Microsoft’s new AI model is capable of delivering 4x more accurate diagnosis compared to human doctors. The company is delivering high-end AI systems that are being designed to handle complex medical cases and it seems the round between AI and humans is being handsomely won by the former. The new AI model from Microsoft called AI Diagnostic Orchestrator or MAI-DxO has used case studies published by the New England Journal of Medicine to test its abilities that are clearly far higher than a human doctor can deliver. 

We have heard about the future of healthcare will involve AI but this is the first time where an actual AI model is massively outperforming even the most experienced doctors in their field. 

Microsoft’s AI Doctor Pitch 

The company’s tests with the AI model is claimed to have got the right diagnosis 85.5 percent of the time, while the group of doctors from the US and UK got it right just 20 percent of the lot. This is why the ‘AI is 4x better than human doctors’ claim is being made. But how have these tests been done and what parameters have been utilised to get the AI doctor pitted against the human doctor? Microsoft gives the full details of these tests which involve some of the rival AI models. 

Microsoft used the o3 model from OpenAI, Claude from Anthropic, Gemini from Google and Llama from Meta among others. The MAI-DxO was involved with all these AI models that made them work like a team of doctors that shared ideas and solved these medical cases. Microsoft pointed out that the MAI-DxO with o3 model delivered the best results. 

That’s not all, reports say this combo was able to solve more cases at a lower cost than the human doctors involved in the comparison tests. 

Fair Contest? 

Using AI models to treat patients will be pathbreaking but the tech does rely on more sources than human doctors, especially those who have crafted their expertise with toil and many years of learning? Microsoft understands these limits of the AI model and states that using the technology for regular health problems could become a challenge. 

The AI model needs to be fed data, research materials written by human doctors. It is again apparent that AI can outperform humans, in this case doctors, but the future will invariably have both involved rather than replacing the other.