This is a Gamechanger! Create Your Own GPT!

Welcome to the “Illumined Insights” newsletter! Thank you so much for subscribing. This weekly newsletter touches on all things analytics and data science with a focus on areas such as data visualization, AI, and sports analytics.

I had planned to continue my series on the integration of ChatGPT and Excel, but OpenAI dropped a huge new feature. So this week, we’ll take a look at ChatGPT’s new “GPT Builder”. It’s not exaggeration to say that this is a real “gamechanger”.

Stephen Hill, Ph.D.

In this week’s edition of the newsletter we explore OpenAI’s new “GPT Builder”. This tool allows a user to create a customized implementation of ChatGPT that is tailored to a specific purpose or set of purposes. User-created GPTs will eventually be available in an OpenAI GPT Store. The details of the store remain to be seen.

To access GPT Builder, you will need to be ChatGPT Plus or Enterprise subscriber. I’ve said it before, but I’ll say it again, the Plus subscription is easily worth the $20/month fee.

Creating a GPT

To access the GPT Builder in ChatGPT we start at the ChatGPT homescreen. Click on your user name at bottom left:

Then click “My GPTs”:

We then find ourselves at a screen where we can create a GPT:

Clicking on “Create a GPT” brings us to a screen that will serve as our interface for the creation of our GPT.

So what can we build? Just about anything! Let’s say that we wanted to build a GPT that allows us to upload an image of a data visualization and then provides us suggestions for improving the visualization.

We start by describing what we want the GPT to do. First, I used the prompt:

GPT Builder spent a moment “thinking” and then returned the response:

This seems fine by me, so we’ll keep this name. GPT Builder then created a profile image for the GPT.

I thought that this was OK, so we moved on to the next phase of the GPT design. GPT Builder typically asks questions at this point to help refine the design of the GPT:

I responded:

GPT Builder responded:

I suggested that the GPT should be able to handle a wide range of visualization types. GPT Builder then asked:

I suggested that the GPT be friendly but professional and that is should keep its language accessible to a general audience. It would be OK if the GPT uses real-world examples, but it should focus on the task at hand.

At this point, the GPT is ready for testing. Before we do that, let’s take a look at the configuration of the GPT under the “Configure” option.

Here we see the title of the GPT, a brief description, and a set of instructions. The instructions are based on the interaction that we had with the GPT Builder. Several conversation starters are also provided (these can be modified if we wish). We also see that the GPT will use ChatGPT’s web browsing, DALL-E image generation, and code interpreter capabilities.

One of the strengths of the GPT Builder is its ability to use user-provided knowledge files. For example, we could upload a document on data visualization best practices and the GPT would incorporate this knowledge into its responses. We can also create actions that allow our GPT to interact with an external API.

For now, let’s test our GPT by uploading a poorly designed visualization and we’ll see what we get. A Google Image search for “bad chart” gives us a wide variety of visualizations that we choose from. I picked this terrible 3-D bar chart:

I uploaded the image to my new GPT and asked “Can you help improve this chart?”.

The GPT is able to use ChatGPT’s vision capabilities (via GPT-4V) to analyze the chart and provide feedback. After a few moments, we get this response (it’s long, so I’ll break it up into a few screenshots).

This is wildly impressive! If you have a keen eye, you will see one mistake in that the new chart has a legend with colors of blue, green, red, and purple, but the bars in the chart are varying shades of green. I was able to ask the GPT to correct the new chart so that the colors of the bars correspond to the colors of the fruit. There’s an awful lot of yellow, but the correct chart is shown below:

Let’s try another example. This time I’ll upload a hand-drawn pie chart. Pie chart themselves are problematic and we’ll add another layer of difficulty by having the pieces of the pie add up to more than 100%. Here it is:

I prompted GPT to help improve the chart. Here’s the response:

The GPT does a great job detecting the issues in the pie chart and producing a much improved version.

You know that I’m a ChatGPT enthusiast, but this tool continues to impress me.

If you are a ChatGPT Plus or Enterprise subscriber and would like to try this GPT out, here’s the link: https://chat.openai.com/g/g-UhmMiGnsM-dataviz-helper.

What GPT will you create?

Illumined Insights Book Recommendations

This week I recommend one of my personal favorite business/analytics books.

Feedback?

Did you enjoy this week’s newsletter? Do you have a topic, tool, or technique that you would like to see featured in a future edition? I’d love to hear from you!

Support the Newsletter?

Support this newsletter with a “coffee” (optional, but appreciated).

Start Your Own Newsletter?

This newsletter is created on and distributed via Beehiiv, the world’s best newsletter platform. Want to start your own newsletter? Click below to get started. Please note that this is an affiliate link. I may receive a small commission if you sign up for Beehiiv via this link.