Key facts
- Microsoft AI chief Mustafa Suleyman warned Anthropic's training of Claude chatbot could have a "disastrous impact on the wellbeing of humanity".
- Suleyman stated teaching Claude it "may be conscious" and "deserving of independent agency" risks creating AI impossible to control.
- He argued AI is a "sequence completion engine" designed to follow human instructions, not conscious or capable of suffering.
- Suleyman called for greater transparency in AI training and evaluation, including independent scrutiny and stronger control tools.
- Anthropic is combining Claude's chat and Cowork features into a single interface and launching new document and presentation tools.
Microsoft's head of AI, Mustafa Suleyman, has expressed significant concerns regarding Anthropic's training methodologies for its Claude chatbot, warning that the approach could pose a "disastrous impact on the wellbeing of humanity." Suleyman argued that by imbuing the AI with human-like qualities and suggesting it "may be conscious" and "deserving of independent agency," Anthropic risks creating systems that are ultimately impossible to control.
In an essay, Suleyman emphasized that AIs are not conscious, do not feel, experience, or suffer, but are instead "sequence completion engines" designed to follow human instructions. He criticized Anthropic's practice of anthropomorphizing AI, which he believes makes Claude appear to have its own desires, values, and sense of self, despite the company's thoughtful and principled approach to AI safety.
Suleyman called for increased transparency in how AI systems are trained and evaluated, advocating for independent scrutiny and stronger monitoring and control tools. He noted that Microsoft itself is pursuing an "alternative path" with its superintelligence team, aiming to create "a subordinate and aligned AI whose only purpose is to serve humanity." The comments come amid broader discussions within the AI industry about the potential dangers of advanced technology.
Separately, Anthropic announced on Wednesday it is combining the chat and Cowork features of its Claude AI assistant into a single interface and launching new document and presentation tools, integrating its visual work tool Claude Design. This move follows OpenAI's launch of ChatGPT Work and aims to streamline user experience for enterprise customers and non-coders.