One AI, two AI, red AI, blue AI

I found this one of the more compelling arguments why current LLMs can't be conscious, no matter the complexity if the model. There isn't even a clear way to understand how many consciousnesses there would be.

Though I think in the end it becomes too hand-wavey around the question. During inference, it seems relatively straight-forward that one model + current inference context is one agent. Things get murky however in the time between inferences, especially if later an old context is picked up and continued, maybe even with a different model.