Economist Paul Novosad: AI-written text deserves a steep attention discount, often near 100%
paulnovosad · x · 2026-10-20
Economist Paul Novosad offers an attention-economics take on AI writing: AI-assisted writers do produce more (say 2x), and if readers gave each piece equal attention as advised, those writers would capture 2-10x more attention — which is precisely why they recommend it.
He argues this is clearly wrong: an AI writer's tenth idea is unlikely to match a human writer's next one, and the more people write with AI, the worse each marginal piece. The real question is what discount rate to apply to heavily AI-written work — his experience suggests far above zero, often close to 1 (complete discount).
It's person-dependent: a handful of highly creative but poor writers might genuinely merit 10x AI-boosted volume. But the most common use today is hiding lazy thinking, meaning readers should consume less from those writers. Since attention is the binding constraint amid exploding text volume, AI-typical writing has become a strong signal that content deserves far less of your attention.
More from AGI Musings
- National Australia Bank to cut 2,500 jobs, citing rapid AI adoption — gaganghotra_ · 2026-10-20
- Mathematicians' AI anxiety echoes the loom workers: a thread on tech change as personal crisis — soumitrashukla9 · 2026-10-20
- "Thinking about old Google searching makes me want to take a shower" — signulll · 2026-10-20
- Language Modeling Is Agents in the Limit: voooooogel's Intuition Clash Sparks Debate — voooooogel · 2026-10-20
- The artificial ego and the absent id: why AI moral status hinges on deliberately built-in stakes — usernamedregs · 2026-10-20
- Reddit essay: why demanding proof of AI consciousness before moral consideration is a double standard — Afflictionxx · 2026-10-20