ai兩年內解決35%的理論科學問題
ai模型在2024年至2026年間解決的理論問題比例從2%提高到35%，預計這一進展將繼續。
a study by mit researchers neil thompson, gabriel manso, and claudio mayrink verdun found that openai models, including gpt-5 and gpt-5.5, solved 35% of theoretical problems in fields like economics, high-energy physics, and quantum physics by mid-2026, up from 2% in 2024. the models were tested on problems from research papers posted online between december 2025 and may 2026, and had to work out solutions for themselves without relying on training data. this progress is expected to continue, potentially transforming these fields. the study suggests that ai may soon be the default way to solve theoretical problems in these areas.