Take-Two’s CEO says AI’s not in the business of making hits, ‘datasets by their very nature are backward looking’, but that doesn’t mean AI can’t be ‘super helpful’

Tech CEOs and other Silicon Valley sickos like to present AI as a gift that levels the playing field, making activities accessible to those who’d previously never have been able to rise to the occasion. But what if I told you that making things easier and making things accessible aren’t always the same thing.

I could tell you that while vibe coding may raise productivity, it rarely ends in a higher quality product, and just because a dog can now make videogames it doesn’t mean anyone has any more chance of creating the next hit game as they did years ago before AI. The tools have been around for a while, but the defining factor is whether people have the time, effort, and passion to put them to use.

But instead of warbling on for the rest of this article I’ll just hand it over to Take-Two Interactive’s CEO Strauss Zelnick, who recently discussed, during a podcast with David Senra, the myths around AI and the ways in which devs can use this technology in a way which is helpful.

“That was the thesis, that with AI anyone can make a videogame,” Zelnick says. “Anyone could make a videogame last week, anyone could make a videogame five years ago, the technology is readily available. You know how many mobile games get released a year? Thousands. You know how many hits get made a year? Zero to five. You know who makes them? Thank you very much, we do.

“You don’t need this new technology to make assets, that already exists, it will be quicker to do it. But speed isn’t the issue. The technology, prior to AI, existed to clone GTA, but it won’t be GTA. It’d be a clone of GTA; clones don’t sell. All hits are by their very nature unexpected. That’s the most important thing to take away. Things that are data-driven in their entirety can’t be unexpected, but that doesn’t mean that AI isn’t super helpful.”

The crux of the issue is that for a videogame to be a hit it has to offer up a new experience. But “datasets by their very nature are backward looking, creativity by its very nature is forward looking”. New ideas can be based on existing data, which is where datasets can help inform new ideas, but there needs to be at least one element that does something different to everything else, otherwise there’s no incentive for players to try it out. Think of games like Palworld, Marvel Rivals, or even Arc Raiders—these games took an existing idea and added something new to it.

“I would love to say that AI will make it easier, quicker, and better to make hits because who would benefit more than we,” Zelnick adds. “We’re in the business already, we own IP.” But the issue still stands that AI creates “derivative property” and that’s simply not as compelling. “AI is really great at asset creation but hit creation isn’t asset creation, asset creation is a necessary but insufficient condition for hit creation.” So no, the next hit videogame won’t come from the hallucinations of an LLM, it’s reserved for the hallucinations of some tired dev.