数学家Andreas Thom在社交媒体上指控OpenAI可能在未经授权的情况下，将他与Gábor Kun关于Gromov soficity猜想的未发表研究用于训练Astra模型1。该模型随后宣称已解决包括这一猜想在内的十大问题1。
Thom是该领域的领先专家，已在这一问题上投入二十年的研究1。除Thom外，Levent Alpöge和Tristan Buckmaster也曾提出类似指控1。这些指控表明，未发表的人类学术工作可能被吸收入AI模型的训练过程，随后被误认为是模型本身的突破性发现1。若属实，这将不是孤立事件，可能构成科学领域的重大学术丑闻1。
OpenAI faces allegations of incorporating unpublished research without permission into its Astra model's training data. Mathematician Andreas Thom posted evidence on Mastodon claiming that OpenAI used an unauthorized conversation involving his work with Gábor Kun on the Gromov soficity conjecture to train the system 1. According to the allegations, Astra subsequently announced it had solved this problem, which Thom identifies as one of ten major questions the model claims to have resolved 1.
Thom, a leading expert who has devoted two decades to this mathematical challenge, is not alone in raising such concerns 1. Other researchers including Levent Alpöge and Tristan Buckmaster have previously made similar accusations against the company 1. The pattern of allegations suggests that unpublished human research may be absorbed into the model and then misrepresented as the model's own breakthrough discoveries 1. If substantiated, this incident would constitute a significant academic scandal in the scientific community, demonstrating that AI systems are assimilating rather than independently discovering mathematical advances 1.
评论
还没有评论，欢迎留下第一条。