The most visited website in the world, Wikipedia, has long been trusted to provide accurate and up-to-date information on any and all topics. Since it is freely accessible and might be altered by anybody, however, its reliability has been called into doubt.
The external links in Wikipedia are crucial since they give independent confirmation of the content. Let’s assume, nevertheless, that there are problems with these citations. In order to increase trust in Wikipedia, AI is currently being utilized to improve the site’s citations.Â
This novel strategy was initially revealed in a study released on October 19 by Nature Machine Intelligence. The research shows that a neural network can efficiently spot questionable sources in Wiki entries. Improve the quality and trustworthiness of Wikipedia as a whole by updating these citations.Â
Samaya AI, an artificial intelligence firm located in London, has stood up to the challenge. Samaya AI’s SIDE (System for Identifying and Diagnosing Errors in References) was built under Fabio Petroni’s direction using neural networks. Our intention in creating SIDE was to check if there was sufficient proof for the assertions made in Wikipedia entries.
When SIDE finds a disparity, it does more than merely sound an alarm. It does an intensive online search for better options. Given that AI technologies like ChatGPT have been criticized for botching citations in the past, the use of AI for this task can be seen as ironic. However, experts believe that artificial intelligence may be used for far more than just chatbots.
Noah Giansiracusa, a researcher in artificial intelligence at Bentley University in nearby Waltham, MA, stressed this point. While he did admit that chatbots had their limits, he did highlight AI’s enormous promise in the area of language modeling. Artificial intelligence may be used to enhance citations, and it does so in a very effective manner. Â
The instruction at SIDE was quite thorough. It was trained using a hand-picked collection of high-quality Wikipedia articles. Products selected here have been reviewed by our moderators and editors and found to be superior in every way. SIDE was taught to recognize when evidence was lacking to support a claim. It may also use sophisticated algorithms to scour the web for reliable resources, which it would then rate according to criteria like usefulness and authority.Â
Petroni’s team probed deeply to determine SIDE’s practical value. During the testing period, they fed SIDE highly rated Wikipedia articles it had never seen before. The findings were not only encouraging, but they also demonstrated the potential efficacy of SIDE. Reference SIDEs for almost 50% of the test cases were included in the published work. The remainder of the issue was resolved expertly by presenting other options that had been thoroughly investigated.Â
The advice of Wikipedia editors is invaluable in this respect. Opinions of frequent Wikipedia contributors were sought. Surprisingly, although 21% of users preferred the references identified by SIDE, 10% preferred the ones already there, and 39% did not like either. Successful implementation of SIDE will depend heavily on how the Wikipedia community responds, says Aleksandra Urman, a computational communication expert at the University of Zurich in Switzerland.Â
As a result of using AI to improve Wikipedia’s citations, the quality of online content has increased significantly. While these findings are promising, it needs to be seen whether or not the Wikipedia community will really accept and make use of tools like SIDE. The future is bright because AI will be used to verify the reliability of our data sources.Â
Journal Reference Â
Stokel-Walker, C. (2023). AI tidies up Wikipedia’s references and boosts reliability. Retrieved from https://www.nature.com/articles/d41586-023-02894-xÂ
