Microsoft is launching a new initiative, the MAI Superintelligence Team, to develop artificial intelligence that can outperform humans in specific fields, beginning with medical diagnostics. Mustafa Suleyman, Microsoft’s AI chief, will lead this significant effort.
Microsoft plans to invest heavily in the project. This move follows similar ventures by other tech giants, such as Meta Platforms and Safe Superintelligence Inc., all aiming for major technical advancements in AI. While some remain skeptical about these ambitious goals without breakthroughs, Microsoft is pushing forward.
Suleyman mentioned that Microsoft AI will continue to recruit top talent from other labs and will also staff the new team with current researchers, including Karen Simonyan as chief scientist. Meta, for example, has offered substantial signing bonuses to attract prominent AI experts.
However, Microsoft’s approach differs significantly. Suleyman explained that the company isn’t pursuing “infinitely capable generalist” AI, unlike some competitors. He doubts we could control completely autonomous, self-improving machines, despite ongoing research into AI safety.
Instead, Microsoft envisions “humanist superintelligence”—technology designed to solve specific problems for human benefit. “Humanism requires us always to ask the question: does this technology serve human interests?” Suleyman stated.
AI experts have long debated the potential dangers of advanced AI alongside issues such as algorithmic bias and trustworthiness. Suleyman aims to focus his team on creating specialized AI models that achieve “superhuman performance” with “virtually no existential risk whatsoever.” He cited examples such as AI that optimizes battery storage or develops new molecules, including AlphaFold, DeepMind’s AI for predicting protein structures. Suleyman himself co-founded DeepMind.
Medical diagnosis has long been an area of interest for AI, and Microsoft has focused on it for some time. Suleyman believes they have a clear path to achieving “medical superintelligence” within the next two to three years. This effort relies on AI that can reason through complex problems, but it will still require breakthroughs. If successful, he believes this AI could “increase our life expectancy and give everybody more healthy years, because we’ll be able to detect preventable diseases much earlier.”
