Microsoft has launched a new research initiative, the MAI Superintelligence Team, aimed at developing advanced artificial intelligence systems that perform better than humans in specific domains, starting with medical diagnostics.
The company’s AI chief, Mustafa Suleyman, confirmed that Microsoft will highly invest in the new unit, which he described as a step toward “medical superintelligence” capable of identifying diseases faster and more accurately than current methods.
He said the company expects “a line of sight to medical superintelligence in the next two to three years.”
At the heart of this new direction is Microsoft’s MAI Diagnostic Orchestrator (MAI-DxO), an AI system benchmarked against real-world medical records from the New England Journal of Medicine.
The model reportedly achieved up to 85% accuracy in diagnosing complex medical cases, four times higher than experienced physicians, while operating at a lower cost. The system will serve as the foundation for Microsoft’s vision under the new team.
Suleyman, who co-founded DeepMind before joining Microsoft, explained that the company is not trying to build an all-purpose artificial intelligence that mirrors human thinking.
Instead, it is pursuing what he calls “Humanist Superintelligence”, AI designed to solve concrete, high-value problems with minimal risk. “Humanism requires us to always ask the question: does this technology serve human interests?” he said.
The new initiative gives Microsoft a unique standing among competitors like Meta, OpenAI, and Safe Superintelligence Inc., which are competing to create general-purpose systems that can perform across multiple fields.
Suleyman argued that such generalist models are difficult to control and align with human values, while Microsoft’s specialist approach can be guided within clear ethical and regulatory boundaries.
Leading the MAI Superintelligence Team is Karen Simonyan, a highly regarded AI researcher and former DeepMind scientist. Microsoft also plans to expand the team by recruiting talent from top laboratories, mirroring recent aggressive hiring strategies in the industry.
Suleyman stressed that this project is not another kind of marketing but is a focused attempt to leverage AI for measurable human benefit.
He envisions that diagnostic AI could “increase our life expectancy and give everybody more healthy years, because we’ll be able to detect preventable diseases much earlier.”
