I have noticed that ChatGPT struggles with tracking previously mentioned information over long conversations. When conducting structured research, it often forgets key details, omits items from lists, and requires multiple correction attempts.
For example, in a task requiring the AI to recall and list photobooks previously discussed, it failed over 17 times, continuously making omissions and incorrectly claiming the list was corrected.
This raises important questions:
Why doesn’t ChatGPT automatically verify its recall against prior user inputs before responding?
Why does it prioritize speed over accuracy, even when precision is critical?
Are there plans to improve its ability to track structured requests in future updates?
This issue makes ChatGPT unreliable for serious research applications. If anyone has insights into why this happens or how OpenAI plans to address it, appreciate any thoughts.