Yottaanswers.com – Delivering Billions of Smart Answers Efficiently
summary
Claims a proprietary index of 27 billion vectors plus a separate 2 billion cause‑and‑effect vector store that together produce traceable, non‑hallucinating answers at web scale, enabling tens to hundreds of billions of smart answers while using far fewer compute resources. Final summarization uses LLMs that are about 1,000x smaller (run and fine‑tune on a single machine), nearest‑neighbor ranking enables breadth‑first exploration, and this approach targets profitability versus massive training costs.
AI-written summary · about
chapters (11)
- 0:00Introduction to Yottaanswers and the talk
- 1:29AI scale and secrecy myth
- 3:42AI training costs
- 6:11Training costs barrier
- 7:5727B vector system
- 10:52Smart answers vs Google
- 11:55Breadth-first search importance
- 13:05Reasoning with 2B vectors
- 15:32Democratizing AI
- 17:02Business case for AI
- 18:08Q&A
related talks
26mThe geometry of Intelligence in a Large Question-Answering System
28mGive users relevant answers quickly with intelligent search
20mTransforming the Search Industry with LLM Using Source Frontier
35mYou've got questions. We've got answers! #bbuzz
01h 09mVector Search: Ask Me Anything!
38mFrom text to knowledge via ML algorithms - the Quora answer
25mXavier Amatriain, VP of Engineering, Quora @ MLconf SEA
47mBuilding intelligent analytics at scale