Sure, TR and Wired are fluff. Except that the article links to an arXiv paper. This is not a standard TR article. Moreover, you can't present scientific research in half a page without making it fluff. That, too, is an interesting information-theoretic problem.

"That would be interesting if they weren't using so poor an approach."

You missed the forest for the trees. In case you didn't notice, the people who wrote the paper are electrical engineers. They used the information-theoretic approach that is used in communications theory to a problem outside the traditional scope of application of the theory. Sure, natural language is hard, but if you start your research by focusing on all the little details you will get nowhere. To me, the paper looks like a first shot at a difficult problem. If you can do better, I would love to hear about it.

"...it sounds like the parameters of the study were determined by someone with very little understanding of how ASL actually works."

That is not the point. The point is that someone who does indeed understand how ASL works can read the paper, find out what is missing and build on it. No one knows everything, and inter-disciplinary work is very hard. Your criticism is too hard, because no one ever built a theory in one single iteration.