AI Firms' Failure to Credit Human Mathematicians Undermines Trust Despite AI's Genuine Mathematical Power
Source: https://www.theguardian.com/profile/editorial. "The Guardian view on AI v mathematicians: humans are still vital to the field, but tech firms refuse to see that | Editorial | The Guardian." September 20, 2026. www.theguardian.com
The Gist
OpenAI claimed its AI solved a famous unsolved math problem, but the AI's work depended heavily on human mathematicians' prior research and effort—research that didn't get proper credit. The Guardian argues that AI companies need to fairly acknowledge and compensate the human experts whose work makes AI's achievements possible, especially since humans are still needed to check and interpret AI's often messy results.
Conclusion
AI companies like OpenAI need to fundamentally change how they attribute, compensate, and collaborate with human experts whose work underpins AI achievements, rather than claiming unilateral credit for AI-generated results.
Premises
- OpenAI claimed its AI agents solved the Navier-Stokes problem, but this relied heavily on digesting and building upon existing human mathematical work
- OpenAI did not give sufficient credit to mathematicians who were close to solving the problem themselves
- There are unresolved concerns that OpenAI may have accessed or benefited from a specific mathematician's (Buckmaster's) unpublished work through its Codex model, which OpenAI has not fully denied
- AI companies rely on human mathematicians to verify, interpret, and determine the usefulness of AI's mathematical outputs, since AI work is often 'sloppy and baffling' without human oversight
- This pattern of using human labor without proper attribution or compensation has created broader alienation and resentment toward AI companies across multiple professional fields, not just mathematics
- Despite legitimate grievances, mathematicians remain notably open to AI collaboration, recognizing its genuine power, suggesting the conflict is about fair treatment rather than rejection of the technology itself
Assumptions
- Proper attribution and compensation for human intellectual contributions to AI systems is both ethically required and practically achievable
- The current framing of AI achievements (e.g., 'AI solved X') obscures the collaborative/derivative nature of the work in a way that is misleading
- Mathematical understanding and clarity (per Thurston) are fundamentally human qualities that cannot be fully replicated by AI, even when AI produces correct results
- Tech companies have both the ability and incentive to change their practices around attribution and human labor, but currently lack sufficient incentive to do so voluntarily
- It is possible and desirable to 'separate the tech from the tech companies' to democratize AI use, as suggested by Guillen