Garcez points to 20 years of neurosymbolic AI research as LLM-plus-tools debate resurfaces

luislamb · x · 2026-09-17

Responding to Eric Topol's Nature piece, Artur d'Avila Garcez shared his paper "Neurosymbolic AI: The 3rd Wave" (arXiv:2012.05876, with Luis Lamb), arguing neurosymbolic AI comprises hybrid systems and neurosymbolic integration built on 20+ years of research, and that there is huge value in LLMs plus tools like knowledge graphs and LEAN.

The paper advocates principled integration of neural learning with symbolic knowledge representation and logical reasoning to address trust, safety, interpretability, and accountability, and outlines promising directions and challenges for the next decade of AI research from a neurosymbolic perspective.

Original post →

More from Research

Research channel →