Reproduce “Multiple Realizability and the Rise of Deep Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sam Whitman McGrath, Jacob Russin
Researchers from Brown University re-examine the multiple realizability thesis in light of deep learning advancements, distinguishing between ontological autonomy and methodological autonomy. They argue that while psychological states are ontologically distinct from their physical bases, studying their implementations, including artificial neural networks, is crucial for cognitive science research.
Key insights
- The multiple realizability thesis correctly implies ontological autonomy (psychological states are not reducible to specific physical types) but does not entail methodological autonomy (studying physical implementations is irrelevant).
- Deep Neural Networks and Large Language Models provide empirical demonstrations of non-biological intelligence, making the study of their physical implementations relevant for understanding cognitive phenomena.
- Analytic distinctions are crucial for guiding empirical research; conflating different types of autonomy has led to misguided research practices in cognitive science.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.