Is AI Demolishing the Scaffolding of the Creative Class?

Opening Dialogue

Shiwen: Yongliang, a report by Apricitas Economics has been widely circulated in the industry recently, stating that over the past four years, more than 200,000 jobs have disappeared in the U.S. creative industries. What do you think about this number?

Yongliang: When I first saw it, I checked the BLS employment data. 200,000 is not a number fabricated for headlines; it’s the result of combining sub-industries such as fiction, illustration, music, film, and publishing. What surprised me was that this unemployment occurred in years when the economy wasn’t declining overall, unlike the complete collapses in 2001 and 2008.

Q1: 200,000 creative jobs disappeared. Why isn’t this a “cyclical unemployment”?

Shiwen: Many people might instinctively think this is due to economic downturns. But the report says that the intensity of job losses in the media sector is comparable to those in 2001 and 2008, yet it wasn’t a general recession. What’s the difference here?

Yongliang: The difference lies in “universality.” In 2001 and 2008, almost all industries were laying off employees—finance, manufacturing, retail, etc. But in recent years, service industries, healthcare, and live entertainment are still hiring. The unemployment is concentrated in the creative sector, which is more like a structural contraction of that industry. In the past, we said creative work was counter-cyclical—advertising budgets might shrink, but content consumption won’t. But now, content consumption exists, just with fewer intermediate personnel needed. Joseph Politano, author of Apricitas Economics, pointed out after comparing BLS data that the loss curve in the media sector has caught up to the previous two recessions, but the macro context is completely different.

Q2: Why are the film and recording industries hit hardest?

Shiwen: There’s a striking number in the report: the film and recording industries lost over 100,000 jobs in three years, accounting for nearly one-third of industry employment. What’s wrong with these industries?

Yongliang: A key term is “deindustrialization.” After the rise of streaming platforms, the production sector was broken up, and mid-to-low-level jobs were cut first. The pandemic made remote collaboration normal, allowing tasks like sound design, editing, and color grading to be outsourced globally. Combined with AI tools for dubbing, scoring, and pre-rendering, many tasks that once required a team working for two weeks can now be done by one person with a few tools. This summer, employment in this industry dropped to the lowest since 2008. This cannot be explained by the failure of a single blockbuster; it’s about how the entire industry is redefining “how many people are needed.”

Shiwen: One-third sounds already quite exaggerated.

Yongliang: Yes. And this one-third is a net loss, meaning it’s harder for newcomers to enter, and senior mid-level employees are also being optimized.

Q3: Digital art is contracting comprehensively, while live entertainment is growing. What does this contrast indicate?

Shiwen: Fiction, images, and music are all laying off employees, but live performances, theme parks, and offline exhibitions are growing. Isn’t this contradictory?

Yongliang: Not at all. It actually explains things well. Digital content is too replicable, and AI has further pushed it toward “infinite supply.” An AI-generated image, a self-composed track, or a novel completed with the help of large language models have almost zero marginal cost. But live experiences sell scarcity: seats, time, location, and the real presence of people. Consumers are willing to pay high prices for scarcity, while digital content is being priced lower in the price war. Thus, capital and employment are shifting from “replicable” to “non-replicable.”

Shiwen: So the definition of the creative class is also changing?

Yongliang: Right. In the past, the creative class made money through copyright and replication, but now that link is becoming thinner. In the future, those who can get premium prices may be creators who can combine on-site work, communities, brands, and experiences.

Q4: The tech industry is also affected by AI coding, how does this relate to unemployment in creative positions?

Shiwen: Many people think AI affects blue-collar or customer service jobs, but reports say tech positions are also losing talent. In 2022, there was a year-over-year increase of nearly 300,000 tech positions, followed by consecutive years of decline. What does this have to do with the creative industry?

Yongliang: The underlying logic is the same: AI coding agents are consuming “moderate complexity” mental labor. In the creative industry, storyboards, scripts, editing, and mixing belong to work that can be automated into modules. The two industries seem unrelated, but they are affected by the same type of jobs: highly educated people engaged in standardized knowledge-based labor.

Shiwen: But aren’t tech companies still hiring a large number of AI talents?

Yongliang: They are hiring top-tier people, while many middle-level technical positions are shrinking. The same is true for the creative industry: top directors and top illustrators still have bargaining power, but many ordinary workers find their skills are rapidly losing value.

Q5: Why are companies reluctant to disclose the truth about AI replacing human labor?

Shiwen: This question is crucial. If AI has really replaced so many jobs, why do we rarely see companies say “we laid off people because of AI”?

Yongliang: There are two very practical considerations. First is legal risk. Many AI training datasets involve copyright disputes. If a company openly says “we replaced original painters/scriptwriters/voice actors with AI,” it’s like providing evidence to plaintiffs that they used copyrighted content to train their models. Second is public opinion and labor relations. Hollywood scriptwriters’ strikes and game company painters’ protests have shown that admitting AI replacement will intensify conflicts. So companies more commonly say “optimizing processes,” “improving efficiency,” “business adjustments,” hiding AI behind the scenes.

Shiwen: Does that mean the actual scale of AI replacement is underestimated?

Yongliang: Most likely. Media like Apricitas, Fortune, and Gizmodo are tracking this phenomenon, but BLS data statistics are lagging, and companies are also reluctant to disclose. The 200,000 we see now may just be the tip of the iceberg.