As Large Language Models become more ubiquitous across domains, it becomes important to examine their inherent limitations critically. This work argues that hallucinations in language models are not just occasional errors but an inevitable feature of these systems. We demonstrate that hallucinations stem from the fundamental mathematical and logical structure of LLMs. It is, therefore, impossible to eliminate them through architectural improvements, dataset enhancements, or fact-checking mechanisms. Our analysis draws on computational theory and Godel's First Incompleteness Theorem, which references the undecidability of problems like the Halting, Emptiness, and Acceptance Problems. We demonstrate that every stage of the LLM process-from training data compilation to fact retrieval, intent classification, and text generation-will have a non-zero probability of producing hallucinations. This work introduces the concept of Structural Hallucination as an intrinsic nature of these systems. By establishing the mathematical certainty of hallucinations, we challenge the prevailing notion that they can be fully mitigated.
https://arxiv.org/abs/2409.05746 Naturalmente, una volta compreso che si tratta di compressioni lossy dei testi usati per programmarli, una presenza di artefatti grosso modo inversamente proporzionale al livello di compressione è ovvia. Ma magari un articolo scritto nella lingua parlata dai believer della AI è utile a far scoppiare un po' di bolle (pustole? :-D) Giacomo
