OpenAI has introduced a new feature called Canvas, designed to enhance the way users interact with ChatGPT, particularly for writing and coding projects. This feature is currently in early beta and is available to ChatGPT Plus and Team users, with plans to extend access to Enterprise and Edu users shortly, and eventually to all ChatGPT Free users once it exits beta.
Canvas provides a separate interface that allows users to collaborate more effectively with ChatGPT on complex projects. It offers a more interactive experience than the traditional chat interface, enabling users to edit text or code directly and receive inline feedback from ChatGPT. The feature includes various shortcuts for tasks such as adjusting writing length, debugging code, and changing reading levels. Canvas also supports version control, allowing users to revert to previous versions of their work.
The coding capabilities of Canvas are particularly notable. Users can benefit from features like:
- Code review
- Bug fixing
- Language translation (e.g., JavaScript, Python)
These functions aim to improve transparency and understanding of code revisions.
OpenAI has trained its GPT-4o model specifically for this collaborative environment. The model can determine when to open a canvas based on the context of the task, such as writing a blog post or handling complex coding tasks. The training process involved automated evaluations and human assessments to ensure high-quality outputs and precise feedback.
This feature has been previously discovered by @btibor91, however, at this moment it seems like the rollout hasn't been started yet in fact.
This development marks a significant update in ChatGPT's visual interface since its launch, reflecting OpenAI's commitment to making AI tools more practical and accessible for varied user needs.