Beyond Hallucinations: The Illusion of Understanding in Large Language Models