Exploring ChatGPT Code Interpreter

In the second week of July, ChatGPT released Code Interpreter to paid users in beta.

So what does it do?

Amongst others, it can:

• Be a calculator (even graphing or financial)

• Analyze data and create charts

• Image to text (OCR)

• Create a video panning an image 

• Basic image edits

• Make QR codes

So I put it to the test. While there is a lot of potential out there, there are still a lot of bugs, and it definitely requires a ton of patience while prompting. But I think the results are pretty damn cool and it shows the next step for generative AI tools.

Word of warning. You need to have patience with this. It's cool because it walks you through its thought process, but it takes a bit longer and requires iterations. I didn't get into the coding capabilities.

It's a Calculator on steroids

Remember those complicated math equations that you had to do by hand or use a graphing calculator or a financial calculator?

Well no longer...

Data analysis and chart-making

I gave ChatGPT a file with the world population by country for every decade until 2022.

Then I started with the easier questions:

Prompt: "make me a graph showing the 10 countries with the highest change from '2000 population' to '2022 population'"

Prompt: "can you make it a more interesting design?'"

Wonder why it did what it did? It told me :) 

OK ChatGPT, now let's do a little more depth:

Prompt: Review this document. Exclude countries with population sizes in 2022 that are in the lowest 10 percentiles.

What insights can you share with me about:

• population growth by country and continent since 2000 should be measured proportionately. Which countries had the highest proportional growth?

• What was the proportional growth by continent?

• Which countries had the largest decay?

• a table of the 10 countries with the highest proportionate growth each decade from 1980 onwards

Wonder why the bottom right graph is empty? ChatGPT explains in #4.

Image to text (OCR)

Yup, take that image and turn it to text. It doesn't have the most amazing OCR out there, but if your image is decent quality it should do a good job. I uploaded an image of this blog's initial intro

Create a video panning an image 

When I make PPTs, I sometimes want to show images that are too wide or too tall for the screen. I can just get ChatGPT to create a scrolling video. At first ChatGPT did not provide the video in the requested ratio, but with some patience I got 6 seconds of pure magic.

Prompt: This image is a panoramic shot.

Help me turn it into a video with aspect ratio X:Y with the image filling the entire video (so the sides are cut off). The video should start centered in the middle of the image.

Then, pan the video smoothly with no sudden jumps as follows:

Start: Center--> Left --> Right --> Center: End

Use the imageio library to help you. Save the frames directly to a video file instead of a list.

Use a frame step of Z pixels. If you need to, crop the image edges so that the size of the image is divisible by the frame step.

Basic image edits

Hey ChatGPT, make the following variations of this image:

1. Blurred

2. Lighten

3. Make in greyscale

The original is the top left

QR Code

Prompt: create a QR code in purple that leads to https://www.penguinstrategies.com/ 

Looking ahead

Well, if you got this far, you have a longer attention span than I do. Hope you enjoyed the read! 

Here is some more information about  generative AI