微軟AI在診斷測試中擊敗醫生
微軟的AI工具診斷準確率高達83%，而人類醫生的準確率僅為36%。這將改變醫療診斷。
microsoft's ai diagnostic tool, powered by openai's gpt-5, was tested against human physicians on medical case reports and emergency department data. the ai achieved 83% diagnostic accuracy vs 36% for generalist physicians and 22% for those without references. the tool also outperformed doctors on emergency department cases, scoring 80% accuracy vs 47% for human physicians. microsoft plans to open-source its benchmark, sdbench. the company aims to develop 'medical superintelligence' that can outperform human-led healthcare.