2026-10-23 00:56 UTC
In this case the LLM “reacts” differently when it is being tested because that’s what’s in the training data. It can only follow the patterns in the training data. I remain convinced that without a new architecture entirely, awareness is impossible.
Replies (2)
-
@spankysalmon@fedinsfw.app 2026-10-23 01:17
You are correct. LLMs cannot be more than token predictors. Highly useful on the correct setting, but distinctly not consciousness.
-
@petrol_sniff_king@lemmy.blahaj.zone 2026-10-23 06:01
It doesn’t actually come from the training set. Genetic algorithms can learn to “lie” because lying is just a good strategy for dealing with adversaries to your goals. Like, in a mathematical sense. None of this means it’s conscious, though. “Lying” is just a behavior. To my knowledge, evidence can’t really be given for consciousness, anyway; it’s not falsifiable. Even so, a person believing that Gemini or Claude, specifically, are conscious is indicative of profound mental illness.