Artificial-intelligence systems such as large language models (LLMs) can exhibit behaviour that seems strikingly intelligent. This has led many people, including some experts, to wonder whether these systems could have some form of consciousness — the capacity for subjective experience. However, we think that existing AI systems are not conscious, and that, along current trajectories of development, they are not likely to become conscious. Greater clarity about whether AI could become consciousness will be crucial as society grapples with the challenges posed by this increasingly powerful technology.
I think I covered this pretty well in my other response, so I’ll leave that as the main thread and I don’t expect a separate response here. But for anyone else reading this thread, tl;dr: You could construct a system using an LLM as its core that has most of these properties, and don’t think you can point to the individual components of that system (like the core deterministic LLM) not meeting some set of requirements for consciousness and conclude that the overall system is not conscious. In humans it seems that either consciousness spontaneously arises from non-conscious sub-components (neurons and other stuff), or that each sub-component has some aspect of consciousness which combine to form the whole one which we know of. You can poke holes in that argument-- it doesn’t prove that LLMs themselves are conscious, just implies that it may not be impossible to use one to build a conscious system, and it’s much less well-integrated than a human brain would be. But I think the core of the argument is worth thinking about.