This study will change how about AI scaling issues in deep learning - the implications for catastrophic losses and overfitting are huge! https://www.reddit.com/user/Otaku_7nfy