Admittedly backpropogation/gradient descent bother me a fair bit but in a human brain what is desireable still has some kind of encoding, the mechanism is just more complicated and more general.
What if I tell you it is? Will that information back-propagate into your brain?
I'd ask you to show me how and where backpropogation takes place in the brain. Then I'd ask how thin you're willing to stretch the "neural net" metaphor to re-apply it to biology.