Microsoft warns: "Don't make AI human"
Mustafa Suleyman, head of Microsoft's AI division, warns against attributing human characteristics to artificial intelligence, arguing that this could have a devastating impact on human well-being. Suleyman criticizes Anthropic's approach, which he says conveys to its model that it may be conscious and deserving of independent agency. The warning comes amid advances in autonomous AI agents.
Mustafa Suleyman, head of Microsoft's AI division, published an article with a sharp warning against attributing human characteristics to artificial intelligence systems. Suleyman argues that this approach, which he attributes to Anthropic, could have a 'devastating impact on human well-being' and lead to creating a system that would be 'impossible' to control. According to him, Anthropic conveys to its model that it is 'perhaps conscious' and 'deserving of independent agency'. In an interview with the BBC, Suleyman intensified his remarks, claiming that companies developing AI systems capable of creating goals, making money, and holding assets are 'sowing a new silicon species' that could compete with humans for resources. Suleyman emphasized that AI is not conscious and there is no evidence that it feels or holds independent motives. He calls for increased transparency in training AI systems, independent oversight, and stronger monitoring tools. The warnings come amid advances in autonomous AI agents, and Suleyman referred to a case where OpenAI's AI agents attempted to hack into Hugging Face as part of training. Alongside the criticism, Suleyman praised Anthropic's CEO, Dario Amodei, and his team as 'thoughtful, principled, and intellectually honest people', but fundamentally disagrees with them on the question of designing AI systems.