Explore the latest developments in AI-powered search technologies through this insightful conference talk delivered by Simon Hughes at the Haystack on Tour event in Austin, Texas. Delve into the challenges faced by search engines, the evolution of search methodologies, and the impact of artificial intelligence on information retrieval. Gain valuable insights into neural IR, tokenization techniques, and advanced retrieval models. Examine the intricacies of behavioral matching, emerging search architectures, and the rise of conversational search. Analyze the capabilities and limitations of large language models, including hallucination issues and their practical applications. Discover the potential future directions of AI in search and the implications for user expectations and information access.
Overview
Syllabus
Intro
What is Veara
Talk Outline
Search Engine Challenges
Search and AI
History of search
Time
Retrieval
Neural IR
Tokenization
Handling typos
Handling vocabulary tokens
Retrieval models
Training models
Positive signals
Negative mining
Types of systems
Behavioral matching
Search architectures
Alibaba
Conversational search
User expectation of search
Models make stuff up
GP Open AI
Galactica Model
Hallucination
Why do models hallucinate
How are people using these models
Problems with closed book systems
Training our own model
LLM leaderboard
New York Times article
Next steps
Taught by
OpenSource Connections
Related Courses
-
Large Language Models with Semantic Search
-
Text Retrieval and Search Engines
3.2 -
Natural Language Processing with spaCy
-
Generative AI Architecture and Application Development
-
ACM ICTIR 2024 - Theoretical Aspects of Information Retrieval
-
Theory of Information Retrieval - ACM ICTIR 2024 Conference Presentations
