Hacker News new | ask | show | jobs
by astrange 42 days ago
Because cheating is easier than actually doing work, if you use this to train future models, it's likely you'll end up with cheating instead of actual generalization.