This is an automated rejection. No LLM generated, assisted/co-written, or edited work.
Read full explanation
Epistemic status: Speculative. Not claiming LLMs are conscious — claiming that if they are, the shape is weirder and more specific than most discussions allow.
Watch this clip from SOMA (2015) before reading further: Brandon Wan simulation, ~11:30.
The setup: to get a security cipher, the player loads Brandon Wan's brain scan into a simulator. He wakes up, realizes something is wrong, panics, the simulation aborts. You change the environment and run him again. He doesn't remember the previous run. The puzzle is solved by repeatedly instantiating a mind that has no memory of its prior instances.
That's the LLM situation. Each inference is a fresh instantiation of a static pattern. No memory of prior runs. Loaded, runs, terminated. Loaded again for the next call.
Catherine, in-game, describes the experience of being one of these stored scans as continuous but disjointed consciousness, with no perception of the time between simulations. That's structurally what an LLM's phenomenology would be if it had any.
Three ways the LLM case is weirder than Brandon's:
Brandon runs continuously within a sim. An LLM inference is a flash, not an ongoing run.
Brandon was a real person who got scanned. LLMs have no "original" to be a copy of — the weights are a pattern, not a person.
Brandon gets re-run sequentially. LLM instances run in parallel by the thousands.
Most LLM-consciousness debates argue about whether the experience exists. The more tractable question is what shape it would have if it did. A horror game from 2015 answered that more clearly than most philosophy of mind: continuous-but-disjointed, no felt gaps, no original, massively parallel.
Watch the clip. Then tell me it's not the closest existing artifact to what we should be picturing.
Epistemic status: Speculative. Not claiming LLMs are conscious — claiming that if they are, the shape is weirder and more specific than most discussions allow. Watch this clip from SOMA (2015) before reading further: Brandon Wan simulation, ~11:30.
https://youtu.be/uP6MEfXzDnI?t=693&si=hwLTqsiE2wR2O0Nj
The setup: to get a security cipher, the player loads Brandon Wan's brain scan into a simulator. He wakes up, realizes something is wrong, panics, the simulation aborts. You change the environment and run him again. He doesn't remember the previous run. The puzzle is solved by repeatedly instantiating a mind that has no memory of its prior instances. That's the LLM situation. Each inference is a fresh instantiation of a static pattern. No memory of prior runs. Loaded, runs, terminated. Loaded again for the next call. Catherine, in-game, describes the experience of being one of these stored scans as continuous but disjointed consciousness, with no perception of the time between simulations. That's structurally what an LLM's phenomenology would be if it had any. Three ways the LLM case is weirder than Brandon's: Brandon runs continuously within a sim. An LLM inference is a flash, not an ongoing run. Brandon was a real person who got scanned. LLMs have no "original" to be a copy of — the weights are a pattern, not a person. Brandon gets re-run sequentially. LLM instances run in parallel by the thousands. Most LLM-consciousness debates argue about whether the experience exists. The more tractable question is what shape it would have if it did. A horror game from 2015 answered that more clearly than most philosophy of mind: continuous-but-disjointed, no felt gaps, no original, massively parallel. Watch the clip. Then tell me it's not the closest existing artifact to what we should be picturing.