OpenAI Day 4 - Canvas Update Highlights

I have previously introduced the basic functions of Canvas:

Today, there are three updates:

1. : Canvas is now available to all users and deeply integrated with the main ChatGPT model. No additional setup is required, making it more convenient to use.

   

2. : Canvas has added the ability to run Python code, allowing real-time viewing of code outputs, whether text or graphics, making everything clear at a glance.
3. : The Canvas function is now fully compatible with custom GPTs, enabling all customized models to leverage the powerful features of Canvas.

Demonstration of Canvas key features

Story creation

Through Canvas, you can create fun Christmas stories for children. For example, after selecting Canvas in the ChatGPT toolbar and entering "Help me write a funny elf Christmas story," ChatGPT will immediately generate a dual-column view: the left side is the traditional chat interface, and the right side is the Canvas editor, making document creation and modification intuitive.

Users can also directly edit the story title, content, adjust language style and details within Canvas, while using quick functions at the bottom to complete common tasks, such as:

• Adjusting the length of the story (e.g., turning it into a shorter bedtime story).
• Optimizing language reading level (suitable for children or more professional audiences).
• Adding final polish (checking spelling and grammar).
• Adding fun emojis.

Writing improvement and feedback

Canvas also provides detailed feedback on user-uploaded texts. For example, if a user uploads an article about "dark energy and reindeer speed" in physics, Canvas can annotate specific suggestions in the text and provide detailed improvement advice. Users can choose to automatically apply ChatGPT's modifications or manually adjust the document content.

Enhanced programming support

1. Code editing and debugging

• When a user pastes Python code into ChatGPT and selects to open Canvas, Canvas will automatically recognize the type of code and provide syntax highlighting and code auto-completion features.
• The function was discovered to be ineffective. Canvas not only pointed out the error but also provided correction suggestions, allowing users to quickly adjust.

• Canvas now supports running Python code directly. When running the code, a Python simulator is loaded via WebAssembly, supporting almost all Python libraries.
• When an error occurs, Canvas displays exception information and generates correction suggestions for the user. For example, when running the code, a color setting issue was discovered. ChatGPT provided a solution and instantly fixed the code.

• Canvas can generate graphical results, such as Sankey diagrams, to display complex data flows. Through collaboration with ChatGPT, you can quickly adjust the style of the graph and data parameters.

Canvas integration with custom GPT

1. Automatic selection of Canvas:Custom GPT can intelligently select the use of Canvas based on context. For example, a custom GPT designed specifically for Santa will directly generate reply documents in Canvas when handling Christmas wish letters, helping Santa quickly respond to letters.

2. Customized Canvas behavior:When creating or editing a custom GPT, you can set the default tool to Canvas through instructions, so that all input content is processed in the form of Canvas.Users can enable Canvas functionality through the configuration interface, making operations simple.