UnknownEssence

joined 1 year ago

[R] "It's not just memorizing the training data" they said: Scalable Extraction of Training Data from (Production) Language Models in c/machinelearning@academy.garden

[–] UnknownEssence@alien.top 1 points 11 months ago

If it is truly memorizing the ENTIRE set of training data, then is it not lossless data compression that is much more efficient than any known compression algorithms?

It has to be lossy compression aka it doesn’t remember its ENTIRE set of training data, word for word.

permalink
fedilink
source

[P] Hey does anyone want to make a million gazillion on a plug-in that filters out ai generated content? in c/machinelearning@academy.garden

[–] UnknownEssence@alien.top 1 points 1 year ago (1 children)

Good luck detecting which content is AI generated. It can’t be done.

permalink
fedilink
source