According to news portal Semafor, one of America's largest newspapers, The Washington Post, was discovered publishing AI podcasts with many serious errors and fake quotes generated by neural networks.
The Washington Post recently announced the launch of a new podcast series created by AI. These episodes will be available on the newspaper's mobile app, allowing users to choose the topic, presenter and episode length. In the future, listeners will even be able to ask questions.
However, less than two days after posting the first article, newspaper staff discovered errors in the AI content. These errors range from minor pronunciation problems to significant changes in the document. The podcasts include misquotes (or fabrications), fabricated comments, and present the opinions of sources as newspaper opinions. This was also noticed by users who left editors confused about the quality of AI's work.
Semafor reports that the editors were alarmed by these findings. In an internal message, they admitted that the content generated using a neural network, whose name was not disclosed, did not meet the publication's standards. The newspaper's head of standards, Karen Penciero, expressed disappointment, saying the errors were “frustrating for everyone involved”.
One of the publication's editors expressed disappointment in Slack. He said = that the Washington Post had damaged its journalistic reputation by publishing not only false information but also fabrications. The editor noted that if the newspaper complies with its principles, it will immediately stop using AI tools to create podcasts.
