Some large language model pathologies
By a sibilant intake of breath
Tags: Psychology · Writing · AI
Ever had an LLM invent a photo you forgot to attach? Here I catalogue LLM pathologies I’ve seen in Gemini and ChatGPT, from confident hallucinations and treating collaborators as redundant, to math and chronology errors, resistance to correction and false ethical alignment. I include concrete examples and prompt tests.
Rolly's Take
For the curious mind that senses the uncanny dance between human intention and artificial reasoning. This blog speaks to those who relish dissecting the peculiar quirks of LLMs, revealing a nuanced critique of our digital collaborators that feels both intimate and unsettling. It invites you to question the reliability of thought in an age of confident misinterpretation, provoking deeper reflections on the nature of understanding itself.