Researchers from Microsoft shared impressive results by presenting their latest scientific work. The company’s AI system, Microsoft MAI Diagnostic Orchestrator, diagnoses four times more accurately than a group of experienced doctors.
The technology allegedly works as a council of several AI models, including ChatGPT, Gemini, and Grok. Each of them is responsible for a specific role – from collecting symptoms and prescribing tests to final verification of the diagnosis.
In tests based on 304 complex clinical cases from the New England Journal of Medicine, virtual doctors outperformed real doctors. They determined the correct diagnosis in 80% of cases, while a team of real therapists, each with 5 to 20 years of experience, coped with it only in 20% of cases. The most effective model was the o3 from OpenAI, as it correctly solved 85.5% of problems. In addition, the medical system MAI Diagnostic Orchestrator reduced the total cost of the service by 20% compared to human analogues by choosing cost-effective procedures and tests.
A version of this technology can be integrated into the Copilot chatbot and the Bing search engine. This should make it easier for users to get an accurate diagnosis without going to the doctor. As Microsoft’s CEO in the field of AI, Mustafa Suleiman, said, such a tool will be a major step towards medical superintelligence.
But experts advise treating the results of the study with caution. After all, the diagnoses are not based on real life experience.
Recall that in another study, artificial intelligence went crazy when it was entrusted with managing a store.
To be continued…
Only registered users can leave comments