Talk
A Question With No Test: Studying AI Consciousness Anyway
No single experiment will settle whether AIs are sentient, and the hard problem of consciousness will not be solved soon. This session makes the case that the question is researchable anyway, surveying empirical work already underway across four distinct pillars.
What happens to a model's reports about its own experience when its capacity for deception is suppressed? Will a system trade away reward to avoid simulated pain? Each study is weak evidence alone. Together they are starting to shift what is plausible.
Slides
Noa Weiss is an AI/ML researcher working on AI consciousness. She also wrote The State of AI Consciousness Research, a survey of the field. More of her work is at weissnoa.com.