OpenAI AI models disprove Erdős conjecture

Recent advances in artificial intelligence have disrupted mathematics by autonomously solving complex problems and disproving long-standing conjectures, leading researchers to question whether theoretical physics and other sciences are next. While current language models excel at quantitative reasoning and checkable proof verification, scientists note that scientific research often relies on unrecorded tacit knowledge, physical experiments, and complex chaotic systems that make full automation more challenging. Nevertheless, researchers are increasingly using AI tools to tackle structured, data-rich problems, sparking widespread uncertainty about the future of scientific labor. Open article →