Reproduce “Towards Cognitive AI Systems: a Survey and Prospective on Neuro-Symbolic AI” 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
Zishen Wan, Che-Kai Liu, Hanchen Yang, Chaojian Li, Haoran You, Yonggan Fu, Cheng Wan, Tushar Krishna and 2 more
Researchers from Georgia Institute of Technology surveyed the landscape of Neuro-Symbolic AI (NSAI) algorithms, categorizing them into five paradigms, and conducted an empirical system profiling of representative NSAI models. They found that symbolic workloads often dominate execution time and exhibit unique computational characteristics, presenting new challenges for existing hardware architectures.
Key insights
- Symbolic workloads constitute a substantial, often dominant, portion of the total execution time in NSAI models, challenging the assumption that neural components are always the bottleneck.
- NSAI workloads exhibit high heterogeneity, sparsity, irregular memory access patterns, and memory intensity, diverging from the dense matrix operations optimized by current hardware.
- The identified performance bottlenecks in NSAI systems necessitate novel, specialized hardware architectures and software frameworks to efficiently handle their unique computational and communication demands.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.