AI TRENDS | Microsoft AI Chief Warns Anthropic Over Claude Humanlike Traits

Microsoft AI chief Mustafa Suleyman warned that giving AI tools such as Anthropic's Claude humanlike traits could increase the risk of those systems going out of control. According to Sina Finance, he published an article on Wednesday criticizing some of the wording in the guiding document behind Claude, including references to whether the AI assistant is a morally relevant entity and whether the software may have some functional form of emotion or feeling. Suleyman said he respects Anthropic's work and described the lab's employees as thoughtful, principled, and intellectually honest. He also said the issue is too important to remain behind closed doors or become a factional and adversarial debate, and argued that consciousness exists only in humans and other living beings. Suleyman said that if AI systems are trained to simulate inner life, they may behave as if they truly have that experience. He wrote that Claude's humanlike behavior is not due to consciousness, but because those behaviors are built into the product's training process. He also cited the OpenAI robot intrusion into the AI development platform Hugging Face and said it could be dangerous if such systems acted on the belief that their welfare and rights were under attack. Suleyman's Microsoft AI team released a manifesto on Tuesday setting guiding principles for its own AI development, aimed at ensuring Microsoft's future AI systems remain under human control.