id Software's former co-owner says they absolutely would've used gen AI back in the day: 'John Carmack would be like making Skynet or something'

Generative AI is a hot-button topic in the industry, and for good reason—it's been kinda destructive, whether that's jacking up hardware prices or putting people out of a job—and while there have been some fairly mundane use cases that seem helpful, it's been hard to escape the sheer sense of mania that's built around its use. Or the bad PR, juiced out of the sheer exhaustion we've all had from executives promising the matrix or something.

So, naturally, it came up in a conversation with Tim Willits, the CCO of Saber Interactive and also the co-owner id Software from 1995-2019. Speaking to PC Gamer's own Joshua Wolens at Gamescom 2026, he notes that: "As most people know, we are experimenting with AI in future games. We don't have it here right now, but we announced [Rideshare Simulator] and I want to clarify things because there's some misconceptions."

If you're wondering why we're talking about a Rideshare sim, here's some context: Saber landed in hot water earlier this month over it, because it seemingly delayed its AI disclosures on Steam for a while, and was exposed as using AI by a former writer who claims they were booted off the team in favour of it.

That same writer was badmouthed by Saber's CEO Matthew Karch, who later apologised. Willits says that Rideshare's main campaign mode and all of its side missions are written by humans, and the AI components are only used for an optional game mode called Free Ride: "We're going to use AI for the passengers, so we can deliver new stories.

"So that is only possible if we use AI for that—this is an experiment. We at Saber want to figure out how we can use this technology, so we're doing it at a small scale to try to learn. We're like everyone else, we're just trying to figure this out. But there is no generative AI in our big titles. No Hellraiser, no Turok, none in Stuntman."

I feel like an asterisk has to hang over that "only possible"—while, yes, churning an infinite stream of dialogue for an optional mode is only possible with generative AI, a free ride mode with enough conversations for it to not grow stale absolutely could've been cobbled together with actual writers and voice actors. Sure, you'd run out eventually, but I don't think anyone's hungry to make Rideshare Simulator their forever game.

Details

Still, Willits emphasises: "We want to experiment, and I do believe that there will be a company—I don't know who—but some game development company will create a game that uses AI in such an amazing way that everyone will be like, 'That's it, that's what we need to do'." He then goes on to joke "that someone needs to come up with the 'Half Life 2 of AI', something that's so amazing that it just launches some type of killer app."

I remain characteristically unconvinced, though in fairness to Willits, I do at least think that—if generative AI does find a use as the core of a videogame—it'll probably be a new genre of its own, rather than trying to create an infinite RPG with infinite NPCs. An act which would just shake out to make something unappealing and meaningless.

The thing that makes me cynical is we keep watching money being burned and datacenters being built precisely because the tech doesn't really seem all that useful for anything other than busywork. When it takes centre-stage, it produces voice acting that's bland, writing that's terrible, and art devoid of direction or memorability, even as it gets more technically proficient.

I'll eat my artificially-generated hat with a side of humble pie if Willits' "Half-Life 2 of AI" game materialises, but a lot of people have swung at that ball and missed, and each time I feel like the bat is a little smaller. Still, Willits says that id Software probably would've used generative AI if it was a thing back in the day—because the studio was always at the cutting edge.

"Yes! Oh, of course. I mean, look at what [id Software] has done. In 1999, when we released Quake 3, it required a graphics card, and we got so much hate for that. People said we were going to ruin the industry. Nobody was going to spend $100 on a video card, and it was the dumbest thing that anyone ever did … I know for sure back in the day if it was available, John Carmack would be like making Skynet or something."

Still, to be fair to Willits, he seems to share my view that the game'll have to prove that generative AI can change the industry itself, rather than via the promises of a CEO somewhere: "The games will speak for themselves, even in our experiment with Rideshare, I encourage people to check it out when it comes out, play it themselves. They may be like 'oh, this is cool! I like this'. And if they don't like it? Fine. And but the games speak for themselves."