Doctors lose to AI, but only when they’re working alone
Microsoft’s AI beats generalists in a simulation, but the real test is yet to come.
The story
Microsoft claims its AI outperforms doctors on complex cases in a simulation, but the study compared it to generalists, not specialists, and the AI's accuracy gains were modest.
Source: techcrunch.com
Our reading
Microsoft’s AI, DxO, claims to outperform doctors on complex cases, but the real kicker is that it only beat generalists, not specialists. The study says it’s 83% accurate, but the doctors it beat were working alone with reference materials, not as teams. That’s like saying a calculator beats a math teacher at arithmetic, but only if the teacher is doing it without a textbook. The AI’s best results came from OpenAI’s models, which Microsoft then slightly improved with its orchestration system. It’s like a chef using a pre-made sauce to slightly enhance a dish, then claiming it’s a revolution. The real irony is that the AI’s biggest gains were on cases that are usually handled by specialists, not the generalists it was tested against. It’s like saying a robot can beat a beginner at chess, then calling it a breakthrough in defeating grandmasters.
