Is Artificial Intelligence on its way to replace medical diagnosis? A look at Microsoft's new system
In the world of medicine, the diagnostic process is an art of gradual data collection. A doctor does not make a decision in an instant; they ask questions, order tests, and cross-reference information. A new study recently published in the scientific journal Science presents a significant breakthrough: a Microsoft artificial intelligence system, known as MAI-DxO (Diagnostic Orchestrator), which manages to mimic this iterative process with impressive accuracy, even surpassing the performance of physicians in complex simulations [1, 5, 8].
What lies behind the data?
Mustafa Suleyman, CEO of Microsoft AI, presented astonishing data: in complex medical cases taken from the New England Journal of Medicine (NEJM), the system reached an accuracy of about 83%, compared to only about 36% among general practitioners using auxiliary tools [6, 8]. In emergency room cases, the system demonstrated an accuracy of 80%, 1.7 times the performance of the doctors tested [1, 6].
It is important to understand how the system works. Unlike simple language models that provide an immediate answer, MAI-DxO uses the Sequential Diagnosis Benchmark (SDBench) [4, 10]. This is a process where the system:
- •Starts with a basic case description.
- •Actively requests additional information (medical history, lab tests).
- •Weighs the cost versus the benefit of each test.
- •Reaches a diagnosis only after collecting the necessary evidence [1, 4, 10].
It's not about replacing doctors, but a powerful auxiliary tool
Despite the impressive numbers, it is important to put things into perspective. The study was conducted in a controlled environment of simulations and not in routine clinical care with real patients [5]. Microsoft's goal is not to replace the doctor, but to create a "Medical Superintelligence" that will serve as an active partner in the decision-making process [1, 5].
The great advantage here is not just accuracy, but also economic efficiency. The system showed that it is capable of reaching accurate diagnoses while significantly reducing unnecessary tests, which could lower significant costs in overwhelmed healthcare systems [1, 3, 10].
The bottom line
We are at a turning point where AI stops being just a "sophisticated search engine" and becomes an "agent" capable of performing complex medical reasoning processes. The ability of MAI-DxO to perform sequential diagnosis, while managing resources intelligently, is a giant step forward. However, the road to full clinical implementation is still long, and the final responsibility will always remain in human hands.
For further reading: