LLM and ML generated translations generate a series of tokens individually. That’s why AI Chatbots hallucinate so often, they decide the next most likely word in a sequence is “No” when the correct answer would be “Yes” and then the rest of the prompt devolves into convincing nonsense.
stephen01king@lemmy.zip 2 months ago
How incorrect is it?
FiniteBanjo@lemmy.today 2 months ago
Sentences are a lot like math problems. An incorrect part changes the entire outcome.
stephen01king@lemmy.zip 2 months ago
Yes, so show me how incorrect is their translation, since you claim it to be incorrect.
FiniteBanjo@lemmy.today 2 months ago
LLM and ML generated translations generate a series of tokens individually. That’s why AI Chatbots hallucinate so often, they decide the next most likely word in a sequence is “No” when the correct answer would be “Yes” and then the rest of the prompt devolves into convincing nonsense.