Sign up to see the future, today<\/p>\n
Can\u2019t-miss innovations from the bleeding edge of science and tech<\/p>\n
Earlier this year, a 23-year-old without any formal mathematics training made headlines<\/a> by claiming he\u2019d used OpenAI\u2019s ChatGPT to solve one of<\/a> the \u201cErd\u0151s problems\u201d \u2014 a database of challenging conjectures left behind by Hungarian mathematician Paul Erd\u0151s.<\/p>\n Then, last month, scholars were taken aback<\/a> when OpenAI claimed its AI had disproved an 80-year-old \u201cunit distance\u201d conjecture, also devised by Erd\u0151s.<\/p>\n \u201cThis marks the first time AI has autonomously solved a prominent open problem central to a field of mathematics,\u201d OpenAI boasted<\/a> at the time. <\/p>\n But whether the frontier AI models powering tools like ChatGPT really represent a major leap in our ability to solve problems that have been plaguing mathematicians for decades remains hotly debated among experts.<\/p>\n In perhaps the strongest public rebuke yet<\/a>, a new declaration signed by over 150 mathematics experts from around the world warned governments not to \u201cbelieve the hype\u201d when it comes to AI\u2019s capabilities to solve complex mathematical problems, throwing cold water on claims of a revolution in the field.<\/p>\n In a statement accompanying the 11-page \u201cLeiden Declaration on AI and Mathematics<\/a>,\u201d International Mathematical Union vice president Ulrike Tillmann argued that AI \u201craises questions that cannot be left unexamined.\u201d<\/p>\n \u201cThe future of mathematical research must be guided by human judgment, fair and transparent practices, and the shared values of the global mathematical community,\u201d Tillmann said.<\/p>\n \u201cThere is currently a strong commercial incentive on the part of the technology industry to overstate the capabilities of their products,\u201d the declaration reads, advising policymakers to \u201cconsult with experts, including mathematicians, in forming policy decisions rather than relying on press releases or popular reporting of mathematical results.\u201d<\/p>\n Worse yet, AI models may produce convincing-sounding solutions that don\u2019t actually withstand scrutiny.<\/p>\n \u201cCurrent automated techniques can produce plausible but unreliable (or even incorrect) arguments which are difficult to distinguish from correct mathematical proofs,\u201d said signee and University of Oxford head of computer science Leslie Ann Goldberg in a statement<\/a>. \u201cThis is a serious problem: research in mathematics (and in mathematical disciplines like theoretical Computer Science) almost always builds on previous research, so it is essential for researchers to know that the results in the literature are correct.\u201d<\/p>\n The declaration highlighted the highly precarious position many academics are finding themselves in. Attracting new funding has proven difficult, while interest in AI continues to soar, often forcing them to endorse the tech at all costs.<\/p>\n \u201cWe recognize that industry has offered lucrative jobs, monetary rewards, computing resources, and intellectually stimulating opportunities that some mathematicians have found attractive,\u201d it reads. \u201cThis has taken place in an era of underfunding of higher education and precarious academic employment.\u201d<\/p>\n The document also noted there were plenty of other reasons to call for regulatory oversight beyond the field of mathematics, noting the AI industry\u2019s \u201cinvolvement in military and mass surveillance programs, development of technologies which promote misinformation and undermine democracy, and environmental costs.\u201d<\/p>\n In short, it\u2019s a ringing denunciation of the persistent hype surrounding AI, and a call for reining in its use that reverberates far beyond the world of mathematics. The broader scientific community has been reeling from a flood of papers that make heavy use of AI<\/a>, risking contaminating the peer-review process with hallucinations.<\/p>\n It\u2019s also a pertinent reminder that AI models are being trained on cutting-edge research, often without sign-off from the original authors.<\/p>\n \u201cMathematicians who never intended to contribute to AI development are having their work used for this purpose without their consent,\u201d Leiden University anthropologist of AI Rodrigo Ochigame, who helped draft the declaration, told Scientific American<\/a>. \u201cI think that\u2019s a deeply concerning situation.\u201d<\/p>\n