logoalt Hacker News

pbhjpbhjyesterday at 6:15 PM0 repliesview on HN

Aye, but do they train on user data in these circumstances or not? If they do, then almost certainly the model was influenced by the input of the allegedly plagiarised material.