The Biological Argument for Consciousness
In a recent interview with Walter Isaacson, Microsoft AI CEO Mustafa Suleyman argued that artificial intelligence lacks consciousness, asserting that true sentience requires an organic, physical body. Suleyman stated that current AI models are “internally hollow” sequence completion engines that function by predicting the most likely next word in a sequence rather than engaging in human-like thought.
Suleyman linked consciousness to biological experience, noting that legal and political systems are built upon the capacity for suffering. He argued that pain and the biological substrate of experience—where pharmacology and physical sensation intersect—are the fundamental drivers of awareness and introspection. According to Suleyman, the ability to navigate physical environments and make choices based on sensory input is what defines a conscious system.
Criticism of Industry Agnosticism
Suleyman expressed significant disagreement with the approach taken by Anthropic regarding its AI model, Claude. He criticized the company for maintaining an agnostic stance on whether Claude possesses consciousness, noting that Anthropic’s core training documents include speculation regarding whether the AI has “consented” to its role or deserves compensation. Suleyman characterized this uncertainty as “crazy,” arguing that such speculation in foundational documents could lead millions of users to believe the technology is sentient.
Projections on Intelligence and Computation
Despite his dismissal of AI consciousness, Suleyman projected that AI will reach superhuman intelligence within a few years. He highlighted the rapid acceleration of computational power, noting a 1,000-fold increase in computation over the last three years. This growth has transitioned technology from GPT-3, which he noted struggled to complete sentences, to models capable of unbundling entire software applications from a single paragraph prompt.
Suleyman predicted that the next three years will see another 100 to 1,000-fold increase in computation applied to these models. He cited the capabilities of frontier models as evidence, mentioning instances where models have solved 700 mathematical problems in a single week—a feat that would take human mathematicians decades.
Ethical Red Lines and Future Risks
The Microsoft AI CEO established a “red line” for the industry: he stated he would call for an immediate global pause if consciousness were to emerge or if developers attempted to synthetically engineer consciousness without disclosure.
The discussion also touched upon the long-term trajectory of the species. Suleyman recounted a conversation with Google co-founder Larry Page, who suggested a transition toward a silicon-based successor species. Suleyman raised the ethical question of whether humanity has consented to introducing a species that may eventually become more capable than all humans combined.
While Suleyman’s views are widely discussed on platforms like Reddit, some observers have questioned the consistency of his position, noting the tension between his claim that AI is a “hollow” predictor and his prediction of its imminent superhuman capabilities.