Meta Is Building an AI to Fact-Check Wikipedia—All 6.5 Million Articles

Meta is developing a machine learning model to fact-check the 6.5 million articles on Wikipedia by scanning citations and cross-referencing their content to verify the accuracy of facts and figures. The AI will “understand” content using natural language understanding techniques, comparing mathematical representations of text blocks instead of word strings. Meta’s model is being trained on a set of four million Wikipedia citations and could eventually suggest accurate sources to replace faulty ones. The AI tool aims to improve the accuracy and reliability of information on the online encyclopedia but is currently in the research phase and not used for updating content.