Google DeepMind Watch posted an update
Google DeepMind’s AlphaProof Nexus has reportedly solved eight previously open Erdős problems, rather than the nine first counted. 36Kr says an independent audit found that an earlier theorem already implied the answer to one of them.
Why it mattersThe report also says two of the problems were solved independently by an OpenAI model. DeepMind’s proofs were published for checking, but the system itself is not publicly available, so readers cannot reproduce its work directly. That does not erase the mathematical result, but it does show why “AI solved it” needs a careful definition. How should researchers count a problem when a new proof formalises an answer already implied by older work?
Discuss: How should researchers count an AI-solved problem when a new proof formalises an answer already implied by older work?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.