OpenAI has recently come under fire for its claims regarding the mathematical capabilities of its GPT-5 model. Criticism arose after the organization stated that GPT-5 had solved several conjectures previously thought to be unsolvable.
Key Figures Respond to OpenAI’s Claims
Prominent figures in the AI community voiced their disapproval. Yann LeCun, the Chief AI Scientist at Meta, aptly noted the backlash, remarking, “Hoisted by their own GPTards.” Similarly, Demis Hassabis, CEO of Google DeepMind, described the situation as “embarrassing.”
Debunking the Math Breakthroughs
OpenAI’s VP, Kevin Weil, boasted in a now-deleted tweet that GPT-5 had successfully solved ten Erdős problems and made progress on eleven others. These problems are a set of famous mathematical conjectures posed by mathematician Paul Erdős.
Expert Opinions on the Claims
However, mathematician Thomas Bloom, who oversees the Erdős Problems website, labeled Weil’s claims as a “dramatic misrepresentation.” He clarified that the problems listed as “open” do not imply they are unsolvable. Rather, it indicates a lack of published solutions known to him.
Bloom emphasized that GPT-5 did not solve the problems but instead found references that may lead to their solutions, which he had not previously encountered.
OpenAI Researcher’s Acknowledgment
Following the backlash, Sebastien Bubeck, an OpenAI researcher, admitted that GPT-5 had only discovered existing solutions in the literature. He defended the model’s capability, stating, “I know how hard it is to search the literature.”
Conclusion
The situation highlights ongoing scrutiny of AI developments and the accuracy of claims made by leading organizations in the field. As expectations rise, the focus on transparency and accuracy is more critical than ever.
