Dan Houser, the cofounder of Grand Theft Auto publisher Rockstar Games, cast AI in gruesome terms last week by likening it to when farmers in the United Kingdom inadvertently spread bovine spongiform encephalopathy — commonly known as mad cow disease — by feeding meal tainted with the flesh of diseased cows to healthy heifers.
“AI is going to eventually eat itself,” said Houser during an interview last week at the “Chris Evans Breakfast Show” on Virgin Radio UK while promoting his science fiction novel “A Better Paradise.” “The [AI] models scour the internet for information, but the internet’s going to get more and more full of information made by the models. So it’s sort of like when we fed cows with cows and got mad cow disease.”
Houser’s pointed remarks are striking in the context of the larger video game industry, where huge numbers of workers have been laid off while CEOs have gone full hog into AI to cut cost, boost profits, and capture the next generation of gamers; case in point, Tim Sweeney, CEO at Epic Games — which makes the smash hit Fortnite — has bet big on AI and even became publicly irate that gaming platform Steam is labeling games that contain AI-generated assets.
But it’s uncertain what the future will look like for an AI-driven gaming industry, and if the most optimistic predictions that it’ll enable dynamic and immersive new types of storytelling will pan out; others are doubtful, expressing fears that the quality of games will suffer and lose an element of the human touch, becoming soulless objects with janky environments inhabited by twitchy NPCs.
That’s what Houser fears: a scenario that mixes together dead internet theory — the notion that the internet is full of AI bots and AI-generated content — and the idea of model collapse, which is when AI models trained on AI-generated content start to degrade in quality. Both could lead to an internet that gets eaten inside out — which isn’t too different from what happens when a cow catches mad cow disease, and starts acting weirdly and losing bodily functions before ultimately dying.
And we haven’t even mentioned the possible legal and ethical questions that the tech could invoke in gaming — like an AI-powered NPC in a video game encouraging a user to kill themselves, which sounds like science fiction but has happened numerous times with consumer-facing chatbots like OpenAI’s ChatGPT.
Strauss Zelnick, the CEO of Take Two Interactive — the parent company of Rockstar, which Houser left in 2020 — took a skeptical tone on AI last year, saying he doubts the tech is “going to make things cheaper, quicker, better, or easier to make hits” and that it should be treated as a tool with the human developer as the ultimate arbiter.
“The machines can’t make the creative decisions for you,” he said.
That echoes Houser’s view, even with his pessimistic opinions about AI eating up the world; he admitted in an earlier interview with the UK’s Channel 4 “Sunday Brunch” show that his new firm Absurd Ventures is “dabbling in using AI.” But in that interview and the one with Virgin Radio UK, Houser cautions that AI isn’t the magic wand that its most ardent proponents claim, and is only useful for certain, narrowly defined tasks.
“The truth is a lot of it’s not as useful as some of the companies would have you believe yet,” he said on Channel 4. “It’s not going to solve all of the problems.”
More on AI: CEO of Fortnite Maker Furious That Steam Is Labeling Games With AI-Generated Assets