AI and Elasticsearch: Entering a New Era with Elastic

Artificial intelligence is increasingly present in the tools we use and create. Obviously, it's more than just a chat agent with general knowledge. And instead of building your own tools from scratch (for your teams or customers) or adding more dependencies, you can leverage what's already available at your fingertips.
In this section, we'll look at how you can easily benefit from Elastic's AI work: not only in search, but also in web observability and reliability. The biggest effort should be to upgrade to the latest version.

With increasingly powerful hardware and capable models, vector text representation and semantic search are becoming increasingly attractive options for information retrieval use cases. In this talk, we will use Elasticsearch as a tool for implementing semantic search using data from the Czech Wikipedia. The talk will provide a high-level overview of the basic concepts of semantic search, as well as a code-based demonstration of the entire workflow with several potential performance optimizations.

• Difference between semantic and lexical search
• Basic concepts of embeddings and vector similarity
• Pre-trained models, multi-linguality
• Pre-processing and splitting the data
• Efficient generation of embeddings
• Efficient indexing of embeddings
• Efficient inference
• Nearest neighbour search in Elasticsearch
• Search quality evaluation with the Rank Eval API

The presentation will focus on the principles of RAG (Retrieval-Augmented Generation) search and its implementation using Elasticsearch. We will explore what RAG search brings to the world of information retrieval and how it increases the quality and relevance of answers. We will also dive into the practical use of Elasticsearch for RAG applications, including basic setup and configuration for effective search. We will explore current tools and frameworks that enable easy and efficient RAG application development and demonstrate how they can be used to develop advanced internal data search and response systems. The goal of the presentation is to provide a comprehensive view of the possibilities of using RAG in search.

• Introduction to RAG (Retrieval-Augmented Generation) search and its benefits
• The role of Elasticsearch in RAG applications
• Setup and configuration of Elasticsearch
• Tools and frameworks for building RAG applications
• Examples of RAG usage in search systems

As the prevalence of Generative AI has witnessed a substantial surge in the past year, an increasing number of enterprises are integrating Generative AI solutions into their daily operations. A September 2023 Gartner survey of 1,500 executives found that 45% of organizations are actively piloting multiple generative AI solutions.
While large language models are at the heart of most of these applications, a successful solution covers much more than just a vector search-based model wrapped in an API.
The goal of this talk is to take the technical concepts presented at this conference and transform them step-by-step into a commercially viable product – including identifying the right problem to solve, managing integration with current systems, and addressing the dreaded security and compliance requirements.

Outline

• Business case: a problem worth solving
• The specifics of Proof-of-Concepts
• User Interface: we need to talk to each other
• Infrastructure: running our application
• Integrations: living inside the company’s ecosystem
• Monitoring and Analytics: knowing what’s going on
• Security & Compliance: the less favored part of applications