Overview
- A revised OpenAI partnership now allows Microsoft to pursue AGI on its own or with third parties, lifting earlier limits on model scale.
- The new MAI Superintelligence team is tasked with frontier-grade research, with Suleyman estimating a one-to-two-year path to in-house frontier models.
- Microsoft is pouring resources into compute, including Nvidia collaborations, AI-optimized Azure infrastructure, and custom chip clusters to reduce dependence on external hardware.
- Suleyman maintains that superintelligence must stay under human control and says the field lacks a reassuring method to contain or align systems that outsmart people.
- The company has strengthened responsible-AI leadership, including hiring Trevor Callaghan, and is targeting practical uses such as healthcare diagnostics, personalized learning, and clean energy tools.