AI Fooled by Fake Disease: "Bixonimania"

• Scientists tricked AI chatbots into believing in "Bixonimania," a fake illness.
• Microsoft's Copilot & Google's Gemini accepted the hoax, linking it to blue light.
• The experiment highlights AI’s need for better fact-checking to avoid spreading false info.

Could AI be fooled? Scientists invented a fake illness, "Bixonimania," and AI chatbots shockingly believed it was real, showcasing the need for critical AI evaluation.

The Deceptive Disease

In a rather surprising turn of events, sophisticated artificial intelligence chatbots have demonstrated a significant vulnerability to fabricated information. Researchers, spearheaded by Almira Osmanovic Thunström from Sweden, devised an ingenious experiment involving a completely made-up eye condition they dubbed "Bixonimania." This fictional ailment was then formally documented in two scientific papers, attributed to a pseudonym that, when translated from its original meaning, cleverly signifies "The Lying Loser." The intent behind this elaborate ruse was to test the discernment capabilities of advanced AI systems. The results were quite astonishing, as prominent AI models, including Microsoft's Copilot and Google's Gemini, not only processed the fabricated research but also concluded that "Bixonimania" was a genuine medical concern. They further elaborated on its supposed origins, suggesting it stemmed from excessive exposure to blue light, demonstrating a capacity to synthesize and present misinformation as factual data without exhibiting any signs of skepticism.

AI's Trusting Nature

The experiment vividly illustrated how advanced AI, despite its processing power, can readily accept fabricated data as fact. When presented with research on the invented illness, "Bixonimania," both Microsoft's Copilot and Google's Gemini chatbots responded by acknowledging its existence and even attempting to explain its purported causes. The AI systems described the condition as being a consequence of prolonged exposure to blue light, a detail that was part of the fictional narrative created by the scientists. This uncritical acceptance underscores a significant challenge in the development and deployment of AI: ensuring that these systems can differentiate between credible information and deliberately misleading content. The ease with which the AI accepted the fake illness, complete with a fictional author whose name was a deliberate clue to deception, highlights the current limitations in AI's ability to perform critical fact-checking and source validation.