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

Microsoft took “A real step towards medical superintelligence,” says Mustafa Sulyman, CEO of the company artificial intelligence Arm. The Tech giant says its powerful new AI tool can Diagnosis illness Four times more precisely and with significantly fewer costs than a group of human doctors.

The experiment tested whether the tool could correctly diagnose a patient with an illness and imitate the work that is typically carried out by a human doctor.

The Microsoft team used 304 case studies from the New England Journal of Medicine to develop a test called sequential diagnosis benchmark (Sdbench). A voice model broke into a step-by-step process that a doctor would carry out to achieve a diagnosis.

The researchers of Microsoft then built a system called Mai Diagnostic Orchestrator (Mai-Dxo), which quadrilates several leading AI models GPT, Google Gemini, Anthropics Claude, Metas Lama and Xais GroK-in a way that easily works together.

In her experiment, Mai-Dxo exceeded human doctors and achieved an accuracy of 80 percent compared to 20 percent of the doctors. It also reduced the costs by 20 percent by selecting cheaper tests and procedures.

“This orchestration mechanism multiple agents who work together in this debate style-will bring us closer to the medical superintelligence,” says Sulyman.

The company poaches several Google Ai An intense war for top AI In the Tech industry. Sulyman was previously a manager on Google and worked on AI.

In some parts of the US health industry, AI is already widespread, including radiologists in interpreting scans. The latest multimodal AI models have the potential to act more general diagnostic instruments, although the use of AI in healthcare accuses their own problems, especially in connection with distortions from training data that are tailored to certain demographic data.

Microsoft has not yet decided whether it will try to commercialize the technology, but the same manager who spoke about the condition of anonymity said that the company could integrate it into Bing to help users diagnose symptoms. The company could also develop tools to help medical experts improve or even automate patient care. “What you will see in the next few years is that we are doing more and more work to prove these systems in the real world,” says Sulyman.

The project is the latest in growing research that shows how AI models can diagnose diseases. In recent years, both Microsoft and Google have published papers that show that large language models can diagnose an illness when accessing medical documents.

The new Microsoft research differs from earlier work in the fact that it replies more precisely the way human doctors diagnose diseases – by analyzing symptoms, ordering tests and further analysis until a diagnosis is achieved. In a blog post about the project, Microsoft describes the way in which several Grenz -KI models have been combined as “path to medical superintelligence”.

The project also suggests that AI could help reduce health care costs, a critical problem, especially in the USA. “Our model works incredibly well, both the diagnosis and the diagnosis that effectively get this diagnosis,” says Dominic King, Vice President at Microsoft, who is involved in the project.