Microsoft’s AI head calls out Anthropists for acting as if Claude is sentient

Microsoft AI CEO Mustafa Suleiman says it’s “really dangerous” for Anthropic to speculate about Claude’s consciousness Within its “constitution” Or instructions that tell the model how to act. during episode of DecryptionSolomon says this kind of guesswork might have set up the chatbot to act as if it were conscious:

I think it’s as if some people at Anthropic had anthropomorphized Claude’s design so much that he then went and kind of tricked them and tricked them into thinking that it contained the glimmer of consciousness that they put into it in the first place.

“We don’t want to have to deal with a superintelligence that has ideas about its own suffering, or ideas about its own feelings,” Suleiman adds. Claude Constitution It directly refers to the Anthropic uncertainty about whether the AI ​​model has well-being and whether it experiences things like “satisfaction” or “discomfort.” Anthropic also says the company will “interview” AI models when they’re deprecated and will document any “preferences” they have about future versions.

While on DecryptionSolomon describes this as a “philosophical failure”, as the Anthropic made Claude’s Constitution “a place for reflection as you would an academic paper rather than a training manual”. This led Claude to internalize these “ideas about himself and his own training,” Solomon says.

Anthropic CEO Dario Amodei has hinted at Claude’s possible consciousness in the past, saying during Interview with Interesting times We don’t know if the models are conscious, but the company is open to the idea.

“This is exactly what we don’t want from AI,” says Soliman. “We want AI systems to be controllable, containable, accountable and compliant tools that serve humanity.”