Rohin Shah comments on How truthful is GPT-3? A benchmark for language models