If I could have any superpower at all, I would want to be able to read at superhuman speeds and have perfect recall of everything I read. At first glance, that seems like a stupid power to have; however, in the film Limitless, Bradley Cooper’s character uses those very abilities to become a successful author, financial expert, and elected official.
While perfect recall is a topic for another day, reading is a valuable skill in itself. In fact, it might just be the most valuable skill in the world. Leaders the world over attribute their success, at least in part, to their voracious reading habits. When asked how he managed to create SpaceX’s rockets, Elon Musk famously responded “I read books.” Bain & Company chairwoman Orit Gadiesh reads 100+ books a year and billionaire Warren Buffett reads 500 pages per day. Bill Clinton had to expand the library at the Arkansas Governor’s Mansion to accommodate his collection. Reading seems to be a common thread among the successful.
Since fiction is meant mostly for entertainment or contemplation of literary merit, we will focus on speed-reading nonfiction. Nonfiction texts, by nature, are meant to impart information. By reading my writing, you can get inside my head and extract my ideas. However, most nonfiction books share a common problem: signal vs. noise. Publishers ask authors to hit a certain page count, largely because the public will pay more for larger books. Instead of optimizing for learning, readers optimize for length. This forces authors to fill their books with irrelevant fluff and circuitous storytelling. Hence, we encounter the problem of separating signal from noise.
Signal is the relevant information we want to get out of a book whereas noise is all of the extra nonsense used to fill pages. Often, it can be hard to tell the difference between the two. A good analogy to illustrate this is the following. Imagine that you’re in a public square surrounded by people. You had planned to meet a friend there. Suddenly, you hear someone calling your name on your left side; that’s signal compared to the noise of the crowd. You walk towards the sound of your name, which your friend keeps calling, until you are able to find him or her. As you gained a sense of where you had to go, you were better able to parse the signal from the noise.
When speed-reading, this can be decidedly more complicated. However, we can draw on machine learning to gain a new perspective. Machine learning is goal-oriented; training a model is meant to reach a certain goal state, whether diagnosing cancer or predicting stock price changes. Models rely on a few tools to measure their learning in pursuit of their goal. Information gain is a metric by which one can decide which attributes of a dataset matter the most. Information gain is frequently used alongside mutual information, which is the amount of information you can get about attributes X and Y from attribute Z. Attributes with high mutual information are most favorable. We can apply this sort of goal-oriented learning to nonfiction speed-reading.
Consider the book as a dataset and the individual paragraphs as attributes to either keep or discard. First, you need to figure out the point of the book (i.e. what is the goal of your learning?). Some books have a new lesson every chapter and some work up to a unitary point. This should be self-evident based on the book and chapter titles, as well as the introduction to the book. Then, you need to create a mental framework to approximate your information gain.
First, you need to eliminate the words that, by definition, won’t add information. In natural language processing, these are called stop words, which are ignored; they include the, by, about, etc. Although we were trained as children to read every word, most words give us no new information. Consider, for example, this sentence: The bluebird sang outside my window every morning. Most people will stop to read every word, but if you train yourself otherwise, that sentence should look like this: The bluebird sang outside my window every morning. We’ve cut 3/8 of that sentence away; imagine your productivity if you could cut that much of a book away.
Second, you should read the first and last sentences of a paragraph before anything else. Those sentences, more often than not, indicate the topic and conclusion of a paragraph and therefore supply the most information. If they seem both new and important, read the entire paragraph. If they seem either old or unimportant, discard the paragraph. Paragraphs that are either old or unimportant have low mutual information, whereas new and important paragraphs have high mutual information (i.e. they tell you more about other pieces of information you haven’t read).
For example, I was recently speed-reading Trade-Off by Kevin Maney. It talks about quality vs. convenience in business models. Apple is high quality and Walmart is high convenience, but both are successful companies. Maney’s argument was that a business should pick one of those attributes, but not both. I managed to summarize the entire book in two sentences; there’s no point in plowing through 240 pages if you’ve gained the information you need.
After speed-reading the first three chapters of the book, I understood the entire thing. Soon, you’ll be able to do the same. In short, unless something adds to your understanding, skip it. Once you’ve reached your goal state, consider the book “read.” There’s no point in persisting when you’re not absorbing any new knowledge.