We asked five leading AI models the same 22 questions every day for 10 weeks running, and the results were... mixed.
Great to see the Independent cover the first stage of @fullfact.org’s new project evaluating LLMs
www.independent.co.uk/news/uk/home...
independent.co.uk
Five AI chatbots were tested on accuracy. The results weren’t good
Responses from leading AI models were prone to being ‘incomplete, lacking context, or poorly sourced’, report finds after testing of Claude, ChatGPT, Gemini, DeepSeek, and Grok