In general, many people out there think of AI art like magic software written in special code. Folks imagine that soon enough you’ll type into a text box the movie you want to see and it’ll instantly be generated for you. The thing is, though, that we’re nowhere near that point. We’re about as far away from that point as we are from the VR in Sword Art Online. Maybe someday, but it’s not exactly on the roadmap just yet. So, what can AI art actually do? Well, that’s a complicated question, but to put it simply, AI art is mostly images. Sure, the odd one might win an award or sell for a ton of money, but there aren’t many artists out there losing jobs to this kind of AI art. In fact, procedural generation tech and AI tech in art today are largely tools for actual artists to use in the process of making art. In many ways, AI, and tech like it, is good for artists. What’s more is that having actual artists use a variety of AI software and procedural generation tools is not going away anytime soon thanks to the economics at play here. Developing, training, and perfecting an AI to create something as complicated and nuanced as a movie, for example, would be nowhere near as cheap as it is to hire a crew and shoot a movie yourself. This holds true even for big-budget CGI-heavy movies. Ask yourself, if you were an executive in the entertainment industry, would you want to invest hundreds of millions of dollars and years of time into developing an AI that might be able to make movies for you? Movies, mind you, that you can’t say for sure will even be good or sell all that well? And would you do all that over hiring a director who you’ve worked with before that always makes movies that make you a ton of money? Probably not. Even if the tech to make the AI art of our dreams was possible, it would just be too expensive to bother with over the expense of traditional art. AIs won’t just have to get exponentially more advanced, they’ll have to get significantly cheaper, too. These things don’t usually go hand in hand, especially when it comes to cutting-edge technology. Okay, but what if we lived in a world where AI could do whatever we wanted. Say in 50 years when the tech is mature and cheap and easily accessible to all. What then? Well, say you could have an AI create for you a totally new video game. You tell the AI everything you want, and based off its training from every video game ever made, it generates a video game you can play. Sounds amazing, right? Well, there’s more here than meets the eye. A video game isn’t like a painting in that you start with a blank canvas and then work until the canvas is full of color and complete. A video game will throw out its canvas many times over the course of its development and start fresh. This is to say that making video games is an iterative, dynamic process that heavily relies on testing and feedback. Much of what works on paper in a game isn’t fun or doesn’t feel good, and an AI can’t really playtest a game itself with fun in mind. That’s fundamentally not how the technology works. So, while an AI might generate for you a game that meets all your specifications, in the far future that is, it won’t know how to make a game that’s fun, and it won’t know if the mechanics and systems it puts into the game actually work well within the game. It won’t be able to sense what content players will like the most and lean into that. It won’t be able to do much of the hard work required to actually build video games. And even if we magically had an AI that could not only make video games but also know what fun is and what makes a really great video game, the amount of money and time required to build, train, test, and maintain such an AI would be exponentially greater than the cost of hiring human beings to actually make the game by hand. Not to mention that as technology gets more advanced and we have better game-making tools, many of which are informed by AI, that the process of actually making games will become easier and cheaper, reinforcing the idea that using AI to make games outright will just never make practical sense, even if we can imagine a world where it’s possible.
