Pattern matching machines seems more appropriate.
For humans?
LLMs do not match patterns. They predict one statistically most likely token (only one!) given a history of some N previously known tokens.
For humans?