You aren’t trying to develop a new strategy, you are trying to select from one or more already existing strategies based on which one is optimal. You do this by finding where the biggest proven payoff is and then emulating the behavior among your potential strategies that results in the highest payoff. It doesn’t generalize well because you ask PA+i what your partner will do against each of the other potential strategies, but only emulate those strategies if PA+i proves that they also would do that against you. In cases where the complexity of the other decision maker were significant, Masquerade would always fall back on its default- try writing a few RPS programs, that need to throw Rock, Paper Scissors against each other and have access to each other’s source code but not random numbers. Payoffs for a win are 3-0, for a tie 2-2.
M(M)=C IFF PA+i proves that M(M)=C. PA+i proves that M(M)=C IFF (PA proves M(M)=C or PA+i proves M(M)≠D)
M(M)=C IFF PA proves that M(M)=C or PA proves M(M)≠D.
I’m probably missing some magic property of PA+i in that simplification, but otherwise it’s the last three lines of your program, given that the sanity check is checking for C,C.
Thank you. You have finally proven to me that Lob’s theorem is as absurd as it sounds on the face of it. If PA+i can prove “If
You have finally proven to me that Lob’s theorem is as absurd as it sounds on the face of it.
When math seems absurd to you, this is generally a signal that you’ve got something interesting to learn, rather than that you’ve got a superior epistemic vantage point to mathematics.
No, seriously, this stuff is awesome and you should check it out. I’m realizing that I need a better grasp of it in order to write it up, so I’m checking out The Logic of Provability on Monday.
Oh, I agree that it’s true in mathematics. I just recognize that things which are proven within a strict construction of mathematics might not apply outside that strict interpretation.
It’s the intersection between Godel and Lob’s theorems that really throws me off- anything which is not false in PA can be proven in PA, including nonce statements, but there are things that are true in PA which cannot be proven in PA, and there is a mechanism for generating such statements.
Consider the statement G: “G cannot be proven in PA”. G is proven in PA+1, because if PA were to have a proof of G, then G would be false and PA would prove something which is false. So far, so good- but PA proves that PA+1 proves G; since the only axiom in PA+1 not in PA is “PA is sound”, PA proves that “PA is sound->G is true”.
If PA is not sound, then PA proves G is true. (trivial, since an unsound system proves everything)
PA is either (sound or not sound).
PA proves (if PA is sound, G is true.)
PA proves that (if PA is not sound, PA proves that G is true)
(PA proves that G is true) within PA implies G is true.
PA proves that G is true If PA is (sound or unsound).
PA proves that G is true.
PA proves that PA is not sound.
I’m not sure if I’m literally only using the axiom “S is either sound or not sound” to prove “S proves that S is not sound”. I’m sure that the principle of explosion requires a few axioms more, and that I’ve made a (common?) fatal error somewhere, but I’m comfortable with calling mathematical constructions “Absurd” without calling them false.
The best response to “X is absurd” is “Reality is absurd”, a point with which I will readily agree.
EDIT: Corrected a couple of major typos resulting in GCE, other errors are probably still present.
Oh, I see the premise now.
You aren’t trying to develop a new strategy, you are trying to select from one or more already existing strategies based on which one is optimal. You do this by finding where the biggest proven payoff is and then emulating the behavior among your potential strategies that results in the highest payoff. It doesn’t generalize well because you ask PA+i what your partner will do against each of the other potential strategies, but only emulate those strategies if PA+i proves that they also would do that against you. In cases where the complexity of the other decision maker were significant, Masquerade would always fall back on its default- try writing a few RPS programs, that need to throw Rock, Paper Scissors against each other and have access to each other’s source code but not random numbers. Payoffs for a win are 3-0, for a tie 2-2.
M(M)=C IFF PA+i proves that M(M)=C. PA+i proves that M(M)=C IFF (PA proves M(M)=C or PA+i proves M(M)≠D)
PA+i proves M(M)≠D IFF (PA proves M(M)≠D or PA+i proves M(M)=C)
M(M)=C IFF PA proves that M(M)=C or PA proves M(M)≠D.
I’m probably missing some magic property of PA+i in that simplification, but otherwise it’s the last three lines of your program, given that the sanity check is checking for C,C.
Thank you. You have finally proven to me that Lob’s theorem is as absurd as it sounds on the face of it. If PA+i can prove “If
When math seems absurd to you, this is generally a signal that you’ve got something interesting to learn, rather than that you’ve got a superior epistemic vantage point to mathematics.
No, seriously, this stuff is awesome and you should check it out. I’m realizing that I need a better grasp of it in order to write it up, so I’m checking out The Logic of Provability on Monday.
Oh, I agree that it’s true in mathematics. I just recognize that things which are proven within a strict construction of mathematics might not apply outside that strict interpretation.
It’s the intersection between Godel and Lob’s theorems that really throws me off- anything which is not false in PA can be proven in PA, including nonce statements, but there are things that are true in PA which cannot be proven in PA, and there is a mechanism for generating such statements.
Consider the statement G: “G cannot be proven in PA”. G is proven in PA+1, because if PA were to have a proof of G, then G would be false and PA would prove something which is false. So far, so good- but PA proves that PA+1 proves G; since the only axiom in PA+1 not in PA is “PA is sound”, PA proves that “PA is sound->G is true”.
If PA is not sound, then PA proves G is true. (trivial, since an unsound system proves everything)
PA is either (sound or not sound).
PA proves (if PA is sound, G is true.)
PA proves that (if PA is not sound, PA proves that G is true)
(PA proves that G is true) within PA implies G is true.
PA proves that G is true If PA is (sound or unsound).
PA proves that G is true.
PA proves that PA is not sound.
I’m not sure if I’m literally only using the axiom “S is either sound or not sound” to prove “S proves that S is not sound”. I’m sure that the principle of explosion requires a few axioms more, and that I’ve made a (common?) fatal error somewhere, but I’m comfortable with calling mathematical constructions “Absurd” without calling them false.
The best response to “X is absurd” is “Reality is absurd”, a point with which I will readily agree.
EDIT: Corrected a couple of major typos resulting in GCE, other errors are probably still present.