ChatGPT is one of the most popular and powerful AI tools available on the internet. Notably, users across the world have used it for multiple purposes ranging from creating content ideas to solving mathematical equations. However, much to one’s dismay; several users who regularly use GPT-4 have claimed that it is getting worse with time. A recent research paper from Stanford University sheds light on the performance of GPT-4, the underlying technology behind ChatGPT Plus. The findings confirm what users have been reporting for several weeks – that the AI chatbot is experiencing a decline in its capabilities. The paper delves into a detailed analysis of GPT-4, comparing it to its predecessor, GPT-3.5. The researchers discovered that the software driving ChatGPT is actually delivering noticeably poorer performance. The researchers, Lingjiao Chen, Matei Zaharia, and James Zou, conducted experiments comparing the language model's releases in March and June. They observed a degradation in performance. For instance, when asked whether 17077 is a prime number, ChatGPT's accuracy plummeted by a staggering 95.2 per cent. In contrast, GPT-3.5, which powers the free version of ChatGPT, saw a significant improvement, with accuracy rising from 7.4 per cent to 86.8 per cent for the same question. Notably, Peter Welinder, OpenAI's VP of Product, responded to these complaints and requested evidence to support the claims.
FAQsQ1. Why is ChatGPT-4's performance declining? A1. The researchers are not sure why ChatGPT-4's performance is declining. They speculate that it may be due to a number of factors, such as changes in the training data, the way that the model is fine-tuned, or the way that it is used.Q2. What is OpenAI's response to
Read more on economictimes.indiatimes.com