If multimodal models produce human-like outputs but report different observations when they observe themselves, this would provide some evidence against their consciousness.

“Right? And the phenomenality, I think, is a representation here in this case. Um, if w- if we look at the, uh, what the LLMs are doing, if, uh, and we extend this into a multimodal model, and they produce, uh, similar, uh, output as us, but they are able to observe themselves and their ob- uh, observations that they report differ from what we are doing, then, uh, we would have some evidence against it.”