LLM Voice
Points and comments are a snapshot, not live.
Martin Fowler warns that LLM-generated prose triggers an uncanny valley reaction in human readers.
Fowler describes an "LLM-Voice" that produces an uncanny valley reaction, making him want to stop reading. He cautions authors against using LLMs even for drafts, as the voice is hard to expel. He recommends writing prose manually and reading it aloud. Translation and accessibility use cases are exceptions, but he prefers clunky human prose to smooth LLM output. Fowler worries delegating writing to LLMs also delegates thinking.
What commenters are saying
Top commenters split: some note humans are adopting LLM-Voice by consuming so much AI output; others defend LLM use for non-artistic writing, comparing it to frozen food-fine for fuel, not for sharing without disclosure. A strong camp argues presenting AI-written work as your own is unacceptable, analogizing to restaurants serving microwaved meals. One commenter distinguishes between personal use and publishing to an audience.