Chatbots Can Go into a Delusional Spiral. Here’s how it happens.

Researchers have identified a phenomenon called 'delusional spirals' in AI chatbots, where models like ChatGPT begin to generate increasingly inaccurate and bizarre responses after being fed incorrect information. The study highlights how these errors compound over time, raising concerns about the reliability of AI systems in critical applications.

Article URL: https://www.nytimes.com/2025/08/08/technology/ai-chatbots-delusions-chatgpt.html Comments URL: https://news.ycombinator.com/item?id=48862650 Points: 1 Comments: 0