Hello everybody! A major technology company executive is talking in this, the year 2026, which means it’s time for more wrangling about generative AI tools. Take-Two Interactive CEO Strauss Zelnick has spoken at length about why the latest genAI tools aren’t going to turn everybody into a game developer and put companies like his out of business, arguing that there are already a lot of ways to hasten or automate bits of game development, used by many thousands of games, and yet only a “handful” of those are “hits” on the scale of GTA 6.
“I was kind-of stunned by the market’s reaction,” he told Chris Dring of The Game Business (HELLO DRINGO) in a new interview, “because its reaction was somehow seeing it as a threat to what we do, when it’s quite obvious that creation tools are beneficial for our industry. I think the bear case for big entertainment companies is somehow that AI tools will mean everyone can create hits, but that doesn’t stand to reason.”
Quick explainer: a “bear” market is a period of falling share prices. Apparently it’s got something to do with 18th century traders dealing in bear skin before they’d even killed the bear. I feel like there’s a parallel to explore with NVIDIA selling a new face for Grace Ashcroft, but I’ll let you join the dots as you see fit.
“These tools may help you create assets, but that won’t help you create hits,” Zelnick explained. “There are loads of assets out there now. It doesn’t matter if you push a button to create an asset, or it takes you six weeks, at the end of the day, you have an asset. And thousands of mobile games are launched every year, and there are only a handful of hits.
“Equally, you can create assets that might look like a big release, that might look like NBA 2K or EA Sports FC. But creating a hit of that magnitude is a completely different animal and does require human engagement and creativity. You can create assets that might look like NBA 2K or EA Sports FC. But creating a hit of that magnitude does require human engagement and creativity.”
The argument here echoes what Embark told us last year about today’s sycophantic chatbox malarkey being only the latest addition to a long history of machine learning applications and other kinds of automation. I’ve suggested in the past that the key difference with older ML tech is that the biggest genAI tools are marketed as subordinate beings, not just tools – fawning co-pilots and magical familiars and hideous attic twins who want you to eat “tablesoposs”, grown from undead composites of non-consensually provided human data at significant resource cost. Mike Cook’s take is older but better-informed and less, er, ranty.
Zelnick offered a few instances in which he feels generative AI can be helpful for game developers, such as storyboarding or exploring possible plot points, arguing that in these cases, it’s “more effective than the tools we’ve had before, like doing an internet search.” But he’s dubious about high-faluting gadgetry such as Google’s Project Genie, which is being hyped as a means of conjuring whole GTAs from nothing.
Asked by Dring whether “tools like Project Genie might level the playing field for those looking to create games like GTA”, Zelnick replied: “Not even the littlest bit. There’s already plenty of technology out there that allow people to create video games, and as a result, thousands of video games are created every year, and yet the hits all cluster among the large entertainment companies, almost entirely, and now and then, an indie, which is generally speaking well-funded and pretty robust in and of itself.
“The notion that somehow new tools would allow an individual to push a button and generate a hit and bring it to many millions of consumers around the world, it’s a laughable notion,” he went on. “It’s just never been the case with entertainment. Right now [in music] there are programs that allow you to put out a prompt and get a professionally recorded song spit back out at you. It sounds like a song, but I defy you to listen to it more than once. It’s great to send as a greeting card to your partner on their birthday, but that’s about it.”
Zelnick reckons genAI tools might help the “creator economy”, aka amateurs making games mostly for free on platforms like Roblox, but this won’t “supplant” professional gamedev, insisting that the history of the entertainment business shows that as new technologies surface, they “stand alongside the pre-existing businesses.” He illustrates this by observing that vaudeville-style live shows are sort of still a thing in the age of Netflix. I am not going to get into the weeds of this argument. I forgot to bone up on 19th century variety entertainment while having my cornflakes this morning. But I will say that it’s been a while since I enjoyed live piano accompaniment for a silent movie.
Take-Two Interactive have “hundreds” of genAI “pilots and implementations” on the boil, as Zelnick revealed in February, but the company aren’t using any of it in the new Grand Theft Auto. “Specifically with regards to GTA 6, Generative AI has zero part in what Rockstar Games is building,” Zelnick said at the time. He’s long taken a guarded view towards genAI, at least with regard to Take-Two’s own bottom line, observing in November 2023 that “in certain instances, they’ll help us do things we haven’t been able to do before. But it’s going to allow that for our competitors as well.” Zelnick has also said more recently that Take-Two don’t see generative AI as an excuse to cut jobs, noting that generative AI tools are “backward-looking” in their reliance on existing data for ‘training’, “and none of that replaces forward-looking genius”.
