New Human-AI Paradigm Proposed for Scalable Mathematical Discovery on arXiv

A recent arXiv paper, 'The Problem Is the Problem: Towards Scalable Mathematical Discovery,' presents a new approach to AI-assisted mathematics. The authors argue that current workflows concentrate human effort on selecting research problems and subsequently reviewing results, making these stages bottlenecks. To address this, they propose a paradigm where human experts provide a research direction rather than a single problem. A system then searches a broad literature corpus for candidate problems, inspired by search and recommender techniques. The paper introduces a workflow called Find, Attempt, and (cut off), emphasizing efficient allocation of scarce reasoning resources.