Written in SCIENCE he
Wikipedia lives and dies by references, which are links to sources that support the information in the encyclopedia online. However, sometimes these references are incorrect. They point to malicious websites, misinformation, or untrustworthy sources.
A study published on October 19 in Nature Machine Intelligence 1 shows how artificial intelligence (AI) can help clean up inaccurate or incomplete reference lists in Wikipedia entries and improve their quality and reliability. Suggests it may be useful.
Fabio Petroni and his colleagues at Samaya AI, based in London, analyze whether the references of Wikipedia and suggest better alternatives to the claims they do support, they developed a neural network-driven system called SIDE. they don’t do that.
“It may seem ironic to use AI to help you date, given that ChatGPT He’s very buggy and crazy about dating. It’s important to remember that he’s not just a chatbot,” says Noah Jansiracusa, an AI researcher at Bentley University in Waltham, Massachusetts.
What are SIDE filters?
SIDE AI filters are advertised on the site and are able to find good references in existing articles of Wikipedia which receive a lot of attention from editors and moderators.
It then uses a validation system to identify referrals to sites that contain low-quality information, testimonials can identify complaints. You can also search for reliable sources on the Internet and consider options for replacing incorrect quotes.
To test the system, Petroni and his colleagues used SIDE to suggest references to featured articles on Wikipedia that they had never seen before. In almost 50 percent of cases, the recommended SIDE references are already cited in the article. For others, I found alternative references.
When we show the results of SIDE To a group of Wikipedia users, 21 percent preferred the quote found by the AI, 10 percent preferred the existing quote, and 39 percent had no preference.
The tool can save editors and moderators time verifying the accuracy of Wikipedia entries, but only if implemented correctly, said Alexandra Ullmann, a computer communications researcher at the University of Zurich in Switzerland. “This system could help detect estimates that may not match,” she says.
Ehrman discovered that users of Wikipedia Those who tested the SIDE system were twice as likely to not prefer any of the references suggested by the AI. “So even in these cases, you’ll probably search online for relevant quotes,” she says.