Monday, May 16, 2016

Writing technical books: the craft of simplifying ideas

I am in the process of writing a fairly broad book on setting up a laboratory for cognitive technology / artificial intelligence. I don't find writing to be easy but I enjoy the process a lot. The main problem that I have is removing unnecessary materials and ideas, leaving just enough so readers can understand the core ideas and experiment with these core ideas using example programs. Unnecessary complexity makes understanding difficult and generally does not help a reader solve their specific problems.

If a reader understands the core ideas then they will know when to apply them. It is easy enough, when working on a project, to dig down as necessary to learn and solve problems but the difficult thing for most people is knowing what ideas and technologies might work.

In my field (artificial intelligence) the rate of progress has accelerated greatly, leading to much complexity and thus increasing difficulty just to "keep up" with new advances. I organize my thoughts by using a rough hierarchy of classes of useful technologies and form a taxonomy by mentally mapping problems / applications to the most appropriate class in this hierarchy. When I read a new paper or listen to a talk on YouTube, I try to place major themes or technologies into this hierarchy and when someone describes a new problem to me, I try to match the problem with the correct classification in my hierarchy of solutions / technologies. Vocabulary is important to me because I organize notes in small text files that might contain a synopsis of a web page with a URI, a business idea, interesting ideas from reading material, etc. Key vocabulary words are the search terms for finding relevant notes.