Wikipedia, the online encyclopedia, is not really a reliable source of information. People will sometimes attempt to edit Wikipedia entries maliciously, although most factual errors are caused by a well-intentioned person getting it wrong.
This is an issue that the Wikimedia Foundation has recently addressed in collaboration with Facebook parent company Meta. The two decided to focus on citations. There are nearly too many footnotes on Wikipedia for the platform’s volunteer editors to check. With the internet adding over 17,000 new articles per month, numerous citations are incomplete, missing, or just incorrect.
Meta created an AI model that can analyze citations at scale to ensure their accuracy. When it comes across a poorly referenced piece, it might also offer alternate citations. Human editors at Wikipedia use common sense and experience to assess citations. When an AI does the same task, it employs a Natural Language Understanding (NLU) transformation model to understand the various relationships between words and phrases inside a text. The system’s knowledge index is Meta’s Sphere database, which contains over 134 million web pages. The model is meant to locate a single source to validate every claim as it checks the citations in an article.
To demonstrate the AI’s capabilities, Meta gave an example of an unfinished citation the model discovered on the Blackfoot Confederacy‘s Wikipedia page. The article mentions Joe Hipp, the first Native American to contend for the WBA World Heavyweight title, in the Notable Blackfoot People section. Hipp or boxing are not mentioned on the linked website. The model discovered a more appropriate reference in a 2015 article from the Great Falls Tribune when searching the Sphere database. The model highlighted the following passage:
“In 1989 at the twilight of his career, [Marvin] Camel fought Joe Hipp of the Blackfeet Nation. Hipp, who became the first Native American to challenge for the world heavyweight championship, said the fight was one of the weirdest of his career.”
The above passage is interesting in that it does not explicitly reference boxing. By its natural language capabilities, Meta’s model discovered an appropriate reference. The technique may one day be useful in addressing Facebook’s misinformation issues. “More generally, we hope that our work can be used to assist fact-checking efforts and increase the general trustworthiness of information online,” the model’s developers stated. Meanwhile, Meta intends to provide a framework for Wikipedia editors to utilize to routinely check and amend footnotes.
- Google Introduces AI Mode: A Game-Changer for Search - March 12, 2025
- Security Researchers Find DeepSeek Security Vulnerabilities - February 6, 2025
- Tech Giants Unite to Tackle Child Safety Online with ROOST - January 11, 2025
