Howard Scott Warshaw designed Yars' Revenge and, less happily, E.T. the Extra-Terrestrial for the Atari 2600. The 1982 game was made in a matter of weeks to reach stores for Christmas, and its quality is still cited as one factor in the crash that took US game revenue from more than $3 billion in 1983 to around $100 million two years later. Unsold cartridges were later dug out of a New Mexico landfill.

At DevGAMM Lisbon this week, Warshaw gave a talk called "From Atari to AI" and drew the comparison himself. A glut of poor games opened the door for the crash, he said, because the machine made publishing cheap and everyone walked through. AI, in his view, does the same thing at a larger scale.

"So way more games came out, and good games were still being made, but they were way overshadowed by the crap," he said. "AI is going to have the same impact on a larger scale, because so many more people can now realise and create a video game than ever could before."

The bottleneck moves to taste

Warshaw's argument is not that AI makes bad games. It is that it removes the cost of finishing one, and finishing was the step that used to filter people out. "Game design is just like movie ideas and novel ideas," he said. "Most people who do it can't do it very well. And what AI is going to do is make that more possible for people who can't do it to do it."

He does not think the result is only noise. The same tool that lets someone ship a weak idea lets someone with a strong one build it without spending years on engine work, and he expects the emphasis in development to shift from technology to design. That, he said, would be good for the industry.

The practical consequence is a filtering problem. "The new challenge we're going to have is to have a way to winnow out the crap and get to the good games." Storefronts already face a version of it: half of the games released on Steam in October disclosed AI use, and most of them went unplayed.

No crash this time

Asked directly whether he expects a repeat of 1983, Warshaw said no. The earlier collapse was not caused by bad games alone but by investors deciding games might be a fad and pulling their money. "The crash just meant that investment left the field and nobody wanted to pay for doing games, because they weren't sure if games were here to stay."

Games are now embedded in daily life, he said, so the industry cannot disappear the way it nearly did. What he expects instead is "a reorganisation, where we need better filters to figure out a way to get rid of the garbage games and not waste time with that."

That is a milder forecast than the one Epic's Tim Sweeney and Romero Games co-founder Brenda Romero have made in recent years, both of whom reached for the 1983 comparison when describing the current wave of layoffs and closures. Romero later said on Bluesky that she meant how the industry feels to work in, not what its revenue would do.

What the filter would have to be

Warshaw was speaking four decades after the crash as someone who has watched curation move from shelf space to algorithms. A shop in 1983 could stock fifty cartridges. A storefront now lists tens of thousands, and the ranking system that decides what a player sees is the only shelf left.

His answer, that the industry needs better ways to surface good work, leaves open who builds them. Studios are already pulling AI voice work after players noticed, which suggests audiences will do some of the filtering themselves, loudly. Whether that scales to a catalogue growing by thousands of titles a month is the question his talk did not answer.