Stanford researchers claim that ChatGPT’s accuracy and performance have decreased over the years
Researchers compared the performance of OpenAI GPT-3.5 with GPT-4.
The honeymoon phase of large language models (LLMs), which was introduced to gain traction in the generative AI field, seems to be over. According to a recent study conducted by researchers from Stanford and UC Berkeley the performance of OpenAI’s LLMs have decreased over time.
Researchers wanted to know if LLMs improved, since they can be updated using data, user feedback and design changes.
Robert Way/iStock.
According to a Stanford and UC Berkeley study, the performance OpenAI’s LLMs have decreased over time.
Source:
https://interestingengineering.com/science/chatgpts-performance-and-accuracy-has-decreased-over-time-stanford-research