Does AI Replace Human Creators?

The rise of generative artificial intelligence systems has raised many difficult questions. Is it legal to train AI systems on copyright-protected work? How trustworthy are the systems? When and how should the use of AI be disclosed? And so forth.

However, the most crucial question is how AI will impact people working in creative fields. This includes writers, artists, and coders whose work is used to train AI systems and whose works most directly compete with AI systems.

Some had argued that AI was a feature that would help creatives be more productive. Others, however, argued that AI would be used to replace human creators and result in less work for humans in those fields.

Though it’s still early to know exactly how AI will impact our society, a recent study published by the Harvard Business Review paints a fairly bleak picture. Authored by Ozge Demirci, Ozge Demirci and Xinrong Zhu, it found that freelance jobs for creatives have declined since the introduction of generative AI and are showing no signs of rebounding.

While it may still be premature to attribute all or even most of this loss to AI, it is clear that AI has not been a boon for creatives, particularly freelancers.

The Basics of the Study

The study analyzed over 1.3 million job posts on a “leading global online freelancing platform” from July 2012 to July 2023. Considering that ChatGPT launched publicly in November 2022, this meant that they collected data 8 months after the launch and 16 months before.

From there, the jobs were grouped into three categories: manual-intensive jobs, automation-prone jobs and image-generating jobs. Manually intensive jobs were intended as the control, as searches and postings for such jobs remained relatively flat.

Writing, coding, and engineering jobs, categorized as automation-prone, had the steepest decline. They dropped 21% compared to manual jobs, and image generation jobs fell 17%.

Writers were the hardest hit, with an over 30% decline in job postings. Coders were next with an over 20% drop, and engineers saw over a 10% decline.

However, those areas with the most significant decline also had the highest search interest after AI systems were launched. According to the authors, this indicates those industries had the highest awareness of AI systems.

The authors encourage companies to be transparent about their use of AI and for concerned individuals to start upskilling or reskilling to learn AI systems. They argue that significant opportunities are also available while some job replacements will take place.

Limitations of the Study

As interesting as the study is, it also has some notable limitations.

The study has a relatively narrow focus. Though it examines a large number of postings, it focuses on freelance postings only and only on one platform. This doesn’t say anything about full-time employees or jobs being offered on other platforms.

Second, while all the areas checked dropped after the introduction of AI systems, they also were already dropping before. For example, the rate of increase in automation-prone jobs peaked in October 2021, a full year before ChatGPT was launched. It had also gone into the negative severam months before ChatGPT launched.

Likewise, the change in image-generation jobs peaked in January 2021, six months before image-generation tools were introduced.

As others have noted, other factors could be involved in these declines. So, it may be early to argue that AI is taking away creative jobs. However, it’s also been pretty apparent that it has been a boon, either.

If AI were a positive for human creators on a macro level, we would expect at least some upward shift in these numbers. Instead, they all fall faster after the introduction of AI. In short, AI has not been a positive for creators working in those fields and, most likely, has been at least somewhat negative.

It’s difficult to say how negative the news is. We are very early in this change. But it’s safe to say that it’s not good, and that may have legal implications down the line.

Legal Implications

While the biggest and most important implications of this are on a social or societal level, there are also legal implications that must be considered.

Human authors and other creators have long argued that training AI systems using their work is a copyright infringement. That issue is the subject of dozens of lawsuits against various AI companies.

AI companies have argued that using copyright-protected training data is not an infringement but a fair use. However, the Andy Warhol ruling in May 2023 blunted some of those arguments.

Before that ruling, transformativeness was the most important consideration when weighing fair use. However, the Warhol ruling reigned in that approach and focused more on other aspects, including economic ones.

One factor is whether the new work replaces or competes directly with the original. This study makes it difficult to argue that AI works do not compete with their source material, at least somewhat.

While this doesn’t mean that AI companies are in trouble, it does point to another obstacle they will have when making their arguments.

Bottom Line

As interesting as the study is, I don’t think it’s definitive. It has too narrow of a focus, too little time has passed, and there are too many other factors to consider.

Still, it’s a canary in the coal mine. It warns that, at the very least, AI systems have not been a net positive for human creators. If we don’t make adjustments, we may be dealing with a significant upheaval for which we are unprepared.

To be clear, there are no easy answers here. As discussed in July, copyright won’t kill AI systems even if rightsholders win. AI is here to stay, for better or worse.

However, we still have plenty of time to shape AI’s role in our society and what role human creators will play.

These are big questions, and they need big answers. However, I don’t expect them to come from AI companies. Instead, they have to come from us and our governments.