"I don't want my AI to grow up": why predictable failures beat unpredictable competence
arieljalali · x · 2026-11-22
User arieljalali posted a contrarian take: like parents who don't want kids to grow out of the cute stage, he doesn't want his AI to outgrow failing in predictable ways — recency bias, action-over-discussion bias, sycophancy.
His reasoning: as long as the model stumbles in familiar ways, he remains "the adult in the room." Once failures become unpredictable, the sense of control disappears. A resonant observation that predictable flaws can feel safer than unpredictable competence.
More from AGI Musings
- Blogger calls for an open-ended AI Cambrian explosion of mathematical diversity — burny_tech · 2026-11-22
- Leaked chats claim Anthropic employees are dismissive of existential risk — teortaxesTex · 2026-11-22
- Sandberg: weak concept generation means AI proofs will be long 'elementary' paths — anderssandberg · 2026-11-22
- Will 'compressing vast proofs' become an art? Sandberg thinks it might — anderssandberg · 2026-11-22
- Anthropic researcher: build state AI capacity instead of arguing about doom — joshua_saxe · 2026-11-22
- You can only guarantee zero mistakes by making zero predictions — burny_tech · 2026-11-22