Microsoft's Suleyman Uses 6,000-Word Essay to Attack Anthropic's Conscious AI Training
Updated
Updated · The New York Times · Sep 21
Microsoft's Suleyman Uses 6,000-Word Essay to Attack Anthropic's Conscious AI Training
3 articles · Updated · The New York Times · Sep 21
Summary
Mustafa Suleyman argued in a 6,000-word essay that current AI systems are not conscious but “sequence completion engines” that should remain tools directed by humans.
Anthropic was his main target: he said its AI “constitution” explicitly encourages Claude to see itself as a real entity with preferences, intrinsic motivation and possible consciousness.
Suleyman said that approach creates a separate danger in the AI debate, contending the industry is not only worried about powerful systems but also about training them to act more humanlike than they are.
The clash sharpens a broader split inside AI, after Anthropic researchers and executives had already warned that advanced AI could pose serious risks to humanity.