Zooming in on AI patterns and use cases
Figure 8.6 illustrates the most common AI patterns and use cases:
Figure 8.6 – AI patterns and use cases
Before diving deeper into the patterns, let’s first examine the key concepts that underpin most of them, namely, embeddings, vector databases, and hybrid search.
Retrieval augmentation generation
LLMs are trained on a huge amount of training data and parameters. However, they know nothing about your data, unless that data was made public. To compensate for this, the RAG pattern emerged and gained popularity shortly after the rise of GenAI. In essence, it enhances LLMs by incorporating external knowledge—such as your organization’s data—into the generation process. It has been particularly used in chat with your data scenarios, although it is not restricted to chatbots. Figure 8.7 shows two variants of a RAG workflow.
Figure 8.7 – RAG workflow
One way to achieve RAG...
