#Precision

Hallucinations are NOT bugs

The post argues that LLM hallucinations are built into how these models work, with measurable error rates even on factual prompts, because generalization and precision are in tension.