Wikimedia Widens Wikidata for AI
The Wikimedia Foundation has launched a new initiative to make its vast knowledge base more user-friendly for artificial intelligence (AI) applications. As part of this effort, it has introduced the Wikidata Embedding Project.
According to Wikimedia, the project will allow even small AI developers to access and use vectorized data from its massive knowledge repository—an opportunity that was previously limited to large technology companies.
The project involves converting nearly 120 million Wikidata entries into numerical vectors, or embeddings. This transformation will make the data easier to process for large language models (LLMs) and other AI systems.
While Wikidata has long been machine-readable, it has not been directly optimized for generative AI. Through the new embedding process, related concepts (such as “dog” and “puppy”) will be positioned closer together in vector space, enabling AI systems to better understand semantic relationships.
The announcement comes at a time when Tesla CEO Elon Musk has unveiled plans to launch Grokipedia, a rival platform to Wikipedia. Many observers see Wikimedia’s latest initiative as a strategic move to safeguard and promote open access to knowledge.
