Neural Annealing: Toward a Neural Theory of Everything (crosspost) — LessWrong