next word, Why is there NO keyboard that integrates one??
Just why?? It’s gonna make typing wayyy easier, and now with the smaller models it won’t even use many tokens?? All the input methods are terrible and always autocorrect last names and, as I also type Chinese, give horrible suggestions of words with the same spelling but different tones. Just why hasn’t anyone done this yet? Is there anything that makes this idea a bad one?
(You can totally run those smaller models locally and who cares about privacy anyways with companies like meta)