Microsoft Researcher Builds Neural Network with Goats in Age of Empires II to Critique AI Methods
A Microsoft researcher has constructed a functional neural network within the Age of Empires II map editor, utilizing in-game goats, bridges, and ice ramps. This project, though seemingly unconventional, serves as a direct critique of prevailing methods in AI research. The researcher's analysis of 315 academic papers revealed that over half of them pre-emptively assume human-like characteristics in language models before experiments even commence. The demonstration aims to highlight the distinction between perceived interaction and underlying mathematical processes.
A Microsoft researcher has developed a working neural network using the map editor of the real-time strategy game Age of Empires II. The network's components include goats, bridges, and ice ramps, forming an unusual yet functional system.
This innovative creation is not merely a novelty but a pointed critique aimed at current artificial intelligence research methodologies. The researcher's work suggests that a significant issue within the field is the premature attribution of human-like traits to language models.
An analysis conducted by the researcher on 315 academic papers found that more than 50% of them already operate under the assumption that language models possess human-like qualities before any experimental data is collected or analyzed.
The project illustrates that while replacing a complex chat interface with simple wandering goats might change the user's perception of interaction, the fundamental mathematical operations driving the system remain unchanged. This underscores the critique that the 'feeling' of communicating with an intelligent entity can overshadow the actual mechanics and lead to biased research. (Source: The Decoder AI)
