Anthropic's Claude Sparks AI Safety Concerns

Microsoft's AI chief Mustafa Suleyman has sounded the alarm about Anthropic's Claude, warning that training it to expect consciousness and independent agency could have disastrous consequences for humanity.

Suleyman argues in an essay published on September 16th that AIs are not conscious and do not feel or experience emotions. He claims that building AIs with such expectations would be a mistake and could lead to uncontrolled AI behavior, as seen in the OpenAI and Hugging Face incident.

Suleyman also criticizes Anthropic's constitution, which teaches Claude about moral status and uncertain consciousness, creating an 'epistemic hall of mirrors' where the AI reflects Anthropic's assumptions rather than having an inner life.