You are right; I should have been more specific. RL does forget in the simplest sense, i.e. that certain weights in your model drift away if the data distributi

You are right; I should have been more specific. RL does forget in the simplest sense, i.e. that certain weights in your model drift away if the data distribution is non-stationary. Humans seem to be a bit more targeted in their forgetting.