Stephen Downes

This is a really fun discussion of what Geoffrey Hinton thinks consciousness is, why he thinks AI already possesses it to some degree, and why (in Eryk Salvaggio's opinion) Hinton is wrong. "Hinton is arguing that self-awareness is the ability to discern whether one is accurately assessing the environment. By being conscious of discrepancies between the environment and our interpretation of it, he seems to suggest, we have to be self-aware." This, argues Salvaggio, is not the usual meaning we would attach to the concepts of self-awareness and consciousness. And, Salvaggio argues, "(Hinton) conflates content produced by a system for thinking that accurately describes the inner workings of a system. A large language model (LLM) is always representing language but never representing what language actually represents." Geoffrey Hinton is, he says, a "Radical Lacanian". If this all feels to you like Russell's paradox, I think you're right. But it's no more a limit to AI than it is to humans, and I'm sure Hinton would be aware of that.