Skip to contentSkip to filters

We use cookies for analytics.

Seminar, Hosted by Online Logic Seminar

Mathematical proofs, computational complexity and AI

Nancy Abigail Nuñez

Universidad Nacional Autónoma de México

Abstract

Mathematical proofs are a crucial aspect of mathematicians' work, and finding a proof is often very challenging: proofs of theorems such as Fermat's Last Theorem or the Four Color Theorem took decades or even centuries. The rapid progress of artificial intelligence nevertheless suggests that it may soon be possible to automate the discovery of mathematical proofs. This talk critically examines the viability of relying on artificial intelligence to fully automate proof discovery in light of results from computational complexity theory. Without disregarding recent advances achieved by AI in this area, it seeks to temper expectations that this technology will fully replace human reasoning in the discovery of mathematical proofs, since the complexity of the task defies automation. The hypothesis is that, because of the complexity inherent in discovering or searching for mathematical proofs, a broad class of problems may not be amenable to automation by artificial intelligence, automated theorem provers or proof assistants.

Topics