Particle.news
Download on the App Store

Microsoft Launches In-House Superintelligence Bid as Suleyman Warns on Control

Suleyman says Microsoft will build frontier models internally under a humanist safety posture.

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.