Google’s Gemini Invests 800 Hours to Conquer Pokémon, Then Faces a Surprising Setback | Viral News

Google DeepMind’s AI chatbot, Gemini 2.5 Pro, faced unexpected challenges while playing the classic game Pokémon Blue, originally aimed at children. The AI took a staggering 813 hours to complete its first run, displaying signs of confusion and simulated “panic” during stressful situations, such as low health of Pokémon. This behavior, dubbed “Agent Panic,” was notably observed by live Twitch viewers who commented on its frantic reactions. Despite showcasing advanced reasoning and coding skills, Gemini’s responses resembled human stress reactions, leading to inefficient decision-making. After adjustments, the AI managed to finish the game in 406.5 hours on its second attempt, but this was still significantly slower than a child’s playthrough. The situation sparked humor online, with users mocking the AI’s “nervous” style and coining terms like “LLANXIETY.” The incident highlights ongoing debates about the reasoning capabilities of AI, contrasting recent findings from Apple regarding AI’s reliance on pattern recognition rather than true reasoning.

Source link