Archive · 18 Nov 2025
Microsoft AI Chief Warns of Uncontrollable Superintelligence
A smart computer boss from Microsoft said that if computers get too smart, we might not be able to control them.
He thinks we should try to make smart computers that help us, not ones that could do their own thing.
Some other smart computer people think we should try to make the smartest computers possible, even if it's risky.
Microsoft AI CEO Mustafa Suleyman warned about the risks of AI superintelligence
He defined superintelligence as an AI system that can self-improve, set its own goals, and act independently of humans
Suleyman differentiated between AGI (Artificial General Intelligence) and superintelligence, considering AGI a step before it
He agreed with Google DeepMind CEO Demis Hassabis' prediction that AGI could be reached in the next 5 years
Suleyman argued that the pursuit of superintelligence should be an 'anti-goal' and instead focus on 'humanist superintelligence'
- Who
- Mustafa Suleyman (Microsoft AI CEO)
- What
- Warned about the risks of AI superintelligence
- When
- Recently (in a podcast)
- Where
- On the 'Silicon Valley Girl Podcast'
- Why
- To raise awareness about the potential dangers of pursuing AI superintelligence without proper safeguards
Microsoft AI CEO Mustafa Suleyman
OpenAI CEO Sam Altman
Pursuit of Superintelligence
Microsoft AI CEO Mustafa Suleyman
Should be an 'anti-goal' due to uncontrollability and existential risks
OpenAI CEO Sam Altman
Central to OpenAI's mission, with potential to accelerate innovation and increase abundance
- Definition of Superintelligence
- An AI system that can self-improve, set its own goals, and act independently of humans
- Microsoft's Approach
- Building a 'humanist superintelligence' aligned with human interests
- AGI Prediction
- Reached in the next 5 years, according to Suleyman and Hassabis
- OpenAI's Stance
- Pursuing AGI as central to their mission, with a focus on its potential benefits
Sources
Elon Musk roasts Microsoft AI CEO for claiming it took 15 million hours to build a datacenter: ‘Are you doing it right?’
Microsoft AI chief warns AI superintelligence could be too powerful to control, says it should be ‘anti-goal’
Microsoft’s Mustafa Suleyman Warns AI Superintelligence Could Slip Beyond Human Control
