In “generative adversarial” models, two elements combine to learn about some aspect of the world: the “generative” bit aims to predict it as precisely as possible; the “adversary” does its best to decide whether what it is looking at is the real world or the output of the generative model. The generative model constantly ups its game to masquerade as the real McCoy; the adversary keeps honing its connoisseurship to distinguish the authentic from the fake. Something similar happens in the brain. The “adversary” in the human brain, charged with reality checking, keeps watch from our huge frontal lobes: Area 10, in particular, at the tip of the frontal cortex, becomes active in tasks requiring us to decide whether items were seen or imagined. It is smaller and less active in people with psychosis than in healthy people, especially so in people with psychosis who hallucinate.
— Read on www.theguardian.com/books/2025/feb/10/the-big-idea-how-do-our-brains-know-whats-real
Leave a Reply