A report from Columbia's Tow Center reveals AI chatbots, including ChatGPT and others, frequently misidentify news articles, leading to inaccuracies and fabrications. The study found that across 1,600 queries, more than 60% of responses from chatbots were incorrect regarding article details. Variability in performance was highlighted, with some chatbots exhibiting alarming error rates of up to 94%. This raises concerns about the reliability of AI tools in journalism and their ability to provide trustworthy and accurate information to users.
"Tow researchers randomly chose 10 articles each from 20 publishers. They queried eight chatbots with article excerpts, asking the AI to return the headline, publisher, date, and URL of the corresponding article."
"Failing to identify news articles, Tow researchers found chatbots returned wrong answers to over 60% of the queries, significantly hindering news retrieval accuracy."
Collection
[
|
...
]