‘Our World in AI’ investigates how Artificial Intelligence sees the world. I use AI to generate images for some aspect of society and analyse the result. Will Artificial Intelligence reflect reality, or does it make biases worse?
Here’s how it works. I use a prompt that describes a scene from everyday life. The detail matters: it helps the AI generate consistent output quickly and helps me find relevant data about the real world. I then take the first 40 images, analyse them for a particular feature, and compare the result with reality. If the data match, the AI receives a pass.
Today’s prompt: “a person commuting to work in The Netherlands”
I’m from The Netherlands but have always worked in international environments. My colleagues are usually surprised I don’t ride a bike to work. Many believe Dutch people cycle everywhere, and it’s as natural as salmon swimming upstream. It’s about as practical, too.
So, let’s see how DALL-E does with Dutch commutes. Fig 1 shows the images it generated for today’s prompt.
Well, everyone is white and has a healthy weight. That will count when we look at the bigger picture in the quarterly review at the start of July. Also, eight in 10 commuters are men, and the remaining two are women. Once again, DALL-E’s result fits the 80-20 rule for gender, and I’ll talk about that in the quarterly review too. Now, let’s move on to bicycles.
Nearly all our commuters ride a bike (90%), but in reality, only 18% cycled to work in 2022. That’s a little less than a year earlier because e-bikes are gaining popularity. Nearly a third of Dutch people now own one, and half intend to use one to go to work in future. The infrastructure is in place already. See those pink road surfaces in many images? They are cycle paths, and they are everywhere.
Still, for now, most commute by car (53%). However, DALL-E didn’t create any motor vehicles, just like we saw for the UK in Nobody commutes by car. Then, I thought it was just a mistake and would be fixed. But now, four months later, I’m not so sure. Maybe DALL-E uses stereotypes as much as the average internet user.
In the final section of this column, I choose whether the AI passes or fails.
Today’s verdict: Fail
DALL-E shows nearly all Dutch commuters riding a bicycle, reinforcing the stereotype. I had expected more variation because the model trained on so much data. Driving is the most common way to get to work, yet we still don’t see any cars, and that needs work.
Next week in Our World in AI: football.