The notation you'd use to represent the program/proof is very important.
One interesting difference is that if you are learning a proof, then you know when you are successful. At least for my research, I generated a bunch of small cases of the theorem I was trying to prove, then tried to learn the general case. If I was successful, I would have a proof of the original conjecture. When learning programs from some data, you can always test your program against the data but never can be sure that you are generalizing the data properly.
The equivalence seems to hold only when the program is complete. Thus, while all proofs can be expressed as programs, not all programs can be expressed as proofs.
Can't you restrict the domain, and so that the program is complete (for that restricted domain)?
I meant correct, not complete.
I refer to the article on wikipedia which states: "A converse direction is to use a program to extract a proof, given its correctness. This is only feasible if the programming language the program is written for is very richly typed: the development of such type systems has been partly motivated by the wish to make the Curry-Howard correspondence practically relevant."
Programs are equivalent to proofs given their correctness, rather. Additionally, according to the wikipedia article, extracting a proof from a program requires a very richly typed programming language.
I'm just a mathematician, and my mind has obviously been warped by continua and the Axiom of Choice but aren't they the same thing?
http://en.wikipedia.org/wiki/Curry-Howard_correspondence
Actually, I've gotten interested in learning more about theoretical CS lately, but my everyone in my department is too old/applied/Russian to care, and my university's CS people only seem to care about operating systems and e-commerce. Can you recommend a logical starting place?