DALL-E Visualizes Gender Divergence

Alice Evans is one of the most interesting scholars speaking to a broad audience today. She has a forthcoming book on the global history of gender inequality, a topic that she explores at length on her podcast and in her newsletter. She’s been on a bit of a media blitz over the past few days, including an appearance on CBS Primetime with John Dickerson, after a recent post on gendered ideological polarization went viral.

Alice is scheduled to visit us during the week of February 26 as a Weiss International Fellow, and will deliver a public lecture, visit classes, have lunches with students and faculty, and participate in a conference being put together by my colleague Anja Benshaul-Tolonen. If you’d like to attend the public lecture you can register using the QR code on this flyer (but please email me or Anja first if you don’t have a Barnard or Columbia affiliation).

We decided to use a fairly conventional poster for the public lecture in the end, but along the way I wanted to see what DALL-E might come up with. The title of the talk is The Great Gender Divergence, by which I think Alice is referring to the fact that progress towards gender equality has been very uneven worldwide. There have been rapid advances in some regions and cultures, but relatively little change or retrogression in others, and she is interested in understanding the determinants of this variation.

So I asked DALL-E to capture this idea, by depicting a path that splits into segments with different slopes, with some people climbing and others descending. The initial results were artistically mediocre and failed to properly capture the idea, but after a few tweaks to the prompt I ended up with this:

Next, I tried to get an image that expressed the same idea using spiral staircases. Again, initial results were disappointing but the algorithm eventually produced something quite interesting:

I shared these images with Alice, who found them impressive and described them as “credibly imperfect.”

This exercise brought a number of questions to mind. Do these images carry any signature that would identify them to experts as being artificially generated? If they were reproduced using oil-on-canvas and displayed among authentic works in a gallery or exhibition would anyone notice the difference? Would people bid on them at auction and gladly display them in their home if they were considered authentic? If so, who would be entitled to the proceeds?

We know that there exist books for sale, especially cookbooks and self-help guides, that contain artificially generated content and have fictitious authors. In fact, as Conspirador Norteño has discovered, “once you find one author with an AI-generated face and potential AI-generated books, Amazon will helpfully direct you to more.” Presumably it is only a matter of time before artificially generated novels are on (real and virtual) bookshelves, and some of these may even be good enough to win literary prizes.

Which leads to yet another question—is the experience of reading a book, or gazing on a painting, or listening to a song affected by whether one believes it to be a human creation in the conventional sense? I suspect that it does, but even that might change as the lines between the artificial and the real become increasingly blurred.