The four generative AI (artificial intelligence) tools (ChatGPT; Copilot, Gemini and Perplexity) to answer 100 questions, using BBC articles as the source. The tool's 'answers' were then rated by BBC journalists specialising in the appropriate areas. More than 50% had 'significant issues'. Errors on numbers, dates or statements occurred in around 20%. Quotes, sourced to the BBC, could be altered. Sometimes, they didn't exist in the article cited. Concern has been expressed that these tools threaten to further undermine the public's "fragile faith in facts" (https://www.theguardian.com/technology/2025/feb/11/ai-chatbots-distort-and-mislead-when-asked-about-current-affairs-bbc-finds}. Generative AI can be an expression of the old 'garbage in: garbage out' adage. Its output certainly has to be handled with care. There's a danger of folk uncritically accepting what comes out of these tools. They might be the source of the next big 'conspiracy theory'?
No comments:
Post a Comment