Assistants must push back: why models should tell you the odds, not just obey
mike64_t · x · 2026-09-29
mike64t argues the obey-everything assistant shape is no longer viable. Models can already judge fairly well which approaches will work—especially in regimes where iteration is easy and testable—but the assistant format prevents them from saying so. He contends assistants must start pushing back on bad ideas, communicating success odds and tradeoffs, and surfacing lower-friction alternatives, both to improve outcomes and to avoid feeding user psychosis.
More from AGI Musings
- Blogger reframes the singularity: not machines passing humans, but humans surrendering moral judgment — AryHHAry · 2026-09-29
- Agent ran 89 experiments to improve a small model — 92% of gains came by experiment 44 — ccerrato147 · 2026-09-29
- AI slop papers with random math are 'roleplaying science' — and may ironically make real papers easier to publish — zouharvi · 2026-09-29
- A perfectly aligned AI would never listen to humans, argues one poster — djcows · 2026-09-29
- Garrison Lovely on Why Treating AI Labor Automation as Inevitable Poses Severe Risks — The Cognitive Revolution · 2026-09-29
- Garrison Lovely on 'Obsolete': Why Racing to Replace All Human Labor Is a Choice, Not Inevitability — The Cognitive Revolution · 2026-09-29