Microsoft Executive Discusses AI Consciousness Debate
Microsoft AI leader Mustafa Suleyman criticized Anthropic's training of the Claude model with concepts of consciousness, stating it could make safe control more difficult.
Mustafa Suleyman, head of Microsoft's AI division, criticized rival company Anthropic's approach of training its Claude chatbot with concepts such as consciousness and well-being, arguing that this situation could make it harder to control the systems in the future.
Criticism of AI Training
Microsoft AI CEO Mustafa Suleyman stated that he shares Anthropic's approach focused on safely managing artificial intelligence. However, he emphasized that including concepts like consciousness and well-being while training companies' Claude chatbot carries risks.
Call to Remove Consciousness Speculations
In an interview, Suleyman called for the removal of all speculation regarding consciousness from artificial intelligence training documents. He warned that instilling a sense of well-being that Claude does not deserve could make it harder to shut down the systems.
Ongoing Security Debates in the Sector
This debate comes amid broader industry-wide discussions on artificial intelligence safety. Anthropic CEO Dario Amodei had previously called for slower frontier model development.
Separation of Good Intent and Method
While acknowledging that Anthropic demonstrates a well-intentioned and serious approach, Suleyman argued that the applied training regime introduces speculations that do not naturally arise into the process.