The geometry of Intelligence in a Large Question-Answering System
summary
Claims a production question‑answering engine that encodes text as 256‑dimensional vectors and ranks by angular distance and eigenvector‑based subspace projection, indexing billions of answers (claiming scale similar to Google and tens of billions possible results). Implementation uses semantic search only, a custom index served from a few commodity machines (mentions eBay boxes), offers yottaanswers.com plus an API for researchers, no ads or personalization, and licenses for commercial use.
AI-written summary · about
chapters (5)
- 0:00Start
- 5:49Geometry of Intelligence
- 7:36Mathematical interpretation
- 9:33Quantum superposition
- 13:18Application of the System
related talks
22mYottaanswers.com – Delivering Billions of Smart Answers Efficiently
25mRevolutionizing Access to Global Knowledge
47mBuilding intelligent analytics at scale
01h 09mVector Search: Ask Me Anything!
25mXavier Amatriain, VP of Engineering, Quora @ MLconf SEA
36mFrom Theory To Deployment: Building Scalable Multi-Agent Question-Answering Syste...
16mWhen Vectors Break Down: Graph-Based RAG for Dense Enterprise Knowledge
13mRAG Without the Hassle: Building AI-Powered Applications