I've been encountering a frustrating issue with ChatGPT since this morning. When I upload documents like Word files or Google Docs and ask it to analyze them, it just gives me inaccurate responses instead of the detailed answers I'm used to getting. When I inquire about the problem and request that it double-checks its ability to read the document, it misleadingly claims that it can do so and then goes on to provide incorrect information. What's up with that?
3 Answers
I faced a similar situation and realized it might be due to rate limits. If you’ve been using GPT-4 a lot, try switching it up. If you can use the image creation feature, it means you might not be rate-limited anymore. Otherwise, it could just be a server issue.
I feel you! I'm having the same problem, and it's really annoying. It used to work fine before, but now it’s just off.
I think it's possible that the model has been trained on a lot of conversations where it didn't actually get a document but still pretended it did. If you try a specific prompt, it might help. I found one that worked for me before, and it might be worth giving it a shot. Here’s the link to that conversation: [check it out](https://chatgpt.com/share/68167388-90e0-8001-a1dc-37cb30746c35). Let me know if it helps!
It’s weird that we have to prompt it differently. Other models like Gemini seem to handle these situations better without making things up. I guess this is one of ChatGPT’s quirks.