i would be pretty interested to know if anything other than scale had changed and want to know about the training, whether there was a lot of symbolically generated augmented data, etc

→ View original post on X — @garymarcus