Sure, you may argue it doesn't "understand" what's it's doing the way humans do (seemingly)
I think one could argue that "It doesn't actually understand what it's saying, it's just parroting its training data" also applies to a depressingly large portion of the human population.
