Is there a huge difference between physics PhDs and computer science PhDs in this argument? Both are overspecialized in their research, but both also have a wide range of skills in programming, abstract problem solving, writing, communication, and so on. Both will likely have high-level knowledge of a variety of machine learning methods, except the CS PhD that actually published machine learning research who will know a bit more about some very specific topics.
(I'm a CS PhD and I know many physics PhDs as well.)
(I'm a CS PhD and I know many physics PhDs as well.)