OpenAI annonce trois prétendues découvertes en mathématiques, en informatique et en physique
· Lecture en 50 mots
En 50 mots
OpenAI a annoncé trois prétendues découvertes en mathématiques, en informatique et en physique, moins d'un mois après une précédente publication de manuscrits. Des experts extérieurs guideront la vérification des travaux d'informatique et de physique. Selon l'une de ces affirmations, l'hypothèse du temps exponentiel (Exponential Time Hypothesis) serait fausse ; elle n'a pas encore été vérifiée.
Pourquoi c'est important
The announcement highlights AI's growing potential to accelerate scientific research and tackle problems that have challenged mathematicians and scientists for decades. If independently validated, these results could reshape theoretical computer science, mathematical analysis and physics-related research. However, the findings remain subject to expert verification, raising important questions about scientific credibility, transparency and equitable access to AI-generated research.
La suite
Independent specialists will examine the mathematical arguments, test their validity and publish explanations. A working group led by Russell Impagliazzo, Ramamohan Paturi and Ryan Williams will review the claimed refutation of the Exponential Time Hypothesis. OpenAI's use of expert “shepherds” will also face scrutiny over concerns about preferential access, researcher selection and scientific independence.
Dates clés
- October 6, 2026: OpenAI releases hundreds of AI-generated mathematical manuscripts.
- October 25, 2026: OpenAI's mathematics advisory panel raises concerns about selective early access to AI-generated discoveries.
- October 26, 2026: OpenAI announces three major research claims and publicly releases mathematical and computer science materials.
- Late 2026 onward: Independent researchers begin examining the proofs and developing explanations. No verification deadline has been announced.
Posez vos questions à Brivfy AI sur cet article
Sources
- The VergeVoir l'original →