AI image generation for Pro-Life images
I thought I'd write a post on some experimentations I've made with Open AI's DALL E2. For those that don't know, it's an AI image generator. It works by taking some textual prompts from the user, before fashioning what seems to be a unique image every time. Not only can you specify what you want it to draw, but you can also specify the style you want it to be drawn in. Now, it seems to have some degree of poetic (or not so poetic) license in how it interprets your prompts. No matter how specific you make it, it will still deviate significantly at times from your instructions. The result of this, though, is some large variation in output.
So, I started first by looking to generate pro life images by generating images of women and the zygote. I won't go into specifics on the prompts I used (due to another project I’m working on), but I got quite specific. Here is the first keeper that was generated.
I was surprised by the quality of this. Compositionally, it is not perfect, as it almost seems to be cut from a larger picture. Perhaps the software isn't aware of the conventions for which features should be in the frame. But it seems to be of reasonable quality, and is certainly better than anything I can draw.
An even more interesting image is included below. As we can see, it isn’t a very detailed image of a woman, but it is an image that an artist could perhaps choose to take forward further themselves.
I then moved on to images of the baby. This generated a number of interesting, if not wholly bodily accurate images. I’ve included one below.
I would classify the images as interesting, although I don’t really connect with them. Perhaps connection is related to intention and workload.
Interestingly, the Catholic Pontifical for Life is currently conducting a review into "emerging technologies", and Pope Francis has said that we must, "ensure that scientific and technological growth is reconciled more and more with a parallel development … in responsibility, values, and conscience." When looking at AI image generators, we must consider the potential, as with other AI's, for job displacement. At the moment, with the low resolution images these AI’s are outputting, the artist's job is safe. But what about the future? Certainly, there will presumably always be a market for physical art created by human hands. But for those specialising in digital art, there may be problems on the horizon in terms of competition, as the creation of art passes into the hands of the masses, as digital cameras, and particularly phones, did for photography. The result will be a more diverse art space, but perhaps a less lucrative one.
However, we must also consider the value that a non-generated piece of art possesses, namely that of intention. There is no outpouring of soul into an AI generated image. Even with the most specific of prompts, the pictures generated are those seen by the computer, not the human. And so physical art, and perhaps art which travels on a medium, such as film photography, will keep a degree of value as being that created by a human (although there is no way to know whether an AI image has been used as a starting point for the picture). Certainly, when I look at AI generated images, it's hard to find a connection, knowing that I am not peering into a human's soul, despite it being an art piece. They are more like Frankenstein images.
With this in mind, it's hard not to wish that AI image generators are banned. What we seem to place value on is human imagination, but with the advent of this technology, we can never be sure whether the image idea execution came from a human, or from a machine. It is important to note, though, that humans have often borrowed or reimagined art ideas from other humans in the past, which means that intentionality is never truly what we imagine it to be. It's just that now, humans can borrow from another source too when creating their physical art.
A further problem comes from the AI generator in that you can ask it to copy a style, or produce variations on an image input. This is perhaps more problematic for the artist, as a style is like a signature for the artist. Often, it defines who they are. Problems arising from this could include "style dilution", where the artist's style becomes saturated into the market before they have reaped any kind of financial reward from it. One way to combat this would be to keep art secret until a body of work is complete, but this also diminishes the artist's earning potential. In a market where any style can be replicated, how do you keep the market hungry for originals?
Despite calls for AI research to be halted or delayed it's unlikely that this technology will remain out of reach forever. Perhaps then, we need to brace ourselves for this "Artificial New World".
So, can this image generator be used to benefit the prolife movement? I think yes. There is relatively little pro-life art, as far as I am aware, and part of the difficulty is in the difficulty of creating art in general. Unfortunately, people are not drawn to highly divisive topics for fear of backlash. In this sense, AI provides an avenue for those that perhaps do not have the time or skill to create art in the traditional sense. Problems, however, are the disconnect people may have with the art generated. Is it going to be able to move people to changing their minds? Is the software good enough to even depict human emotion accurately? On the latter point, I would say no. The software seems to have difficulty depicting highly realistic human emotions. It also seems to have a propensity for inaccurate bodily depictions, particularly the face, which may conjure negative perspectives for people.
Still, others may have better luck than me in dealing with this software, so I have created a Facebook Group for AI Pro-life images, if anyone is interested in contributing. It's called ProLife AI