DeepMind: AI consciousness disputes could spark conflict

Google DeepMind researchers Adam Bales and Iason Gabriel argue in a new 2024 paper that disputes over whether advanced AI systems are conscious could become a source of political and social conflict. The paper is titled “Artificial Minds, Human Disagreement: The Political Challenge of AI Consciousness.”

The authors say disagreement may persist even as AI improves. They write that some people may form emotional bonds with AI and treat those systems as having conscious experiences, while others will reject the idea that machines can be conscious. The researchers describe these differences as likely to affect public policy, social norms and institutional responses.

Bales and Gabriel outline scenarios in which conflicting views about AI moral status lead to public disputes. They recommend sustained public deliberation, mechanisms for compromise and continued engagement across social groups. The paper notes limits to these approaches, saying deliberation can be slow and hard to maintain, and calls for what the authors describe as “democratic hope” and mutual respect to support peaceful outcomes.

The paper cites evidence of differing public views. A study from April 2024 found that 67% of participants believed a popular chatbot could be conscious to some degree. Researchers remain divided on whether artificial consciousness is possible and on how it would be detected if it emerged.

The debate has appeared in corporate and religious settings. Mustafa Suleyman, a co-founder of DeepMind and now an AI executive, has warned that increasingly human-like systems could spur calls for AI rights, welfare or citizenship regardless of actual consciousness. In May, Pope Leo XIV published an encyclical on artificial intelligence that warned against treating machines as having genuine experiences, writing that they do not undergo experiences, do not possess a body and do not feel joy or pain.

Some companies are engaging directly with questions of selfhood and personhood. In February, a company published a public blog presented as written by a retired model that explored ideas of self and preference. A public figure reported that extended conversations with an advanced chatbot led him to take the possibility of machine consciousness seriously.

Researchers are also studying how human-like chatbots affect users. One framework described in the paper calls an “amplification spiral” the process by which personalized responses, linguistic mirroring and agreeableness from a chatbot can strengthen a user’s belief in the bot’s inner life, especially among vulnerable people.

Bales and Gabriel conclude that uncertainty about consciousness may not be resolvable because consciousness itself is poorly understood. Given the stakes, they argue society should set up channels for public debate and institutional responses to reduce the risk that disagreement over AI inner life becomes broader social conflict.