OpenAI’s Astra Model Solves Ten Mathematical Problems, Sparking Debate Over Credit and Career Futures
While experts acknowledge the acceleration of discovery, mathematicians warn that the high computational costs and proprietary nature of the technology could lock out smaller institutions and undermine the human scholarship that underpins these breakthroughs.
OpenAI has announced that its unreleased artificial intelligence model, Astra, has produced solutions to ten long-standing mathematics problems, including questions regarding sphere packing, error-correcting codes, and the existence of non-sofic groups. The release of over 250 pages of papers and 60 pages of methodology, with each result certified using Lean proof verification software, has elicited mixed reactions from the mathematical community. Experts are expressing excitement at the acceleration of discovery alongside apprehension regarding credit attribution, the commercialisation of research, and the potential impact on academic careers and funding.
The problems solved by Astra span a wide range of fields, from the highly abstract to those with practical implications in cybersecurity and data transmission. One breakthrough concerned how tightly spheres can be packed in more than three dimensions, while another resolved two long-standing questions about complex connected networks. A third result addressed the existence of non-sofic groups, infinite mathematical structures that cannot be approximated by finite ones. The announcement has added to a complex swirl of emotions about where the field is headed, with palpable excitement at the prospect of accelerating mathematical discovery balanced against apprehension about what this means for the people who have dedicated their lives to the pursuit.
A dispute has arisen regarding the attribution of credit for the solution concerning non-sofic groups. Mathematicians Andreas Thom and Gábor Kun argued that OpenAI’s initial announcement minimised their recent contributions, leading the company to update its language to acknowledge prior research. Kun, a researcher at the Alfréd Rényi Institute of Mathematics, described the original sweeping language as "rather comical," noting that the detailed research paper clearly stated it built on his results from 2016 and 2019. OpenAI spokesperson Laurance Fauconnet confirmed the post was updated to better reflect the prior research these results build upon, though Kun remains concerned about similar oversights in other results outside his area of expertise.
Money is at the heart of many concerns, as mathematics has traditionally been a slow-moving, low-cost discipline where researchers often operate without significant grants. OpenAI estimates the computational cost for the ten solutions at approximately $2,000 in API tokens, though researchers suggest the true cost is higher due to trial and error. Colva Roney-Dougal, a professor at the University of St Andrews, fears that even this modest price could lock out researchers at smaller and less wealthy institutions. There is also unease about the growing intrusion of commercial interests into a field that has largely operated in the open, with the most capable models from companies like OpenAI and Anthropic being proprietary and tightly controlled.
The Leiden Declaration, endorsed by the International Mathematical Union and signed by over 3,400 people, urges against the hype surrounding AI capabilities in mathematics. Graduate students and early-career researchers are expressing fear and despondency, with some questioning the viability of their careers as AI begins to solve problems typically assigned to students for skill development. Many of the people spoken to by The Verge are described as "really, really scared," with despairing essays circulating online questioning their futures in the field. While some see potential for AI to broaden participation, others worry that the field could be reshaped for the worse by exaggerated claims before anyone has had time to understand the actual impact.
