ChatGPT's Declining Accuracy Raises Concerns Among Experts

A study conducted by Stanford University found that OpenAI's AI chatbot, ChatGPT, experienced significant performance fluctuations over a few months. The study compared two versions of the technology, GPT-3.5 and GPT-4, in tasks such as solving math problems, generating software code, and visual reasoning. The results showed that the accuracy of GPT-4 in solving math problems dropped from 97.6% to 2.4% within three months, while GPT-3.5 showed the opposite trend. The study also revealed that the models' ability to explain their reasoning and engage with sensitive questions decreased over time. The findings highlight the unpredictable effects of changes in the model and the need for continuous monitoring of performance.