“Magic.”
Months ago, that’s what I told the AdExchanger editorial team it felt like to play around with ChatGPT, DALL-E and other generative AI tools.
But while generative AI can feel like magic – instantaneous, effortless, able to leapfrog human abilities in a single bound! – many of the tech’s wondrous outputs are built on the backs of creatives who have dedicated their lives to their craft.
After attending the Federal Trade Commission’s (FTC) virtual roundtable about the impact of generative AI on creative fields last Wednesday, all I can wonder when I see an AI-generated creation is whose work it’s based on.
Creators contend that large language model (LLM) makers are taking an “ask for forgiveness, not permission” approach, unleashing the web-trawling tech to ingest, digest and regurgitate original creative content. These companies have lifted text, images, audio and video from creators – and given them nothing in return. Creators are angry, shocked and worried as they see their work appropriated without their knowledge.
It’s particularly galling when programs like ChatGPT suggest users request art “in the style of” the creators they’re ripping off.
“In today’s reckless, careless rush to launch new generative AI products, we are seeing what is euphemistically referred to as AI ‘training’ or ‘learning,’ but which is in fact illegal copying of artistic works on a massive scale,” said FTC roundtable participant Jen Jacobsen, executive director of the Artists Rights Alliance.
Dollar signs in their eyes
Perhaps the idea of recognizing creators, paying them or offering them the chance to consent fades to background noise when you’re locked in a race to secure billions in investments from VCs to build out your startup and rake in obscene profits from companies eager to use your au courant tools.
And the race is only beginning: A June Bloomberg Intelligence report estimated the gen AI market will grow to $1.3 trillion by 2032.
It stands to reason that when many creators pull back gen AI’s curtain, what they see is naked greed and exploitation of the most intensely personal aspects of their identities. Their voices. Their time spent honing their skills. Their bodies of work.
Their actual bodies.
Models are increasingly asked to consent to full-body scans that generate 3D models of their body or face, according to FTC roundtable participant Sara Ziff, founder and executive director of the Model Alliance. It’s a troubling request, Ziff notes, given the rise in deepfakes.
Tim Friedlander, president and founder of the National Association of Voice Actors, said that it takes “three seconds of source audio to create a realistic voice clone,” which can “deceive consumers into believing that a trusted voice is communicating with them.” Exhibit A: Recently, Tom Hanks, popular YouTuber MrBeast and morning show host Gayle King all had to tell followers to ignore scam ads featuring their unauthorized AI-generated likenesses.
Some in the modeling community also worry about another form of fakery: that companies will use AI-generated models to support DEI initiatives.
“There is a real risk that AI may be used to deceive investors and consumers into believing that a company engages in fair and equitable hiring practices and is diverse and inclusive when they are not,” Ziff said.
Arguably, this type of deception has already happened. Levi’s faced serious blowback this year when it said it would use AI-generated models to represent a range of skin tones and body types instead of hiring diverse human models.
What we owe to each other
As a number of FTC roundtable participants mentioned, the data sets AI systems train on are deeply flawed. For instance, Neil Clarke, founder and editor of sci-fi and fantasy-focused Clarkesworld Magazine, pointed out that the Books3 data set contains pirated copies of at least 183,000 copyrighted books. Because of YouTube and TikTok videos and blog posts that “make false claims about earning riches from ChatGPT,” including through ad revenue, Clarkesworld is drowning in AI-generated spam short story submissions.
“We reached out to YouTube and others in the hopes of having these videos taken down. But no one ever responded,” Clarke said.
The concept artist and illustrator Karla Ortiz, who worked on many Marvel movies to develop their look and feel, said the entirety of her work lives in Laion-5B, a data set containing 5.8 billion text and image pairs alongside private medical records, nonconsensual pornography, images of children and social media images. Ortiz’s name has also been used thousands of times in image-generation prompts as people seek to recreate her distinctive style.
The initial rollout of gen AI models has already harmed creative communities, including artists, photographers, writers, editors, translators, composers, musicians, actors, models and software developers. And the advertising industry can’t be Switzerland on this issue. After all, many marketers have spent the better part of the past year experimenting with these tools for everything from brainstorming ideas to writing ad copy to generating creative versions of campaigns.
At the very least, those in the advertising industry should reflect on whether they’re perpetrating the exploitation of creators. Better yet, stand with creators in demanding that commercial AI models limit themselves to opt-in data sets and reveal where their data comes from as well as how it’s being used. Active and enthusiastic consent is a must.
Only by protecting artists can society fulfill – and fully benefit from – the wondrous potential of generative AI.
If brands, agencies and other marketing pros instead choose to replace creatives with the very tools that scraped their data and stole their work, sure, they may see short-term efficiency gains.
But in the long run, the collective decision to turn our backs on creators could produce a society where no one aspires to create anymore. And all we’d have left is a banal procession of copies of copies of what might have once contained a spark of genius.
“Data-Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.
For more articles by Hana Yoo, click here.
