the Harvard professor mentions that current peer-reviewed journal articles that get published now get an average of SEVEN total reads. What good is that doing anyone, unless is it some super obscure but important fact that is later picked up on?

That is a (somewhat) shocking figure if true. Such data are only sporadically available (for instance in that some journals publish download stats for .pdf versions of publications), and a better indicator might be citation metrics. For each time a publication is cited it has presumably been read at least several times.

With the increasing emphasis on citation rates in scientific career progression any publication that is not ultimately cited at least a couple of times per year was demonstrably not worth the effort of writing it in the first place. The scientific journals all know this of course and prefer to accept only the (anticipated to be) highest-impact manuscripts submitted to them. The lowest-impact stuff thus ends up filling journals which few people read and again in terms of career progression there is little reward in writing stuff that ends up in such journals (better to bin the manuscript and try again).