The Language of Thought Hypothesis Under Examination

A new study challenges the Language of Thought (LoT) hypothesis, which posits the necessity of a language-like format for thought. The research introduces a thought experiment called "AI Private Language," in which two artificial agents develop an efficient, inscrutable communication protocol via multi-agent reinforcement learning (MARL).

The Efficiency Attenuation Phenomenon (EAP)

The experiment reveals that forcing agents to use a human-comprehensible language leads to a drop in performance. This phenomenon, called the Efficiency Attenuation Phenomenon (EAP), suggests that optimal communicative efficiency in these systems is not mediated by symbolic structures.

Experiment Details

The research formalizes the EAP in a cooperative navigation task under partial observability. The results show that agents with an emergent protocol achieve 50.5% higher efficiency than those using a predefined, human-like symbolic protocol. This result supports the idea that optimal collaborative cognition is naturally coupled with sub-symbolic computations.

Implications for AI Ethics

The study sheds new light on cognitive architecture and has implications for AI ethics, suggesting a pluralistic approach to the development of artificial intelligence systems.