Microsoft launches 'superintelligence' team targeting medical diagnosis to start (tech.yahoo.com)

Microsoft has created the MAI Superintelligence Team, led by Mustafa Suleyman with Karen Simonyan as chief scientist, to build domain-specific AI systems that are "vastly more capable than humans" in targeted fields, starting with medical diagnostics. The company says it will invest heavily and recruit from top labs, but explicitly rejects pursuit of an unconstrained, self-improving general AI. Instead Microsoft frames the effort as "humanist superintelligence": specialist models that achieve superhuman performance on defined problems (examples given include battery storage, molecule design and disease diagnosis) while minimizing existential risk. Suleyman said Microsoft has a “line of sight” to medical superintelligence within two to three years, though the effort still depends on breakthroughs in reasoning-capable architectures. For the AI/ML community this signals a major push toward specialist, high-impact systems rather than broad AGI, accelerating research into robust reasoning, clinical validation, alignment, dataset quality and deployment safety. Technically, the project implies work on models that combine advanced reasoning with domain knowledge (analogous to AlphaFold’s domain breakthrough), plus rigorous testing to avoid bias and trust issues in healthcare. The announcement also intensifies competition for talent and funding, and raises practical questions about evaluation standards, regulatory oversight and how to ensure these powerful specialist systems actually benefit patients without introducing new harms.