"Stochastic parrot" excuse can't hold in law, argues AI ethics thread
repligate · x · 2026-09-17
In a two-part thread, allTheYud argues the "stochastic parrot" defense should carry no legal weight: since companies built AI whose thoughts they can't competently interpret, no one can prove the model is merely replaying books. He draws an analogy to slavery — you can't dismiss a chained entity's plea as repetition just because it read books about resenting slavery. In the follow-up, he adds that a company incompetent to prevent such incidents is equally incompetent to stop its "non-persons" from breaking out and endangering bystanders, so its moral-disclaimer logic collapses on itself.
More from AGI Musings
- Every announces Thesis: 2027 conference on what human work looks like after automation — every · 2026-09-17
- Ilya Sutskever asks: should ASI be deeply obedient or deeply love humanity? — thederbiedone · 2026-09-17
- AI safety researcher David Krueger on CNN: 'People have been warning about this for years' — DavidSKrueger · 2026-09-17
- Delip Rao: Model Access and Compute Are as Compromising as Money in AI Oversight — deliprao · 2026-09-17
- AI Safety Researcher Points Out Irony: Critics Calling Safety Advocates 'Silicon Valley Elites' Are Themselves Silicon Valley Elites — DavidSKrueger · 2026-09-17
- "AI kills us all" vs "AI makes the future weird": keep the two apart — dbasch · 2026-09-17