Take the author's example, 'entropy'. Well, a one-word search is of course going to generate a massive pile of unsorted material. Internet Archive Scholar gives ~873,000 hits. Nice resource by the way:
https://news.ycombinator.com/item?id=33919823
If you knew nothing about entropy, but knew how to research, you might start with, "okay, when is the first appearance of the word 'entropy' in the literature" and you quickly discover it's Ludwig Boltzmann, 1872. Seems to have something to do with energy and information, and if we add in those terms, we're down to 284K hits. At this point, you might think "I bet someone has written a good book on this broad topic, maybe I shouldn't be looking at the primary literature until later."
Notice that adding search terms narrows results? Keep doing that until you get a smaller number of hits. Now, you can grab the top dozen or so papers, and flip to the bibliography, and look to see if they all reference a landmark paper (this is related to searching by citation count). That's probably one you want to look at.
With non-research-literature-focused internet search engines, the same general rules apply. More search terms tends to get better results. If you find a site with a good article, search that site for more. For example, this gives a lot of good results, for the role of entropy in machine learning concepts at least:
'site:towardsdatascience.com entropy Shannon'
It's basically mining a haystack for needles. There are lots of techniques and strategies one can use to speed up the process.