The Economics of Generative AI in Creative Industries

Artificial intelligence (AI) software stands to revolutionize the creative industries by taking commands written in common language and turning them into high-quality outputs.  This allows laypersons - not computer-savvy professionals - to get computers and software to create sophisticated and unique products.  Alarm has been raised over the effects of AI on many jobs, especially those that are more easily automated.  Previously, these jobs had been insulated due to software’s inability to accept common vernacular as input or deliver such vernacular as output…but no longer.  

The artistic and journalistic fields are considered particularly vulnerable, as these practitioners frequently use computers already to generate their work.  They often produce work similar to existing work, which gives AI tremendous banks of material on which to draw for examples.  Questions abound as these workers, and the public, wait to see whether human endeavor in art, music, and journalism starts to be replaced by AI-generated output.

Economics Behind AI 

Cost Structures

While AI products may not be more demanded by consumers than human-created products, costs per unit will likely be significantly lower.  While a human artist, musician, author, or journalist must receive revenue commensurate with his or her time, talent, and skill, a computer only needs to cover its average total cost.  Although computer servers can use tremendous amounts of electricity, they last for many thousands of hours and can generate millions of images and documents.  Thus, the marginal cost of AI production is very low.

Paywalls

However, costs for using generative AI may increase in the near future as the technology becomes established as creating sellable output.  Currently, there are freemium versions of AI available to the general public, with users having to purchase subscription packages to generate more sophisticated output.  As AI improves, it is likely that more AI capabilities will be moved behind paywalls, requiring users to purchase subscriptions.  This will be done to generate more revenue for AI companies, but may also help human artists remain more economically competitive.  As the marginal cost of AI work increases due to paywalls, more human artists will remain viable as substitutes.

Job Polarization

Popular Artists Generate More Output

AI may cause mass unemployment in certain sectors, but not all workers will be similarly affected.  Those who are already popular or established may benefit tremendously from AI, which will allow them to create more output under their name.  If their name has “weight” and can be used in nonprice competition (i.e., marketing), they will likely see an increase in revenue as they can use AI to produce more units of art, music, writing, journalism, etc.  This may lead to the top tier of artists in each field increasing their market share substantially, boosted by AI.  Overall, this could create a relative homogenization of art due to a decrease in the number of competing substitutes.

Lesser-Known Artists Face Unemployment

Unfortunately for artists who do not already have an established following, and are thus seen as more of a perfect substitute for each other, the AI revolution is likely to cause a significant reduction in revenue.  If your name does not make you a differentiated product, consumers have little incentive in not replacing your labor with AI, which has a lower marginal cost.  Artists who work for design companies, such as advertisers, may face mass unemployment due to their relatively anonymous work being seen as easily substitutable by AI.

Government Regulation and Intellectual Property

When mass unemployment begins, in any sector, there will almost certainly be lawsuits alleging that AI is infringing on the intellectual property of human creators.  Because AI “learns” by analyzing existing content, displaced human workers will argue that AI has infringed on their intellectual property without providing compensation.  This could spark landmark (policy-changing) court decisions regarding what AI companies can do in regard to using data sets to “train” AI or what companies can do regarding replacing human workers.

Royalties

If AI is designed to replace specific human workers, using formats developed by those human workers, should those human workers receive royalties?  Content creators of all kinds have arguably created frameworks, rubrics, and protocols for different types of work, and will argue that having AI use those frameworks constitutes a violation of intellectual property laws.  While employers likely cannot be forced to keep human workers on the payroll instead of AI subscriptions, they could theoretically be made by courts to pay royalties (a percentage of revenue).

Market Failure: Compensation Funds for Content Creators?

A worst case unemployment scenario, often called an “AI unemployment apocalypse,” will likely result in intense public pressure to create compensation funds for affected workers.  These unemployed content creators may be those who have not worked long enough to be considered eligible for royalties, and thus receive no revenue share.  Immense regulatory challenges exist under any compensation fund scenario, especially length of benefits.  How long can unemployed content creators expect to receive financial assistance?  At what point will former content creators be expected to retrain for new careers?

Regulatory Limits on AI

To prevent the layoffs of content creators, state and national governments may experiment with regulations.  This could include requiring companies to retain human staffers to review AI output for acceptability, thus shifting content creators to editors and reviewers.  Ideally, this could preserve jobs while allowing faster content generating, benefiting both consumers and taxpayers.  However, wages would likely decrease for former content creators as employers attempt to rebrand their new jobs as requiring less skill and effort.  Potentially, this could spark a regulatory tug-of-war between governments and companies as governments attempt to preserve wages in the face of AI efficiency and companies attempt to lower them to reduce costs and maximize profits.

Government agencies, especially massive consumers like public education institutions, could decide not to purchase works created with a certain amount of generative AI.  This would incentivize publishers, especially textbook companies, to retain human writers and artists in order to be eligible for lucrative contracts with public schools.  However, this is not a foolproof solution, as it would be difficult to enforce - companies would likely try to surpass any caps on generative AI usage (or ease below the floor on any human employee requirements) in order to reduce costs.  Thus, any government regulations on AI and/or human employee percentages will be perpetually challenged by cost-cutting firms, which will argue that government auditors have erred in declaring them out of compliance.