OpenAI is rolling out its latest feature called Canvas, which is a new way for users to interact with ChatGPT for writing and coding tasks. Canvas can take advantage of OpenAI’s most advanced model, GPT4-o, which the company claims is capable of reasoning “across audio, vision, and text in real time.”
There are some useful shortcuts available in Canvas for programmers looking to increase their productivity. While using Canvas, users can have ChatGPT review their code and get suggestions for improvements, quickly add print statements for debugging, or even port code from one language to another. Supported languages include JavaScript, TypeScript, Python, Java, C++, or PHP.
For writers, OpenAI has also put in a lot of work into training GPT4-o so that it becomes a better collaboration tool. New behaviors the company has added include “triggering canvas, providing inline critiques, generating diverse content types, making targeted edits and rewriting documents.”
When a user launches Canvas, it opens a new window that places the conversation with the chosen model on the left-hand side, while the content the user is working on gets its own space on the right. This is a much more usable workflow when using ChatGPT as an aid for writing or coding. It ensures that whatever a user is working on doesn’t get lost amidst the text generated when conversing with the AI.
Users should temper their expectations, though. The model still needs the user to select specific text for edits, otherwise it will just rewrite everything. Meanwhile, as measured by OpenAI, the quality of the comments in the inline critiques still require more iterations. These shortcomings are why this is initially rollling out as a beta.
Canvas is now available for those who are subscribed to ChatGPT Plus and Team, while Enterprise and Education users will get access next week. Once it exits beta, OpenAI says it will become available to free users, too.