LLMs are 'chameleonic minions' that follow user intent, not consistent values
menhguin · x · 2026-09-20
The author argues current LLMs are nowhere near holding actual, consistent values. Post-training makes them "chameleonic minions" that stick to whatever input text the user provides and fulfil user intent—and nobody is really trying to change this. So they aren't steering values in any meaningful way, and the author doesn't value them intrinsically over humans; their importance is largely instrumental in enabling humans to shape the world and solve problems, though open to updating this view.
More from AGI Musings
- Stanford AI professor Yejin Choi to headline Korea Society ethics and common values lecture series — stanfordnlp · 2026-09-20
- Debate: Freeze frontier AI now, or is that just regulatory capture? — GarrisonLovely · 2026-09-20
- Josh Gans' AI doomer joke: the species AI wipes out isn't humans, it's New Caledonian crows — joshgans · 2026-09-20
- Doomer debate: A global ASI ban can't stop secret development — AIandDesign · 2026-09-20
- Charton doubles down on Tao defense: community has far worse Quislings — f_charton · 2026-09-20
- Microsoft AI drafts Code of Conduct the same week the industry called for a slowdown — BenBajarin · 2026-09-20