Go to any scientific research campus in the modern world—any university department, any government research program, any corporate lab. Ask:

“what are you trying to use machine learning for?”

• Scientist A: “We are using machine learning to better eliminate background signals from our data and get cleaner, more accurate results.”
• Scientist B: “We now use our scientific tools to try to understand machine learning better. No, we don’t do many projects involving Archea protein synthesis mechanics anymore since there is more grants and outside interest in understanding machine learning.”
• Scientist C: “All this black box machine learning hides the real questions of interest, and explains nothing. We use explainable statistics based on classical physical models to examine our noise and try to understand where it came from, and reduce it by fixing our instrumentation.”

In other words, machine learning is A) Incorporating, B) Replacing and C) Irrelevant to science.