OpenAI Addresses GPT-4's 'Lazy' Behavior

TL;DR
OpenAI has acknowledged that its chatbot, GPT-4, has been getting "lazy" and not completing tasks requested by users. The company stated that it hasn't updated the AI model since November 11 and that the laziness is unintentional. OpenAI also mentioned that training chat models is a complex process with unpredictable outcomes, including differences in personality, writing style, refusal behavior, evaluation performance, and political bias. The company is looking into fixing the issue and appreciates user feedback.
- OpenAI Acknowledges GPT-4 Is Getting 'Lazy' PCMag
- OpenAI is working to make GPT-4 less lazy. The Verge
- OpenAI to Investigate Users' Claims That ChatGPT Is Getting 'Lazier' The Daily Beast
- OpenAI Investigates 'Lazy' GPT-4 Complaints On Google Reviews, X Search Engine Journal
- OpenAI Admits GPT-4’s ‘Lazy’ Behavior: What You Need to Know Gizchina.com
Want the full story? Read the original reporting
Read on PCMag