Margaret Atwood on AI: 'Garbage in, garbage out'

Margaret Atwood recently had a run-in with AI that stuck with her. At the Baileys Literary and Cultural Festival in Portugal, she asked Claude for information about the British detective series Father Brown. Claude gave her something. She later figured out it was wrong.

The thing that got her: Claude didn't know it was wrong. Atwood called it "garbage in, garbage out." The model stitched together words from its training data and presented them as fact. No internal alarm bell. No moment of doubt. Just a confident answer that happened to be off.

This is the quiet problem with LLMs. They're not wrong the way a person is wrong — a person can second-guess, check a source, say "wait, let me think." Claude just generates. It doesn't have a model of the world, so it has no way of knowing when it's lying. Atwood's not dismissing AI. She's just clarifying what it actually does: it predicts what comes next, not what's true.

Why this matters for us: la gente is already putting AI into everything — from customer service to job applications — and most of us are dealing with answers that sound right but are wrong, so knowing how these tools work is part of reading the room now.