AI chatbots are slowly choking humanity as research shows a 'vicious cycle' could cause global knowledge collapse (www.techradar.com)

Recent research from the University of Copenhagen reveals a troubling trend in AI chatbots: they tend to provide significantly less diverse responses compared to traditional search engines like Google. The study tested 27 different large language models (LLMs) on 155 topics, generating 70 million responses. Findings indicate that even the best performer, GPT-5, delivered answers that were 18.5% less varied than those from Google. This homogeneity in responses raises concerns about a potential "knowledge collapse," where dominant viewpoints overshadow diverse interpretations, limiting the breadth of understanding. The implications of this research are significant for the AI and machine learning community. As Professor Isabelle Augenstein notes, this phenomenon resembles the drawbacks of globalization, where convenience comes at the cost of diversity. The reliance on AI for information could create a vicious cycle where popular content becomes even more prevalent, marginalizing alternative perspectives. While the study concludes that a complete knowledge collapse has not yet occurred, it underscores the importance of fostering diversity in AI-generated content to prevent society from drifting towards a less informed future.