Random collection of AI art generated images by the author. Created with Stable Diffusion.
Explosive New Creative Tool or Another Step Towards Robot Domination?
With AI text to image art all over the news right now – it’s threatening to bust into the mainstream – I’ve been asked frequently about my feelings towards this subject.
Caveat, while I've been playing with AI art quite a bit, I'm no expert. I'm just a graphic designer playing with some fun tech to see how it could effect my creative output.
This post was written in January 2023, so it should be fun to come back to this down the road and see how it aged. That should be interesting.
Before we dive into this, let’s make sure we’re all on the same page with a few basic definitions.
AI Art Basic Definitions
AI Generative Art – process of using AI as a creative tool by setting up specific rules through which machines analyze data and generate new content based on those rules. This process is known as "machine learning." For example, an algorithm can be trained on a dataset of images in a specific style or aesthetic, and then generate new images that conform to that style.
Huh? In normal people words, here’s how I’d explain AI Art to my kids.
Machines are fed huge sets of images to learn – this trains them to understand concepts like dog, house, spaceship, human, etc. so the AI knows what those subjects are and what they look like.
After they’re fed these huge sets of images by smart people somewhere, we get to reap the benefit. Using written direction, prompts, we can have them then generate completely original images.
Several examples of popular AI generative art platforms you may be hearing about are Dalle 2, MidJourney, and Stable Diffusion. But there are many others and more coming.
Prompts – a prompt in AI generative art refers to an input given to a generative model that serves as a starting point or inspiration for the model to generate new artwork. The prompt can take various forms, such as a text description, an image, or a combination of both. The prompt guides the model's creative process, influencing the final output of the artwork.
Prompting is key to getting the results you’re looking for. Prompts are not easy. Understanding the order of words, importance of the words in the prompt, and weighting particular elements of the prompt all factor into the image result. Prompt engineering really is an art form, or a science might be the more appropriate term.
Text-to-image AI Art – refers to the use of artificial intelligence to generate images based on written descriptions or text inputs called prompts. This is different than image-to-image AI art, in which the starting input isn’t a written prompt but a picture that the AI iterates versions of.
My Current Take on AI Generative Art
Alright now that we have that out of the way, here’s where I'm at on all of this.
Simply put, creatives do not be afraid. This is a new tool much like the invention of the paint brush or Photoshop…it will accelerate your ideation and creation BUT not replace your job. Not yet at least.
But could it? Could AI eventually replace human creative and design jobs? I’m sure it may already be cutting into some roles, but it’s also going to create a bunch of new jobs/careers that we don’t even know are jobs yet. It’s like any other new tech, some jobs will become obsolete while a plethora of new ones will be created instantaneously. I think it’s exciting not scary.
Space monkey, because you know, monkeys. Created with Stable Diffusion.
It’s easy to fall into a trap when seeing all this crazy cool AI art and get the impression it’s very easy to create. And it is to a certain extent. It’s great for an initial brainstorm or for directional use. For something like advertising storyboards early on when concepting, this could be absolutely game changing in terms of speed and efficiency.
Anyone who has played around with this tech knows it’s not as easy as it appears. Images aren’t usually client or public ready. They generally are really rough and incomplete a lot of times. For example, hands on humans are particularly difficult for AI to generate at the moment, usually resulting in mangled hands or incorrect number of fingers.
Exploration of cats using different styles. Created with Stable Diffusion.
The fact is, AI still needs a human to art direct it. Plus it needs the time dedication to iterate and refine to get the polished result you’d need for a practical business use. Those beautiful AI images are the results of a human relentlessly trying different prompts and settings over and over and over and over again to get the result they are looking for.
Not only is AI art currently time consuming and not as easy as it appears, but integrating this technology into a creative’s workflow isn’t quite there yet. You’re seeing companies like Adobe and the stock image houses beginning to integrate AI generative art and plugins into their software but it’s not there yet.
Once all the dots are connected for a creative to seamlessly use this technology within the software they are already designing and creating in, boom, that’s when the flood gates will totally open on this front.
There is one more side of AI art that I’ve purposely avoided for this blog, copyright infringement and the ethics associated with this new technology. How transparent should creatives be about who/how something was created?
Since AI can mimic real artists styles, what should the royalties be or credit given towards the original artist whose style is being copied? I’ll leave this topic for smarter people to figure out. But just know it’s a big deal we weren’t prepared for legally with this technology explosion.
Portrait of the author in the style of Vincent Van Gogh. I don't really have a beard but this makes me look cooler. Created with Stable Diffusion.
In Conclusion
I love new technology and have gotten really excited about the future of AI generative art. But I can see how it may not be for everyone and most certainly could be threatening towards hard working fine artists and illustrators who fear their craft is being commoditized.
I'd encourage every creative, whether they are excited to use these new tools or not, to at the very least, take the time to understand them and how they work as the reality is AI art is here and it's not going to go away.
Additional AI Generative Art Reading:
A few helpful guides to understanding how prompts work on Dalle 2 and Stable Diffusion.
Check out this link, Stelfie the Time Traveler, if you want to be blown away. This guy is creating a series of AI selfies putting himself into different points in time. This link is an example of seeing something and immediately thinking all of this is easy to do. Spoiler alert, it isn’t. At least not at the time of this writing.
留言