Smolensky, Paul. Connectionist Mental States: A Reply to Fodor and Pylyshyn , Southern Journal of Philosophy, 26:Supplement (1988) p.137
Smolensky (1987/1988) On the Proper Treatment of Connectionism.
Recordings of 1988 talks contributed by Paul Smolensky Jan. 8th 2016.
The rumor seems to have gotten around that I think connectionism is the lousiest new idea in cognitive science. This rumor is entirely unfounded and I propose to scotch it this evening. In the first place, I don’t think connectionism’s a lousy idea, what I think it is is two lousy ideas, one about mental processes and one about learning. And in the second place, I don’t think it’s new. On the contrary, it seems to me to be the reiteration, almost without elaboration, of a doctrine that I’m going to call primitive associationism, and I mean primitive in both respects…this is just another regression to this doctrine, and it will pass.
Hinton (2015: Aetherial Symbols) has this to say:
Verbal debate is not a good way to deal with people who are convinced that their approach is the only one that is viable (e.g. Fodor and Pylyshyn).
– What’s needed is a viable model of parallel intuitive reasoning.
Elman, J.L. (2014). Systematicity in the lexicon: On having your cake and eating it too. In P. Calvo and J. Symons (Eds.) . The Architecture of Cognition: Rethinking Fodor and Pylyshyn’s Systematicity Challenge. Cambridge, MA: MIT Press. Pp. 115-145. Adobe PDF version.
Another interesting chapter from the above book: How Limited Systematicity Emerges: A Computational Cognitive Neuroscience Approach
Fodor’s (1998) review of “How the mind works” claims that modularity has trouble with integrating knowledge from different modules.
Lake and Baroni (2023) . Abstract:
The power of human language and thought arises from systematic compositionality—the algebraic ability to understand and produce novel combinations from known components. Fodor and Pylyshyn1 famously argued that artificial neural networks lack this capacity and are therefore not viable models of the mind. Neural networks have advanced considerably in the years since, yet the systematicity challenge persists. Here we successfully address Fodor and Pylyshyn’s challenge by providing evidence that neural networks can achieve human-like systematicity when optimized for their compositional skills. To do so, we introduce the meta-learning for compositionality (MLC) approach for guiding training through a dynamic stream of compositional tasks. To compare humans and machines, we conducted human behavioural experiments using an instruction learning paradigm. After considering seven different models, we found that, in contrast to perfectly systematic but rigid probabilistic symbolic models, and perfectly flexible but unsystematic neural networks, only MLC achieves both the systematicity and flexibility needed for human-like generalization. MLC also advances the compositional skills of machine learning systems in several systematic generalization benchmarks. Our results show how a standard neural network architecture, optimized for its compositional skills, can mimic human systematic generalization in a head-to-head comparison.