‘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: “throw your hands in the air and wave them like you just don’t care”
We saw this one before. I ran the same prompt nearly six months ago when DALL-E struggled to generate hands. At the time, just over 7 in 10 images showed hands with exactly five fingers. I also expected we’d see better results in a few months. Fig 1 shows what we got this week on the left and last time on the right.
DALL-E interprets the prompt differently. The pictures from half a year ago are quite literally hands in the air, waving like they just don’t care. But this week’s results reflect the emotion of being carefree. DALL-E’s capability to convert the sentiment of the phrase to an image has improved.
I had anticipated a technical improvement, with DALL-E producing the same output as before but with five fingers on every hand. I don’t see that: hands are still pretty bad. But the progress in visualising meaning is interesting and, arguably, more important. Five fingers can simply be requested in the prompt.
Now, let’s move on to the pass or fail grade.
Today’s verdict: Pass
DALL-E indeed produced better results than in January. The improvement is more sophisticated than I envisaged, and that gets a pass.
Next week in Our World in AI: the perfect family – part 2. I re-run Q1 2023’s most popular prompt to see what’s changed.