Microsoft has claimed to have developed an AI tool that successfully diagnoses patients with four times the accuracy of human doctors.
The technology MAI Diagnostic Orchestrator (MAI-DxO) combines several AI language models, including ChatGPT, Gemini from Google, Claude from Anthropic, Llama from Meta, and Grok from xAI.
During the system tests, researchers used 304 real case studies published in the New England Journal of Medicine that involved rare diseases considered particularly difficult to diagnose.
These cases were turned into a series of scenarios, where the AI had to discover the disease in the same way a doctor would, analyzing symptoms, requesting tests, and narrowing down possibilities step by step.
According to the study, the AI proved to be capable of mimicking a team of doctors working together, sharing opinions and discussing symptoms before reaching a diagnosis.
The researchers said the AI correctly diagnosed the cases with 80% accuracy. In comparison, a team of human doctors diagnosed the patients with only 20% accuracy.
But that’s not all: the AI also managed to reduce the cost of diagnosis by about 20%, choosing more affordable tests and avoiding unnecessary procedures.
These findings were detailed in a recent post on Microsoft’s blog titled “Introducing the MAI Dx Orchestrator: Our path to medical superintelligence.”
In the post, the company positioned the technology as a significant step forward in diagnostic automation but emphasized that its goal is “to help, not replace.”
“This is a real step towards medical superintelligence. This research is just the first step of a long and exciting journey,” said Mustafa Suleyman, CEO of Microsoft AI.
“We’re excited to continue testing and learning with our healthcare partners in search of better and more accessible care for people everywhere,” he added.
Photo and video: Unsplash. This content was created with the help of AI and reviewed by the editorial team.
