The idea that "Q*" represents a fundamental breakthrough toward the goal of "AGI" seems overblown to me, as the supposed breakthrough is limited to closed-domai

The idea that "Q*" represents a fundamental breakthrough toward the goal of "AGI" seems overblown to me, as the supposed breakthrough is limited to closed-domain reasoning problems. Defining a reward function that can be evaluated in an automated way is straightforward for narrow tasks like math or chess. But most domains lack an objective measure of quality or truth, much less one that could be evaluated without human involvement.