I wasn’t even planning on writing a blog post this week let alone one on such a topic as art generated by artificial intelligence, but as it is part of my industry and I’ve seen so much outrage from my peers (many who are friends) I thought I’d do a little digging and put something a little more comprehensive together than, “Support artists! Denounce technology!”
I’m going to try and keep all of this as brief, simple, and informative as I possibly can, but I’m also going to try and approach this from a (slightly) less biased angle.
What is A.I., and is it Bad?
You don’t have to be relatively well informed to know what AI is. It covers all aspects of our lives from taking care of minor tasks we don’t even think about to the stuff of science fiction nightmares requiring Will Smith to show us that even machines can—and in fact do—love.
Recently I finished reading Yuval Noah Harari’s Homo Deus: A Brief History of Tomorrow, a follow up to his bestseller Sapiens: A Brief History of Humankind. Harari writes a lot about AI and its benefits. Without going into too much detail and wasting time here, the three big takeaways for now are that:
AI is actually a really good thing that can drastically help with all sorts of things to improve life everywhere for everyone
Fundamentally, human life as we continue to understand how it works is not that different from how AI learns, adapts, and grows itself
It’s becoming clearer that creative things people believed AI could never reproduce aren’t far off and many areas like classical music are actually surpassing humans in quality, structure, and beauty
I know most people will not believe any of what I have just written, and that’s completely fine. I would never insist you take my word (or anyone else’s) on anything as gospel at–a–glance. I would strongly encourage you to do your own research though. The point is, however, that whether we like it or not, the world is going to continue to change, it always has, and it’s certainly not going to stop because a few of us don’t like the idea of being replaced by anyone or anything. Automation has been changing how we do our jobs and live our lives for centuries, and the whole process has continued to increase at breakneck speed, especially since the 1980s. In other words, AI is most definitely here to stay, and it’s probably better we figure out how to change with it, rather than stubbornly anchor ourselves against it until the next generation sees us as living fossils who refuse to accept change.
But we’re not here for a lecture on science or philosphy from some Muppet–loving Jersey boy whose blog readership doesn’t extend far past his own family and close friends. So let’s instead talk about AI generated art, specifically the latest fad known as Lensa or “Magic Avatars” because there’s a lot to unpack and it gets complicated.
Avatar Insanity or High Art?
Remember Bitmojis? I hated Bitmojis when they first debuted. Not because of how well or poorly they’re drawn (depending on your tastes), but because I didn’t come up with the idea first. For those not familiar, Bitmojis aren’t too dissimilar from modern day avatars you create like Meta’s more CGI looking Facebook avatars or Apple’s Memojis. My argument at the time was also that they took away opportunities from artists like myself to create illustrations for profit. On the flip side, they give people who do not possess the skill to draw an opportunity to express themselves creatively quite literally.
Bitmoji, Facebook, and Memoji avatars of yours truly, and one I drew myself (2017)
The point is that there will always be new technology to engage users and experience something they would otherwise have to commission an artist for. I don’t know a single person that has ever created a digital avatar using some type of technology and then claim that their “artwork” was anything other than a fun opportunity to represent themselves online or even just to “jump on the bandwagon.”
Now I recognize this is a slippery slope that could lead to something more nuanced down the road, but for the time being, let’s all try and remember a social media avatar is not the same as a portrait or caricature that you would give as a gift, hang in your living room, or rock on the side of your 1988 conversion van as you drive to your next gig.
AI Art Stink
The main focus of this article and the commotion that it’s caused is a program called Lensa by Prisma AI. At a glance, Lensa is a pretty standard photo and video app that lets you take and edit media with a variety of different features. It’s most popular of course are the filters that “turn your photos into works of art in the style of famous artists” and transform them “with popular art styles - anime, cartoons, sketches, watercolors...” all by using artificial intelligence. The app is free but offers premium monthly or annual subscriptions.
Okay, so nothing too new there. I am very far from having my finger on the pulse of any social media platform, but even I’m familiar with these types of things enough to know there are tons of them. I’ve even used some of them before myself, so why is this app getting artists in particular so upset?
The Lensa Learning Problem
Not too long ago (like literally just several months ago) Dall–E 2 debuted to slightly different fanfare. Dall–E 2 is also an AI art generator but it seems it’s absolutely ludicrous creations were more comical than threatening (for the most part). Hence, many people looked at Dall–E’s attempts as technological proof that a computer could never imitate the skill of an experienced artist. Or that optimistically, it’s still a ways off.
Dall–E (and Lensa) uses Apple’s TrueDepth API which most iPhone users know as the same technology that allows them to unlock their phones just by looking at it. Dall–E 2 learned how to create its images by studying tons of information to create images based on text prompts. This AI learning technique is known as Stable Diffusion.
This is tricky, but basically when AI uses Stable Diffusion, it’s not just learning to recognize features and characteristics of someone’s art, but essentially manipulating and reproducing elements from that art. So the issue then is that artists are accusing AI of using this learning technique by having Lensa specifically build its creations from existing art without permission from the artists it’s emulating. Now it’s hard to find sources to corroborate some of the allegations I’ve read, but many artists have actively accused Prisma AI of stealing their art specifically, requesting the company stop doing so repeatedly, then Prisma AI allegedly refusing and even cyber bullying them about it. There is compelling evidence to lend credibility to these claims where remnants of the artists signatures from original works are sometimes still visible in the AI generated art.
Darker Secrets
Okay, maybe I’m coming across a little too unbiased by praising what AI could (or hopefully should) be even though I’ve stated in the past how infuriating and hurtful art theft can be. So let’s look at how one of the more nefarious problems with Lensa isn’t even allegations of art theft, but how it depicts your pictures when they become “art.”
A big problem with AI in general is that it learns its lessons from its creators, and even modern day AI—which is touted as pure—has repeatedly displayed racism, nepotism, and sexism. Lensa has lightened skin tones of people of color, struggled with (re)producing Asian features, and sexualized women and children. Other AI art generators have done similar unwarranted things like taken on macabre tones when “crossbreeding” images that did not previously convey violence.
I want to be clear: these are not alarmist warnings that AI will rise up and destroy humanity. Artificial intelligence reflects the zeitgeist, which includes everything from cute and fun to morally questionable. In other words, whether it’s an art–stealing bot, a perverted algorithm, or even an altruistic ghost in the machine, it’s all taking its cue from us.
Now What?
So where do we go from here? As artists, we tend to react more emotionally because, you know, suffering is kind of “our thing” (until artificial intelligence corners the market on angst too). But like I mentioned before, technology is going to keep moving forward regardless of how we feel about it and that’s not necessarily a bad thing. A knee jerk reaction is to call out for regulation but that’s way easier said than done for two big reasons.
While there has been initiative in congress to try and moderate how things like AI continue to develop, bureaucracy—love it or loathe it—purposely moves slowly to make sure it’s covering all its bases properly (and that is an exceptionally gracious and arguably naive platitude). On top of that, it’s an antiquated system that’s literally hundreds of years old. Technology moves ludicrously faster, meaning that by the time well intentioned and thought out legislation finally passes even in the best of circumstances, the applied science behind that technology is usually obsolete, meaning any government progress was all for nothing.
Art © DaveHulteenDesign.com
The second problem is that government officials aren’t exactly young entrepreneurs anymore who even understand the technology they’re hoping to regulate, prioritize, or control. So when you ask older people using an even older system to help answer these questions, you eventually have to consider if the whole process wouldn’t be better served by the AI you want them to regulate in the first place! Understand though that this is not an endorsement to willfully hand the keys over to tech bros like Elon Musk or Sam Bankman–Fried. It’s pretty clear that being rich does not equal being responsible… or smart… or ethical… or competent… or sane.
So if creative people only know how to get upset over it, and our leaders only know how to politicize it, as usual, it all comes down to you, the user. AI really can do incredible things, but moving forward, it’s up to us to decide how we’ll use it.
And this is something everyone really needs to learn how to do better. Consider if that neat new AI avatar is worth the likes verses its privacy policy. Yeah, that’s a whole other kettle of fish because one thing practically no one understands is what kind of personal information you allow software developers access to when you agree to terms and services.
Like almost every app, Lensa uses legalese to ensure you maintain the rights to your photos, but then vaguely explains they have the right to use those photos to independently aid in research, development, and improving new and existing products.
This is not a dystopian outlook from a conspiracy theorist either. Your personal data is way more important to all of these developers than what kind of review you leave them on the App Store. A great rule of thumb when you download an app is its cost. If the app is free, then you’re the product that’s for sale.
Sources
I did a fair amount of research for this post, so if you’d like to read a bit more in depth on all of this, please check out these articles:
Lensa AI app: What to know about the self portrait generator by Meera Navlakha
Mashable
Understanding the impact of automation on workers, jobs, and wages by Harry J. Holzer
Brookings
Prisma is coming to Android, but there's a way to get it sooner by Stan Schroeder
Mashable
Careful — Lensa is Using Your Photos to Train Their AI by Shanti Escalante-De Mattei
ARTnews
Lensa, the AI portrait app, has soared in popularity. But many artists question the ethics of AI art by Morgan Sung
NBC News
‘Magic Avatar’ App Lensa Generated Nudes From My Childhood Photos by Olivia Snow
Wired
If you’d like to track what some artists are saying, a lot of insight on Lensa’s theft I read came from Jon Lam on Instagram. He credited Lauryn Ipsum for the discovery of remnants of artist’s signatures on AI creations. Karla Ortiz is helping lead the fight for artists rights.