(Credit: OpenAI)
OpenAI is ready to release a new ChatGPT interface, dubbed Canvas, that could help computer programmers or writers achieve more fluid workflows.
Canvas opens up a separate window, parallel to ChatGPT's regular chat window, that can be used as a workspace for writing or coding projects. Rather than having to re-do a prompt every time you want a different output, users of Canvas will be able to highlight the lines of text or code produced by ChatGPT and edit them directly. In addition, Canvas will include a menu of shortcuts that allows you to ask ChatGPT to adjust your writing length, debug your code, change the reading level, and even suggest appropriate emojis.
For coders, Canvas also has “Port to a language” shortcut that can translate your code into JavaScript, TypeScript, Python, Java, C++, or PHP.
Canvas will open automatically when ChatGPT detects a scenario where it might be helpful. You can also include “use canvas” in your prompt to open up the tool and put it to work on an existing project. Users can also restore previous versions of their work by using the back button.
The new tool, which could be good news for many programmers, comes as ChatGPT is starting to gain some approval from the wider academic community about its coding capabilities—at least in some respects. A study published in June found GPT-3.5 was able to solve easy coding problems in some of the most popular programming languages at least 89% of the time, though its success rate fell to just 40% in the case of hard problems.
One of the researchers noted that though ChatGPT showed reasonable levels of proficiency at correcting other people's code, it sometimes “fails to understand the meaning of questions."
Canvas will become available to GPT Plus and Teams users on Thursday, while Enterprise and Edu users will have to wait until next week.
The news comes after OpenAI rolled out GPT-o1, codenamed "Strawberry," aimed at researchers, scientists, and developers who need to solve coding and math problems. OpenAI claims the tool achieved comparable results to PhD students on many physics, chemistry, and biology tasks.