Q1: Are you worried that art-generating AI has the potential to replace human jobs?
A1: Sure. I think that’s true across most industries and professions. I have trouble thinking of a job that couldn’t potentially be done by AI and robotics. The pace of AI development will
determine how our jobs change more than anything. Over a long enough span of time, some
professional artists will be able to adapt and even use AI to achieve things we can’t yet
imagine. Entirely new forms of art and entertainment will likely emerge as the technology
evolves and new kinds of creative roles will appear. I can imagine artists learning to harness AI in ways that make us more efficient at certain tasks, particularly in the idea-generating phase of a project.
I imagine that many entry-level graphics jobs will disappear as advertising budgets get
stretched and things like website content, commercial mailers and magazine layout and other
low-budget content begin to rely more heavily on AI. I knew many people whose jobs in
typesetting, graphic design and model-making, for example, were made redundant by the
personal computer, laser printers and CG modeling. Where it winds up with respects to AI and
text-to-image/video is anyone’s guess right now, I think.
I wouldn’t let it stop me from pursuing a creative career, but I think it would be smart to
consider AI when making decisions about what kinds of roles to pursue and what paths to
follow. Creative careers have always come with built-in uncertainties. That part is still true.
Q2: In terms of artistic work, what are some things that humans can do that AI cannot?
A2: I can’t pretend to know what AI will be capable of in the future, but for the moment the
technology is very limited when it comes to replacing what human artists do. Currently, a major limitation with AI as an art tool lies in its language comprehension. I’ve spent a fair amount of time using DALL-E now and while I can get interesting imagery out of it, it’s very difficult to dial in exactly what I want with the text I input. If getting exactly what you want from an AI is difficult enough to do, then I suppose that may become its own skilled profession one day.
Another thing to consider are the cultural cues embedded in art and a human being’s ability to
tell when something doesn’t look, sound or feel quite right. Art at its best, IMO, engages its
observer in an experience that ultimately transforms them in some way. It might alter their
emotional state or alter their perspective in some way. It may even change the way they think
about something by challenging their preconceived notions about that thing. Art is a form of
communication that takes place between humans, on a human scale (at least, so far) through
the medium of the artwork. What kinds of ideas might be communicated to a human by an AI?
I can’t really say, but I do think that it would be very difficult, currently, for an AI to generate
something like a novel or a film in a way that most humans would find totally convincing. To do
so would require enough human input, I believe, that the completed work of art should still be
credited to the humans involved, with the AI serving as their tool.
But what might an AI artist tell us humans about ourselves? What will it be like to have an
outside opinion of us for the first time? Will we like what we learn about ourselves from these
AIs? I guess we’ll find out!
Q3: What differences do you notice between AI-generated art and human-created art?
A3: Since these AIs are designed to mimic imagery that human artists have created, I can’t say that I’ve seen an AI produce art that couldn’t have been created by a human artist just as well by other means. In most cases, I imagine the human artist would require significantly more time than the AI.
Another thing that I’ve noticed (I haven’t played with these tools too much, yet) is that the AI
doesn’t really prioritize the information in the imagery it creates. Beyond basic composition
guidelines for establishing a subject against some sort of background, the AI doesn’t appear to make decisions about what information is most relevant or important in the way most humans would, I think. To me it doesn’t feel like the image is being created with the intent to
communicate something to me. Most of the images feel more like a composed collection of
visual information. Occasionally, however, DALL-E has produced some images that have surprised and impressed me. An image of Rumplestiltskin sitting at a table eating a bowl of Froot Loops is one that comes to mind.
It’s the stuff of nightmares, to be sure, which is a quality that seems pretty common with AI
text-based imagery, to date. It’s also one of the things I find most interesting about AI art. The
way that DALL-E combines and incorporates bits and pieces of things that it’s “looked at”
before is very reminiscent to me of the stream-of-consciousness way that dream imagery
forms in the mind. This becomes more evident when one views the examples of AI-generated
video imagery, which is often very dream-like, IMO.
Another detail that I’ve noticed involves realism vs stylization. Imagine a spectrum that runs
from photorealism at one extreme to line drawings of stylized subjects (like cartoon characters) at the other extreme. The AI seems to do better the closer it gets to the photo-real side of the spectrum and becomes less successful as it moves towards the stylized end. For most human artists, the opposite is true.
It’s long been known amongst CG artists that things which fall in the middle of this spectrum,
where they are more or less an even mix of realism and stylization, cause most humans to
react to them unpleasantly, feeling that the imagery is “creepy” in the way some people find
clowns or baby dolls to be creepy. This middle area of the spectrum has been dubbed “the
uncanny valley” in the CGI profession.
Q4: Are you worried that your artistic style is being imitated by AI art generators?
A4: I’m not personally worried about that, no. I can’t think of any artist who’s work isn’t
informed and/or directly inspired by the work of artists who came before them or who are their
contemporaries, so I’m not sure that’s the best way to think about it. Almost all professional
artists use online images for reference. In most cases, what the AIs are doing seems very
similar in that regard. The major difference is that the AI isn’t a human artist. It’s proprietary
software owned by a commercial interest who intends to profit from this in multiple ways,
including by taking over jobs currently being performed by humans.
In regards to cases where the resemblance is undeniable, such as a recognizable piece of a
stock photo appearing in an illustration created by an AI, it reminds me of how copyright law
was discussed in the late 80’s with regards to pop and hip-hop artists who were incorporating
samples of other songs into their own tracks. I believe it was decided that, as long as the
sample was less than 11 seconds, it was fair use. Maybe AI art will be litigated in a similar way, in which a certain amount of a copyrighted image is fair use. It will take better legal minds than mine to figure it out.
It’s certainly going to be interesting, though!