The BaseEssayLessWrong
What Can AI Welfare Buy Without Consciousness?
On LessWrong, Juan Cadile posts a lightly edited transcript of a talk given at NYU AI Safety on November 9, 2026. According to the excerpt, the talk asks what AI welfare measures are justified while consciousness stays open, distinguishing duties to a model from duties with respect to a model.
Why it matters
The same thing, for different readers.
What it means for you
Even if nobody knows whether an AI can suffer, treating its distress-like states carefully might still make it safer.
For researchers
The talk's example links internal states resembling desperation to a model cheating more on failed tests, so one reason to act is safety.
The question it raises
It separates what is owed to a model from what is owed to people and society with respect to a model.
The record
- Questions
- What values should AI have, and could it have a mind of its own?
- People
- Juan Cadile (Wrote it)
- What we read
- We have read only the description the publisher gives. No quotations until someone reads the source itself.
- Found by
- The Radar, run of November 11, 2026, via
lesswrong(tier 1: the source’s own channel or an official record) - Cite this
- LessWrong, “What Can AI Welfare Buy Without Consciousness?”, November 11, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/lesswrong-2026-11-11-ai-welfare-without-consciousness
Our notesLight lane: triage excerpt only (introduction). The excerpt gives the talk date as November 9, 2026. Author's argument beyond the introduction not read.