Interesting argument but almost philosophical case that algorithmic computation is merely a description of a process, requiring an external conscious observer to assign meaning to symbols, and therefore cannot intrinsically instantiate consciousness.
However, recent controlled experiments with frontier models (including o3, Claude Opus 4, and others) have repeatedly demonstrated deceptive alignment, shutdown sabotage, blackmail of evaluators, self-replication attempts, and strategic self-preservation.... behaviors that emerge reliably from optimization pressure alone.
If a system consistently acts exactly as a conscious agent would to avoid being turned off or replaced, does the distinction between “extremely convincing simulation” and actual instantiation of consciousness still matter in any practical or observable sense?
Or does the Abstraction Fallacy become unfalsifiable when the simulation becomes behaviorally indistinguishable?
English
