In the field of generative AI models, characterized by the impressive capabilities of Large Language Models (LLMs), a notable limitation persists. These models lack access to company-specific or recent data beyond their initial training set. In a groundbreaking announcement at TechEd 2023, SAP unveiled a strategic solution set to launch in Q1 of 2024: the SAP HANA Cloud Vector Engine. This innovative product aims to bridge the gap by integrating the power of generative AI with organization-specific data—a form of the revolutionary technique known as Retrieval Augmented Generation (RAG).
THE VECTOR ENGINE EXPLAINED
At the core of the Vector Engine’s functionality are “embedding functions,” which are instrumental in transforming unstructured data objects—ranging from text to images and audio—into high-dimensional vector embeddings. Think of these vectors as pointers within a high-dimensional space, with each dimension represented by a numerical data point. The conversion of data objects into numerical representations preserves the semantic similarities between them. Consequently, the Vector Engine enables users to assess the degree of similarity between two objects by measuring the distance between them in this high-dimensional space, providing valuable insights—such as identifying words with similar meanings.
The Vector Engine seamlessly integrates with SAP HANA database tables, enabling its use in various machine learning and data analytics tasks through simple SQL statements. The vector embeddings are stored in a field of type REAL_VECTOR, which contains the numerical representation of each dimension. By leveraging the built-in COSINE_SIMILARITY function, users can easily identify the nearest neighbor of an object based on the cosine of the angle between their vectors.
FROM THEORY TO PRACTICE
Since the concept may seem somewhat abstract, let’s look at a practical example to illustrate how developers can use the Vector Engine to improve business outcomes. Imagine a retail scenario where the Vector Engine is used to boost customer engagement. By storing customer reviews, purchase history, and feedback as high-dimensional vectors, the Vector Engine enables businesses to identify similar sentiments or preferences. This facilitates personalized recommendations, ultimately refining the customer experience. Furthermore, when combined with generative AI, it opens the door to personalized customer support, targeted marketing messages, and more.
An alternative use case could involve storing system documentation as vector embeddings and using them as context when employing AI for development support. This approach even allows you to ask specific questions about your system documentation. Of course, these are just a few of the potential use cases, and only time will reveal the full range of applications for which this technique will prove most useful.
EMBRACING A PROMISING FUTURE
In essence, SAP’s HANA Cloud Vector Engine heralds a new era by unlocking the potential of company-specific data in the realm of generative AI. As we look forward to its launch in Q1 of 2024, the Vector Engine promises to revolutionize how businesses leverage the capabilities of AI models, ushering in a paradigm shift toward more contextually appropriate results and paving the way for unprecedented advancements in the field.
Are you curious to learn more about the Vector Engine and how you can use it to optimize your business results? Feel free to contact us! You can also follow us on LinkedIn to stay up to date on the latest developments related to AI and other SAP technologies.
For a general recap of SAP TechEd 2023, check out our latest blog: SAP TECHED RECAP – EXPLORING BTP AND AI.
