OpenAI announced the achievement of significant scientific breakthroughs in the fields of mathematics, computer science, and physics, asserting that it is adopting a new approach to sharing these results with the scientific community. However, this step was met with widespread criticism from academics who expressed their dissatisfaction with the company's continued imposition of restrictions on access to its research.
The "Shepherds" Strategy and Scientific Results
In its new approach, OpenAI relies on employing "shepherds" from external experts to determine how results are shared. Launch strategies have varied; in the field of mathematics, the company released the AI-generated outputs directly, while in computer science, it decided to form an external group to interpret the results, and in physics, the launch has been entirely postponed pending guidance from external experts.
The announced results include solutions to complex problems, such as the "Kakeya conjecture" in mathematics, research on the "Standard Model" in physics, in addition to a claim of refuting the "Exponential Time Hypothesis" in computer science, which is a fundamental hypothesis related to P vs NP problems.
Why Does This Approach Anger Researchers?
Despite experts acknowledging the importance of the scientific results, they have strongly criticized the "selection" mechanism followed by the company. Researchers believe that limiting access to the results to a selected group of scientists bolsters the power of major corporations and creates a gap in access to scientific knowledge.
The "Advisory Group for Mathematics and Artificial Intelligence" (AGMAI) has warned against this practice, emphasizing that selecting specific individuals to receive the results first is unnecessary and risks causing bias and unjustified delay. The group called on the company to make the outputs available to the public immediately, especially since the company stated that formal verification processes (Lean formalization) have already been completed.
What Does This Mean for the Scientific Community?
Academics point out that the problem goes beyond mere announcement mechanisms; some believe the real challenge lies in the speed at which scientific results are generated by AI without clear human contribution, making it difficult for the scientific community to read these research papers and verify their validity. Researchers emphasize that making results available openly and transparently is not just a matter of cognitive democracy, but a necessity to ensure the quality of science and accelerate the safe and reliable adoption of new tools.