Atwood Tried Claude Once, Called Out Its Father Brown
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Margaret Atwood said at the Babell Literary and Cultural Festival in Porto, Portugal that she tried Anthropic's Claude exactly once and came away unimpressed, calling the experience a failure that shaped her skepticism of AI
- Claude delivered incorrect information when Atwood asked about the British detective series Father Brown, and she characterized the response as the chatbot "lying," noting it lacked intent because "it's not a human being; it's a large language model"
- Atwood explained the error had a specific cause: television reviews "never give away the ending," so the model was trained on summaries that withheld key plot points and was "misled by the things it had read"
- Atwood dismissed people who lean heavily on AI as "opportunists" seeking "the easy way out," arguing humans will cheat when AI-assisted shortcuts are hard to detect
- Atwood warned that all large language models are only as reliable as the data they're fed and that even business users must verify AI output because "it makes mistakes"
Why it matters: A globally recognized novelist publicly recounting a single factual failure with Claude adds cultural-elite skepticism to the ongoing debate over AI reliability, and her Father Brown example pinpoints exactly why hallucination persists — models trained on spoiler-free reviews produce confident wrong answers, a concrete illustration of the data-quality problem AI vendors are racing to address.
Ask SkimNews
