so uh who wants to bet that anthropic’s fucking gigantic riemann hypothesis proof and their “rumored” P=NP proof (lol, is math just gta6 now?) are all just accidental implementations of the Falso proof system: https://sigtbd.csail.mit.edu/pubs/2016/paper9.pdf
it’s amazing how these parody papers from 10 years ago read exactly like AI company press releases about how their model just mathematically proved that their company is profitable, just in time to juice their IPO
falso is very funny because most proof assistants assume they’re not being operated by a malicious asshole intentionally making shit up for funding or a bunch of malicious assholes who made a making shit up for funding machine
like, it’s not that difficult to find an “exploit” in a proof assistant if you’re looking for one, cause a proof assistant assumes you’re using it to assist you in finding a proof that you will later read and verify
I’m uhh not sure why the team behind lean doesn’t seem to know the above by the way, I’m not a proof assistants person (I’ve only implemented one by typing it in from The Little Prover) but if you know the thinnest amount about how they work, something like Falso follows very directly from that knowledge. which is what made the joke funny in 2016
anyone pretending lean isn’t fully AI-brained can miss me with that shit by the way, I’ve seen these guys talk and they’re all in on garbage
@zzt wait what, a "P==NP" Proof?