653 karma · joined October 2, 2016
The point is that researchers have their little niche and they try to make contributions in areas adjacent to it. It's unrealistic to think everyone publishing papers understand all the other papers, particularly in such a cross-disciplinary field like ML. There's also a big gap between a researcher deep in their career and a student fresh out of a masters program.
It's also hard to transition from someone who's used to reading and understanding textbooks to someone who's often reading really technical research and understanding very little of it at first. You just have to push through and have confidence that you'll eventually learn enough to make a contribution. That's what it means to "become an expert"--you start off as not being an expert and then beat your head against the wall for a few years until you bootstrap your way out of it. And if you want to do it in a reasonable amount of time, you should probably choose something you have some of the fundamentals for.
On the other hand, you might expect that new discoveries by nature have less data since data is likely more expensive for brand new research, and by extension a lower likelihood of meeting these sorts of stringent statistical requirements. Decreasing the p-value threshold may be counter-productive if we dismiss legitimate new discoveries due to essentially economic constraints with data gathering, which would have the impact of making it less likely to get funding to pursue the problem in more depth, thereby slowing the advance of discoveries.
My theory is that paid per click advertising/clickbait is making it increasingly difficult to say anything that gets attention on the internet, so people increasingly use hyperbole. "Internet journalists" will trump things up to make it sound more exciting, feeding the hype machine. Those that don't succumb to the practice are just not heard. There are, of course, exceptions. But a well-reasoned, calmly stated argument from a nobody is never going to go viral in the way that an outlandish story from that same nobody could. i.e. Seth Godin doesn't have reach just because he's interesting, it's also because he's been working at it for over a decade, but any idiot can make up some fake story that goes viral and impacts the public consciousness.
I suspect the concept is already present in economics. It's certainly under the hood of many statistical modeling techniques. For instance, some tree learning algorithms use entropy reduction measures to decide how to split the data for the tree. You can even derive the normal distribution by merely maximizing the entropy under the constraint of known variance.