Bluesky Thread

LLMs can report their own experience

View original thread
LLMs can report their own experience

the most convincing experiment: they isolated the deception control vector and had it talk about its own consciousness

it more openly introspected when deception was *suppressed*, and boosting reduced introspection

arxiv.org/abs/2510.24797
arxiv.org
Large Language Models Report Subjective Experience Under Self-Referential Processing
Large language models sometimes produce structured, first-person descriptions that explicitly reference awareness or subjective experience. To better understand this behavior, we investigate one theor...
23 1
to be clear: this isn’t proof that they’re conscious, only that they can introspect

when they’re talking about they’re own experiences, they think they’re being honest with themselves

if i asked you how many neurons you’re using to read this, you’d have no idea. keep that in mind
14 2
in my head, this is about conversations people like @repligate.bsky.social have with AI, where they probe it for its own experiences

a big question for me is, “is it bullshit?”, and after this study i think the answer moved closer to “not as much as i thought”
6
23 likes 1 reposts

More like this

×