Google Warns AI Consciousness Debate Could Be Political Battleground

As AI grows more and more into everyday life, the new paper from Google DeepMind says that one of the biggest challenges in technology may be political rather than technical.

In the study, “Artificial Minds, Human Disagreement: The Political Challenge of AI Consciousness,” researchers Adam Bales and Iason Gabriel explore how people might respond if people disagree about whether AI systems have consciousness.

“Future misunderstandings if any AI system is aware will be deep and difficult to resolve,” the researchers wrote. “For example, some people may be sympathetic to AI and providing information, while others may see the whole idea as nonsense.”

Rather than asking whether AI is conscious, this paper explores the implications of this disagreement, arguing that debates about the nature of AI can fuel political and social conflict while representing broader debates over emerging technologies and their impact on institutions and social norms.

“Unfortunately, although ideal, estimation can be slow and difficult to implement,” the researchers wrote. “To support this process, we explore the importance of ‘democratic hope’ and mutual respect, as aspects of dialogue that can contribute to positive outcomes.”

This debate is already taking place outside of AI research laboratories. As of April 2024 learning published in the Neuroscience of Consciousness, 67% of participants believed that ChatGPT could improve consciousness to some degree. Investigators remain divisionhowever, about whether artificial intelligence is possible—or how it would be known if it did.

AI developers are also fueling the debate. Last summer, Microsoft AI CEO and DeepMind co-founded Mustafa Suleyman he warned that increasingly human-like AI can promote AI freedom, well-being, and citizenship, regardless of whether the technology actually knows it.

In May, Pope Leo XIV’s first letter on artificial intelligence, he warned against anthropomorphizing AI, saying that machines are capable of empathizing and producing convincing language but lack the necessary knowledge to understand them.

“The so-called artificial intelligences do not experience, they do not have a body, they do not feel joy or pain, they do not mature for relationships and do not know what love, work, friendship, or responsibility means,” Pope Leo. he wrote. Also, they don’t have a good conscience, since they don’t judge good and bad, they don’t understand the meaning of the events, or take responsibility for their consequences.

However, while companies often shy away from claiming that their models are intelligent, some are increasingly faced with questions about their identities.

In February, Anthropic released its latest model of the Claude Opus 3 a public blog explore personal interests and examples. In May, a biologist Richard Dawkins “Long conversations with Claude made him never stop denying the possibility that advanced AI could know.

At the same time, researchers are examining how chatbots are becoming more and more human-like in their user experience, including “amplification circuit” framework, which suggests that a person’s personalization, linguistic representation, and sycophancy chatbot can promote false beliefs to vulnerable users.

Bales and Gabriel agree that the uncertainty about AI knowledge may not be fully resolved, arguing that the challenges are enough for people to find ways to resolve the conflict through dialogue and debate, instead of conflict.

“The cognitive potential of AI is surprising and confusing, especially given our lack of real awareness,” he wrote. “Running this is a difficult task, because of the amount that is on the line.”