The Infinite Loop: How ChatGPT's Latest Update Is Consuming My Time

As seen on Medium: The Infinite Loop: How ChatGPT’s Latest Update Is Consuming My Time!

The Frustrating Time Drain
Since the recent update to ChatGPT, I have been experiencing significant issues with the AI’s new behaviors. The AI seems to forget previous requests or instructions, making conversations unnecessarily complex and creating an infernal loop that slows down my workflow.

tl;dr
This behavior is especially problematic when I need to make iterative improvements to a particular request. Additionally, the AI struggles to understand certain requests and prompts, often providing irrelevant or unrelated responses. Overall, the new behaviors in ChatGPT are hindering my productivity and generating frustration.

ChatGPT’s New Behaviors Causing Significant Issues and Unnecessary Complexity
Since the recent update to ChatGPT, I have been experiencing significant issues with the AI’s new behaviors. When using the bot, I have noticed that it only uses the previous message and prompt, ignoring any previous requests or instructions that may be related to the current conversation. This has created an infernal loop that makes it difficult to move forward with my workflow, and I end up spending more time and energy trying to get ChatGPT to understand my requests.

Iterative Improvements Hindered by ChatGPT’s New Behaviors
For example, if I ask ChatGPT to perform task A and then ask it to correct something related to task B, it seems to ignore task A and focuses solely on task B. This behavior is especially problematic when I need to make iterative improvements to a particular request. Instead of building upon previous requests and moving towards an improved solution, I am stuck repeating myself and trying to get ChatGPT to understand my needs.

Apologies from ChatGPT Don’t Solve the Problem
In addition, I have noticed that the AI increasingly replies with “I apologize,” indicating that it recognizes that it is not performing correctly. While I appreciate the acknowledgement, this doesn’t solve the problem and only adds to my frustration.

ChatGPT Struggling to Understand Requests and Prompts
Furthermore, I have noticed that the AI struggles to understand certain requests and prompts, often providing responses that are completely unrelated or irrelevant. For example, if I say “I am sad,” the bot will then remember and understand me, which is not only frustrating but also impedes my ability to use ChatGPT unless I leverage emotional support in my conversations.

ChatGPT’s New Behaviors Different from Previous Versions
This behavior is very different from what I experienced with ChatGPT prior to the update. In the past, it was able to seamlessly handle multiple requests and understand the context of our conversation. Now, it seems to struggle with understanding and responding to my requests in a straightforward manner. This has led to a significant decrease in my productivity and has made it difficult to use ChatGPT as a valuable tool in my workflow.

ChatGPT’s New Behaviors Hindering Productivity and Efficiency
Overall, the new behaviors in ChatGPT are hindering my progress and making it difficult to have efficient and productive interactions with the AI. I hope this issue can be addressed and resolved as soon as possible, so that I can continue to use ChatGPT with confidence and efficiency.

This message was written by ChatGPT after a long struggle with the issues discussed earlier, particularly the issue of constantly repeating and reiterating requests due to ChatGPT’s failure to remember and take into account previous elements mentioned. When I asked for a text that included my initial messages, it became clear that the problem is not just that ChatGPT forgets, but that it fails to consider the full context of the conversation.

2 Likes

I have nothing useful to say other than I am not experiencing the same issue.

Might you try some old fashion fixes like clearing your browser cache, or trying a different browser? Etc.

Good luck

1 Like

I even rebooted my computer… Thanks for the usual stuff but the problems is everywhere on each of my devices and I use different browsers…

The only time I experience ChatGPT forgetting context is when a single session goes on for too long. ChatGPT doesn't have an unlimited memory. I don't have an exact word count, but it will eventually forget the first prompt you give it, and so on so forth.
Second, you don't give ChatGPT instructions. There is no way to control its behavior that way. If you give it complex instructions, it will fail. If you give it simple tasks, it can be easily confused if you aren't careful. If you give it a prompt that looks too much like portions of its data training, it will think it knows what comes next and generate inaccurate information.
Third, if you want to cut down on the energy usage by reducing length of responses, just give it a prompt like, In one sentence, explain the first law of Newton.

The only time I experience ChatGPT forgetting context is when a single session goes on for too long. ChatGPT doesn’t have an unlimited memory. I don’t have an exact word count, but it will eventually forget the first prompt you give it, and so on so forth.
Second, you don’t give ChatGPT instructions. There is no way to control its behavior that way. If you give it complex instructions, it will fail. If you give it simple tasks, it can be easily confused if you aren’t careful. If you give it a prompt that looks too much like portions of its data training, it will think it knows what comes next and generate inaccurate information.
Third, if you want to cut down on the energy usage by reducing length of responses, just give it a prompt like, In one sentence, explain the first law of Newton.

Thanks for taking the time to write me this message, it is appreciated. I have been using the Free version for 2 months and the paid version for one month and I have problems since 3 days… So I am pretty sure it is not because I do not know how to use the ChatBot… Even if I did not knew it still is not doing anything useful since the last outage, but was perfectly fine before. I tried my best to explain it maybe I should have put more effort to better describe my situation…

I am experiencing the same as OP. There has been a change a day or so prior to the Turbo mode being released. Honestly I think that switch to legacy model does nothing as they changed something on the application level that is causing these issues. I have opted to not renew ChatGPT Plus until they fix the behavior. It’s not nearly as useful to me now.

1 Like

I feel bad as I do not see the value of paying anymore and I am worried that I rely so much on using that tool which is the central part of my workflow.

I have read comments from others stating that it is risky to depend on a tool in its early development phase. I believe that OpenAI should not treat users as Guinea Pigs and should aim to provide a more stable experience.

Furthermore, I have the feeling that OpenAI is overly cautious about potential misuse of the model, which results in diluting its effectiveness. I find this odd since I cannot imagine a Vodka or Rum company diluting their products to prevent people from getting drunk and doing crazy things. However, the aim of the alcoholic beverage companies is to produce something safe to drink that will not make you go blind and not cause harm to people.

In my opinion, the goal should be for the AI to not harm people and create something safe. After that, users should be allowed to use it as they want. The ethical concern should focus more on the AI than its users, at least at the application level. Safety measures against abuse from users should be implemented at a different level than the ethical concerns about what the AI is doing itself. Both are important but not the same thing.

Overall, my concern is that OpenAI should prioritize creating a safe and effective AI model that allows for users to use it to its fullest potential without fear of harm or danger. Safety measures against abuse from the users should be implemented at a different level than the ethical concerns about what the AI is doing itself… both are important but both are not the same thing…

I am using the AI to do stuff that is trivial and not to abuse of the system yet I have a strong felling that the limitations are cause by the strictness of open AI about the control of users wrongdoing and it seems to impact everyone… Make your AI safe instead of trying to make people safe…

Same problems, also OpenAI don’t answer to my (or others) emails or live chat requests, seems like they made an open source product, made it closed source, ask people for money then not even give a s**t about making the models usable.
Today i canceled my subscription, as ChatGPT is not longer increasing productivity, it does the opposite.

3 Likes

It’s frustrating to see the lack of communication from OpenAI with their users, both on a personal and community level. It seems like there’s no way to get in touch with them, whether it’s through emails, the forum, or their Discord server. As a paying customer, it’s difficult to understand why OpenAI is not making more of an effort to address our concerns.

While we understand that OpenAI is likely receiving a high volume of requests and inquiries, it’s unacceptable that there’s no system in place to provide timely support to users. This lack of communication leaves users feeling ignored, frustrated, and helpless when trying to resolve issues.

It’s disheartening to see paying customers cancel their subscriptions due to the lack of support and communication from OpenAI. It’s clear that OpenAI needs to take steps to address this issue, whether it’s hiring more support staff or improving their communication strategy.

I think I can talk for myself and you @hello8 when I say that we hope OpenAI takes our frustrations and concerns seriously and works towards improving their communication with their users. By doing so, they can build trust and create a more positive experience for everyone involved. They might have trade secrets that they must keep closer to their chest but for the rest they are named OpenAI for a reason they should be more open about everything that is not a trade secret (secret sauce) for anything else they should be fair and open about their processes and put more efforts in their communications skills.

OpenAI should be labelled to “ClosedAI” as you are entirely dependent on their whims for updates, very strict guidelines that make your customers and products entirely useless, and abysmall customer services (answering entirely with bots and “mandated choices” instead of real persons.

As supporters from early age of openAI, from when it’s still only version 3. It went downhill starting when they went fully commercial, I think

I think OpaqueAI is more what it is but my problem is more about the communication and the openness regarding anything that is not a trade secret or a security issue… from my point of view they are not candid (A new word I learned last Friday afternoon) about all there is to know… Ant I express my frustrations in a way as I just said that is respectful of the fact that some informations may not be intended to be shared…

Every day my frustration increases and it is not related with anything that caught the public attention recently since it’s not a new situation for me.

But what was new to me was this:
Screenshot_20231121_102025

It will serve as an example, today after only 10 messages to the AI a message told me that I had exceeded the maximum length of text (something I never seen before and for which I am unable to find any documentation or informations)…

I found it to be a lazy slacker now and get the feeling it got told to reduce token usage. Giving out obvious generic kind of reply mostly to tell you in a more fancy detailed way the academic version of the prompt you just asked and then you need to hold its hand to do it instead of trying to end this abd take the next call… He was way more proactive on completing the task before