A new study suggests that artificial intelligence models are rapidly improving their ability to solve complex theoretical problems across various scientific disciplines. Researchers tested several OpenAI models, including GPT-5.5, on problems from fields like probability, statistics, and quantum physics, finding that AI capabilities have jumped from solving 2% to 35% of these problems in just two years. This advancement mirrors the disruption AI has already caused in mathematics, raising questions about the future role of human scientists in theoretical research and problem-solving. AI
IMPACT AI's increasing ability to solve theoretical problems could accelerate scientific discovery but also necessitates new methods for verification and understanding.
RANK_REASON The cluster discusses a study on AI's capability in solving scientific theory problems, which falls under research. [lever_c_demoted from research: ic=1 ai=1.0]
- .anthropic
- Claudio Mayrink Verdun
- Gabriel Manso-Díaz
- GPT-5
- GPT-5.1
- GPT-5.4
- GPT-5.5
- MIT
- Neil Thompson
- New York University
- o3
- .openai
- Openai O1 System Card
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →