Angelina YangFounder and Host, TwoSetAI · AI builder for visibility and discoverability

Hallucination is Just a Selection Error

19 Aug 2026 · 1 min

Watch on YouTube

0:34

In this brief segment, the speaker argues that hallucinations in generative models are not mysterious failures but simply selection errors arising from an extremely large discrete output space. Because the set of possible outputs is so vast, the model’s choice can appear as if it is inventing information. By constraining the output to a truly discrete, lower‑dimensional set, the behavior becomes deterministic and the phenomenon of hallucination disappears. The core claim is that hallucination is a perceptual artifact of high‑dimensional output spaces rather than a fundamental flaw in the model’s reasoning.

Chapters

Will hallucination be a problem here? Hallucination is an interesting word. Hallucination only matters on the output itself. So if you bound the output into discrete outputs, then hallucination is actually just a selection error. It's just such large discreteness that it doesn't feel discrete. And so if we reduce that dimensionality down to something that truly does feel discrete, then that's how determinism will work. Hallucinations only feel like a hallucination because the possible set of outputs is so large that it feels like it's just making things up.