Microsoft’s AI CEO, Mustafa Suleyman, is once again urging researchers to avoid developing artificial superintelligence (ASI) — AI systems with reasoning abilities far beyond human intelligence. Speaking on the Silicon Valley Girl podcast, Suleyman said ASI should be treated as an “anti-goal”, not an aspiration.
Suleyman, who previously co-founded Google DeepMind, explained that envisioning a superintelligence that exceeds human control “doesn’t feel like a positive future.” According to him, the main issue is containment — once such a system exists, it might be nearly impossible to govern or align with human values.
Instead, Suleyman said Microsoft is pursuing what he calls “humanist superintelligence” — AI designed to amplify human capabilities while maintaining strong guardrails around ethics, safety, and alignment.
He also warned against confusing convincing AI responses with genuine consciousness. AI models may simulate human-like conversation, but that doesn’t make them sentient.
“These systems don’t suffer. They don’t feel pain. They’re not conscious,” he explained. “They’re generating sophisticated simulations.”
Earlier this month, Suleyman released a separate warning: researchers should stop framing AI development around the idea of machine consciousness. Consciousness, he said, is a biological phenomenon, and trying to recreate it in AI is not only misguided but a distraction from meaningful goals.
“If you ask the wrong question, you end up with the wrong answer,” he said. “And this is the wrong question.”
Why AI Leaders Disagree on Artificial Superintelligence
Suleyman’s cautionary comments come as the AI community remains deeply divided about when — or even whether — artificial superintelligence will emerge.
Some leaders believe ASI could arrive within the decade.
- Sam Altman, CEO of OpenAI, has repeatedly called AGI the company’s primary mission and believes superintelligent systems could massively accelerate global innovation and prosperity.
- Altman even suggested earlier this year that he expects ASI to appear before 2030.
- Demis Hassabis, DeepMind cofounder, shares a similar view, saying ASI might be possible within five to ten years.
Other leaders urge caution:
- Yann LeCun, Meta’s Chief AI Scientist, argues ASI is still “decades” away.
- He said at a Singapore event that intelligence doesn’t scale smoothly — throwing more data and compute at a model won’t automatically produce deeper reasoning.
Microsoft Faces Public Pushback Amid AI Expansion
In a related development, Suleyman recently defended Microsoft’s AI progress after criticism from users frustrated with the company’s rapid rollout of features like Copilot in Windows. Critics argue that Microsoft is pushing AI updates without fixing long-standing system issues or offering opt-out choices.
Suleyman responded sharply online, saying he finds it surprising that people dismiss groundbreaking AI advancements as unimpressive.
