Round Table Discussion
summary
Panelists debate concrete paths for Postgres to serve AI: adoption of PG Vector today but need for new index types and GPU/TPU support for higher dimensionality; examples include Milvus-style native GPU acceleration and PGML/PGAI integrations; evolutionary genetic query optimizer from 1996 already exists while modern automation for index and server-parameter tuning runs in production; retrieval augmented generation (RAG) and two NLQ strategies—send retrieved data to an LLM or generate SQL—show 60–85% accuracy in tests and argue for small 300–500M models on edge.
AI-written summary · about
chapters (8)
- 0:00Introduction and panel
- 2:08Postgres and AI challenges
- 6:33Open source vs proprietary
- 11:06Ecosystem and evolution
- 14:19Query tuning and automation
- 19:06Community education and role
- 24:26Future hardware and design
- 30:27Natural language and conclusion
related talks
50mFrom Transactions to Intelligence: Postgres for Open AI & Analyti
34mAnalytics and AI with PostgreSQL - Mission Possible!
27mPutting AI in Production
25mPostgres and the Artificial Intelligence Landscape
35mPostgres in Artificial Intelligence
01h 00mPostgreSQL and Artificial intelligence
34mAnalytics and AI with PostgreSQL - Mission Possible!
53mModern PostgreSQL for AI...