The British Broadcasting Corporation has developed an investigation into the representation of news content by artificial intelligence assistants and discovered this in Tool responses contain 51 % on major problems.
a wall At least 91 % of the answers that contain some problems, 19 % of the answers created from artificial intelligenceIncluding incorrect numbers and dates, and 13 % of the aforementioned materials have been changed or not in the original sources BBC.
“This inaccuracy can cause the public in error and distort their understanding of critical issues.”Read in the study.
From the studied chatbots, Google Gemini had the highest error rateWith 46 % of their discriminatory answers with great accuracy problems, with errors that include incorrect citing official data and important dates, which were distorted with the BBC content in 34 % of cases.
There is no case of IA from Microsoft, Copilot event, 27 % distortion of cases, confusion in 17 % and ChatGPT by 15 %.
In many cases, AI assistants selectively cited the BBC articles, which provides incomplete novels that distort the original, so “This inaccuracy has traces in the real world, where the public can make decisions based on incorrect information.”.
Looking at this context, the BBC suggests three basic steps for the future Artificial intelligence companies must cooperate with editors to ensure accurate representation of the news content.
The organization may also be necessary to maintain the safety of the mediaOn the pretext that regular reviews and reviews should be made, as well as considering the creation of instructions for news resulting from artificial intelligence.
The study was based on the analysis of 100 news associated with news, artificial intelligence assistants to reach the BBC content during the investigation period, and journalists evaluated the answers based on accuracy, appointment of sources, neutrality, editing, context and representing the BBC content.
