Are AI models doomed to always hallucinate?