3 ms·
I didn't read the SEP article, but is the language of thought hypothesis perhaps already verified by our ability to do some sort of logic on neural network embe
by cubefox 2y ago
I didn't read the SEP article, but is the language of thought hypothesis perhaps already verified by our ability to do some sort of logic on neural network embeddings? E.g. queen-royal=woman.
- foldr 2y agoFundamentally LoT is a hypothesis about human (and possibly animal) cognition, not a hypothesis about what artificial neural network models can or can't do. But you can see Fodor and Pylyshyn's thoughts on 'connectionist' models (as they were known at the time) in this extensive article from 1988: https://ruccs.rutgers.edu/images/personal-zenon-pylyshyn/proseminars/Proseminar13/ConnectionistArchitecture.pdf https://ruccs.rutgers.edu/images/personal-zenon-pylyshyn/pro... The discussion on p. 14 is particularly relevant.