It’s a > ~80% match with my models, which I have independently derived from the mechanisms of individual human cognition (that also can be thought of as having two components that work in an adversarial setup), so I guess that counts as additional evidence for both of us.
I guess positive feedback should also have a direction, and in this case I would spend most of my norm on reinforcing the way of thinking that you used to generate this post: noticing a deep correspondence between abstract objects in different domains, and drawing useful conclusions from it.
Hint: if you make a single connection ML social norms, you have two ways to draw useful conclusions. If you add human cognition to the mix, you form a triangle with three bidirectional correspondences, and that gives you six ways to generate useful cross-domain insights. [This generalizes to thinking efficiently about anything.]
Excellent post, congratulations!
It’s a > ~80% match with my models, which I have independently derived from the mechanisms of individual human cognition (that also can be thought of as having two components that work in an adversarial setup), so I guess that counts as additional evidence for both of us.
I guess positive feedback should also have a direction, and in this case I would spend most of my norm on reinforcing the way of thinking that you used to generate this post: noticing a deep correspondence between abstract objects in different domains, and drawing useful conclusions from it.
Hint: if you make a single connection ML social norms, you have two ways to draw useful conclusions. If you add human cognition to the mix, you form a triangle with three bidirectional correspondences, and that gives you six ways to generate useful cross-domain insights. [This generalizes to thinking efficiently about anything.]