A user disputes complexity-theory arguments that “true” AI is impossible

The user objects to using computational complexity to declare “true” AI impossible. Their example is learning a linear classifier to minimize classification error: they say it is computationally hard even to distinguish whether 51% or 99% accuracy is achievable. The cited paper in SIAM Journal on Computing describes a key contrast: the noise-free problem can be solved efficiently, while even a tiny amount of worst-case noise can make it intractable.