Brands caught using AI continue to face rebound from consumers. The speculation sparked online protests when it featured an AI-generated model in an ad that appeared on Vogue last month.
So, even outside of the obvious mistakes made by AI tools, some artists say that their clients simply want a human touch to distinguish themselves from the growing pool of content generated online in AI.
For Todd Van Linda, a Florida illustrator and comic artist, AI art can be easily identified by the specific telltale contradictions of details and then by the plasticin effect that defines images generated in AI across different styles.
“I can look at the work and tell you not only that it's AI, but that it's what descriptions were used to generate it,” Van Linda said. “Especially when it comes to independent authors, they don't want anything to do with it because it's so stylistic. That's obvious. They seem to have stopped at Walmart to get a cover of their book bargain.”
The author will come to him, he said. Because they know that the art generated by AI cannot capture the hyperspecific “atmosphere” of their individual stories. Often, his clients can only give them a rough idea of what they want. It is then Van Linda's job to decipher Van Linda's preferences and create something that elicits the exact feeling that each client is trying to evoke from his art.
Van Linda also said he would be approached by people who want to “fix” the art generated in their AI, but he is now avoiding those jobs as he has usually found that those clients are not willing to pay him what he believes his labour is worth.
“There's more work to correct these images than starting with clean paper and doing it right, because what they have is a collection of generality inconsistencies that don't really follow what they're trying to do,” he said. “But they don't want to spend more money, so they try to squeeze square pegs into round holes.”
Low wages from clients who already make AI tools cheaper have impacted gig workers across the industry, including technical ones such as coding. In the case of India-based web and app developer Harsh Kumar, many of his clients say they have already invested much of their budget in “vibe coding” tools that can't provide the results they want.
But others have noticed that firing for human developers is worth a headache that will save them from trying to acquire AI assistants to fix their own “Crappy Code.” Kumar said it brings him an atmospheric website or app that often brings to him a system that his clients often have unstable or completely unusable.
His projects include fixing AI-driven support chatbots that gave customers inaccurate answers, and rebuilding AI content recommendation systems that have leaked details of sensitive systems due to their lack of safety, frequently crashing, rebuilding unrelated recommendations and revealing data.
“AI can increase productivity, but it cannot completely replace humans,” Kumar said. “I'm still convinced that humans are needed for long-term projects. At the end of the day, humans were the ones who developed AI.”
