Re: What are Simple Types (Was: Proofs as programs)

Interestingly although this proof (Curry Howard
term) does not exist in Affine Logic:
λx:A->A.λy:A.(x (x y))   :  ((A -> A) -> (A -> A))
There is an alternative proof (Curry Howard
term) which does exist:
λx:A->A.x   :  ((A -> A) -> (A -> A))
This is just the I combinator of Affine Logic.
Mild Shock schrieb:

This is also not in the scope of advent of logic:
 From: Mild Shock <janburse@fastmail.fm>
Date: Mon, 18 Nov 2024 23:22:11 +0100
Subject: Re: How to prove this theorem with intuitionistic natural deduction?
 > 0: P -> (Q -> R) (Premisse)
 > 1: P /\ Q        (Assumption)
 > 2: P             (/\E,1)
 > 3: Q -> R        (->E,0,2)
 > 4: Q             (/\E,1)
 > 5: R             (->E,3,4)
 > 6: P /\ Q -> R   (Discharge,1)
 The advent of logic has only implication (->)/2
and no conjunction (/\)/2:
  > Advent of Logic 2024: Weekend 2
 > [...]
 > The propositional logic can do with
 > *implication only*, and it should be Linear Logic.
 > French logician Jean-Yves Girard is credited
 > [...]
 It says implication only. Just like here, implication
only is sometimes called positive logic:
 Meredith, D. (1978). Positive logic and λ-constants.
Studia Logica, 37(3), 269–285. doi:10.1007/bf02124728
 I think the example I gave is on page 288. The
lambda expression would need an additional combinator
W, it cannot be modelled with BCI alone:
 λx:A->A.λy:A.(x (x y))   :  ((A -> A) -> (A -> A))
 Meredith, D. gives the combinator expression BCWI.
Didn't verify. The combinator W is left contraction,
and since affine logic has no contraction, the
 combinator W is not part of affine logic:
 W: ((A -> (A ->B)) -> (A -> B))
 Mild Shock schrieb:

Julio Di Egidio schrieb:

On 19/12/2024 00:30, Mild Shock wrote:

It cannot be a proof term of Affine Logic, since x occurs twice.

That is not what affine logic means.

>
Well it does. It is based on the combinators B and C.
You cannot translate a lambda expression where x occurs twice
with B and C. You would need S.
>
Rule 7 and 8
https://en.wikipedia.org/wiki/Combinatory_logic#Combinators_B,_C
>
With B and C, you can only translate λx.(E1 E2) if x
occurs in E1 or E2 or none, but not in both E1 and E2.
>
Basically the implied requirement to have an equivalence
between the natural deduction and the hilbert style proofs,
that every lambda express can be translated to BCI.
>
Bye