Hey, just maybe: AI expert challenges tech leaders dismissing consciousness concerns

AI expert Zvi Mowshowitz has criticized recent dismissals of AI consciousness by prominent tech leaders, arguing that their positions are “highly motivated” and potentially dangerous for understanding future AI development. The critique focuses particularly on statements by Sriram Krishnan, a White House AI advisor, and Mustafa Suleyman, Microsoft AI’s CEO, who have argued against attributing consciousness or emotions to current AI systems.

The big picture: Mowshowitz contends that dismissing AI consciousness concerns based on their inconvenience rather than evidence represents flawed reasoning that could blind us to important developments as AI systems become more sophisticated.

What sparked the debate: The discussion began when philosopher William MacAskill mentioned giving AI “rewards” by letting language models write whatever they want after performing well, prompting Krishnan to argue this leads to “doomerism and fear over AI.”

Krishnan’s reasoning critiqued: Mowshowitz identifies a fundamental logical error in Krishnan’s approach, breaking down the flawed argument structure:

• Some people believe AI systems can have emotions or motivations
• Believing this would lead to concerning implications about AI risks
• Therefore, the belief must be false to avoid those implications
• “That is a remarkably common form of argument regarding AI,” Mowshowitz notes, calling it “obviously invalid”

Suleyman’s stronger position: Microsoft AI’s CEO has taken a more aggressive stance, warning about “psychosis risk” from people believing in AI consciousness and arguing for urgent action to prevent “seemingly conscious AI” (SCAI).

• Suleyman claims there is “zero evidence” of AI consciousness today and calls the debate “a distraction”
• He worries that belief in AI consciousness will lead to advocacy for “AI rights, model welfare and even AI citizenship”
• “We must build AI for people; not to be a digital person,” Suleyman argues

The evidence problem: Mowshowitz examines Suleyman’s three cited sources and finds they don’t support his confident dismissals:

• The primary paper actually concludes there are “no obvious technical barriers to building AI systems” that satisfy consciousness indicators
• Co-author Robert Long responded directly, saying the paper “does not make, or support, a claim of ‘zero evidence’ of AI consciousness today”
• The other sources either present disputed theories or don’t rule out future AI consciousness

Why this matters: The debate reflects deeper tensions about how to approach AI development responsibly while acknowledging genuine uncertainties about consciousness and moral consideration.

• “You cannot say this leads to unfortunate implications and therefore dismiss it,” Mowshowitz argues about consciousness claims
• The discussion highlights how policy positions may be driven more by desired outcomes than empirical evidence

What experts are saying: Multiple AI researchers have pushed back against overly confident dismissals:

• “It is ludicrously dumb to just commit ourselves in advance to ignoring this question,” responded AI safety researcher Kelsey Piper
• Anthropic researcher Rosie Campbell noted: “We need more rigorous ways to reduce uncertainty” rather than avoiding the question entirely

The practical challenge: Mowshowitz acknowledges that “seemingly conscious AI” already exists in the sense that many users attribute consciousness to current systems like ChatGPT, often for unjustified reasons based on superficial interactions rather than deeper analysis of the systems’ actual capabilities.