This blog post is one I have wrangled with for some time. I created the actual word doc months ago. The topic of artificial intelligence, or AI as it’s more commonly known, is everywhere and it is evolving day by day. Just writing this article felt like trying to hit a moving target. AI is divisive and controversial, and the only thing we can be sure of is that it’s not going away any time soon.
So, what is AI exactly? I asked ChatGPT himself (herself? Itself?) and this was the reply:
Simple enough, right? At a glance, it seems that AI should overall be a positive thing. A computer system that helps us and makes our lives easier? Sign me up! But as I’m sure you’re already aware, it’s much more complicated than that. But why is AI a source of controversy? (Outside of the environmental concerns.) And how does that affect us? Here are my thoughts:
AI: Is it Ethical?
Why do people hate Artifical Intelligence, and particularly, AI-generated images? Essentially, AI does not ‘create’, rather, it ‘generates’. In the case of AI art, it copies from art that can be found online. The extent to which AI generated images copy art already on the internet can be seen in the fact that AI art has been found with logos and/or artist signatures of the art it ‘borrows’ from. This has raised copyright concerns, not to mention the questions of whether it’s even ethical to use AI at all. All you have to do is type ‘I want a chibi with a big bow in her hair’ and an AI will mishmosh a bunch of art available on the internet to create an image which you can essentially call your own. Without an existing database of art that is being sampled, the AI would not be able to generate ‘art’ for you.
But is this such a bad thing? A friend of mine, who introduced me to AI, had an interesting response to this concern. A musician himself, when I asked if it was ethical for AI to draw from existing works we humans have created, he answered that essentially, that’s what humans do when creating anyway. For example, the creator of Sailor Moon, Naoko Takeuchi, was famously inspired by the high fashion runways of Christian Dior and Mugler, referencing them in many of her character costume designs. And then, when I draw, my own art often references Sailor Moon, not to mention other media I have been exposed to. As a teen I traced Sailor Moon art so many times it’s engrained in my muscle memory.
But what AI does is less drawing from inspiration, and more copy-pasting existing things together. That being said, copy-pasting things together is still a form of art – collage, for example. People have long made art using newspaper clippings, magazines, and bits of found and recycled materials. Andy Warhol, in his famous Marylin Monroe pop art portrait, did not even own copyrights to the image nor permission to use it. He did get in trouble for it, however, it’s generally accepted as an important piece of art history regardless. People definitely wouldn’t have accepted that from AI. But human beings have been doing it themselves. The act of stealing art is inherently human!
As humans we draw from the world, the art we have seen and the media we have consumed. So does AI, but I think we can agree that it is not quite the same. In many ways it is less transformative and more an emulation. Moving forward, it is important that artists have the authority to decide whether or not their works are included in AI data sets. Adobe Firefly, for example, is an AI image generator which uses “a dataset of licensed content, such as Adobe Stock, along with public domain content”. An AI that draws only from images that are approved for AI usage is the ideal solution. People should not steal art, and neither should AI.
AI: Is it Art?
I’ll tell you right off I don’t think anyone can ever truly have a definitive answer for this question. Still, I asked our good friend, ChatGPT for his thoughts on what ‘art’ is:
Then I asked him whether or not AI can create “art”:
Art is something which is created that can evoke a reaction from those who experience it. AI art can do this. Although AI references an existing data base, humans do that to (we create art that is shaped by the art we consume), and although AI pieces what’s in the database together, humans do that as well. ‘Is it good art?’, and ‘Is it valid art?’ are questions that would take you even further down the rabbit hole.
I guess the question I’m thinking is: what would it take for us to consider what AI creates to be ‘real’ art? What would be the parameters? How transformative would the work have to be? But then, does that destroy the point? Humans have created art installments from such things as a pile of candies guests are encouraged to take from, to a ‘Museum of Broken Relationships‘ featuring ordinary items left behind by broken relationships. To de-legitimize any kind of art by saying ‘it’s not art’ can sometimes be a slippery slope. If that’s not art, what else is not art?
However, what I can be sure of is this: whilst AI generated images may constitute art, people who type prompts into AI may not be artists. If your only contribution is having typed ‘a rainbow fairy holding a kitten’ then surely the AI is the artist of that piece, not you.
AI: Is it any good?
Not long ago, ChatGPT was hailed as the revolutionary AI poised to transform society as we know it. Social media overflowed with tips and strategies on using ChatGPT for business growth and social media presence. If you didn’t use AI, you’d be left in the dust. However, the initial excitement has since faded as its limitations have become apparent. Limited in scope, prone to making errors, and often biased, ChatGPT is less useful than it seemed at first.
Although it serves as a time-saving tool for simple tasks, it falls short when it comes to more complex ones. It streamlines activities such as summarizing articles, generating hashtags, and breaking down complex tasks. However, it functions more as a virtual assistant than a comprehensive solution. While capable of assisting, its abilities are limited. It was helpful, for example, while writing this article, but could it have written the article itself? Well, yes, technically speaking. But not a very good one.
Artificial intelligence draws on existing information and images, as we have mentioned previously. It cannot (yet) create or innovate in the way a human can. You can ask an artificial intelligence to come up with a four-day posting schedule for your social media account, but the content it comes up with would be roughly the same as what you’d find by doing a quick google search.
AI art, similarly, is good at generating things that are referential. The more expected or ‘basic’ the prompt is, the better the result. Once you get into the ‘generating ideas’ stage, that’s when the AI begins to struggle. There’s this amazing TikTok in which the creator tries to get ChatGPT 4.0 to create a picture of a bowl of ramen without any chopsticks in it, to hilarious failure. (Go watch that one, seriously)
To further explore my thoughts on AI art and its capabilities, I have been messing around with Adobe Firefly. Now, I am aware that Adobe Firefly is not the most sophisticated AI art generator out there. However, it is free and claims to use only stock images as its data set, so I’ll be using it for the examples in this article.
So, I tried making some mock-ups for coloring page flat lays. I put in a simple prompt of ‘a mockup of a white sheet of A4 paper in a flat lay set up, on a table decorated in an Alice in Wonderland style tea party’. This worked pretty well. Basically, it’s a bunch of stock images moshed up together. Simple enough. If you zoom in, however, you can start to see where it gets crunchy. The teapot has two spouts and some of the cookies look like eldritch horrors.
Input a prompt like ‘chibi cat girl’ and it also gives a pretty decent result. A cute chibi with cat ears. Done. As generic and soulless as you can get.
The problem arose when I wanted something a little more creative and unique. Input a prompt that doesn’t have a million existing ideas out there already, and it starts to go pear-shaped. I input some prompts to try and get a half-human, half-caterpillar creature (thinking about some art for my Alice in Wonderland coloring book) and the results are considerably worse. That’s because there are a lot more images of cat girls out there than chibi caterpillars. I even tried DreamUp, DeviantArt’s AI art generator. Nothing came close to the idea I had in my head and it was a waste of time overall. What this really demonstrates to me is that AI generated work is derivative more than it is transformative.
On the left, Adobe Firefly’s efforts. On the right. DeviantArt’s DreamUp.
Wake up babe, new eldritch horrors just dropped.
If you want to see what I actually ended up creating, check out my Instagram!
AI: Will it Replace Us?
AI gets more sophisticated as time goes on, there’s no denying it. It can now create portraits that barely 61% of people can identify as AI-generated, which was unimaginable before (although it’s better at generating white people because its data sets are inherently biased). AI has absolutely put people out of jobs. Then again, so did the Industrial Revolution. Whenever there is a jump in technology there is usually a time of turmoil and change, followed by assimilation. Mass-production has made products more affordable and accessible for people around the world, but it also put a lot of people out of work as many tasks became automated by machines. Despite this, handmade products still exist and traditional artisans are still out there.
Back to the topic at hand: so you’ve asked AI to generate your ramen bowl without chopsticks, and it didn’t work. And you asked for a teapot and got a two-spouted horror instead. Now, if you are a proficient artist, this will not present much of a problem. You can use your art skills to clean up the image and fix the error. You can touch up the lighting, change the colours, or do anything else you want to do. Similar to how I used ChatGPT as an assistant to help me write this article, an AI-generated image can give you something to work from. It can make your life easier and save some time.
On the other hand, if you’re not an artist, these issues in AI-generated art can present a significant problem. Giving the AI consecutive prompts can often end in the image looking more uncanny valley as the prompts keep coming. The more the AI messes with the initial prompt, the more jumbled and incoherent it becomes. It can also be very difficult to get AI to understand *exactly* what it is you want or the vibe you’re going for. Putting in a prompt that exactly conveys what you want and then getting what you want is not as easy as it seems.
Take this image below, and think of how you might convey this to an AI; how would you generate exactly this image? Have a go at generating it in your AI of choice, see what happens. There is a far cry between ‘cute chibi girl’ and ‘a girl with long blond hair tied in two pigtails wearing a magical girl uniform smiling at the viewer, while in the background a man in a tuxedo and a mask carries the same girl, now in a princess dress, floating in front of a giant crescent moon’.
(Image copyright © Naoko Takeuchi)
Moving forward (Does it even matter?)
Whether AI is good or bad, ethical or unethical is a moot point. It’s not going away. But at the moment it seems that the people most passionate about AI are people who really, really don’t want to pay humans for their work and would rather a robot do everything for them, for free, instead.
Practically every website now has an ‘AI virtual assistant’. And they don’t exist because they are somehow better or more efficient than human live chat agents – I am not exaggerating when I say that no bot has ever provided me with anything I could not have already figured out myself by just looking through the website help section. No, they exist because they don’t need to be paid a salary. At best, AI chat bots are a sort of miniboss you need to defeat to reach the actual human customer service agent at the top of the tallest tower. Because of that, human live chat agents are still around, but who knows when or if the technology will get good enough to put them out of a job as well.
Change is scary, and this affects so many of us. I believe that eventually AI will fully assimilate and exist alongside human-made creations and creators. As regulations come into place and AI generated content becomes more transparently labelled, things may start looking up. Maybe some companies will choose to generate all their content rather than hiring humans. Maybe a studio like Netflix will create a series completely created by AI. Will that lead to quality content? Only time will tell. But I doubt that AI will take over and lead to a future where only AI-generated content exists. Human-created content will always be around and there will always be people passionate about art created by humans.
So what can we do? We can fight for more regulations concerning AI-generated content. Some bigger corporations like Disney have started the battle when it’s been shown that AI generators were infringing on trademarks (don’t anger The Mouse is all I’m saying). Some artists are winning legal copyright battles. We really have no idea what the future holds, but we can help shape the future through our action and speaking up for our beliefs and our rights. We can boycott AI that is unethical or built off of stolen work, and we do not have to support any work we don’t believe in (that hypothetical Netlflix AI series? Don’t watch it).
Still, dismissing AI entirely as a waste of space seems counter-intuitive to me. As small business owners, artists and content creators, we can use the tools available to us to make us more efficient. I can’t afford to hire website designers, ghost writers, editors, digital assistants, photographers and more. Once in a while I use Fiverr to outsource some real human help, but even that is a stretch on my budget. But I can use AI art to help me run my business and make my life easier. AI has been really helpful for helping me sort out SEO for my website, researching hashtags for my content, summarizing my blog posts into Instagram carousel posts, generating email drafts, writing terms and conditions for my commissioned art, and creating better, more informative blurbs for my coloring books. And I think that’s an overall net positive. Looking to the future, we can only stand by and let technology take its course and see where we end up.
Did you experiment with any AI after reading my article? Let me know in the comments below!
-Yams
