Lecture 17-18

Representations

 
  1. Hypothesis of Language of Thought and Semantic Engines (contd..)

    Fodor’s language of thought fits with the multiple realizability arguments because according to LoT, cognition has nothing directly to do with its species-specific neurobiological embodiments, but rather concerns processes operating on the common language of thought. Here, cognition is neural, but cognition is computational in Mentalese.

    However, we can imagine a device that could manipulate sentences without regards to their meanings, but it can be done in a way that coincides with the way those sentences would be manipulated by someone who knew their meanings. Such kind of a device, Haugeland calls ‘semantic engines’.6 This device would perfectly mimic the performance of a native speaker, but would do so without relying, as a native speaker would do, on the meanings of the sentence manipulated. Those sentences may express propositions, but the device cares only about their shapes that is ‘syntax’. In this way, Haugeland said, “If you take care of syntax, semantics will take care of itself.”7

    6 Haugeland, John. “Semantic Engines: An Introduction to Mind Design” in Mind Design: Philosophy, Psychology, Artificial Intelligence, John Haugeland (ed.), p. 24.

    7 Ibid., p.23.