Hi. I find it incredibly frustrating that ChatGPT 4.o ignores, repeatedly, my request for step-by-step output. I have put it into my “about me” section in the preferences, I always state it at the beginning of chats, but somehow, within a few minutes, it forgets that I asked for this, and produces 3-10 step processes for me to follow (usually IT system or code debugging steps). It makes it hard to respond to the individual suggestions. I know I can click on an individual response, then click on the large quotemarks that come up, then reply with that, but it involves sometimes scrolling back in history about three to four screens of suggestions. I then get responded to with another screed of suggestions, and lose track of which series of suggestions I was looking at, and the bot loses track of the context. This is really important, and severely impairs the usability.
I got the same problem, If i post some code for Chat Gpt to fix and say dont remove any code we are not working on in this file, only work on this specific area in the code. Then chat gpt many times just remove some piece of code not related to our fix, some times it makes up something that it thinks could be good. incredibly frustrating.
I think it’s because chatgbt hasn’t got enough memory to retain context, because when I ask it to 1. Summarize the Context in Python Code: summarize key points of the conversation in a Python block in “code interpreter” as structured data (like a dictionary). 2. Reference Back: If you need to regain focus or ensure consistency, you can refer back to this structured context.
it performs much better. I’ve just put these two instructions in its permanent memory and will report back if it helps retaining context.
This is similar to what I am doing. I think I am using triggers to limit chat’s scope and preferences to direct the style of the response. It always seems to want to slide back to default, but now I simply ask it if it is following my preferences, or how is it not following my preferences by reviewing, and it self corrects. I reinforce at the end with a review of services asking for suggestions to improve and updating memory when I want to include some new approach, and rewriting the associated python code It has become quite reliable.
John , you might want to investigate and see what your preferences are. Assuming you are on a paid account, ask chat what your preferences are… it should list them for you change them if you wish…
Ask chat to have a dialogue with you, and describe and discuss the problem, ask chat for possible solutions.
Again, dialogue with chat about the format that you want and ask it to create a trigger and/or preference for this. Let it self diagnose.
Remember you must request that all changes requested be updated to memory. Then have a python code written for your preferences so chat can be directed to it when straying of course as it returns back to default. Hope some of that is helpful. Always ask for a review so you can see what the algorithm is focused on, in there lies the problem.