OpenAI and the main artificial intelligence laboratories have ended up at the center of a harsh formal stance by 25 Fields medalsthe most prestigious recognition in the discipline. Leading international figures, including Terence Tao, Alessio Figalli and Peter Scholze, have published a joint position statement to contest the tendency to transform the solution of complex problems into a mere commercial battleground. According to the signatories, the current rush to solve theoretical puzzles just to exhibit performance benchmarks risks compromising the very nature of the researchseparating the purposes of companies from those of the scientific community.
The disagreements have worsened following direct controversies that arose in US universities. Tristan Buckmaster, an academic at New York University, accused Sam Altman’s company of having pressured him not to mention the contribution of a researcher linked to Anthropic in solving a relevant problem, also hypothesizing that the work sessions carried out on Codex served the company to generate a proprietary demonstration through a weekend of intensive computation. Widespread discontent then led the company to cancel its sponsorship of an academic event at the California Institute of Technologyin response to concerns raised by researchers at the institute.
In the text drawn up by the signatories, formally called the joint declaration, the academics clarify that the resolution of a single question represents only a verification tool for building new theoretical knowledge, not the ultimate goal. Announcing answers hastily, without publishing detailed steps, without isolating the underlying methodological principles and omitting references to previous studies, triggers serious questions related to attribution and plagiarism. Without the development and metabolization work guaranteed by researchers, artificially generated solutions remain lifeless statements which risk breaking the chain of transmission of knowledge between generations of scholars.
From the value of demonstration to the risk of academic opacity
The most urgent fear expressed by experts concerns the compromise of the traditional transparency that animates the academic community. When frontier laboratories have the ability to identify a promising lead and invest tens of millions of dollars in pure computing power to get ahead of the trailblazers, the primary incentive inevitably shifts to confidentiality and defense of insights. The virtuous practice of sharing drafts and discussing open hypotheses risks coming to a halt in the face of I suspect that inference platforms are being used to cannibalize the work of others.
The position follows the recommendations already made in June by the Leiden Declaration working group, and extends the alarm to any activity with a high intellectual content. Years of advanced training are not only used to prepare a formal response downstream, but to develop critical sensitivity necessary to formulate new questions. If automated systems bypass the cognitive path to deliver exclusively the final product, science risks losing along the way the profound meaning for which those same problems were conceived.
