[ahartmetz]

I'd be very surprised if biological brains couldn't do better than fancy gradient descent, which is what the current standard seems to be for artificial neuronal networks.