kaal:claim:6421319-023
Probabilistic language models hallucinate because they are trained to predict statistically likely token sequences rather than to verify propositional truth, so the error is intrinsic to the substrate: plausibility and truth are orthogonal properties in high dimensional token space.
Source quote, verbatim
Probabilistic large language models hallucinate because they are trained to predict statistically likely token sequences, not to verify propositional truth.
From
Cite as
Holds when
Classification
mechanismsupport: arguedfailure: Plausibility Truth Orthogonalityfamily: ai-model-and-training-failureinstitutional-design
Verify