Microsoft scientist proves it is possible to create artificial consciousness within Age of Empires II

Modern research often analyzes whether large language models possess signs of empathy or consciousness. In his new work, Microsoft and York University researcher Adrian de Winter challenged this approach. The author demonstrated that it is possible to build both a simple neural network and an analog of a universal Turing machine within the 1999 strategy game Age of Empires II. This was reported by Ixbt.com reports .

The researcher built a perceptron, the basis of neural networks, using the game's scenario editor. Instead of electronic signals, game objects were used: sheep and villagers moved along predefined paths, acting as logical elements. In the Turing machine analog, various buildings represented symbols on the computational tape, while game economy parameters represented the system state.

Adrian de Winter notes that many modern tests are based on the implicit belief that language models possess human traits. If the same computational principles are transferred to the Age of Empires II environment, the mathematical essence of the process remains unchanged, yet the sense of "intelligence" vanishes. This shows that traits attributed to AI depend not on the model itself, but on how humans perceive it through a convenient interface.

The scientist proposes applying the "Null Assumption" approach. According to it, one should not attribute psychological traits to AI until reliable evidence emerges. Phrases like "I think" or "I feel" used by bots like ChatGPT can create false impressions and emotional attachment in users.