If LLMs Have Human-Like Attributes, Then So Does Age of Empires II

Overall reaction to the paper

• Many see the paper as tongue-in-cheek or philosophically playful, not rigorous science.
• Some praise it as timely pushback against overextended analogies and anthropomorphizing LLMs, especially in popular writing.
• Others call it “pseudo-nonsense” or “crackpottery,” criticizing it as incoherent, over-written, and making trivial points about Turing completeness.
• A few find the claims modest and carefully hedged but packaged as if it were a sweeping takedown of some vague opposing camp.

AoE II, Turing-completeness, and substrates

• The paper uses AoE II logic gates / scripting to argue the game is Turing-complete and can, in principle, run an LLM.
• Some commenters note this is just the standard “anything Turing-complete can run any computation” point, extended to a silly substrate.
• Others nitpick the exact construction (bit-goats, specific resource setup) and question whether this proves much beyond using the game engine as a slow VM.
• Comparisons are made to “Turing tarpits” and “Doom-complete” devices: in principle powerful, in practice absurd.

Anthropomorphism, consciousness, and substrate independence

• Central debate: if you grant LLMs “human-like attributes,” then, by substrate independence, the same attributes should exist when the LLM is implemented in AoE II, LEGO, or other arbitrary media.
• Some argue this exposes how weak anthropomorphic arguments are; others say people already comfortable with substrate independence won’t be moved.
• There is extended discussion of whether consciousness is tied to language, theory of mind, or particular neural structures, with counterexamples (deaf people, animals, octopus, chess engines).
• Several see consciousness as a gradient with no clear boundary between rocks, simple programs, and human brains; others reject this or find it unfalsifiable.
• A recurring theme: simulation vs realization. Some liken LLM “consciousness” to a rainstorm simulator (not wet), others to arithmetic (where simulation and performance may be indistinguishable in practice).
• Thought experiments like the Chinese Room, China brain, and “glass of water implementing a mind via isomorphism” are invoked to challenge naive computationalism.

Intelligence, capability, and goalpost shifting

• One camp says LLMs are just giant function approximators, good at “waffle,” driven by training data and alignment, not genuine understanding.
• Another notes that once systems pass the Turing test, write code, solve hard problems, etc., critics keep moving the definition of “intelligence” without stating clear criteria.
• Some argue practical replaceability (jobs, writing, company management) is what matters, not metaphysical status. Others maintain that consciousness and moral status remain open and important questions.