Yes. From a quick scan of the paper, it includes a formal proof in Lean4. That said, it is very long and complicated, with lots of steps in the chain (as you might expect) so it would need to be checked carefully to ensure it proves what it claims to prove.
Lean uses Curry-Howard correspondence, so how proofs work is you declare your propositions as types and then your proof is actually a recipe that goes from things that have already been established and finishes by instantiating that type. The guarantees there are very strong - if you succeed in instantiating the type you have definitely proved something. The question is whether you have proved the thing you said you have. So here scanning the proof (it’s like 100 pages and I am sick so definitely sub-par intellectually) they use category theory to embed the problem, so the proof is actually a proof of the properties of this embedding. So if there is a problem with the proof, my guess would be that it would lie in the embedding not being exactly representative of the problem somehow.
It seems a pretty serious attempt though- it’s not just some random crank paper.
Thank you! This is the kind of comment I hoped to see.
I'm betting it was published in a hurry. I know I would hit "publish" within 24 hours of creating such a result, and would hope it would go wide. I'd publish to arxiv before getting clearance to release the code. I bet that's what happened here.
I appreciate your explanation of the Curry-Howard correspondance. I was familiar with it, but not with Lean in particular. I'd heard of Lean, but didn't know how it worked.
You are most welcome. Reading it a little more, the summary of the proof is they embed computations as a particular sort of category and then use homological algebra to show that computations in P have a certain property[1] and computations in NP have a different property[2], and they say they go on to demonstrate these properties are mutually incompatible, thus proving P != NP.
I don't know homological algebra at all and only the very basics of category theory and while the (107-page) proof gives a lot of background it would take more time for me to get myself up to speed than I can really afford to spend right now. But that's the gist.
The fact that they have formalized the proof should mean it will be quicker to verify whether or not this is indeed it.
[1] which they call "contractible computational complexes (Hn(L) = 0 for all n > 0)."
[2] which they call "non-trivial homology (H1(SAT)̸ = 0)"
Lean uses Curry-Howard correspondence, so how proofs work is you declare your propositions as types and then your proof is actually a recipe that goes from things that have already been established and finishes by instantiating that type. The guarantees there are very strong - if you succeed in instantiating the type you have definitely proved something. The question is whether you have proved the thing you said you have. So here scanning the proof (it’s like 100 pages and I am sick so definitely sub-par intellectually) they use category theory to embed the problem, so the proof is actually a proof of the properties of this embedding. So if there is a problem with the proof, my guess would be that it would lie in the embedding not being exactly representative of the problem somehow.
It seems a pretty serious attempt though- it’s not just some random crank paper.