☆ Yσɠƚԋσʂ ☆@lemmy.ml to Memes@lemmy.mlEnglish · 1 month agoBased if truelemmy.mlimagemessage-square10linkfedilinkarrow-up1107arrow-down16
arrow-up1101arrow-down1imageBased if truelemmy.ml☆ Yσɠƚԋσʂ ☆@lemmy.ml to Memes@lemmy.mlEnglish · 1 month agomessage-square10linkfedilink
minus-squareExotiqueMatter@lemmygrad.mllinkfedilinkarrow-up6·1 month agoAlmost surely false. I seriously doubt it’s possible to train a modern multy-billions parameters LLM with less than 2 dozen million prompts, even if it’s 16M each, let alone if it’s 16M combined.
minus-square☆ Yσɠƚԋσʂ ☆@lemmy.mlOPlinkfedilinkarrow-up5·1 month agoyeah I’m very skeptical here as well
Almost surely false. I seriously doubt it’s possible to train a modern multy-billions parameters LLM with less than 2 dozen million prompts, even if it’s 16M each, let alone if it’s 16M combined.
yeah I’m very skeptical here as well