You said: "training NNs (for any purpose) using evolution works well". I gave you an example of a purpose where it does not work well. So, let me ask you again

You said: "training NNs (for any purpose) using evolution works well". I gave you an example of a purpose where it does not work well. So, let me ask you again: can you give an example of evolutionary methods that work well when applied to training NNs, other than this breakthrough by OpenAI, which only works for RL?