Greetings and salutations, folks.

Welcome back to another edition of FridAI. I’m your host, Mike King, CMO of AIPRM, and what I want to talk to you about today is the power of Code Interpreter.

What is Code Interpreter?

If you don’t know what Code Interpreter is, it’s a native ChatGPT plugin that allows you to upload and download files and basically perform a series of programming-driven tasks. It doesn’t require you to code; it does the coding for you based on what you ask it to do.

In this example, I’ve uploaded an image and I say, “Convert this into an MP4 video with a slow zoom and make it 10 seconds.” In the background, it writes a series of Python statements and executes on them until it completes the goal that you wanted it to do.

Sometimes it can run into problems. For example, it said my file was too big, it down-sampled it, and then was able to produce the video that you just saw.

Use Cases for Code Interpreter

There are a ton of use cases for it. You can do data analysis, data visualization, make animated gifs, do some simple video editing, file conversions, make QR codes (as if anyone needs to do that), allegedly it does math, and it can also do things like document QA. I’m sure there are other use cases that we haven’t come up with. But if you’re creative enough, you’ll come up with some cool things that you can have it do for you.

Enabling Code Interpreter

To enable the Code Interpreter, you’ve got to have a ChatGPT plus account. Go to your settings, click beta features, and then enable the Code Interpreter. You’re off to the races.

Exploratory Data Analysis

Here’s an example of me doing exploratory data analysis. I had a dataset in a CSV file that was literally just called “dataset”. I had no idea what was in there. So I said, “Hey, I don’t know what’s in this dataset. Make some sense of it and then visualize it for me.” It goes through, gives me some explanations of the different columns, and then asks if I want a distribution of search volume. I said, “Sure, sounds good.” And then it visualized it for me using one of the various Python packages for Matplotlib.

Document Review

In another example, I did a document review and asked, “Here’s a document I wrote. Can you run a grammar check on it and then export it as a PDF?” You can see it went through and asked some questions, ran into some issues with file formatting, figured it out, and then gave me the PDF to download.

Code Interpreter Limitations

However, there are several limitations with this toolset. It’s not an actual data analyst, so it can be very wrong and is also subject to the same incorrect answers that ChatGPT can generally give you. It’s Python-only, only half a gig maximum on the file size, limited to external Python packages it can use, can’t access the internet, connect to databases, has limited CPU power, and there are a few different time limits you might run into.

Despite that, it’s a very powerful tool and something you can get a lot of value out of if you don’t have someone to write the code for you. It can figure out based on what you’ve asked for.

That’s all I’ve got for this week, and as always, at AIPRM, there’s a prompt for that.

If you haven’t already downloaded our free Chrome extension, you definitely should. Otherwise, feel free to let me know in the comments what you want to hear about next time on FridAI.

Later!

Join in the conversation over at our Community Forum on AIPRM.