You could call it a machine learning machine. This brings up a very interesting question. "Machine learning" is a term composed of two words that when combined in that order result in a logical contradiction (i.e. if a machine could learn it would no longer be a machine). Does the additional word "machine" added to the term rescue it from from the realms of logical impossibility? Analogous to how a double negative (~, ~) becomes a possible outcome in logical operations. i.e. is a "machine learning machine" logically possible unlike machine learning which is obviously not logically possible. I would argue it is still not possible since the word "machine" when appended to the term "machine learning" still implies a machine is actively involved, and in fact required, for the learning to happen. It is a different machine perhaps from the original machine learning machine but still a machine nonetheless. Since a machine, no matter if it is machine learning or a machine learning machine, is not capable of learning, a machine learning machine is still a logical impossibility, just as is machine learning. Only of a human or of some non human animals can we say they are capable of learning, and neither a machine nor a machine learning machine are (logically) capable of learning since they are obviously not animals nor non human animals but in fact inanimate abiological objects. Rocks cannot learn and neither can machines or machine learning machines. A rock is as capable of learning as a machine is, that is to say not capable of it, just as no abiologial object known to man is capable of learning. Interesting thought experiment though. lol.