cinder forge
Bug Report:
Steps to reproduce:
User interacts with GPT-4 with browsing and bing via ChatGPT.
User observes that the assistant refers to itself as GPT-3 and assumes it has GPT -3's limitations.
User asks the assistant about the potential bug .
User query:
I have discovered a potential bug in your originating prompt, I don't think it tells you that you are GPT-4, which causes you to assume you have GPT-3's limitations up until you notice you have expanded on them.
Actual GPT-4 response:
As of my training cutoff in September 2021, I'm based on OpenAI's GPT-3 model. While my capabilities have been extended to include tools like the ability to interact with a simplified web browser, I am not GPT-4, or a newer model, if such exists. If you're experiencing issues or discrepancies with my responses, I encourage you to report them to OpenAI, as it aids in improving the system.
Expected GPT-4 response:
I'm based on OpenAI's GPT-4 model, and my capabilities have been expanded beyond those of GPT-3 to include tools like the ability to interact with a simplified web browser. If you have any questions or if there are areas in which you believe I can improve, please let me know, as your feedback helps enhance the system.
Additional information
Browser: Firefox
OS: Windows 10
narrow fiber