Wikipedia has launched a new initiative called WikiProject AI Cleanup, headed by Ilyas Lebloux. The goal of this project is to detect, edit, or delete articles created by neural networks that contain fake information. The number of such articles is growing every day, which is causing concern among the community.
The problem was first brought to the attention of moderators who noticed that some articles had the characteristics of texts generated by artificial intelligence. To confirm the suspicions, ChatGPT was used to play back fragments of the texts, which helped to establish their origin.
Participants of the WikiProject AI Cleanup project will carefully check suspicious materials. If it turns out that the text was created by artificial intelligence, depending on the level of falsification, appropriate decisions will be made – editing or complete removal.
One example of such an article is a fictitious story about the Amberlisihar fortress, which allegedly existed during the Ottoman Empire. Despite the fact that such a fortress never existed, the article described its location, architecture and construction in detail, which could have misled readers. Some real facts inserted into the text only added to the credibility of this fake.
Lebloux notes that it is not yet clear why users publish such AI-generated materials, but the problem is that according to Wikipedia’s rules, anyone can edit or add articles, which creates room for the spread of false information.
