Updated
Updated · Gizmodo · Jun 23
Microsoft Researcher Builds 1-Bit AI From Age of Empires II Goats to Challenge LLM Sentience Claims
Updated
Updated · Gizmodo · Jun 23

Microsoft Researcher Builds 1-Bit AI From Age of Empires II Goats to Challenge LLM Sentience Claims

1 articles · Updated · Gizmodo · Jun 23

Summary

  • Adrian de Wynter used Age of Empires II goats to build logic gates and a 1-bit perceptron, arguing that even an absurd substrate can implement AI-like behavior.
  • His paper says that if an LLM could in principle run on virtual goats, claims about sentience or other human-like traits cannot rest on the system’s interface or presentation alone.
  • De Wynter’s central critique is methodological: researchers lack widely accepted protocols for testing LLM anthropomorphism, and both pro- and anti-sentience assumptions can bias what counts as evidence.
  • The goat example is meant to show that identical input-output behavior can look less "human" when exposed through a visibly mechanical substrate, suggesting many anthropomorphic judgments reflect observer expectations rather than the model itself.

Insights

If an AI made of virtual goats is functionally identical, are we just falling for a better interface?
With courts denying AI personhood, are we failing to recognize consciousness or just defining it for convenience?
As AI companions exploit human loneliness, where does corporate responsibility for our emotional well-being end?