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 neurona

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.