Is Credit Assignment Used for Learning?

The current l s t m, the current architectures are not able to capture the presumab yo, you're thinking of very long term. Whereas humans seem to be able to do credit assignment through essentaly, arbitrary times. So in which way do you think artificial neny works? The current recurrent nets are doing fairly good job for sequences with dozens or hundreds of time stabs. And then it gets to harder and harder, and depending on what you have to remember and so on as you consider longer durations.