If you want to go more meta, I can point out that your comment is pattern matching to training models and applying them to humans vs. thinking from first principles and realizing that humans and models learn differently. That's exactly the issue with applying functional truths as truths.