Making a video games require effort, repeated work. How could it not be possible? Developers are in the world-building business, so it’s easy to understand why game companies would be interested in creating AI. With computers doing the tedious work, a small team can scratch a map the size of San Andreas. Crunch becomes a thing of the past; games are released in a completed state. A new generation is coming.
There are, at least, two problems related to this matter. First, there is the idea of the hype itself – recalling the gold rush in the crypto / Web3 / metaverse – that, consciously or not, seems to be considering the creation of artists’ work as a way forward.
Second, there is a difference between these words and reality. Back in November, when DALL-E was appearing everywhere, the industrial giant Andreessen Horowitz posted a long-term analysis on its website showing a “game changer for AI” that could do everything from shortening development time to changing the types of titles being created. The following month, Andreessen partner Jonathan Lai wrote a Twitter thread comment on “Cyberpunk where more of the world/text was created, causing devs to shift from product development to higher-level tasks such as storytelling and innovation” and hypothesized that AI could help make games “better + faster + cheaper”. In the end, what Lai mentioned was filled and answers so angry that he wrote a the second thread admitting that “there are many problems to be solved.”
“I’ve seen, of course, false claims about things that are about to happen,” said Patrick Mills, senior director at CD Projekt Red, a software developer. Cyberpunk 2077. “I’ve seen people say that AI can build Night City, for example. I think we’re a long way from that.”
Even those who advocate artificial AI in video games think that many interesting stories about machine learning in the industry are going away. “It’s ridiculous,” said Julian Togelius, director of NYU’s Game Innovation Lab, who has written extensively on the topic. “Sometimes it feels like the worst kind of crypto bros that let the crypto ship sink, and then they come here and they’re like, ‘Generative AI: Start the hype machine.’
It’s not that native AI can’t or shouldn’t be used in game development, Togelius explains. It’s that people don’t realize what they can do. Of course, AI can create some kind of tools or write dialogues, but compared to text or images, the design is interesting. You can forgive the generators that produce a face with cute ears or old lines. But the game’s broken level, no matter how magical it seems, is pointless. He said: “It’s a scam, you have to throw it away or fix it manually.”
Basically—and Togelius has talked to several manufacturers—no one wants generators that work less than 100 percent of the time. They make games unplayable, ruining entire titles. “That’s why it’s so difficult to take reproductive AI that is difficult to control and put it in there,” he says.
