7 ms·
>OpenAI doesn’t disclose what data it used to train GPT-4, its most advanced large language model, but lawyers for Sancton say ChatGPT divulged the secret. “In
by Gregaros 3y ago
>OpenAI doesn’t disclose what data it used to train GPT-4, its most advanced large language model, but lawyers for Sancton say ChatGPT divulged the secret. “In the early days after its release, however, ChatGPT, in response to an inquiry, confirmed: “Yes, Julian Sancton’s book ‘Madhouse at the End of the Earth’ is included in my training data,” the lawsuit reads.
This is evidence of absolutely nothing as GPT cannot reflect over what is or is not in it’s training set. I’m shocked this is reported uncritically - here the impartiality of ‘just the facts’ distorts more than it enlightens.
- jacquesm 3y agoAgreed that it is evidence of nothing but it is a copyright violation and should be dealt with accordingly. After all, the writer and rights holders did not agree to the making of the copy that placed the book in the training set.
- Gregaros 3y agoWe don’t know (and they don’t either) that the text of this book appeared in the training set.
- polski-g 3y agoThey do agree to sell books. If I can write a commercial book after reading hers, using her book to train my creativity, it's legal. The legality doesn't go away because I write software to train it's creativity. It's not copying. If you asked GPT to recite page 27, it couldn't.
- ratg13 3y agoAgreed that your quote is evidence of nothing, but it is absolutely possible to ask the AI a number of questions to determine if it has seen the source material or not. Something like this: https://www.reddit.com/r/ChatGPT/comments/17prrqe/who_needs_sonic_when_youve_got_originality/ https://www.reddit.com/r/ChatGPT/comments/17prrqe/who_needs_... Have you read the whole lawsuit to determine that they did not dive deeper?
- cyanydeez 3y agothat's bs. you absolutely can prove things are in there.
- daynthelife 3y agoYes, but not by directly asking it.