He doesn’t think asking ChatGPT to make a better GTA will work, though.
Audio player loading…
Over the past few years of listening to gaming executives talk about the industry during earnings reports, I’ve found Take-Two CEO Strauss Zelnick to be one of the more circumspect bosses in the business. He’s been relaxed about the rise of Xbox Game Pass, noting whenever he’s asked about it that subscriptions are still a small part of the business, and he didn’t declare NFTs to be the future of commerce five minutes after finding out about them. It’s a low bar, but it’s mildly refreshing to hear a tech exec respond to buzzwords with something other than a golden retriever’s unconditional enthusiasm. So it went on Monday when Zelnick was asked during an investors call what he thought of advances in AI tech, at least at the start of his response.
“You know I’m the first person to be skeptical of other people’s hype,” said Zelnick. “And I would like to note that AI stands for ‘artificial intelligence’ and there is no such thing as artificial intelligence.”
The CEO thinks some of the hopes and fears AI has inspired are overblown, reasoning, for instance, that the handheld calculator didn’t stop kids from learning math, so writing bots like ChatGPT won’t mean the end of essays. And he doesn’t think Take-Two’s studios, which include Rockstar, Firaxis, Cloud Chamber (the new BioShock developer), and Hangar 13, are in any danger of being replaced by a bot that spits out games.
“And no, [AI is] not going to allow someone to say, ‘Please develop the competitor to Grand Theft Auto that’s better than Grand Theft Auto,’ and then they will just send it out and ship it digitally and then that will be that,'” said Zelnick. “People will try, but that won’t happen.”
That tempering aside, though, Zelnick does think AI research is a big deal, and he’s interested in using it to make games. Rather than making development cheaper overall, AI tools will “just raise the bar” for the industry, he says.
“We are ushering in a very exciting era of new tools,” said Zelnick on the call, “and they’re going to allow our teams and our competitors’ teams to do very interesting things more efficiently, so we’re going to want to do more. We’re going to want to be even more creative.”
Even in this research stage, controversial AI image generators have been used to produce inspiration for game art and even to generate assets directly, but Zelnick is certainly also referring to less sensational applications for machine learning algorithms, the strength of which is that they can be harnessed to solve all kinds of problems, from image upscaling to self-driving car navigation. As one example, Ubisoft already uses a machine learning-powered animation tool called Anything World for prototyping.
Another interesting application for machine learning is in the training of AI opponents—like Google’s AI StarCraft player—although that’s also opened up uncharted territory for cheaters, something Rocket League players recently learned firsthand (opens in new tab). There tends to be a catch with each interesting advancement in the machine learning space, but for better or worse, even sensible ol’ Zelnick thinks we’re on the fringe of a new era. Not one that will replace this era’s big development teams, but one that will see them to do more, as he put it.