AI agents are still an experiment: ship-first playbook pushes the cost onto users
gerardsans · x · 2026-09-12
Replying to farzyness, gerardsans sharply criticizes the current AI agent wave: labs repeat the chatbot playbook — ship first, spend little on safety, dismiss warnings — leaving the public to pay for misinformation, hallucinations and psychosis. Key points: agents get too much autonomy with no basic controls; multi-day runs are great for token billing but mean errors, loops and dead ends for users; businesses report little upside. Labs are outsourcing their experiment's cost to customers, with weak harnesses and no accountability.
More from AGI Musings
- Dario's three-step global coordination plan sparks skepticism — and open-source worries — tokenbender · 2026-09-13
- Mollick: OpenAI and Anthropic hint at early recursive self-improvement in rare statements — emollick · 2026-09-13
- The smartest AI safety and e/acc people are far more nuanced offline than on Twitter — danfaggella · 2026-09-13
- Sam Altman backs Dario's frontier pacing call, pledges independent evaluators — but logits access history resurfaces — MaziyarPanahi · 2026-09-13
- Researchers: AI developers must show real-life benefits to earn trust — dhadfieldmenell · 2026-09-13
- Altman and Musk appear to side with Dario on frontier AI pacing — Hesamation · 2026-09-13