ChatGPT may have so many bugs because there isn't a dedicated place in the forum to report bugs

ChatGPT may have so many bugs because there isn’t a dedicated place in the forum to report bugs.

The title kinda says it all. I’m a bit frustrated with OpenAI at the moment. grrrr…


There is a whole category for these issues:

I fully understand that bugs are annoying and I wish OpenAI would set different priorities but it appears that fixing bugs is not what most people want. Instead people want new features. So that’s what we get, mostly.

Not sure why, but that’s the situation.


Indeed, seeing this title, someone might try to switch it to a ChatGPT tag.

Then, topics tagged with ChatGPT often end up being ignored by everyone and eventually users are directed to go to, or they cause disputes among users and developers.

However, if OpenAI were providing adequate support, there would be no need to create such topics in the first place.

It’s clear that they are failing to communicate sufficiently with users and developers, which is quite disappointing.

Because it is LLM, many things that go wrong are not necessarily caused by GPT itself. Even developers make mistakes. Moreover, it is a model that provides services and has tens of millions of users. It also learns from human behavior. And there are mistakes from the behavior that most people have learned from.

[post removed by user because it’s not constructive. user empathizes with OP]

I can understand from the first message. You might think that’s all, but my meaning is much stronger than that. And I’m not just criticizing AI for its mistakes, I was comparing they to low-level companies that stupidly use automated messages to answer basic questions. Or even teaching GPTs to properly deny banned activity, they couldn’t imagine that responding like this would damage their own image. It was a stupid mistake, unlike your reasoning.

If you understand it carefully, you’ll find that GPT offers more freedom than its broad counterpart. In addition, OpenAI have to deal with the big-brained owners of the work who act like saw the team stuffing their books into GPT. So OpenAI has to be careful. Moreover, if you look closely, ChatGPT is frequently featured in bad news. Until I saw an abnormality and reported this problem.

They could start an infinite poll where people can click which one thy are experiencing or new ones to indicate importance and frequency


I think that’s a great idea! Even if they’re not bugs (i.e. features), they’d give people a place to vent their frustrations and an indication of how pertinent the issue is.

The cool thing is that polls can even include links where people can discuss the actual issues

0 voters

and then the thread itself could be used to discuss what items should be added to the poll… we’d just need to check if votes persist when the poll is edited.

edit: it doesn’t work. the poll is reset when edited :confused: I guess the disourse folks would need to implement it first

one for chatgpt and one for the API might be good :thinking:

for me focus in problam from team than GPT working performance. like limitation message. it has effect to gpt too. User not massage for say hello or good bye to GPT. or do some thing more than working . no more message for reward modeling .

Hey y’all

I’d just like to mention that OpenAI will pay you money to submit bugs over here:

You’re always more than welcome to submit a bug concerning ChatGPT, by making a topic here on the forum in the Bugs category.

The reason why you don’t see a lot of topics about these, is because they get unlisted from public view when they’re fixed :hugs:


I really appreciate the thought behind the polling idea mentioned by @Brecht.corbeel & @Diet, but this is unfortunately not as helpful as you may think.

It’s not a competition; ideally, we want all bugs to be fixed. Creating a poll will just make it harder for OpenAI to fix them as they provide much less information than when people write a post/topic about it.

Bug reports are like symptoms of underlying issues. When they are fixed, the topics about them get unlisted because new bug reports with the same symptoms will be caused by a different underlying issue. OpenAI won’t be able to discover new bugs if people keep reporting their symptoms on topics where the issue is already fixed.

If you could edit polls and keep the votes, then you could also just swap the choices to manipulate the results :wink:


You make a very good point! It may slow OpenAI down.


Thank you!

I think it’s important to note that this forum is not actively monitored for issues in the same way that bugcrowd or are. It’s much better to submit bugs and get account or billing related support there.

The main purpose of this forum is to give developers who are using the OpenAI API a place to find solutions to their problems, exchange ideas, share their experiences, and collaborate on projects. It serves as a community hub where members can learn from each other and stay informed about updates and best practices.

1 Like

I believe that would be called a lazy implementation :wink:

What I mean is a smart detailed one, that has ones that fade out or are less frequently upvoted fade down to the bottom, in a smart way, so you know what they bugs are that are becoming frequent.

Maybe indeed, something someone encounters as an exploit might not end up at the top of the list

Considering the unprecedented scope of OpenAI’s functionality and the speed at which it is evolving, it is impossible to cover everything.
So, it’s up to us to adapt or go to another platform which, if it is a cutting edge one, will have the same issues.
We are seeing history in making and need to gear up.

In terms of management, bughunter is not an excuse for refusing to support bugs on behalf of end users.

Although enduser problems have different origins. That doesn’t mean it’s not important. Providing a solution to this problem In addition to solving problems, we must also pass on knowledge and understanding in dealing with them, but OpenAI has many problems in this area. and also other management problems. In the end, the impact will come back, either directly or indirectly.

I’m an enduser, but I’d venture to say that none of the best people at OpenAI has solved every enduser problem. Enduser problems shouldn’t be overlooked. They’re not as simple as they seem.

A post was merged into an existing topic: GPT-4 problems (I don’t use GPT-4 though)