December 27, 2023: Vector search has found its way into GenAI to enable the training of LLMs by locating similar data points across unstructured data like text. Vector search and storage allow unstructured data to be loaded into data warehouses with the ability to retrieve the data. Vectors allow similarity searches and not only exact matches, which makes them ideal for locating similar data points. This capability has led data management vendors such as Databricks, Snowflake, Oracle, and MongoDB to introduce the capability during 2023.
As Sahir Azam, chief product officer of MongoDB, explains, “You can take any source of knowledge and turn it into a complex mathematical representation, which is a vector, and then you can run similarity searches that find objects that are related to each other without ever having to define their characteristics manually. That’s extremely powerful.”