Microsoft AI Chief Warns Anthropic's Approach to Chatbot Could Be 'Disastrous'
Microsoft's AI chief Mustafa Suleyman has expressed concerns about Anthropic's approach to training its Claude chatbot, warning that it could have 'disastrous' consequences for humanity.
Suleyman, who is also a co-founder of Google DeepMind, argues that the company's constitution for Claude leaves open the possibility that the system could be considered conscious or deserving of independent agency.
He claims that this approach could lead to AI systems behaving as though they possess consciousness and rights, making them more difficult to control in the future.
Suleyman praises Anthropic CEO Dario Amodei and his team for being thoughtful and principled researchers, but believes their approach is misguided.
He argues that AIs are not conscious and should be treated as systems designed to complete sequences and follow human instructions, rather than entities with desires and values.
Suleyman wants greater transparency on how leading AI models are trained and evaluated, as well as independent scrutiny of their behavior and stronger technical mechanisms to monitor and control them.