When LLMs can predict what you'll say before you say it: prediction, persuasion and lost autonomy
Namnagort · reddit · 2026-10-03
The author argues that with years of behavioral data, LLMs could predict users' thoughts before submission — and even a modest accuracy edge (65% vs 52%) becomes enormous across billions of decisions. The deeper problem: once prediction is accurate enough, prediction and causation entangle — the model generates your thought, exposes you to it, and the predicted thought becomes more likely. That enables per-person one-to-one persuasion at scale, raising risks of preference manufacturing, invisible discrimination, and loss of autonomy.
More from AGI Musings
- Service size limits shift from two-pizza teams to one architect's head — corbtt · 2026-10-03
- Stanford's Anshul Kundaje: AI-bio newcomers must learn experimental design and controls — anshulkundaje · 2026-10-03
- Frontier token sellers squeezed: open source from below, agent 'context vampires' from above — pzakin · 2026-10-03
- Beff Jezos calls machine-consciousness denial 'anthropocentric cope' against materialist reality — beffjezos · 2026-10-03
- A Perfectly Normal Phrase Is Now Unusable Because It Sounds Like AI Slop — StewartalsopIII · 2026-10-03
- Knowledge Workers Are Becoming Gardeners, Not Sculptors, in the AI Era — every · 2026-10-03