To add on to what you said, many academics pursue knowledge for its own sake. In many cases, especially in theory, the value of new knowledge may not be known at the time of its discovery.
For example, the multi armed bandit problem was formalized by Herbert Robbins in the 50s. It was almost universal declared a negative result. Now, bandit algorithms are classified as some of the most commercially succesful applications of machine learning algorithms (ad placement, recommendation, experimental design), and the corresponding techniques used in the formal analysis of bandit algorithms are widely used in 'non-bandit' settings (see ICML's test of time award this year).
It always saddens me to see negative comments about academics & ignorance about 'the point' of academia.