@indif4ent yes. one very consistent effect, for instance, is what i've called 'cowardice' - the model learns to not even attempt things it's less likely to do perfectly. which is maybe good if all…

Posters: repligate. repligate

2 likes recorded by the source corpus.