Employing Humans to Clean up AI Slop

Lisa Carstens, a graphic designer, is busier than ever because to the same technology that was supposed to drive her out of business.

A significant amount of Carstens’s day is spent working with companies and individual clients who want to correct their unsuccessful attempts at AI-generated logos. Carstens is a lifelong freelancer located in Spain.

She frequently receives images from customers that are strewn with illegible text and messy lines, and when enlarged beyond a certain scale, they appear to be a jumble of pixels.

Some individuals come to you upset because they couldn’t do the task themselves using AI, while others are aware that AI isn’t flawless, Carstens added. Additionally, you sort of need to have empathy. You do not wish for them to feel foolish. Then it has to be fixed.

For a lot of gig workers, the problem has changed the employment market. Some claim they have discovered new employment as a result of AI’s incompetences, despite the general fear that AI is displacing labor across industries: Authors are urged to improve ChatGPT’s writing. AI photos that are crooked are being fixed by artists. Even programmers are expected to update glitchy applications created by AI assistants.

More outsourced workers have been replaced by AI than permanent employees, according to a new MIT study. However, 95% of companies’ generative AI experiments provide no return on investment, according to the study.

The paper claims that talent, regulations, and infrastructure are not the main obstacles to growth. It’s learning. The majority of GenAI systems retain no input, adjust to changing circumstances, or get better with time.

The AI-generated logos that customers submit Carstens are occasionally so well-designed that they just need minor adjustments from her. Sometimes, though, Carstens has completely redo the logo while staying faithful to the AI-generated design in order to provide a high-quality outcome. This process frequently takes longer than if she had created the design herself.

Many freelancers claim that correcting AI’s errors is not their ideal career path because it typically pays less than traditional jobs in their field. However, others claim that it helps cover their bills.

Kiesha Richardson, a freelance writer, stated that the only thing one can do is to learn and adjust. Additionally, some of my coworkers are determined not to use AI. However, I say, “I need money.” I’ll take it.

Located in Georgia, Richardson noted that half of her current clients engage her to edit or rewrite their AI-generated articles that “don’t look remotely human at all.”

Some of the issues are simpler to fix, such as AI’s excessive use of em dashes, even when alternative punctuation would be more appropriate, or its obvious preference for terms like “embark,” “deep dive,” and “delve.” In addition to these peculiarities, Richardson stated that creating an article necessitates conducting her own study on the subject because AI frequently provides general answers that don’t fully address queries as a person would.

She said that firms sometimes offer lower money for these projects because they are assumed to be less difficult, but many customers don’t realize the effort that goes into rewriting a badly written AI post. “But it can take as much creativity and thought as writing the entire article myself to make AI sound more human,” she added.

According to Richardson, “I am a little worried because people are using AI to cut costs, and one of those costs is my pay.” They discover, however, that they are unable to function without human assistance. Since AI isn’t providing them with the stuff they desire, perhaps we’ll be around for a little while longer.

According to recent data from freelance job platforms Upwork, Freelancer, and Fiverr, demand for a variety of creative work increased this year, and clients are increasingly seeking people who can collaborate with AI technologies without completely relying on or rejecting them. This is in addition to the fact that businesses are having difficulty determining how to handle AI.

AI is already automating monotonous and low-skilled activities, according to data from Upwork, but the site is witnessing an increase in demand for more complicated tasks, including creative art direction or content strategy. In the last six months, Fiverr reported a 250% increase in demand for specialized activities related to web design and book illustration, ranging from “Shopify website design” to “watercolor children story book illustration.”

In a similar vein, Freelancer reported a spike in demand for human labor this year in the fields of writing, branding, design, and video production, including requests for emotionally charged material such as “heartfelt speeches.”

Matt Barrie, CEO of Freelancer, remarked of the practice, “I mean, the fastest way to get dumped is to send a love letter to your girlfriend or boyfriend and use ChatGPT to write it.” Additionally, the same holds true for brands. When AI is fully utilized to make anything, the market is aware of it and reacts viscerally right away.

Consumers have continued to criticize brands that have been found to use AI. When an AI-generated woman was used in a Vogue commercial last month, Guess caused a stir on the internet.

Therefore, some artists claim that their clients only want a human touch to set themselves apart from the increasing amount of AI-generated content available online, even if there are no evident errors created by AI technologies.

Todd Van Linda, a Florida-based illustrator and comic book artist, believes that AI art is easily identifiable, if not by specific features that are incongruous, then by the plasticine effect that distinguishes AI-generated images in a variety of styles.

According to Van Linda, “I can look at a piece and not only tell that it’s AI, but I can tell you what descriptor they used to generate it.” They don’t want anything to do with it, especially when it comes to independent authors, since it’s so formulaic and clear. They seem to have gone at Walmart to get a cheap book cover.

He claimed that writers seek his assistance because they are aware that AI-generated artwork is unable to convey the hyperspecific “vibe” of their unique narrative. His clients frequently simply have a vague understanding of what they desire. Van Linda’s next task is to figure out what each customer wants and make something that evokes the precise emotion they want their artwork to elicit.

According to Van Linda, he is also contacted by individuals who want him to “fix” their AI-generated artwork, but he now steers clear of those assignments as he has come to the realization that these clients are usually less prepared to pay him what he thinks his work is worth.

“They have a mismatched collection of generalities that really don’t follow what they’re trying to do,” he said, adding that correcting the images would require more labor than beginning from scratch and doing it correctly. However, because they don’t want to spend any more money, they’re attempting to fit the square peg into the round hole.

Clients that have previously cut costs on AI technologies have resulted in low pay for gig workers in a variety of sectors, even more specialized ones like coding. Many of Harsh Kumar’s clients, who is headquartered in India, claim that they have previously spent a significant portion of their budget on “vibe coding” solutions that have failed to provide the desired outcomes.

However, he said that more people are coming to the conclusion that hiring a human developer is worthwhile due to the hassles avoided when attempting to have an AI helper update its own “poor code.” According to Kumar, he frequently receives vibe-coded websites or applications from his clients, which lead to erratic or completely unworkable systems.

Among his projects have been the repair of an AI-powered support chatbot that provided customers with incorrect answers and occasionally revealed private system information because of inadequate security measures, as well as the reconstruction of an AI content recommendation system that frequently crashed, made pointless suggestions, and exposed private information.

“AI can’t completely replace humans, but it may increase productivity,” Kumar stated. “I am certain that long-term initiatives will require human labor. Humans were ultimately responsible for the development of AI.

Source link