Vector Engine as a Booster for AI

At the start of TechEd 2023, SAP Chief Technology Officer Jürgen Müller said he was excited to announce one of the most important, if not the most important, enhancement to the Hana database platform: the ability to use vectors as objects on the database platform—a Hana sensation that goes beyond Large Language Models (LLMs), which merely use deep learning algorithms to summarize and organize large amounts of data.

Jürgen Müller’s argument at Tech-Ed 2023 in Bangalore was not wrong. Large language models can usually only capture the past. They are trained on existing data, usually extracted from the internet, and immediate, real-time response based on operational data is difficult to achieve. The SAP Hana database, however, has been responsible for real-time results for many years—now with vectors!

LLM, vectors, and data fabric

SAP clarified and expanded the announcement made at TechEd 2023 earlier this year. New capabilities in the SAP Datasphere solution, including generative AI, are designed to optimize business planning through simplified data landscapes and intuitive work with data. "Capturing data technologically and using it to support decisions is becoming increasingly important. At the same time, AI based on high-quality data is revolutionizing all aspects of a company," said Jürgen Müller, Chief Technology Officer and member of the SAP Executive Board. "Our latest innovations in SAP Datasphere and the expanded partnership with Collibra are a critical step in enabling customers to drive their intelligent business transformation with data."

At the heart of the recent announcements stands a data fabric for managing business data. A data fabric, as defined by Gartner analysts, is an architecture that ensures that data is not only used as a resource, but also as the critical foundation for strategic initiatives. The innovations and partnerships announced will enable organizations to deliver meaningful data to data users while preserving business context and logic.

Thomas Failer, SAP partner and CEO of Data Migration International, said: "We want to be the leading provider of intelligent data management and an intelligent enterprise data fabric by 2027. We have defined milestones along the way!" The extended functions in the area of retention management, i.e. the retrieval of personal data and the generation of business objects with the help of AI, should be available to customers in the spring of 2024 with version 11 of the DMI platform for information management, JiVS IMP. The data fabric market is currently highly competitive: "We decided to use SAP Datasphere for a modern business data fabric architecture in conjunction with our S/4 Hana landscape. This allows us to effectively build a modern data analytics platform that enables business self-service based on trusted data and data models," said Achim Welter, Senior Director ERP, Digital and IT Strategy, Hershey's.

Euclidean distance

With Datasphere, SAP customers can simplify their data landscapes with an end-to-end view of their data. As a result, they can adapt more quickly to market changes and make better decisions. From the new AI assistant and vector database capabilities that ensure business context is preserved in generative AI results, to the new Knowledge Graph that enables insight into complex data and pattern recognition, SAP innovations are helping customers take advantage of their data to the fullest extent.

The Euclidean distance of multidimensional vectors is an excellent way to cluster objects, as SAP Chief Technology Officer Jürgen Müller noted in his press statement from Bangalore, India, where SAP Tech-Ed took place in 2023. Hana is now able to calculate Euclidean distances, allowing it to find similarities in large amounts of data.

The new Knowledge Graph in SAP Datasphere, in addition to Hana, enables organizations to discover hidden insights and patterns in their applications and systems. This enables both technical and business users to better understand the relationships between data, metadata, and business processes, improving the effectiveness of machine learning and large language models. Clustering, or finding similarities, can often be an enlightening way to look at large amounts of data. The trick is based on calculating Euclidean distance and can be done without Hana as a simple mental calculation. According to SAP Chief Technology Officer Jürgen Müller, the supposed sensation of Hana and Datasphere is the ability to use vectors as objects on the database platform. In traditional Euclidean mathematics, vectors are not exactly a new and exciting sensation. With the Hana database platform’s current tools, any first-year computer science student can implement a few simple vector functions.

What is a vector? In a coordinate system with an x-axis and a y-axis, you can select any two points—if you connect these points with the shortest possible straight line and add an arrow at the end, you have a directed graph or vector in two-dimensional space. It is also easy to imagine a vector in three-dimensional space (x-, y- and z-axis), e.g. a pencil lying on a table. The end and the tip of the pencil can be precisely defined as points in space. So, the pencil would be the vector. Now we move on to higher dimensions, which are difficult to visualize (for example, a four-dimensional cube would have a three-dimensional shadow), but are still easy to compute in higher dimensions. Joule, the SAP wizard based on generative artificial intelligence, is now available in the SAP Analytics Cloud solution to automate the creation and development of reports, dashboards, plans, and more. The capabilities of the SAP Hana Cloud Vector Engine combine the power of large language models with relevant business data, ensuring that the business context is preserved in the results of generative AI.

Hana Cloud Vector Engine

What SAP Chief Technology Officer Jürgen Müller presented at TechEd 2023 in Bangalore is of great practical importance. Many Hana customers will appreciate the language extension into higher dimensions. However, the math presented is quite traditional and familiar. It is strange to present Euclidean distance as a milestone in Hana development, but for SAP customers, the capabilities of the SAP Hana Cloud Vector Engine are an important step toward business generative AI.

The Euclidean distances of vector engine graphs can be used as similarity measures for clustering. They can be used to create excellent maps with regions (clusters) that provide direct information about preferences, similarities, and trends. About ten years ago, the E3 Publishing House and Professor Alfred Taudes of the Vienna University of Economics and Business Administration created exactly such a Hana map based on a survey of the SAP community.

sap.com