Comment: I’m a bird photographer. I’m not professional—just an amateur. I’m keen, though. The other week I stood in a mud puddle for almost an hour waiting for an owl to open its eyes so I could get a good shot of it. A friend pointed out to me that I could get a very similar image by asking Stable Diffusion XL to generate one. If he had been just a little crueller, he could have said the AI could generate a much better image—sharper, brighter, and clearer.
Illustrators around the world are upset about the use of AI to generate images used online, in advertising, publishing, and films. Of course, no one wants to lose their job to a computer (or to anything or anyone else). The problem, however, is not just about losing jobs. AI threatens our relationships to our work and to those very human activities—our hobbies.
We admire effort in others and want it for ourselves. We challenge ourselves to do things. Doing things the easy way can be unrewarding and boring. Even in our work, we want to challenge ourselves. Rarely is the product of work the only thing that we value of it.
Just setting machines to produce those things that we need and want will cause us to lose something very important about ourselves. That is not to say we shouldn’t use technology to improve lives for people who find things too much of a struggle. But removing struggle altogether will make life very boring indeed.
When someone does something good, we should praise them. We distinguish, however, between those actions that are merely good and those that are supererogatory. A supererogatory act is one that took extra effort. It is something that one was not obligated to do (at least not in the way that it was done), because it was difficult. In our commonsense morality, we think that supererogatory actions are better than ones that are easy to do. Effort is a good thing.
At work, I’m an academic. My colleagues and I are worried about students’ use of AI programs to write essays for them. What would be good, however, is not the complete elimination of these programs from students’ work, but rather an understanding of how to use them to produce better work and to make other aspects of their work better.
The internet has made finding academic research on a topic much easier and has allowed academics to access much more material. Sadly, this has largely eliminated the job of the research assistant, but it has allowed academics to spend more time on other aspects of doing research—the creative aspects in particular. This is what we have to learn to do with artificial intelligence software. We need to be able to use it as a tool but not allow it to become the real author of the work.
We admire a thing more if we think it was made by hand. If you visit a friend and are served some homemade bread, and it is nice, we are rightly more impressed than if we are given a loaf that tastes the same and is made in a factory. We value the handmade bread more. It was understandable that Karl Marx equated the value of a product with the labour time it took to make it. The labour theory of value led to a problematic formal economic theory, but we can see how it is motivated by our intuitions about the value of effort.
Automation and mass-production have made available to us things we could not afford otherwise. And it has made some jobs safer. AI promises to accelerate this process. But we need to take control of this evolution. One concern is the elimination of human jobs. Another has to do with the loss of an aspect of our own humanity.
To return to the case of my bird photography, I use AI-assisted programs to finish photographs. I use the latest version of Adobe Photoshop, which uses AI methods to sharpen images and reduce noise. It also has a wonderful tool for selecting the subject of the photograph, which allows one, for example, to darken or blur the background. Other photofinishing software now does very similar things. But one still has to stand in mud puddles and wait for an owl to open its eyes. We can see the human talent embodied in a good bird photograph over a mediocre one. And one can admire the effort taken to take a photograph.
What we need, when we are judging our own work or hobbies and those of others, are criteria for distinguishing when something is used as a tool and when it is in fact doing the work for us. Clearly, typing into an AI program a request for a photo-realistic picture of an owl, a painting-like picture done in the style of Vermeer, or to produce a book-length manuscript on the nature of logical entailment is going beyond the use of it as a mere tool. Finding useful criteria that tell us when we have allowed the machines to take over is going to be difficult.
What is crucial is that we have to be able to attribute the ideas incorporated in a piece of work to the human author of it. If we find a work in which one author takes the ideas of another and dresses them up as his own, we call that plagiarism and punish it. But what counts as an idea and what counts as mere presentation can be very difficult in certain cases to determine. Deliberations about this can be arduous, but as I have been arguing, difficulty by itself is not always a bad thing.
This article was originally published on Newsroom.
Edwin Mares is a philosophy professor at Te Herenga Waka—Victoria University of Wellington.