OpenAI's latest math breakthroughs commit research misconduct, experts say

OpenAI's latest AI-generated math breakthroughs have sparked controversy as experts claim the company is engaging in research misconduct. The company unveiled ten artificial-intelligence-generated math results during internal development and testing of its next major large language model, called Astra. Each result resolved or progressed a different longstanding open problem of significant interest to the mathematical community.

However, upon closer examination of the nearly 250-page paper, many experts have found instances where preexisting ideas from recent mathematical literature were incorporated without proper citation, contradicting OpenAI's initial press release claims of novelty. Steven Miller, a mathematician at Yeshiva University, argues that OpenAI's approach to the ball-packing problem in a 1,000-dimensional space exhibits systematic plagiarism, as the LLM-generated proof relies on a mathematical argument first presented in a 2016 paper by Miller.

Another result in group theory also appears to be missing proper attribution, despite OpenAI's claims of originality. These instances of alleged research misconduct have led to frustration among mathematicians, who question whether OpenAI is merely assembling puzzle pieces from existing knowledge rather than making genuine intellectual leaps.

OpenAI maintains that they are meeting the same standards expected of human mathematicians and plan to make small updates to their paper this week. However, some in the mathematical community are growing impatient with OpenAI's apparent disregard for academic norms, and are calling for the company to be held to the same academic standards as their human counterparts.