Comment by sieongioetnio

I don't think it's obviously solvable. All current approaches are plainly incapable of introspection. These GPTs don't understand their own "minds" half as well as we understand them, and we don't understand them very well.