Microsoft says AI system better than doctors at diagnosing complex health conditions | Artificial intelligence (AI)

Microsoft has revealed the details of an artificial intelligence system that works better than human doctors at complex health diagnostics, creating a “path to medical superintendent”.

The company’s AI unit, led by the pioneer of British technology Mustafa Suleyman, has developed a system that imitates a panel of expert doctors who attack the cases “diagnostically complex and intellectually demanding”.

Microsoft said that when associated with the AI ​​Advanced O3 model from Openai, its approach has “resolved” more than eight of the 10 case studies specially chosen for the diagnostic challenge. When these case studies have been judged on practical doctors – who did not have access to colleagues, textbooks or chatbots – the precision rate was two out of 10.

Microsoft said it was also a cheaper option than using human doctors because it was more efficient during order tests.

Despite the demonstration of the potential cost savings in his research, Microsoft played the implications on work, saying that she thought that AI would complete the roles of doctors rather than replacing them.

“Their clinical roles are much wider than simply making a diagnosis. They must sail in ambiguity and strengthen confidence with patients and their families in a way that AI is not set up to do,” wrote the company in a blog post announcing research, which is subject to a peer review.

However, the use of the slogan “Path to Medical SUPERITELLIGENT” increases the prospect of a radical change on the health care market. While artificial general intelligence (AGA) refers to systems that correspond to human cognitive capacities to a given task, superintelligence is an equally theoretical term referring to a system that goes beyond human intellectual performance at all levels.

Explaining the justification for research, Microsoft has raised doubt about the capacity of AI to mark exceptionally well on examination of medical licenses in the United States, a key test to obtain a medical license in the United States. He said that multiple choice tests have favored the memorization of responses on an in -depth understanding of a subject, which could help “overestimate” the competence of an AI model.

Microsoft said he was developing a system that, like a real world clinician, took steps step by step – such as asking specific questions and asking for diagnostic tests – to reach a final diagnosis. For example, a patient with symptoms of coughing and fever may require blood tests and pulmonary radiography before the doctor arrived during a pneumonia diagnosis.

The new Microsoft approach uses complex case studies of the New England Journal of Medicine (NEJM).

The Suleyman team has transformed more than 300 of these studies into “interactive cases challenges” which he used to test his approach. Microsoft’s approach used existing AI models, including those produced by the Chatgpt developer, OpenAi, Meta by Mark Zuckerberg, Anthropic, Elon Musk’s Grok and Google’s Gemini.

Microsoft then used a tailor -made agent type AI system called “diagnostic orchestrator” to work with a given model on the tests to be controlled and what the diagnosis could be. The orchestrator indeed imitates a panel of doctors, which then offers the diagnosis.

Microsoft said that when it was associated with the Openai advanced O3 model, it has “resolved” more than eight of the 10 NEJM case studies – compared to a success rate of two out of 10 for human doctors.

Microsoft said that his approach was able to handle a “extent and a depth of expertise” which went beyond individual doctors because he could run several medical disciplines.

He added: “The scaling of this level of reasoning – and beyond – has the potential to reshape health care. The AI ​​could allow patients to heal the routine aspects of care and provide clinicians with advanced decision support for complex cases. ”

Microsoft recognized that his work was not ready for clinical use. Additional tests are necessary on his “orchestrator” to assess his performance on more common symptoms, for example.