Here is the translated content:

OpenAI has recently released an application called CANVAS, which is optimized for applications for the first time. The main change of canvas is to switch from a dialogue-based interaction like ChatGPT to a document or code-centered exchange.

How does OpenAI introduce canvas and what work has OpenAI done? Canvas is based on the GPT 4o model, and currently, canvas functions are only available to Plus users and Teams users of ChatGPT.

What functions does canvas provide? You can directly edit text or code, a shortcut menu is provided to add emoticons, and there’s also a function to refine and modify the reading level of this document. It also has a history version management function.

To make this canvas function, what optimizations did OpenAI perform on the model? Canvas’s model is based on GPT 4o, which knows when to open a canvas, whether it’s for local modifications or rewriting the entire document or code.

In actual use, I think we can optimize it by preserving the previous code after editing the code, so users can compare. Only after the user agrees, replace the old code with the new one. This may be more in line with the habits of developers.

Translation

OpenAI最新推出的应用叫CANVAS,首次针对应用进行优化。canvas主要的一个最大的改变就是将我们以前与ChatGPT以一种对话为主的方式改变为以文档或者代码为中心的交流方式。 OpenAI是如何介绍canvas以及OpenAI都做了一些什么样的工作呢?canvas是基于GPT 4o模型的,目前canvas功能只针对ChatGPT的plus用户和teams用户开放。 canvas提供了什么样的功能呢?你可以直接编辑文本或者代码,还有一个快捷键的菜单,可以添加表情、最后润色修改这个文档的阅读水平等。它还有一个历史版本的管理功能。 为了制作这个canvas功能,OpenAI都对模型进行了什么样的优化呢?canvas的模型是基于GPT 4o的,这个模型知道什么时候需要打开一个canvas,是针对局部的地方进行修改还是针对整个文档或者代码进行重写。 在实际使用中,我觉得可以优化的地方就是在代码修改之后,可以保留以前的代码,方便用户来做一个对比。在用户点击同意之后,再将旧的代码给替换掉,这样可能更符合开发人员的习惯。 这个canvas功能还是非常实用非常棒的一个功能。 #### Reference: https://www.youtube.com/watch?v=AMyuvkwasY8
<
Previous Post
OpenAI DevDay 2024
>
Next Post
Nature: LLM scaling law