Recent studies have unveiled that ChatGPT is showing signs of performance degradation after just six months of operation — the accuracy of its answers has dropped by 2.4%.
A group of American scientists from Stanford and Berkeley universities conducted an in-depth analysis of ChatGPT’s performance. They found that the neural network, which could once solve mathematical problems with a 98% success rate, has started to degrade. Its performance over six months fell by 2.4%, and problems were also discovered in code generation and responses to sensitive questions. The researchers published their data on GitHub.
The main reason behind this phenomenon might be the restructuring of the chatbot to cater to a growing audience. ChatGPT now answers questions much faster, but at the expense of accuracy. Official representatives from OpenAI deny any intentional degradation in the AI chatbot’s performance.
The degradation of ChatGPT’s performance poses serious questions about the sustainability and reliability of artificial intelligence. If a leading AI like ChatGPT can degrade over time, what does it mean for other AI systems? The tech industry must address these concerns and find solutions to ensure the continued growth and development of AI technologies.