LLMs do not think. They are an interactive, static snapshot of the statistical distribution of words (or tokens) in a large body of text.
If you catch yourself again and again being amazed by the latest magic trick from silicon valley and wondering if there might not yet be some kind of ghost in that machine, inoculate yourself with the following wisdom:
https://softwarecrisis.dev/letters/ai-bird-brains-silicon-valley/
