“Current state-of-the-art AI models like Gemini are prone to errors because their underlying technology is based on predicting the next word.”