Cleo Nardo comments on Shortform

[Cleo Nardo]

I think that SGD isn’t sample efficient enough to solve continual learning