After the latest GPT-4 update, it no longer correctly understands the term “canvas” in the context of editing documents and code. When I attempt to use this term or ask the model to write in the canvas, it responds by saying it doesn’t understand what I mean. This applies to both document editing and code placement within the canvas.
Additionally, the same issue occurs with the internet search functionality. The model no longer recognizes requests to search the internet, a task it previously handled without issue.
Steps to reproduce:
- Ask the model to work or edit in the canvas.
- The model responds with an error or says it doesn’t understand the request.
- Attempt to ask for an internet search.
- The model fails to perform the search or states it doesn’t understand how to do it.
Expected behavior: The model should correctly interpret requests involving the canvas (for editing) and perform internet searches.
Actual behavior: The model no longer understands the term “canvas” in the editing context and fails to perform internet searches.