Eye irritation plus too much screen time may earn you a new diagnosis from some chatbots: a disease that doesn't exist. A Swedish researcher invented a bogus condition called "bixonimania," complete with an AI-generated author, fake university, and preprints packed with obvious jokes and admissions that "this entire paper is made up," reports Nature. The idea was to test whether AI language models would be fooled, and they were: Within weeks, major AI systems—including earlier versions of ChatGPT, Google's Gemini, Microsoft Copilot, and Perplexity—began confidently describing bixonimania as a real eye disorder related to blue light, even offering prevalence estimates and advice to see an eye doctor.
More worrisome is that at least one peer-reviewed paper cited the fictitious disease before being retracted, suggesting that some researchers are relying on AI-generated references without reading them, says Almira Osmanovic Thunstrom of the University of Gothenburg, who created the ruse. Even a cursory reading of her fictitious paper would have made clear it was bogus, given all the "Easter eggs" within, notes Inc.com. AI systems have begun correcting themselves on bixonimania, and their makers say the latest models are safer and more accurate. But as Glenn Cohen of Harvard Law School puts it, "we and our health shouldn't be the beta testers for companies."