Navier–Stokes Lost in Translation (arxiv.org)

A recent article titled "Navier-Stokes Lost in Translation" highlights critical challenges in the autoformalisation of mathematical proofs, particularly those generated by AI, such as OpenAI's proof on the blow-up of solutions to the Navier-Stokes equations. The study reveals that while AI can translate natural language (NL) mathematical statements into formal languages like Lean, this process often lacks semantic fidelity. The translation can introduce ambiguities that render the formal verification of the original NL argument unreliable, as evidenced by instances where AI mistranslations resulted in formal proofs that diverged from the intended meaning. This finding is significant for the AI/ML community as it underscores the limitations of current autoformalisation techniques, especially given that ensuring semantically faithful translations is deemed more complex than even the Halting problem. This research compels mathematicians and AI developers to reconsider the efficacy of using AI-generated proofs in formal settings, emphasizing the need for improved AI systems that can better navigate and resolve the intricacies of mathematical language. Such advancements could enhance both the rigor and trustworthiness of AI-assisted mathematical work in the future.