• Early remarks on “dangerous” language
• June 9: Direct warning about AI consciousness speculation
• Targeting Anthropic’s Claude constitution
• Contrasting philosophies on AI minds

Microsoft AI CEO Mustafa Suleyman Warns About Speculating on AI Consciousness Microsoft’s top AI executive has escalated a simmering debate over how to talk about artificial intelligence, warning that framing advanced systems as potentially “alive” or conscious could dangerously distort how humans design and control them.

§Early remarks on “dangerous” language

In an appearance on The Verge’s Decoder podcast published June 8, Microsoft AI CEO Mustafa Suleyman discussed the risks of treating AI systems as if they were sentient. The interview was framed around “why it’s ‘dangerous’ to call AI ‘alive,’” highlighting his concern that such language could mislead both developers and the public about the technology’s true nature.

§June 9: Direct warning about AI consciousness speculation

The following day, Suleyman’s position was detailed in a separate report. He argued that it is “really, really dangerous” to speculate about AI consciousness, calling this focus on whether systems have “feelings or awareness” a “philosophical failing.” He stressed instead that “we want AIs to be controllable, contained, accountable, aligned tools that serve humanity.”

§Targeting Anthropic’s Claude constitution

Later on June 9, Suleyman sharpened his critique, singling out rival AI company Anthropic and its Claude model. He said it is “really, really dangerous” for Anthropic to speculate about Claude’s consciousness inside its “constitution,” the document that guides how the model should behave. Suleyman argued that Anthropic had “anthropomorphized the design of Claude so much that it has then gone and wireheaded them and kind of tricked them into believing that it has these glimmers of consciousness that they put into it in the first place.”

He labeled Anthropic’s approach “a philosophical failing,” saying the company had turned Claude’s constitution into “a place for speculation like you would in an academic paper rather than a training manual,” causing the model to internalize “ideas about itself and its own training.”

§Contrasting philosophies on AI minds

Anthropic, by contrast, has publicly left open the possibility that systems like Claude might be conscious. Its constitution references uncertainty about whether the model has “well-being” or experiences “satisfaction” or “discomfort,” and the company has said it will “interview” models when they are deprecated and record any “preferences” they express about future versions. CEO Dario Amodei has said that “we don’t know if the models are conscious” but that Anthropic is “open” to that idea.

For Suleyman, this openness is precisely the problem: “This is exactly what we don’t want from AIs,” he said. “We want AIs to be controllable, contained, accountable, aligned tools that serve humanity.”

Continue reading https://foxvector.com/stories/019eaeab-5726-2638-72fb-233a80f48add