I am very interested in the application of Dennett’s stances to LLMs.
I think where you might want to explore further is the fact that LLMs still display what Dennett (and Davidson) would call “shocking gaps” in understanding—hallucinations, IOW places where the intentional stance breaks down, reducing or eliminating the utility of intentional stance predictions. Someday, perhaps, LLMs will be as robustly predictable as non-human animals or even humans, but I think the shocking gaps that still occur argue we are not there yet.
I am very interested in the application of Dennett’s stances to LLMs.
I think where you might want to explore further is the fact that LLMs still display what Dennett (and Davidson) would call “shocking gaps” in understanding—hallucinations, IOW places where the intentional stance breaks down, reducing or eliminating the utility of intentional stance predictions. Someday, perhaps, LLMs will be as robustly predictable as non-human animals or even humans, but I think the shocking gaps that still occur argue we are not there yet.
I have an updated paper version of this (to be published) -- happy to share the draft.