AI is here – time to throw away our search engines?
summary
Claims that AI simply replaces search get tested with concrete techniques: use vector embeddings for multimodal data, apply BM25 + rerank by computing embeddings on the fly to mix lexical and semantic scores, add MMR and evaluation aggregation to avoid repetitive agent hops, and quantify costs with quantization and token‑burn estimates. Practical tools mentioned include Hornet, Quadrant, OpenSearch profile API, CLIP and vector DBs, plus experimental ideas like wormhole vectors and weighted fusion instead of blind reciprocal‑rank fusion.
AI-written summary · about
chapters (8)
- 0:00Introduction and panel setup
- 1:11Panelist introductions
- 3:19Discussion theme and questions
- 4:22Old school search techniques in AI systems
- 11:46Supercharging search engines with AI
- 19:30Fear of AI replacing search roles
- 28:28Combining old and new search techniques
- 39:46Summary and closing remarks
related talks
45mTomorrow’s SEO: How AI is Reshaping Search
37mBest AI Tools for Web Developers / How to Use AI in Web Development
21mEvaluating AI Search: A Practical Framework for Augmented AI Systems
28mUse OpenSearch as Your Vector Search Database To Optimize GenAI Accuracy!
31mAI Search for Marketplaces: Experiences & Tech
24mAI and the Enterprise
44mAI in Digital Marketing: Benefits and Use Cases
28mAgentic AI