You are misrepresenting the article, which is very clear about the tests.

And it's remarkably valuable to show people that AI cannot be trusted with things they understand, so they are less likely to use it to research things they don't understand.