AI

“Consensual hallucination” is a term coined by author William Gibson in his 1984 novel Neuromancer to describe cyberspace, defining it as a “graphical representation of data abstracted from the banks of every computer in the human system”. It refers to a shared, artificial reality that billions of people agree to act as if it is real. The term highlights the strange, shared reality of the digital age, where a non-existent space (cyberspace) has profound effects on the real world.

1984-2024ish – the “cyberscape” exposed many avatars, following tech evolutions, trends and hypes. From hyperfiction to cybalterns, from automation to augmentation to agency, the effects impact everyday in ever surprising modes.

“AI as Consensual Hallucination” alludes to the collaborative, often unverified, nature of generative AI outputs, where users and models participate in creating plausible-sounding information, but …. While AI hallucination refers to AI generating inaccurate or “surreal” data, treating it as “consensual” implies human reliance on these outputs for speed or creativity, often blurring the line between fact and fiction. Some researchers propose “AI mirage” over “hallucination” to emphasize that these errors are not conscious “delirium” but structural products of training data. Structural and infrastructural are the dimensions of this cyberscape that are to be researched.

Welcome to the illusions of immersion and projection in mixed realities.

More (in French) in the IA-Lab series in 2026, co-org by LaScam, SACD, MEDAA, Bela.