Why ‘Age of Empires II’ Proves AI Isn’t Sentient

In a widely shared critique of the notion that large language models possess consciousness, science fiction author Ted Chiang invited readers to reflect on something far more mundane: Microsoft Word. “Being open to the possibility that LLMs are conscious is the same as being open to the possibility that Microsoft Word is conscious,” Chiang argued. He went further, suggesting that if we entertain such ideas, we must also accept that “multiple distinct consciousnesses are dormant in every Word document containing a conversational transcript, and that they are awakened every time the document is loaded.” His point was blunt: “Should you consider the possibility that every time you open a Word document, you are bringing multiple conscious interlocutors into existence, and every time you close one, you snuff their existence out? No. Contemplating that scenario is not a good use of your time.”

Now consider a Microsoft AI researcher who recently took that very kind of contemplation quite seriously. Instead of Word documents, though, he turned his attention to Age of Empires II, the legendary real-time strategy game. His goal was not to prove the game was sentient, but rather to demonstrate just how absurd the claim would be. He built a basic neural network inside the game itself, using nothing more than digital goats, to illustrate his point. The experiment serves as a playful yet pointed reminder that even sophisticated behavior within a system does not imply consciousness. Just as a goat herder in the game follows programmed rules, so too do AI models follow statistical patterns, no matter how convincingly human their outputs may seem.