Heuristic about the what science can tell us about systems that we learn and that learn us back. I’m writing about this elsewhere and will dump the results in when I have finished with that.
What the success of machine learning tells us about the structure of the world. Non-arbitrage versus hypothesis tests. Adversarial versus random noise. Economic systems. The bit before attaining equilibrium. Homogenaeity versus local regularities. Our tendency to look for universals.
Awaiting filing
Scale effects in economics are a subject of the 2019 Nobel Memorial economics prize. Internal versus external validity in economics. Daniel Lakens reviews (Yarkoni 2019) on generalisability in psychology.
Stylised comparison
Static world | Adaptive world | |
---|---|---|
People are… | the objects | the subjects |
Individual behaviour emerges from… | fixed personalities that we can discover | networks and group dynamics |
Society is… | a system with static rules | a game with rules made up by players as they go along |
We should… | learn the truth and act upon it | guess a better next move than the other actors |
Learning means… | accepting a hypothesis based on P-value | betting on the best portfolio of actions |
As pertains to AI
Embedded agency and other framings of the Machine Intelligence Research Institute taxonomise some related ideas from the perspective of building intelligences that interact with the world which seem to interact with these.
No comments yet. Why not leave one?