Renowned game developer Ken Levine has expressed his views on generative AI, his latest game, which involves high stakes, and why he believes we still do not fully understand the essence of video games.
In an interview with GI.biz, Levine mentioned that he regards generative AI as a “highly significant” tool that should not be underestimated. However, he pointed out that it has numerous limitations in its current form, including a lack of continuity.
“When you observe Sora, the ChatGPT video generator, you witness a woman strolling down the street in a beautiful scene—but if she were to pivot and walk backward, it wouldn’t recall where she had been,” he stated. “It does not currently grasp continuity, though that could change.”
While many worry about AI replacing human jobs, video game leader Andrew Wilson from EA believes that will occur, Levine contends that AI still has a long way to go when it comes to creative storytelling.
“So despite all the apprehensions about AI, have you witnessed it generate a solid 20-page screenplay yet? From scene to scene? It does not know how to perform that,” he remarked.
Currently, the beneficial aspects of AI for game development are primarily centered around technical functions, Levine noted. “For example, training your bug database to ascertain how many bugs you have in specific circumstances. But what it cannot accomplish is to narrate a truly engaging story with a three-act structure, or even relate multiple scenes. It becomes extremely perplexed,” he said.
Levine indicated that no generative AI was utilized in crafting his newest game, Judas, apart from elements concerning bug databases.
“At this moment, I’m not overly impressed regarding game development. I’m sure there will be advancements [in the future], but I’m not particularly concerned about it yet in the context of ‘It’s coming to take everyone’s jobs,'” he stated.