Raw LLM Responses

I'll give it a go: Based on the design and function of an LLM, it explicitly doesn't meet Jaynes' description of consciousness, no? Jaynes proposed that the generation of language was functionally the moment consciousness was invented, and this has overlap with the Chomskian idea of Generative Grammar i.e. that humans have a genetic predisposition to generate grammars and by extension, languages. (in general linguistics in the 50s - 70s was super invested in this idea that language and consciousness or the ability to comprehend are inexorably linked). If the generation of grammar and language is the marker of consciousness then LLMs very explicitly are not conscious under Jaynes' description. An LLM "generates" grammar only as dictated by human description, and only functions because it must rely on an expansive history of human language from which to mimic. Semantically it isn't the same as the "generation" linguists talk about, including that there is still debate over how much of humans' predisposition for language is genetic. As a side note, the view that language is the window to consciousness is linked with the Sapir-Whorf hypothesis that language is effectively both the tool for understanding the world and the limit of understanding (e.g. if you don't know the word "blue" you cannot comprehend it as different from any other colour because you have no word for it). Sapir-Whorf has had a lot of impact, and informs a lot of modern linguistic theory, but as a view of how language actually works is considered archaic and fairly disproven as an accurate description for how language interacts with comprehension of the world around you. Tl;dr Jaynes' view proposed that human language is a reflection of consciousness, but LLMs are only imitators of language and so could only be imitations of that consciousness. Anything further is dipping into OP's point, that you are seeing LLMs work and mistaking it for thought and human generation of language, when it's only a