Microsoft Says Its New AI System Diagnosed Patients 4 Times More Accurately Than Human Doctors

Microsoft took “A real step towards medical superintendent,” explains Mustafa Suleyman, CEO of society artificial intelligence arm. The technology giant says that its new powerful AI tool can Diagnosing the disease Four times more precisely and much cheaper than a panel of human doctors.

Experience has tested if the tool could properly diagnose a patient with a disease, imitating the work generally done by a human doctor.

Microsoft’s team used 304 case studies from the New England Journal of Medicine to design a test called sequential diagnostic benchmark (SDBENCH). A language model has broken each case in a step by step process that a doctor would make in order to reach a diagnosis.

Microsoft researchers then built a system called Mai -Diagnostic Orchestratetor (May -Dxo) which questions several main AI models – including Openai GPT, Google Gemini, Anthropic Claude, Meta’s Llama, and XAI Grok – so as to imitate several human experts working together.

In their experience, Mai-Dxo surpassed human doctors, reaching 80% precision compared to 20% of doctors. It has also reduced costs by 20% by selecting less expensive tests and procedures.

“This orchestration mechanism – Moustiple agents who work together in this style of debate chain – this is what will bring us closer to medical superintendent,” says Suleyman.

The company poache several researchers from Google IA to help in effort – but another sign of An intensive war for AI expertise in the technology industry. Suleyman was previously a Google executive working on AI.

AI is already widely used in parts of the American health care industry, in particular by helping radiologists to interpret analyzes. The latest multimodal AI models have the potential to act as more general diagnostic tools, although the use of AI in health care raises its own problems, in particular linked to the bias of training data which are biased towards a particular demography.

Microsoft has not yet decided if it is trying to market the technology, but the same manager, which spoke on condition of anonymity, said that the company could integrate it into Bing to help users diagnose conditions. The company could also develop tools to help medical experts improve or even automate patient care. “What you will see over the next two years is that we are doing more and more work proving these systems in the real world,” explains Suleyman.

The project is the last in a growing set of research showing how AI models can diagnose the disease. In recent years, Microsoft and Google have published articles showing that large languages ​​models can accurately diagnose a disease when they have access to medical records.

The new Microsoft research differs from previous work in that it reproduces more precisely the way in which human doctors diagnose the disease – by analyzing the symptoms, by ordering tests and by carrying out a more in -depth analysis until a diagnosis is reached. Microsoft describes the way in which he combined several border AI models such as “a path to medical superintendent”, in a blog article on the project today.

The project also suggests that AI could help reduce health care costs, a critical problem, especially in the United States. “Our model works incredibly well, both to go to the diagnosis and to go to this very profitable diagnosis,” explains Dominic King, vice-president of Microsoft who is involved in the project.