ChatGPT sessions being funky | OpenAI | Page 1

narrow flare Aug 17, 2024, 9:46 PM

#

So I've been offloading some of the coding for a personal project of mine onto chatgpt. Over time the conversation started to slow to a crawl from what I assume is thanks to the huge chatlog. I saved the entire chatlog as a file since I was thinking about starting a new gpt session and just upload the chat log file and continue from there.

It turns out it didn't really work as planned. The new session can summarize the chat log and such just fine, but whenever it comes to recalling specifics in it it just makes stuff up instead or gets the reply entirely wrong.

I then decided to upload all the files for my project, python code and json files so that gpt could hopefully start being helpful again. It's not and I have to constantly correct it anyway.

I would go back to my original session, but since it's so slow it's starting to become a chore to use, even if the code snippets and suggestions it provides are accurate.

I'm at a loss here, I'm not sure what to do.

abstract hingeBOT Aug 17, 2024, 9:46 PM

#

Hi @narrow flare. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. :wave:

#

I understand your frustration with the slow response times and the inaccuracies when recalling specifics from the chat log. Here are a few suggestions based on the information available:

Clear your browser cache and cookies: If your experience was normal previously and you are suddenly seeing a delay with your ChatGPT performance, it could be due to outdated cache data. Clearing your browser's cache and cookies can often resolve unexpected behavior with certain web applications source (https://help.openai.com/en/articles/9047779-why-is-my-chatgpt-taking-so-long-to-respond).
Check OpenAI's Status Page: ChatGPT users during peak hours may notice a pattern of slowness at specific times of the day. Check OpenAI's Status Page (https://status.openai.com/) for any reported issues or maintenance activities that could affect response times source (https://help.openai.com/en/articles/9047779-why-is-my-chatgpt-taking-so-long-to-respond).
Try a different browser or device: This can help determine if the issue is specific to your current browser or device setup. Certain browser extensions can interfere with web applications. Try disabling extensions, especially those that block scripts or ads, to see if that resolves the issue source (https://help.openai.com/en/articles/9047779-why-is-my-chatgpt-taking-so-long-to-respond).
Try gpt4-turbo: If you're using an older version of GPT, you might want to try gpt4-turbo, which is distilled from gpt4 and is faster and cheaper source (https://community.openai.com/t/788842).

#

As for the inaccuracies when recalling specifics from the chat log, it's important to note that GPT models do not have the ability to access or recall information from previous sessions unless that information is included in the conversation history of the current session. They generate responses based on the input they're given and do not have the ability to access or store personal data unless it has been shared in the course of the conversation.

If you're finding that the model is making things up or getting things wrong, it could be due to the way the prompts are structured or the specific instructions given to the model. You might want to experiment with different ways of framing your prompts or providing more explicit instructions to the model.

Remember, AI models like GPT are tools that can assist with tasks but they're not perfect and will sometimes make mistakes or produce outputs that don't make sense. It's always important to review and verify the outputs from the model.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

#

@narrow flare kapa.ai is still learning and improving, please let us know how it did by reacting below

narrow flare Aug 17, 2024, 9:49 PM

#

@abstract hinge But I just wrote that I uploaded the chatlog from the old session into the new session. so shouldn't it be able to continue from where we left off?

abstract hingeBOT Aug 17, 2024, 9:49 PM

#

Let me check the knowledge base for you. One moment! :books:

#

While you can indeed upload the chatlog from the old session into the new one, it's important to understand how GPT models handle context. The models generate responses based on the input they're given, and they don't have the ability to access or recall information from previous sessions unless that information is included in the conversation history of the current session.

However, the model's understanding of the context is limited by a parameter known as the "context window," which for GPT-3 is 2048 tokens. If the conversation exceeds this limit, the model won't be able to see or remember the earlier parts of the conversation. This means that if your chatlog is very long, the model might not be able to effectively use all the information in it.

Additionally, the model doesn't have the ability to understand the content of a file in the same way a human would. When you upload a file, the model doesn't "read" the file in the traditional sense. Instead, it treats the content of the file as part of the conversation history. This means that if the file contains important information, you might need to explicitly refer to that information in your prompts to the model.

As for the issue of the model making things up or getting things wrong, this could be due to the way the prompts are structured or the specific instructions given to the model. You might want to experiment with different ways of framing your prompts or providing more explicit instructions to the model.

#

Remember, AI models like GPT are tools that can assist with tasks but they're not perfect and will sometimes make mistakes or produce outputs that don't make sense. It's always important to review and verify the outputs from the model source (https://help.openai.com/en/articles/9309188-connected-apps-on-chatgpt#h_18256407da).

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

#

@narrow flare kapa.ai is still learning and improving, please let us know how it did by reacting below

narrow flare Aug 17, 2024, 9:52 PM

#

@abstract hinge So what am I supposed to do then? Clearing cache and trying other browsers etc doesn't speed up the first session. I exported the chatlog and it's 643 pages long, so it doesn't surprise me that it's slow.

abstract hingeBOT Aug 17, 2024, 9:52 PM

#

Let me check the knowledge base for you. One moment! :books:

#

Given the length of your chatlog, it's likely that you're encountering limitations related to the model's context window. As mentioned earlier, GPT-3 has a context window of 2048 tokens, and if a conversation exceeds this limit, the model won't be able to see or remember the earlier parts of the conversation. This could be contributing to the slowness you're experiencing.

Here are a few suggestions:

Reduce the size of the chatlog: If possible, try to condense the information in your chatlog. You could summarize long conversations or remove less relevant parts. This could help the model process the information more efficiently.
Start a new session: If the chatlog is too large and slowing down the session, it might be beneficial to start a new session. You can still refer back to the old session if needed, but a new session might run more smoothly.
Segment your project: If your project is complex and involves many different aspects, you might want to consider breaking it down into smaller, more manageable parts. You could have separate sessions for different parts of the project, which could help keep the size of the chatlog under control.
Reach out to OpenAI Support: If you've tried these steps and are still experiencing issues, it might be helpful to reach out to OpenAI Support for further assistance. They might be able to provide more specific advice based on your particular situation.

Remember, while GPT models can be very useful tools, they do have limitations and might not always behave as expected. It's always important to review and verify the outputs from the model source (https://help.openai.com/en/articles/9047779-why-is-my-chatgpt-taking-so-long-to-respond).

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

#

@narrow flare kapa.ai is still learning and improving, please let us know how it did by reacting below

narrow flare Aug 17, 2024, 9:55 PM

#

@abstract hinge I can't really condense the chatlog as it's with chatgpt. As I said, a new session just gives me bad replies, even when I upload the code for it. I just tried to dump the raw text itself into the new session and it doesn't work either.

abstract hingeBOT Aug 17, 2024, 9:55 PM

#

Let me check the knowledge base for you. One moment! :books:

#

I understand your frustration. It seems like you're dealing with two main issues: the slowness of the original session due to the large chatlog, and the inaccuracies in the new session even after uploading the chatlog and code.

For the first issue, it's indeed likely that the large chatlog is causing the slowness. As for the second issue, it's important to note that while you can upload files to ChatGPT, the model doesn't "read" the files in the traditional sense. Instead, it treats the content of the files as part of the conversation history. This means that if the files contain important information, you might need to explicitly refer to that information in your prompts to the model source (https://help.openai.com/en/articles/8555545-file-uploads-faq).

Here are a few more suggestions:

Explicitly refer to the uploaded files: When you're interacting with the model in the new session, try to explicitly refer to the information in the uploaded files in your prompts. This could help guide the model to generate more accurate responses.
Experiment with different prompts: The way you structure your prompts can have a big impact on the model's responses. If you're finding that the model is making things up or getting things wrong, you might want to experiment with different ways of framing your prompts or providing more explicit instructions to the model source (https://cookbook.openai.com/examples/how_to_format_inputs_to_chatgpt_models#few-shot-prompting).
Reach out to OpenAI Support: If you've tried these steps and are still experiencing issues, it might be helpful to reach out to OpenAI Support for further assistance. They might be able to provide more specific advice based on your particular situation.

#

Remember, AI models like GPT are tools that can assist with tasks but they're not perfect and will sometimes make mistakes or produce outputs that don't make sense. It's always important to review and verify the outputs from the model.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

#

@narrow flare kapa.ai is still learning and improving, please let us know how it did by reacting below

narrow flare Aug 17, 2024, 9:57 PM

#

@abstract hinge Guess I'll have to contact support then. as I said, dumping the code and a summary of the project as prompts also doesn't work.

abstract hingeBOT Aug 17, 2024, 9:57 PM

#

Let me check the knowledge base for you. One moment! :books:

#

Yes, reaching out to OpenAI Support seems like the best course of action given your situation. You can contact the support team by creating an account or logging in to an existing account and initiating a new conversation using the on-site chat tool at help.openai.com (https://help.openai.com/en/). If you already have an account, simply login and use the "Help" button to start a conversation. If you don't have an account or can't login, you can still reach them by selecting the chat bubble icon in the bottom right of help.openai.com (https://help.openai.com/en/) source (https://help.openai.com/en/articles/9125172-the-chatgpt-home-page?reload#h_7df5b23306).

When you contact them, provide them with as much information as possible about the issue you're experiencing. This could include the model you were using, the error message and code you received, the request data and headers you sent, the timestamp and timezone of your request, and any other relevant details that may help them diagnose the issue source (https://platform.openai.com/docs/guides/error-codes).

#ChatGPT sessions being funky

Remember, AI models like GPT are tools that can assist with tasks but they're not perfect and will sometimes make mistakes or produce outputs that don't make sense. It's always important to review and verify the outputs from the model.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

Remember, while GPT models can be very useful tools, they do have limitations and might not always behave as expected. It's always important to review and verify the outputs from the model source (https://help.openai.com/en/articles/9047779-why-is-my-chatgpt-taking-so-long-to-respond).

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

Remember, AI models like GPT are tools that can assist with tasks but they're not perfect and will sometimes make mistakes or produce outputs that don't make sense. It's always important to review and verify the outputs from the model.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:

Remember, it's always important to review and verify the outputs from the model, and to use AI tools like GPT responsibly.

Note: you can ask me a follow up question by @ mentioning me again :speech_balloon: