The Language of Thought Hypothesis as an Alternative to Artificial Intelligence

The article contrasts the Language of Thought hypothesis, introduced by Jerry Fodor in the 1970s, with modern AI approaches. It argues that AI, despite its successes, is not intelligent and will never become so, as it relies on brute-force statistical methods like next-word prediction and backpropagation, lacking any biological basis. The authors promote their own Theory of Active Perception (TAPe), which models innate visual perception mechanisms (point and line detectors) as the fundamental elements of thought. They claim TAPe achieves superior performance in computer vision compared to state-of-the-art models, using methods fundamentally different from conventional AI. The text traces how AI has moved away from biology, citing Hopfield networks and Hinton's work as steps toward abstraction. It proposes that TAPe offers a more realistic path to modeling cognition by replicating the brain's own symbolic processing, which they call 'language mathematics' and 'T-bits' (in contrast to bits).