June 7, 2026
Castle Siege of the AI Ego
If LLMs Have Human-Like Attributes, Then So Does Age of Empires II
Researchers say a medieval war game can look as “human” as AI, and commenters are losing it
TLDR: A new paper argues that if people insist chatbots have human-like qualities, then a game like Age of Empires II can be framed that way too. Commenters loved the absurdity, with jokes about in-game logic gates and sharp pushback against what they see as over-the-top AI hype.
A fresh paper just kicked the AI discourse straight into the castle moat: the authors argue that if people keep claiming large language models — the text-generating systems behind today’s chatbot boom — have human-like traits such as “understanding” or “morality,” then you could play the same game with Age of Empires II, the classic medieval strategy game. Yes, really. Their point is less “the game is alive” and more “people may be reading too much into behavior that can show up in lots of systems.” And the crowd? Absolutely delighted, confused, and a little feral.
The biggest scream-laugh moment came from the line promising to prove Age of Empires II is “functionally- and Turing-complete,” which is academic-speak for “surprisingly powerful in a computer-science way.” One commenter basically melted on contact: “You had me at the NAND gate in AoE II’s editor.” Another just reposted the wildest line in the paper like it was a mic drop. That was the fun side.
But the thread also had real eye-rolling. One commenter said they’re seeing more and more grand, AI-philosophy essays that list things human brains do, find something vaguely similar in chatbots, and then declare victory. Another went full poetry-slam doompost, comparing language models to the Grand Canyon: vast, impressive, but not actually thinking — just channeling a flow. So the vibe is split between “this satire is brilliant” and “thank you, finally, someone called out the overhype.” In other words: nerds found a paper with a medieval game, philosophy bait, and anti-AI swagger, and naturally the comments became the real battlefield.
Key Points
- •The paper argues that human-like attributes often assigned to large language models may be empirically non-unique rather than specific to LLMs.
- •The authors build and train a simple neural network on *Age of Empires II* to support their argument about substrate-dependent interpretation.
- •It claims that some observed properties may remain constant across substrates, while interpretations of those properties can change with representation.
- •The paper says empirically grounded discussion of anthropomorphic attributes requires explicit measurement criteria to avoid interpretive ambiguity.
- •The authors propose a null assumption of LLM non-uniqueness and state that *Age of Empires II* is functionally complete and Turing complete.