AI Enhanced Graph

Here is an interesting use of gen AI for data plots, which also serves as a warning.

I was using Python to investigate a dataset, stored in a pandas dataframe, and noted a complex N:N relationship between two of the columns.

To dive deeper, I plotted a network graph with networkx and matplotlib, and the results were... Well, let's call it "programmer art".

As well as being ugly, this visual gives no context for the data. It would need a key to explain what the red and blue nodes represent.

To make the visual more communicative and more appealing would probably take me longer than plotting it in the first place.

So instead I asked ChatGPT to do it, turning the red nodes into people and the blue nodes into office buildings.

Now, even without further explanation, it is possible to imagine what this graph might represent. It is also slightly more attractive and - most importantly - more engaging. And it took all of two minutes to create.

Warning!

Although AI's ability to transform a visual quickly is astonishing, it has a serious drawback: it makes mistakes.

If you pay close attention, two nearby nodes have merged into one and several edges have gone missing. If you needed your graph to be perfectly accurate, so that individual nodes and edges could be examined, it would not be suitable to use AI in this way.

Nevertheless, this process can be used in the right circumstances. I only needed to make a general point about the data's structure, and would not be looking at individual relationships in detail. In this case, it is better to have a clearer and more engaging visual than to have perfect accuracy, and AI is an efficient way to get there.