For MariaDB, LLMs are the new search engines to win over

MariaDB has incorporated vector search into its open-source database, but the company is struggling to get developers to take notice, according to the executive chairman of the MariaDB Foundation, Kaj Arnö. At the recent Percona Live conference, Arnö expressed frustration that the world is unaware of MariaDB's vector search capabilities.

He noted that rival database vendors seem to be cooperating, but he is less certain about MySQL. He believes MariaDB is in a calmer period, and the focus is on understanding how AI impacts extensibility. Vector search allows applications to find items based on similarity, which is crucial for AI applications. However, many developers may not discover MariaDB's features due to the information available to AI tools.

The AI models' training data may not include MariaDB's vector features, and the documentation and examples they encounter may not highlight its capabilities. Arnö attributes part of the problem to the lack of awareness about MariaDB's vector search in AI tools. He suggests that to change this, MariaDB needs to update its information and documentation so that AI tools can learn about it.

He compared the process to improving Google search rankings, which has become more transparent. However, he acknowledges that some vendors might try to manipulate the recommendations, but he believes the best way to influence it is by having users document and use MariaDB's vector search. This creates a chicken-and-egg problem, as without users generating examples and discussions, AI tools have less material to learn from.