Twenty-five leading mathematicians have signed an open letter saying that artificial intelligence companies are trying to compete with each other in solving well-known mathematical problems, thereby jeopardizing the intellectual work of scientists. All of the original signers of the letter are recipients of the Fields Medal, the most prestigious award in mathematics.
Image credit: Brecht Corbeel / unsplash.com
This week, New York University professor Tristan Buckmaster accused OpenAI of pressuring him not to attribute authorship of one of his papers to an Anthropic colleague who worked with him on an important mathematical problem. Buckmaster also wants to know whether OpenAI can use the data uploaded by scientists to Codex and then use lengthy calculations to derive its own breakthrough proof.
On Thursday, OpenAI withdrew its sponsorship of Caltech math events. Previously, the company’s work had been criticized by researchers at the institute.
The ability of artificial intelligence models to solve critical unsolved mathematical problems can bring huge benefits to humanity. However, the authors of the letter believe that this will only happen if the solutions found can be understood and explained by representatives of the mathematical community and ultimately by the rest of the world.
«Such decisions are often announced hastily. There is simply no time to properly formalize the results of the work, highlight the main methods and ideas, and refer to related work of other scientists. As with any creative profession, this raises serious questions about authorship and plagiarism. Furthermore, without the involvement of mathematicians, who should develop new ideas and incorporate them into the mathematical canon, ideas born with the help of artificial intelligence will never gain full life and the most important chain of knowledge transfer between mathematicians will be lostthe letter reads. TechCrunch authors note that OpenAI’s proof has not been independently verified.
Against this backdrop, other mathematicians began to show a more cautious attitude. Increasingly, they are wondering whether OpenAI can use information from interactions with Codex to create new artificial intelligence models. There are real concerns that the culture of open research is under threat. If a leading artificial intelligence lab finds a promising path to discovery, it may spend tens of millions of dollars running language models to beat the original researchers to be the first to come up with a proof. This dynamic will push the scientific community toward greater secrecy.
The letter from the mathematicians follows the Leiden Declaration issued by a working group of scientists in June. The paper also examines how evidence derived from large language models could change the work of mathematicians and provides recommendations for scientists, academic institutions and authorities.
The value of mathematics lies not only in the proof itself and in the recognition of authorship. Also important is the intellectual environment around them, in which students learn, new questions and ideas emerge, and the results of research become part of the overall body of human knowledge.
«The challenges facing today’s mathematics community are similar to those in other scientific and creative professions. They point to a problem that all of humanity may face: How to ensure that as artificial intelligence changes the way we work, we don’t lose the original purpose of work.“, the mathematician’s letter reads.
If you find an error, select it with your mouse and press CTRL+ENTER.