Wikimedia, the non -profit organization behind Wikipedia and sister sites like Wikimedia Commons and Wikidata, simply made AI models to draw from its massive knowledge base.
Wikimedia Deutschland, the German organization of the organization, has published a new resource called the Wikidata integration project. It takes approximately 120 million open data stored in Wikidata and converts them into a simpler format for large languages ​​to use.
Even if the structured data from Wikidata is already readable by machine, it has not been directly compatible with generative AI systems, which are designed to operate with natural language.
The new project translates Wikidata into vectors, which are essentially digital coordinates that show how different declarations are linked to each other.
Think about it as a map where terms closely linked as “dog” and “puppy” come together, while unrelated terms like “dog” and “bank account” are much more distant. This helps AI systems to understand the terms in the context and to treat them more effectively in natural language.
The project is designed to provide information on better quality AI that leads to more reliable responses, Wikimedia Deutschland said in a press release. He said most AI systems are currently based on opaque data sets.
A secondary objective is to level the playing field. By making Wikidata available for free, Wikimedia says it hopes that small IA companies will be able to compete with technology giants that would otherwise have the resources to vectorize the data themselves.
“The launch of the incorporation project shows that a powerful AI does not have to be controlled by a handful of companies – it can be developed openly and in collaboration,” said Philippe Saadé, Wikidata AI project director, in a press release.
Wikimedia Deutschland has been working on the project since September 2024 in collaboration with Jina AI, who has built the incorporation system that transforms Wikidata entries into vectors, and IBM IBM, which stores these vectors in its database.
On the other hand, the Liberation landed just a day after Elon Musk went to X to announce that he is building a Wikipedia rival called Grokipedia.
“We build Grokipedia @xai,” Musk wrote on Tuesday. “Will be a massive improvement compared to Wikipedia. Frankly, this is a necessary step towards the XAI lens to understand the universe.”
Musk ridiculed Wikipedia as “wokpedia” and complained that there is no alternative aligned with more right -wing views. He also republished Larry Sanger, the co -founder of Wikipedia, who left in 2002 and has since tried to launch several competing projects. Sanger, a long -standing critic from Wikipedia from the right, recently published on X that Wikipedia has become too globalist, academic, secular and progressive.
Musk’s offer to build a rival encyclopedia stored with its favorite facts simply emphasizes why Wikimedia launched its own AI project in the first place. While AI continues to become current, the quality and the data on which these systems are based could potentially have an influence on what millions of people believe to be true.
