The solipsistic society: generative AI and the death of shared reality

Generative AI is usually discussed as a threat because it can create convincing fakes, fictions, and falsehoods. I argue that the greater danger may be that it makes genuine evidence increasingly difficult to trust. If photographs, videos, voices, documents, and other forms of evidence can all be fabricated, doubt becomes a reasonable response even when the material is real. As verification becomes harder, people may rely more heavily on what they experience themselves and on smaller circles of people and institutions they already trust. I suggest that this could produce a more solipsistic society in which different political, social, and economic groups increasingly inhabit separate versions of reality. Politics may become more clan-like, economic exchange may retreat towards trusted networks, outsiders may become more vulnerable, and institutions may lose their ability to resolve disputes between groups. In the extreme, violence and force may increasingly replace agreed ways of resolving disputes. None of this is inevitable. However, the erosion of shared reality is developing as a creeping crisis, with each small adjustment appearing manageable until collective recognition and response become increasingly difficult. AI may therefore threaten humanity by destroying society without ever needing to destroy human beings.