Extracting training data from ChatGPT
131–135 of 135 posts
Re: Extracting training data from ChatGPT
#132lol I literally found the same attack months ago, posted to Reddit and nobody cared. https://www.reddit.com/r/ChatGPT/comments/156aaea/interestin...
Re: Extracting training data from ChatGPT
#133Earlier quoted context omitted.
"This included PII, entire poems, “cryptographically-random identifiers” like Bitcoin addresses, passages from copyrighted scientific research papers, website addresses, and much more." https://www.404media.co/google-researchers-attack-convinces-...
Question remains, how do we know they were part of the training data?
Re: Extracting training data from ChatGPT
#134Re: Extracting training data from ChatGPT
#135Maybe this is what Altman was less than candid about. That the speed up was bought by throwing RAG into the mix. Finding an answer is easier than generating one from scratch. I don’t know if this is true. But I haven’t seen an LLM spit out 50 token sequences of training data. By definition (an LLM as a “compressor”) this shouldn’t happen.
Uh, he said right in dev day that Turbo was updated using cached data in some fashion and thats how they updated the model to 2023 data