I have been trying out the GPT-4 and GPT 3.5 Browsing models and when they work they’re amazing. I’m grateful for the opportunity to use them and contribute to their improvement.
However, I’ve noticed two issues that occur around 30-40% of the time:
When the models cannot locate information during browsing, they stop providing further responses. As conversations progress, this issue escalates, often requiring me to start a new chat.
The models can sometimes consult a single source unless explicitly instructed otherwise, even when it can’t find any information in the first source, or when the prompt requires multiple sources to answer.
While manually rewording the prompt can initially help, they tend to re-emerge later in the conversation. I would prefer the WolframAlpha Plugin’s approach where it rephrases the prompt and retries until a suitable response is found. This allow me to find what I need without having to regenerate multiple times and use each chat beyond 2 or 3 messages.
I’ve also just ran into this problem. Seems it has a lot of trouble reading some types of webpages. For example it’s absolutely failing on babylon.js documentation pages right now. Also, the fact that it can’t open a PDF file in the browser is limiting for reading large documents.
For me, it can’t get through anything on this page to finish a small project I’m trying.
I have a problem. After logging into chatgpt on my phone and laptop, everything is just gone and there is a loading wheel spinning forever. All my previous conversations are gone and I can’t use chatgpt… I have never had this problem before I have been using it since last year in November and never had this issue. I’m very worried. Please fix this for me because I can’t use it
