Ontologies Are Back: UC Berkeley Professor Explains How They Guardrail AI Agents
Latent Space · rss · 2026-07-30
At the AI Engineer World’s Fair, UC Berkeley professor Frank Coyle argued that while LLMs excel at probabilistic reasoning, autonomous agents need "ontologies" as logical guardrails to be truly reliable. He calls this convergence of probability and rules "neurosymbolic AI."
Ontologies, essentially graphs describing data structures, offer several advantages for agentic systems:
- Ready to use: Established Semantic Web standards like Schema.org are already in LLM training data, allowing developers to prompt for them directly.
- Bounding loops: In loop engineering, ontologies act as a bounded set of rules around an unbounded LLM loop, keeping agents on track.
- Thin Agents: Neo4j CEO Emil Eifrem proposed an ontology-based semantic layer, enabling a shift from thick agents with manually wired data sources to "thin agents."
While ontology maintenance remains a challenge, developers suggest agents could autonomously update definitions during runtime. As the industry shifts away from undisciplined "vibe coding," traditional web ontologies are experiencing a revival in AI engineering.
More from coding & agent
- tuicr: A Rust-based Code Review TUI with Vim Keybindings — agavra · 2026-07-30
- Neo4j Launches 2-Hour Free Course on Building Graph-Backed Agents — JeremyCMorgan · 2026-07-30
- 'Agreement is a Bug': Forcing 11 Claude Code Agents to Disagree — w1kke · 2026-07-30
- Sakana AI's Agent Secures 5th Place in OSINT CTF Using Fugu-ultra 1.1 — SakanaAILabs · 2026-07-30
- Kimi K3 Builds $31M Hotel Project: Slashes MEP Modeling Costs by 77% — NandoDF · 2026-07-30
- Uncle Bob Says He Doesn't Read Code Generated by AI Agents — blaizedsouza · 2026-07-30