Completely broken right now

What the heck is going on with ChatGPT? I have been trying for hours to get it to just do one thing, and it will forget what we are talking about every single time it sends new info. I just asked it to write me a powershell script that opens a webpage. It started writing, but did not “CODE BLOCK” the text, so I stopped it, and asked it to try again but make sure to code block, and it apologized and then started writing python code for a fibonacci sequence. I have no idea! I’ve never asked for anything like that. I told it that it was wrong, and it was writing the wrong thing, and please correct it and write out my powershell script, and this is what it wrote to me:


I apologize for that mistake. Let me try to provide the code block properly for you. Here is an example of how to print “Hello, World!” in Python:

print(“Hello, World!”)

Is this what you were looking for?

Seems to work “OK” here, however, did not validate the ChatGPT reply (as required):

Well, I did state that it started writing it out. My problem is that it was inconsistent in the same conversation. It wrote the script out for the webpage thing, and then I asked it to write it out again and it started with some weird python script about the Fibonacci sequence. Then I asked it to stop writing that code into revert back to the code that it was writing me before for PowerShell and it printed out that weird hello world thing. That’s when I gave up and came here and wrote this post.

My hope was that somebody important would see that something’s wrong and they can go fix it. It seems to forget what we’re talking about in every single time it responds. It’s not remembering even the previous thing that it said. Sometimes it does, sometimes it doesn’t! Right now I’m having a massive issue where it just seems to almost never remember what I’m saying from one comment to the next.

I just went in and asked chatgpt to write me a powershell scripts that browses to a website and does a post request. I provided it with the post request necessary and it began writing out the code. It stopped half way through and I asked it to continue. It continued, and stopped again. I asked it to continue and it said that it was done, and that it was all the code. I explained to it that it didn’t even have ending brackets or anything needed to actually startt he program, and it assured me that the code was complete.

I then ran the code, and provided the errors and asked it to fix the code. It took that code, which was 46 lines long, and supposed to be about 60 lines long, and condensed it down to code that was nothing like what I gave it, and down to 15 lines. It took all of the functionality out of the program that it had just given me in the same conversation and still gave me something that wouldn’t work.

It does not seem to be working properly. Below is how the conversation just ended between the two of us before I came here to update.


ME: Are you not able to see that this code has nothing to do with anything that we have talked about so far? Can you please tell me the first thing I said in this conversation?

CHATGPT: The first thing you said in this conversation is: “How to create a program in PowerShell that opens a web page in a GUI window?”

*** This is not at all close to what I asked it to do. I didn’t even say “how to” anywhere.

ChatGPT is a text completion / prediction engine based on a sophisticated LLM and deep learning.

ChatGPT is not designed as an expert system to write “perfect code”. It is a language completion engine.

I write code with ChatGPT and the VSC CoPilot codex based on OpenAI every day. The code is like a “first draft” and is rarely production quality, especially for many lines of code.

It is in the terms and conditions of OpenAI that all output must be verified and is not production quality.

I am confused as to what you are saying. I have been using ChatGPT for a few weeks now and it has never worked this way. It doesn’t remember what I am saying from one statement to another. It forgets completely. If I ask it to write me a powershell script, and it breaks half way through, and I tell it to continue, and it continues with python code, I wouldn’t consider that working. I know it’s an LLM. It hasn’t worked this way for weeks.

For weeks now, it has at least remembered what it has said in the same conversation. It could remember what I said in the conversation. Now it forgets what I say right after it sends me a message sometimes. I asked ChatGPT 3 questions earlier and after that it had completely forgotten my first question and it had not even been 500 words between the two of us. This is not the way it worked before! Another time today, I asked it to revise the code that it just sent me, and it failed to remember what it sent me and started writing completely different code that had nothing to do with my code or what it had just sent me.

Not to mention that ChatGPT says on their own website “While ChatGPT is able to remember what the user has said earlier in the conversation, there is a limit to how much information it can retain. The model is able to reference up to approximately 3000 words (or 4000 tokens) from the current conversation - any information beyond that is not stored.”. My problem is not the coding. I sent it one word “continue” earlier, and it started sending my code that had nothing to do with the code that it had just sent me. It was completely different code for a completely different project. Since the word “continue” is not 3000 words, I wouldn’t imagine that I did anything to confuse it that much.