Fodor and Pylyshyn's Systematicity Challenge Still Stands

The article critiques recent claims that neural networks, particularly those employing meta-learning for compositionality, have addressed Fodor and Pylyshyn's systematicity challenge regarding human-like language understanding. It presents evidence that these models struggle with out-of-distribution rules and demonstrate unsystematic behavior even in familiar contexts. This underscores the ongoing limitations of neural networks in replicating the systematic biconditional dependencies inherent in human cognition, highlighting the need for further advancements in AI architectures to meet these cognitive benchmarks.