My background is that I started out as a scrappy unix sysadmin, and video engineer (film and TV). My approach has always been to focus on getting a "good" result that is guaranteed to work vs say a "moonshot" approach. When I worked in Live Television, often I was called to fix a problem with an expensive system (100k-1million) that needed an answer in an hour before they went on air. I think this shaped the way I work on problems today. My first approach is to get something working, then iterate from there.
The criticism of this approach could be that this is in direct contrast to Google. They mention "moonshots live in the gray area between audacious technology and pure science fiction". I am glad people are thinking that way, and working on those problems, but I think more like Scotty on Star Trek. I like to be given hard tasks that need to get solved in hours. So this is really the thinking behind the book, how can show people ways to get quick results in AI vs moonshot approaches. I think both are very valid approaches, but I gravitate toward the pragmatic version.
After some pecan pie, you might want to cleanse your palatte with this tiny ad:
Devious Experiments for a Truly Passive Greenhouse!