Hacker News new | past | comments | ask | show | jobs | submit
They’re probably training on outputs of existing models.
loading story #42774243
loading story #42772928