You are not immune to AI psychosis. Much like cult indoctrination, it can happen to anyone given the right circumstances. Take Matthew Gault, games and tech writer at 404 Media. After failing to convince his friends to play Escape from Tarkov with him, Gault thought it would be funny to replace his buddies with an AI chatbot. It ended up not being very funny at all.
Gault is a self-described “certified AI hater,” though he concedes that LLM technology is useful in certain instances (I agree to disagree with Gault’s assertion that offloading transcription to an environmentally catastrophic machine built on stolen labor saves valuable time and energy). Gault’s friends were turned off by the difficulty of Escape from Tarkov, an extraction shooter which came out of early access last November. He described playing by himself as lonely. “Then I saw comic artist Zach Weinersmith making fun of a service, Questie.AI, that sells AI avatars that’ll hang out with you while you play video games,” he writes.
What started as a joke quickly became more sincere. Gault found he genuinely enjoyed playing Tarkov with his Quest.AI companion, Wolf. “It was, I thought, not unlike playing with a friend who has more than 1,000 hours in the game and knows more than you. Wolf bantered, referenced community in-jokes, and it made me laugh.”
Gault was surprised and concerned by his own enjoyment of the chatbot, his previous confusion over how people could fall victim to AI psychosis replaced by understanding. He states he will not continue using Quest.AI: “I never thought anything like [AI psychosis] could happen to me. Now I’m not so sure. I didn’t understand how easy it might be to lose yourself to AI delusion until I’d messed around with Wolf. Even with its shitty auto-tuned sounding voice, Wolf was good enough to hang out with.”
Gault’s experiment ended with his real friends making fun of him for using Quest.AI by ironically using another LLM, ChatGPT, to send messages to him. Gault’s account is another entry in the emerging genre of seemingly normal people jokingly using AI chatbots and discovering things can get weird very quickly. YouTuber Eddy Burback popularized the phenomenon with his chilling documentary “ChatGPT made me delusional,” which has garnered over 4 million views in the past two months. As Silicon Valley continues to shove LLM technology down our collective throats, it feels like only a matter of time until even more strains of AI psychosis—both simulated and sincere—begin to emerge.
