OpenAI Next-generation AI model Astra’s mathematical results were accused of plagiarism

IT House reported on August 8 that the technology media scientificamerican published a blog post on August 6, reporting that multiple mathematicians had accused OpenAI of academic misconduct, saying that the mathematical results of its AI model Astra did not give due signatures and acknowledgments to existing research results. OpenAI released a 249-page PDF file on July 31, detailing its next-generation major model Astra (internal version), which has made significant progress in multiple fields such as high-dimensional sphere filling, quantum game theory, group theory, and extreme combinatorics. If calculated based on the Sol API rate, the total cost of tokens required to find solutions to these problems is approximately US$2,000 (IT House Note: The current exchange rate is approximately RMB 13,519). However, after the paper was released, many scholars accused OpenAI of academic misconduct after review. 2 core results do not properly reference established ideas in the recent mathematical literature. Steven Miller, a mathematician at Yeshiva University, points out that Astra presents a core mathematical argument for the sphere-binding problem in spaces of 1,000 or more dimensions. This argument first appeared in a paper published by Miller and his collaborators in 2016, but was presented by Astra as its own work. Another result resolves a long-standing question about Soficity in group theory, proving that there is at least one group that lacks this property. Cambridge mathematician Francesco Fournier-Facio and Dresden University of Technology mathematician Andreas Thom discovered that Astra's key mathematical steps combine ideas from two papers in 2016 and 2019. In the face of doubts, OpenAI spokesperson made a statement