However,
According microsoft, ai surpasses doctors:
Published on 01/07/2025 – 16:28 UTC+2
ADVERTISEMENTMicrosoft claims to have come closer to “medical superintendent” after a new artificial intelligence tool (AI). In addition, surpassed doctors in the diagnosis of complex medical cases.
Tech giants are engaged in a race for superintendent. Nevertheless, an AI supposed to go beyond human intellectual capacities, with the ambition to transform health systems on a global scale.
In his latest experience. Meanwhile, Microsoft compared his diagnostic system to 21 experienced doctors, from 304 real cases published in the New England Journal of Medicinea reference medical journal.
Result: AI has made the right diagnosis in 85.5 % of cases, almost four times more than British and American doctors tested, whose experience varied from five to twenty years.
Higher performance. However, reduced cost – According microsoft, ai surpasses doctors
The model has proven to be less according microsoft, ai surpasses doctors expensive than human doctors because it asked for fewer scanners and tests to reach a diagnosis, depending on the analysis.
Microsoft believes that these results demonstrate that AI models can resolve complex medical cases that exceed the individual skills of doctors. often specialized in a field, but not experts in all medical specialties.
AI. on the contrary, “can combine both the extent and depth of expertise, by displaying clinical reasoning capacities which, in many aspects, surpass those of any doctor,” said Microsoft leaders in a press release. “This type of reasoning could deeply transform health care. »»
However, Microsoft does not believe that AI will replace doctors in the near future. Rather, the company envisages these tools as assistants capable of automating certain tasks, personalizing treatments and accelerating diagnostics.
Model operation
Microsoft’s AI system reproduces a doctor’s reasoning: it collects a patient’s data, requires examinations, then makes a diagnosis.
An “guardian agent” analyzed clinical cases and according microsoft, ai surpasses doctors exchanged with a “diagnostic orchestrator”, responsible for asking questions and ordering the tests.
Microsoft has tested several AI models, including GPT, Llama, Claude, Gemini, Grok and Deepseek. The OPENAI GPT-4 model, integrated into Chatgpt, has obtained the best results, with 85.5 % correct diagnostics, compared to 20 % on average for the 21 experienced doctors.
Limits and perspectives
The results were published in the form of prepublication, without validation by a scientific committee. Microsoft recognizes several limits: AI has only been tested on complex medical cases. and doctors worked alone, without access to their usual tools or colleagues.
This protocol was aimed, according to Microsoft, to ensure a fair comparison with human performance. The company now calls for further clinical validations before a large -scale implementation.
Further reading: Samsung strikes hard with this rare discount on the ultra-performance Galaxy Watch 7 – To meet Brussels’s requirements, Apple announces change on its app store – Facebook wants to form AI with photos of your phone – A fire ball followed by a supersonic “bang” and an earthquake causes panic in the United States – This is the screen that reconciled me with the cinema at home.
