Scallop, Viara, and the Neuro-Symbolic Path to Smarter AI

We dive into Scallop, a Datalog-based declarative language that acts as a scalable symbolic reasoning engine, capable of discrete, probabilistic, and differentiable reasoning, and how it integrates with PyTorch workflows. We unpack Viara, which adds probabilistic relational reasoning for foundation models (GPT, CLIP, SAM), using foreign predicates and foreign attributes to call external models. Tune in for practical patterns like semantic parsing and structured text extraction, and a discussion on why neuro-symbolic AI could unlock deeper understanding and generalization.

Note: This podcast was AI-generated, and sometimes AI can make mistakes. Please double-check any critical information.

Sponsored by Embersilk LLC