Choosing colors can seem like a burden when you just want to make something quick, which is why some people just use default color schemes. But the few extra minutes is worth it.

For The Pudding, Russell Samora and Michelle Pera-McGhee gave generative AI a serious try to see how much effort is required and the quality of the results. There was output. It wasn’t the best. However:

It’s sort of like comparing a woodworking artisan’s table to one from IKEA. The artisans invest immense time and effort into their high-quality pieces, while IKEA produces things quickly and cheaply, and most people probably can’t tell the difference (or don’t care). Which is kind of sad for us artisans. With AI, we can expect a rise in superficially appealing but low-quality content. But that doesn’t mean there’s no place for craftsmanship. We still find meaning in the bespoke, at doing all the little things right, and in creating things that feel like they have a real person at the other end of it. And we can only hope that others do too.

As they say, this the worst that generative AI will be. It’ll only get better and the chart-making will only get easier. But, analysis, narrative, and the finer details that fit perfectly with a dataset are a lot harder to replicate, because the process is more fluid.

Moving forward, the part of the process that makes data less cold and mechanical is where to focus our energy on.

To demonstrate the level of uncertainty in using climate forecasting models, Bloomberg compared two models that estimate the same type of flood risk in the same geographic area.

In the abstract, maybe the differences don’t seem like such a big deal. It’s a thought exercise. But when governments create policies and insurance companies set rates based on the estimates, the choices behind the forecasts weigh more heavily.

Noah Lyles won the men’s 100-meter, but he started as the slowest runner in the final. The New York Times superimposed shaded circles on the track, in combination with composite photos of the runners, to show how it happened.

The color of the circles indicate speed rank, as opposed to run speed. Everyone ran extremely fast, so showing actual speed probably wouldn’t do much good.

People have suggested that the Olympics should have regular people run, swim, and perform alongside the athletes to highlight the extremes. Seems like a good idea. Anything to set a visual baseline since fast against fast or strong against strong can make the extremes seem less extreme.

In 1992, when the Dream Team dominated basketball in the Olympics, the best players in the NBA were all from the United States. The league has grown more international since then. For Sportradar, Todd Whitehead shows the shift in where the best players come from and who they played for in the Olympics.

I’m pretty sure Steph Curry moved up a couple notches after his performance in the gold medal game.

As the 2024 Summer Olympics wrap up, medal trackers will fade from homepages for a couple years. You’ve probably seen a list or five by now where each row represents a country and four columns show the counts for gold, silver, bronze, and the country’s total. It’s a straightforward view that shows what most people are looking for.

However, I like the wrinkles that add a little something to the counts. The premise is that countries can rank differently based on criteria other than total medals, which highlights smaller countries or maybe just gives you a way to toot your own country’s horn.

Bloomberg’s table, shown above, lets you sort by each medal, but also per million population and $100 billion GDP. For example, Grenada, Dominica, and Saint Lucia bubble to the top when you consider their small populations.

Reuters sorts by total gold medals alongside stacked bars to show the relative distributions:

The Washington Post provides a few angles on the same page. There’s a table you can toggle to sort by total or gold medals only, which looks as you’d expect. I like the comparison against the Tokyo Olympics to see if there was an improvement this year. They call it over- and under-performance, but I think I’d just say better or worse than last time.

My favorite view is still this Josh Katz classic for NYT’s Upshot, which they’ve updated each Olympics since 2018. Apply the importance of each medal yourself and create your own rankings. I suspect the heatmaps might go over the head of a healthy proportion of readers, but I’m glad they bring it back.

Update: The Upshot also made a list with a dropdown to order countries 23 different ways, in case you can’t decide how to judge who won the Olympics.

Generative AI things are often unsettling, but the playful uses I’m into. Dries Depoorter created a selfie coach with OpenAI to process video and ElevenLabs to provide a voice. Demo below:

The code is on GitHub so that you too can be a selfie master.

This is the 300th weekly issue of The Process. Have things changed? Do we visualize data differently now? Where is this headed?

Dragons are pretty big, I guess. (Please let me know if you know who made this.)

Update: See the original by Siosin (thanks, Charlotte).

Michael Correll describes two kinds of bar charts in the world. The first kind shows counts where you can apply a visual metaphor of stacking things. The more you stack, the higher the bar gets. The second kind shows aggregates, such as mean and median.

Correll argues you should only use bar charts with stackable values. Otherwise, use something else.

This approach seems too extreme to me. Use bar charts where you see fit, which may or may not be to show aggregates. But the premise, which gets lost in bar chart minutiae, I can get behind, which is that bar charts are not always better and that you’re allowed to visualize complexity.

I’m hearing murmurs that the Venn diagram is making a comeback. Six Seconds made a pair-wise matrix to show the emotions that stem from combining the emotions from the Inside Out movie.

There’s a science to getting thousands of people to wait in line without things getting out of control. The Wall Street Journal spoke to crowd expert Brett Little to explain the different types of queues and how to strategically funnel people away from the venue when an even finishes.

The process of making fentanyl might not be as complicated as one would hope. However, you might be surprised to know that illegal producers aren’t always consistent and can be sloppy, which is why tens of thousands of people die from overdoses. For Reuters, Daisy Chung, Laura Gottesdiener, and Drazen Jorgic explain the process of getting ingredients, producing the drug, and shipping to the United States.

Bad charts and maps are common, but there’s usually a better response than pointing and laughing. Not all the time, but usually.

The Data Provenance Initiative audited 14,000 web domains to see how sites currently restrict scraping for the purpose of adding to AI datasets like C4, RefinedWeb, and Dolma. Sites are putting up more barriers:

Our longitudinal analyses show that in a single year (2023-2024) there has been a rapid crescendo of data restrictions from web sources, rendering ~5%+ of all tokens in C4, or 28%+ of the most actively maintained, critical sources in C4, fully restricted from use. For Terms of Service crawling restrictions, a full 45% of C4 is now restricted. If respected or enforced, these restrictions are rapidly biasing the diversity, freshness, and scaling laws for general-purpose AI systems. We hope to illustrate the emerging crisis in data consent, foreclosing much of the open web, not only for commercial AI, but non-commercial AI and academic purposes.

Bots used to be a welcome thing to see in your web analytics, because it meant that your site was indexed by a search engine. Real people could find your site. However, bots for the purpose of generative AI take everything and those who run sites don’t get much, if anything, in return. The decline in data availability seems warranted.

Aaron Koelker printed a six-foot long map on receipt printer earlier this year. He put it in a route sheet holder for more practical usage. Seems like a good end-of-world product, if you’re into that sort of thing.

Victoria Rose Richards uses embroidery to depict aerial views and landscapes. The above was inspired by a local creek:

I was once again directly inspired by my local landscape, copying the shape of the nearby estuary for this creek. Rather than an estuary though, I wanted to depict a receded river bed with dry mud and creeping plants starting to form across the base. Completed with grass, fresh crops and sprinkles of birds.

It’s one of two pieces still available in Richards’ shop.

Reuters, with illustrations by Catherine Tai, has visual guides to all of the Olympic sports. (Trampoline is tucked into the Gymnastics category in case you’re looking.) For each sport, there’s a schedule, a leading illustration, and a set of visuals that show you important moves or terminology.

It’s very good, especially for the sports you might not be familiar with.