ChatGPT Code Interpreter: A Deeper Dive

ChatGPT Code Interpreter: 3rd Party ChatGPT Plugins are arriving

In a nutshell, OpenAI is currently experimenting with the integration of code execution and third-party plugins in ChatGPT.

The organization has revealed plans to develop plugins for ChatGPT to enhance its functionality. These plugins encompass a variety of third-party tools, including WolframAlpha and OpenTable, as well as our browsing plugin and Code Interpreter.

The latter can generate code, execute it, manage file uploads and downloads (ranging from CSV data to images), and assess the output, all within the ChatGPT interface.

At present, the Code Interpreter supports Python and a limited set of libraries. I’ve had the opportunity to experiment with the Code Interpreter, and it’s been fascinating to see its capabilities, even with basic libraries.

ChatGPT Code Interpreter Use Cases

The Code Interpreter not only generates code but can also analyze the output and utilize it in another function. This allows for the chaining of different code sections, where the output of one section can be used as input for another.

For instance, the gif shown above was created by using the Code Interpreter to generate a maze, transform the maze into blocks, find the exit using an algorithm, stylize it to resemble popular games, and finally generate a gif.

In the past, using ChatGPT to create code required moving the output to a different environment for testing. Now, a significant amount of development can be done within ChatGPT without leaving the user interface.

Coupled with its ability to analyze data, such as spreadsheets, and perform functions like graphing, ChatGPT becomes a potent tool for both coding and research.

Here are some examples of the random experiments I conducted with the Code Interpreter plugin in ChatGPT:

Music Creation with ChatGPT Code Interpreter

Upon realizing that ChatGPT can understand a lot of mathematical data and the Code Interpreter can generate downloadable files, I asked ChatGPT to generate a Shepard tone (a sound illusion that seems like it’s constantly ascending in pitch). ChatGPT visualized the tone in a graph and then output it as a wav file using the Code Interpreter.

Image Creation with ChatGPT Code Interpreter

In the blog post on ChatGPT plugins, we demonstrated some cool uses of Python libraries for image manipulation. This piqued my curiosity about what else could be done with basic libraries.

Generative Animation with ChatGPT Code Interpreter: Possible?

The ability of the Code Interpreter to generate data and output it in various forms, including animated gifs, made me wonder about its limits. I started by asking it to create a bouncing block, then a snowstorm, and finally a session of the Game of Life.

Machine Learning with ChatGPT Code Interpreter: Baby Steps

While ChatGPT can’t load any machine learning libraries into the Code Interpreter (currently), we can do some basic statistical text generation using n-grams.

What else I can do with ChatGPT Code Interpreter?

While there have been some interesting examples of GPT-4 generating SVGs, I was curious to see what ChatGPT and the Code Interpreter could do with more than simple block images. I asked it to draw a visualization, and it created something that resembled a visualization.

Conclusion

These are just a few of my explorations with ChatGPT and the Code Interpreter plugin. As more people gain access, I’m confident we’re going to witness some truly astounding applications.

In essence, the explorations with ChatGPT and the Code Interpreter plugin have opened up a world of possibilities. As more people gain access to these tools, we can expect to see a wide range of innovative applications that could revolutionize various fields, from music and art to education and research.