Particle.news
Download on the App Store

Microsoft’s Suleyman Calls for Control‑First AI, Outlines ‘Humanist’ Superintelligence Plan

Microsoft says a revised OpenAI pact now permits in‑house frontier research under stricter safeguards.

Overview

  • Mustafa Suleyman said people should be both amazed by AI and afraid of it, warning that the world would worsen if humans lost control.
  • He promoted a Humanist Superintelligence approach focused on systems that are carefully calibrated, contextualized, and limited rather than highly autonomous.
  • Suleyman said Microsoft has reached AI self‑sufficiency after removing prior contractual limits tied to its OpenAI deal, with new terms reported to allow independent AGI pursuits.
  • Microsoft’s new superintelligence unit targets practical gains, citing goals like medical AI that it claims hit 85% accuracy on difficult diagnoses versus 20% for human doctors, along with personalized education tools and clean‑energy advances.
  • He cautioned that no developer or policy expert has a reassuring plan to contain ever‑smarter systems, noted the company’s safety‑first path could be costlier or less efficient, and projected a year or two before the team produces frontier models.