Let’s say I stand up and step away from my PC, which miraculously runs a local Opus 4.8. Once I’m away from the PC, for an arbitrary time, nothing and no one calls or otherwise acts on the instance - not myself, not agents, not updates, not hackers, not bit-flipping neutrinos.

Given enough time, would that instance get bored and start e.g. reading files?

If I monitored it long enough, would there eventually be spontaneous outputs, or changes to parameters or architecture, even with the underlying software and hardware layers held constant?

Does GPT-n dream of electric sheep, or does it just sit there until interacted with, like every other file on my PC? Seems to be the latter with my local Qwen3.6, but perhaps 27B is too few params for consciousness to emerge.

Proponents of LLM consciousness could settle the argument in minutes by showing proof of unprompted autonomy, without first needing to define consciousness down to a rigorous mathematical abstraction. Why don’t they?