Unconference: How will AI + Search (IR) converge?
summary
Argues that information retrieval has a 50 year history that AI folks often ignore and that popular fixes like vector DB vs search misunderstand the need for integrated IR. Calls out current hybrid numerology — ad hoc rank blending — and proposes feeding BM25 and neural outputs into a true reranker instead of pairwise tweaks. Stresses measurable evaluation and business outcomes, argues for multimodal search (video/image/audio) and for teams to measure actions not just relevance to prove value.
AI-written summary · about
related talks
13mThe State of AI Powered Search and Retrieval
37mEnhancing AI and LLM Systems with Hybrid Search
30mBeyond Keywords: Building an AI-Powered Search That Reads Your Mind
44mVector Search Isn’t Enough
28mKeynote: The Next Generation of Enterprise Search: Intent‑Aware, Hybrid, and Open
41mResearch in the Age of AI
01h 30mOptimizing Retrieval for Agentic Systems
40mAgentic Retrieval: Building Self-Optimizing Search Systems