Age of Empires II matches LLMs in functional completeness
The main argument of the paper is that the attributes of LLMs may not be as unique as previously thought, and that any entity in a sufficiently powerf
TL;DR
- Adrian de Wynter's research paper "If LLMs Have Human-Like Attributes, Then So Does Age of Empires II" challenges the assumption of anthropomorphic attributes in large language models (LLMs).
- The paper proposes a 'null' assumption, where one assumes LLM non-uniqueness instead of assuming anthropomorphic attributes.
- The research shows that Age of Empires II is functionally- and Turing-complete, presenting similar attributes to LLMs.
- The paper was submitted on May 29, 2026, and last revised on June 11, 2026.
Adrian de Wynter's paper presents a unique perspective on the attributes of LLMs, suggesting that they may not be as unique as previously thought. The research focuses on the idea that any entity in a sufficiently powerful substrate could present similar attributes. The paper proposes a new approach to understanding LLMs, one that assumes non-uniqueness instead of anthropomorphic attributes.
What the data shows
The data presented in the paper shows that Age of Empires II, a videogame, can be used to demonstrate the non-uniqueness of LLM attributes. By building and training a simple neural network on the game, the researchers were able to show that the game presents similar attributes to LLMs. The paper also discusses the idea that the interpretation of perceived behavior may change with the substrate, highlighting the importance of explicit measurement criteria.
What this means for ai readers
For AI readers, this research suggests that the attributes of LLMs may not be as unique as previously thought. The idea that any entity in a sufficiently powerful substrate could present similar attributes challenges the assumption of anthropomorphic attributes in LLMs. This could have significant implications for the development and understanding of AI systems, as it suggests that the focus should be on the substrate and measurement criteria rather than the attributes themselves.
What to do right now
Readers can start by exploring the paper and its findings, and considering the implications of the research for their own understanding of AI systems. The paper proposes a 'null' assumption, which could be a useful approach for those looking to develop and understand AI systems. Additionally, readers can consider the idea that the interpretation of perceived behavior may change with the substrate, and think about how this might impact their own work with AI systems.
Bottom line
The research presented in "If LLMs Have Human-Like Attributes, Then So Does Age of Empires II" challenges the assumption of anthropomorphic attributes in LLMs and presents a new approach to understanding AI systems. The paper shows that Age of Empires II is functionally- and Turing-complete, and that the attributes of LLMs may not be as unique as previously thought. By considering the implications of this research, readers can gain a deeper understanding of AI systems and the importance of explicit measurement criteria.
Frequently asked questions
Q: What is the main argument of the paper?
The main argument of the paper is that the attributes of LLMs may not be as unique as previously thought, and that any entity in a sufficiently powerful substrate could present similar attributes.
Q: What is the 'null' assumption proposed in the paper?
The 'null' assumption proposed in the paper is the idea that one should assume LLM non-uniqueness instead of assuming anthropomorphic attributes when setting up an experiment.
Q: What is the significance of Age of Empires II in the paper?
Age of Empires II is used in the paper to demonstrate the non-uniqueness of LLM attributes, by building and training a simple neural network on the game and showing that it presents similar attributes to LLMs.
Q: What are the implications of the research for AI systems?
The research suggests that the focus should be on the substrate and measurement criteria rather than the attributes themselves, and that the interpretation of perceived behavior may change with the substrate.
