Generative AI tools are the Clipart of the 2020s, and that’s a good thing
Here is an actual statement from 1998:
“Clipart collections will be the end of designers.”
To understand that statement, you need to remember (or imagine, depending on your age) a world where no internet image search existed. Stock photos came as print catalogs that you would order disks from, and later as CD collections. Getting stock images for your digital media was a time consuming and expensive thing.
And then clipart came along: Cheap, pre-made images and illustrations used to illustrate any medium, paid as collections. By the mid-1990s, there were millions of clipart collections available, prompting Microsoft to integrate their own collections directly into Microsoft Office.
They were EVERYWHERE.
It seems laughable today, but people believed that these collections would replace the need for designers. Because who needs designers when you can have everything on a CD for cheap?
And today, I read statements like this:
“Generative AI will be the end of [designers | writers | developers].”
It just seems remarkably familiar.
Case study: Games
Generative art is nothing new in games. It has been used under the term “procedural generation” in game development since the 1980s, for example to create a vast number of star systems for Elite. Procedural generation was an effective way to save memory in early computing and came in many different forms, some simple, some very elaborate.
Since then, procedural generation has been periodically “rediscovered”. Here is for example Will Wright talking about it in 2008 for his game Spore:
But the point is: While generative art can be used to solve a specific set of problems in game development, it’s not a silver bullet. It’s actually quite bad when used generically:
The quality of the generated assets will not change this, even if paired with generated storytelling. Game developers know this.
Instead, procedural generation was integrated into engines and tools to allow developers, artists and designers to be better: Faster workflows, new mechanics, the ability to create at a much larger scale.
Generative art in games might not be clipart, but they are Clippy.
Note that this didn’t mean that game development got cheaper. At this point, the games industry is a Red Queens Race: If a tool makes art generation 10% more efficient, then development doesn’t get 10% cheaper. It means that game studios will still use the same amount of money and resources to make the games 10% better / bigger / longer / complex. In fact, the overall cost of making games has been steadily increasing.
On the democratization of art
But clipart does have a purpose. Of course, those 1990s clipart collections look silly and dated now, but back then they allowed me and many others to make our digital creations much more personal: School documents, invitations, diaries, notebooks, work documents. They made creating digital documents a little more fun and engaging. Not everything looked better, but it did look more interesting.
What clipart did was to democratize digital art. Make it accessible to everybody. And it made the web more accessible through platforms like Geocities that combined web design with clipart. It turned out that designers didn’t lose their jobs, but it elevated everybody’s creativity.
And so here we are with ChatGPT, DALL-E, Midway and others. They will allow everybody access to more, better, and cheaper media (text, image, video, audio) to extend their digital creations — professionals and everybody else. It’s the next evolution of clipart. And that’s amazing and important.
But we must also acknowledge that the original clipart collections were a simple business model: You paid designers to create art for you, then sold it. Today, we need a discussion about how this will work with generative AI, specifically with the usage of copyrighted training material without obtaining a license.
As for the breathless headlines about people and industries losing their jobs, here is a fun experiment: Every time you see the words “Generative AI” in an article, replace it with “clipart generator” and see if it still makes sense.