On the proper treatment of connectionism

By Paul Smolensky. 1988

Subfields: Natural Language Processing, Philosophy

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

In Philosophy

Citations in the corpus

Environmental drivers of systematicity and generalization in a situated agent (2019, background): “TRODUCTION Since the earliest days of research on neural networks, a recurring point of debate is whether neural networks exhibit generalisation beyond their training experience, in a systematic way (Smolensky, 1988; Fodor & Pylyshyn, 1988; Marcus, 1998; McClelland et al., 1987). This debate has”
On the Binding Problem in Artificial Neural Networks (2020, background): “Indeed, the search for an adequate mechanism for binding (in one form or another) is a longstanding problem, not just in neuroscience and cognitive psychology, but also in machine learning (Smolensky, 1987, 1988; Sun, 1992).”
Routing Networks and the Challenges of Modular and Compositional Computation (2019, background): “…details of compositional computation have been highly debated [Fodor, 1975, 1983, Fodor and Pylyshyn, 1988, Touretzky and Hinton, 1988, Smolensky, 1988, i.a.].1 Among others, there are two notable early attempts to build modular systems with a cognitive focus: Miikkulainen [1993]…”
Text Modular Networks: Learning to Decompose Tasks in the Language of Existing Models (2021, neutral): “The resulting sequence of sub-questions and their answers provides a human-interpretable description of the model’s neuro-symbolic reasoning (McCarthy, 1988; Smolensky, 1988), as illustrated in Fig.”
Rethinking eliminative connectionism. (1998, background): “Others have suggested that eliminative connectionist models are more biologically plausible than symbolic models, but the back-propagation algorithm that eliminative connectionist models typically rely on is clearly not biologically plausible (Crick & Asunama, 1986; Smolensky, 1988).”
Inductive biases for deep learning of higher-level cognition (2020, background): “This is where the notion of systematic generalization becomes interesting [19,23,139,140].”
Solving the Black Box Problem: A Normative Framework for Explainable Artificial Intelligence (2019, background): “…discussions of connectionist modeling methods in cognitive science suggest that neural networks and similarly high-dimensional systems are very likely to exhibit this kind of distributed responsibility (Smolensky 1988), and thus, are likely to deploy distributed representations (Clark 1993).”
Control Processes and Self-Organization as Complementary Principles Underlying Behavior (2002, background): “Given enough repetition, control of an activity may drop out of consciousness altogether (Smolensky, 1988).”
Creative Thought as a non Darwinian Evolutionary Process (2005, background): “Each location is sensitive to a broad range of subsymbolic microfeatures (Smolensky, 1988), or values of them (e.”
Grandmother cells and localist representations: a review of current thinking (2017, background): “The term “sub-symbolic” (Smolensky, 1988) was commonly used to highlight that individual units cannot be interpreted by themselves, and this was considered a major break from previous theorising, and key to understanding how cognition is implemented in brain.”
Mapping Part-Whole Hierarchies into Connectionist Networks (1990, background): “Given a parallel network, some inferences can be performed very efficiently by simply allowing the network to settle down into a stable state [28].”
Entrepreneurial Imaginativeness in New Venture Ideation (2018, unknown): “…1980; Dietvorst, Verbeke, Bagozzi, Yoon, Smits, & van der Lugt, 2009; Gough, 1979; Heppner & Petersen, 1982; Kirton, 1976; Paal & Bereczkei, 2007; Spiro & Weitz, 1990; Tellegen & Atkinson, 1974; Zhou & George, 2001) to aid in the generation of items for creative, social, and practical…”
Distributed representations for extended syntactic transformation (2000, background): “1 Whether or not the level of systematicity actually achieved in connectionist systems should be regarded to be a cognitively plausible demonstration of systematicity has been the matter of much debate (cf. Smolensky 1987, 1988, Fodor and McLaughlin 1990, van Gelder 1990, Blank et al .”
Hybrid neural systems: from simple coupling to fully integrated neural networks (1999, background): “The work of researchers such as Feldman [19, 20], Ajjanagadde [4, 5], Smolensky [82] and Shastri [79] has given important insights into why neural networks are suitable for implementing higher cognitive functions usually associated with symbolic processes.”
Frequency Effects on Syntactic Rule Learning in Transformers (2021, background): “Many natural language phenomena are best described as the product of applying rules to abstract symbols, without access to the content of these symbols (Smolensky, 1988; Fodor and Pylyshyn, 1988).”
Dual-Process Models in Social and Cognitive Psychology: Conceptual Integration and Links to Underlying Memory Systems (2000, unknown): “In this article, we proposed a model involving two processing modes that we label associative and rule-based (Sloman, 1996; Smolensky, 1988), which draw in different ways on two underlying memory systems .”