Artificial Stupidity: Entering the AI Image Generation Generation

A drawing of Peter reaching into a monitor to touch the chin of a surprised looking character that is AI generated with multiple arms and hands.

As a graphic designer who’s been around the block more than once, I’ve seen technology trends come and go. Well mostly just come, and evolve – the good ones stick around.

For instance, I’m into my 3rd decade of Photoshop, a tool that at one point was considered “cheating”. Even back then photos became suspect as admissible court evidence since they were so easily tampered with. But nothing has quite captured the imagination in our industry – and induced as many head-scratching moments – as AI image generation. 

Way back in 2022 (a long time in high tech years) I had a client who started sending me these glorious fantasy paintings to put on T-shirts for his Tech Death Metal band. So when I asked him where the heck he was getting them. He responded by quickly sending me an invitation to join the beta group of early adopters of Midjourney, one of the pioneers of AI image generation. I’ll admit it was a struggle navigating through a Discord community full of 20-something basement gamer nerds and trying to get usable images from it. I must admit that even from the get-go it was lots of fun, and mind-blowing, before it became part of our everyday existence. We were told it’s the future, a digital muse that can conjure masterpieces from a few keystrokes. And as it turned out, it is, sometimes. But even now, it’s more like a wildly enthusiastic but slightly confused server at the local diner who brings you your soup with a fork in it.

Don’t get me wrong, my colleagues and I are all about incorporating AI into our workflow, but frustrations still abound. Especially when we’re now expected to do things better, and quicker (and therefore cheaper) than ever. Just enter a short request, click, and Bob’s your uncle. All that and half the time I’m left either cursing or laughing at my monitor, while client expectations continue to ramp up. 

Let’s review the hilarious and often bizarre artifacts regurgitated by our new AI assistants. We’ve all seen them: the six-fingered hands, the unreadable or mis-spelled typography that looks like a cat walked across the keyboard, and the eerily perfect skin that makes human subjects look like they’ve been cryogenically frozen. 

I also run an online merch store that caters to musicians and music fans, and I’m finding that I need to be really careful with this AI stuff. Ask for a vintage Gibson Les Paul, and you might get a lovely looking guitar but with 7 or 8 tuning pegs (instead of 6), a dead give-away to a guitar player. See the devil isn’t at the Crossroads, it’s in the details, especially when it comes to complex but familiar objects. They are often just…off.

The real challenge for professionals, however, lies in the lack of control. It has caused me to go back to my beloved Photoshop time and time again, using up many of those hours I thought I saved by using Artificial Intelligence that turned out to be Stupidity. And it’s not only about quality control… as many of you know consistency is king in branding and marketing. Up until recently getting an AI to produce the same looking character for a children’s book page after page, or a consistent series of product shots for an ad campaign, was like playing the slots; it’s a different pull every time, and you might get lucky, but you sure can’t bank on it. The same prompt could yield wildly different results, and precise editing was a pipe dream. Forget about “moving that logo a little to the left.” 

The good news is it’s getting better and fast. AI apps are adding functionality at a break-neck pace that includes the ability to upload character and reference images to help solve the consistency issues, along with much better post-generation image editing.

From what I’m seeing, the same growing pains and advancements, are evident across the AI landscape. AI-generated video used to feel like a bad acid trip, and AI research can still be riddled with inaccuracies because of poor data quality, but the film industry and academia are still increasingly embracing it. 

I’m still all-in too, but I refuse to think of it as anything more than a powerful, although sometimes idiotic tool. It’s a fantastic starting point for brainstorming and iteration. I would say one of the most valuable benefits to us creatives is that It almost eradicates the classic fear of facing the blank page. But it’s not a replacement for experience, the discerning eye, the creative spirit, and the critical thinking of a human professional. It may now be a collaborator, but not a replacement. And I’m still not ready to give up my job to a machine that thinks a Les Paul has 7 strings. Details, details!

That being said, I know what you’re thinking… did he use AI to write this blog? The answer sadly is yes, but I had to edit the bejesus out of the stupid thing.

Picture of Peter Perko

Peter Perko

Creative Direction, Design

Share:

HOW CAN WE HELP?