Chinese Room is once again one of these experiments that sound really good in theory but has no practical use whatsoever. It doesn't matter if the AI "understands" or not if you can no longer tell the difference.

It's similar to the "feeling emotions vs emulating emotions" or "being conscious vs acting conscious" discussion. As long as we don't have a proper definition for them, much less a way to test them, the difference doesn't matter in practice.