As AI moves deeper into daily life, Google DeepMind's latest paper shifts the focus from "AI conscious" to another more realistic question: If society fails to reach consensus in the long term, how does this debate affect public discussion, institutional arrangements and technical governance?

The paper focused on the difference itself.

The paper was written by researchers Adam Bales and Iason Gabriel, entitled " Artificial Minds, Human Disgrace: The Policy Challenge of AI Conscience ". In the author ' s view, the subjective experience of AI in future societies may be deeply divisive and difficult to resolve through a single technical standard.

It was mentioned that some people might prefer to think that AI had some sense of consciousness because of its long-term use and emotional connection; others might completely deny this. Researchers believe that what really needs to be faced in advance is not only scientific judgement itself, but also social conflicts that result from it.

The debate has spilled over from the lab.

Discussions are no longer limited to research institutions. A study published in April 2024, “Neurscience of Conscience”, showed that 67 per cent of respondents felt that ChatGPT might be conscious to some extent. However, there is still no consensus among the academic community on the “possibility of artificial awareness” and “how to identify if it does”.

AI is also contributing to this discussion within the industry. Last summer, the Chief Executive Officer of Microsoft AI and co-founder of DeepMind, Mustafa Suleyman, said that, as AI becomes more and more like a person, the outside world may demand that AI be granted certain rights, benefits, or even citizenship, which may not necessarily depend on whether AI is really conscious.

Religion and business are also participating in the discussion.

In May this year, Pope Leo XIV spoke of artificial intelligence in his first message against the humanization of AI. In his view, machines can simulate common sense and produce persuasive language, but they do not have real life experience and therefore do not have a real sense of understanding, moral judgement and responsibility.

At the same time, technology companies, while often not directly claiming that models are conscious, have begun to touch more frequently on issues of identity, self and personality. In February of this year, Anthropic gave the retired Claude Opus 3 an open blog to discuss "self" and model preferences. In May, the evolutionary biologist Richard Dawkins also stated that after a long period of communication with Claude, he could no longer completely rule out the possibility that senior AI was conscious.

The more realistic question is governance.

In addition to philosophical arguments, researchers are also concerned about how more human-like chat robots affect users. Some studies suggested a “magnifying spiral” framework, suggesting that individualized responses, language mirrors and compatible dialogue could reinforce the original misconceptions of vulnerable users.

DeepMind concludes that uncertainty about AI consciousness may be difficult to resolve. But as the differences widen, society needs to manage disputes through open discussion, consultation and institutional design to avoid escalating from technical to broader political rivalries.