Related question: Any thoughts as to why mathematical reasoning (Proof generation, or a trained sensitivity to mathematical beauty/quality.) still hasn’t fallen to generative models? Is it just that it’s not the kind of thing that has one core secret that we ought to have discovered by now. Is it improving at a certain slow rate similar to robotics and all we can do is wait for more heuristics to be found so that it will one day work well enough.
Oh, and are there mathematical centaurs, yet? People who, instead of striding around using intuitions about what human mathematicians (or themselves) will be able to prove if they sit down and work on it, instead navigate through intuitions about what their machines can prove? (and then prove a lot faster) I guess if machines can’t prove as much as human mathematicians yet, that would be a pretty sad existence! And the state of the art would constantly be changing and invalidating their intuitions!
Related question: Any thoughts as to why mathematical reasoning (Proof generation, or a trained sensitivity to mathematical beauty/quality.) still hasn’t fallen to generative models? Is it just that it’s not the kind of thing that has one core secret that we ought to have discovered by now. Is it improving at a certain slow rate similar to robotics and all we can do is wait for more heuristics to be found so that it will one day work well enough.
Oh, and are there mathematical centaurs, yet? People who, instead of striding around using intuitions about what human mathematicians (or themselves) will be able to prove if they sit down and work on it, instead navigate through intuitions about what their machines can prove? (and then prove a lot faster)
I guess if machines can’t prove as much as human mathematicians yet, that would be a pretty sad existence! And the state of the art would constantly be changing and invalidating their intuitions!