Author: Melanie Mudge
DALL-E may be the talk of the town when it comes to image generation, but let’s be real: It’s not all sunshine and rainbows with this AI wizardry. Sure, it can conjure up a pineapple wearing a top hat, but what about the potential downsides? What about the fact that it’s essentially playing God with pixels and could have us all out of a job? In this article, we’re going to peel back the curtain and take a closer look at some of the cons of DALL-E, or as we like to call it, the “Pineapple Top Hat Tyrant.” So buckle up, buttercups, and let’s get into it.
Oh, hey, hi. It’s us, Scribely. Could you tell that the previous paragraph was written by an AI? Yep, we asked a text-generating AI to critique the same company’s image-generating AI. (Don’t think about it too much, it might melt your mind.) If you couldn’t tell, you’re not alone. And if you could, how?
AI technology is getting so advanced, it’s able to produce words and pixels that are almost indistinguishable from human-produced content. But as ChatGPT pointed out, it’s not all sunshine and rainbows in AI land.
In case you missed it, in Part 1 we examined some of the benefits of AI art generators and how they might be helpful in an artist's toolkit. Here in Part 2, we’ll look at the flipside and analyze some of AI’s downsides, especially when it comes to creative professions like photography and graphic design.
Art without Permission
We spoke with Kimmy Gerhart, a brand manager and graphic designer who stumbled across DALL-E when she noticed many Twitter accounts posting DALL-E images they had generated from prompts with unlikely mashups of topics. She found some to be “genuinely hilarious” and became curious. But her curiosity quickly turned to unease.
One of the biggest red flags Kimmy saw with AI image generators like DALL-E is how they “train” the algorithms. The technology doesn’t work unless it has a huge database of images to draw on (accidental pun, but we’re here for it), so the companies behind these types of AI have been feeding them millions of images for years—without any of the image creators’ permission. “Is the art coming out of them ethical?” she wonders.
While the ethics have yet to be decided in the court of public opinion, it seems that the legality is already being determined. Kimmy says that she was recently sent a sound bite about “a class action lawsuit against a handful of image generators that were creating ‘derivative works’ by appropriating images they did not have consent to use.” Of course, the technology comes with so many new problems that we’ll be curious to see how courts handle such murky waters.
Art or Facsimile?
Beyond how the AIs are built, Kimmy is also concerned about what they’ll eventually be able to do. “What if DALL-E gets so good that it could produce something nearly perfect to the original?” she muses.
Her fear is not unfounded. Though the nascent technology still has a long way to go, it’s improving constantly, all with the goal of being able to create any image as desired in seconds. If that goal is achieved, which seems very possible, why pay hundreds or thousands of dollars for art when you can create exactly what you want with DALL-E as long as you can describe it with words?
On one hand, DALL-E and other AIs like it seem to be making art accessible to everyone, not just those with deep pockets, which is a good thing. Art should be available to all humans. But on the other hand, it begs the questions: Is it really art?
Lest we get entirely out of our scope—we’re an accessibility solutions company, after all—it’s worth at least putting the questions out there. If an AI is simply using what’s in its database to produce pixels, is that truly creation? Humans draw on others’ work for inspiration all the time, so is this akin to that or simply mimicry at best and stealing at worst? Will AI ever be able to produce something original and new? Or will it always need humans to direct and guide it?
“I kind of feel like,” says Kimmy, “no matter how good DALL-E gets, it will never be able to replicate the human touch. It’s a machine, and art is a human experience.”
Competing with Free
Kimmy’s final concern is how these AIs will impact her ability to earn a living: “That feels like a scary reality for many artists right now, wondering when AI will make their livelihoods obsolete.”
It is possible that her profession could be completely co-opted in the future. Say Kimmy spends 5 paid hours creating digital art for the homepage of a client’s website. Once that art is on their site, any AI can find it in a web crawl and add it to its database unbeknownst to her. Then the AIs can generate images based upon and very similar to—if not exactly like—her original creations, all in a matter of seconds…and for free (at least for now). It’s not a stretch to imagine her client wondering why they paid her when they could’ve had DALL-E make it for free.
Luckily, the technology is not yet this advanced, and there are detectors out there that can quickly and easily determine if something was generated by an AI. But with all their improvements, it seems to be only a matter of time before the AIs overcome the obstacles.
If and when they do, we’ll all be forced to decide where we stand. If DALL-E becomes capable of doing everything Kimmy does as a graphic designer, and DALL-E’s work is just as good as or better than Kimmy’s work, would you still be willing to hire Kimmy for the job, let alone pay a fair wage for the work? Or would you opt for more bang for your buck by buying an AI-as-a-service subscription (did we just invent AaaS?)? It won’t be an easy decision in our increasingly price-driven, cost-minimizing world.
It’s Still In Our Hands
Ultimately, Kimmy isn’t opposed to AI image generators if they work through some of these ethical shortcomings: “It would be really cool if there was an image generator that worked solely off open-source imagery. I would wholeheartedly support that.” Her point is a good one. As the technology’s builders, we humans are in charge. We have the ability to choose where it goes and how it gets there if we ask the right questions now, before the technology becomes entrenched in our everyday lives. Maybe before we unleash the proverbial beast, we should learn from our past and allow more voices into the conversation first. If we do that, DALL-E and others can be molded into the helpful tools they have the potential to be.
