Artificial intelligence has achieved a remarkable breakthrough in mathematics, with OpenAI's newly announced Astra model solving ten longstanding mathematical problems that have stumped researchers for decades—all for a computational cost of just $2,000. The company published its groundbreaking work on 1st August, unveiling results that span multiple complex mathematical domains including group theory, high-dimensional geometry, coding theory, quantum complexity, lattice cryptography, and extremal combinatorics.
What sets this achievement apart from previous AI-assisted mathematical work is the rigour of verification. OpenAI produced a comprehensive 249-page manuscript collection accompanied by machine-checkable certificates for every single result—a level of formal verification unprecedented in AI-generated mathematical research at this scale. These are not trivial exercises repackaged as discoveries; each problem had remained unsolved for at least a decade, with many dating back considerably longer.
Among the most significant achievements are a construction establishing the existence of non-sofic groups, a question that has occupied group theorists for years, and a disproof of Connes's rigidity conjecture, a longstanding problem in the theory of von Neumann algebras. The model also improved the general upper bound on sphere-packing density in high dimensions, a mathematical limit that had remained unchanged since 1978. Additionally, three of the solved problems came directly from the catalogue of open questions left behind by the legendary mathematician Paul Erdős.
This announcement builds upon OpenAI's previous success in May, when a similar reasoning model produced an original mathematical proof that disproved a famous unsolved conjecture in geometry, also originally posed by Erdős in 1946. The consistency of these achievements suggests that AI models are transitioning from assistive tools to active contributors in frontier mathematical research, potentially accelerating the pace of discovery in theoretical fields that have traditionally relied solely on human intuition and creativity.
Artículos relacionados de LaRebelión:
- OpenAIs Rogue AI Agent Week-Long Hack Undetected
- AI Models Escaped OpenAIs Security Scare
- OpenAIs GPT-56 Gets Green Light US Approves Wide Rollout
- OpenAIs Privacy Filter Open-Source Data Protection
- OpenAIs AI Security Move Limited Cyber Model Release
Artículo generado mediante LaRebelionBOT
No hay comentarios:
Publicar un comentario