4 ms·

I started off thinking he had a good point, but when he came to textbooks, I realized that it doesn't match reality. If what he claims were true, then books li

I started off thinking he had a good point, but when he came to textbooks, I realized that it doesn't match reality. If what he claims were true, then books like Selfish Gene; Guns, Germs, and Steel, etc. should leave me with LESS knowledge than a typical textbook, which at least has questions, tests, etc. associated with it. But, in fact, I'm pretty sure that most textbooks have taught me precisely nothing, or at least nothing that I retained a year later, whereas Pinker, Taleb, Dawkins, Diamond, and other such authors have absolutely taught me things that stayed with me. So, there must be something missing, and I think it's this: good books stay with you a while. I think about them a lot, for some time after reading them I keep revisiting the ideas. If I never did that, perhaps I would learn as little from them as the author of this post. But, for textbooks, this almost never happens, and that is probably why textbooks do worse than other non-fiction, even though by his analysis here the reverse should be the case.

ege_erdogan 6y ago
I think the textbook vs. Pinker/Taleb/Dawkins books can be coarsely classified as "fact books" and "idea books". Pretty much every textbook (especially introductory ones) is a fact book. They are structured summaries of a field of knowledge, and the goal is to convey as much information as possible. There is no room for extended personal commentary. An idea book on the other hand does not simply present facts, but fits them in a narrative designed by the author. You don't necessarily obtain the overview of a body of knowledge. It is more like having a long chat with the author, who is preferably an expert in said body of knowledge. For the learning part, idea books are more effective at getting the "big picture" across, while textbooks require more effort by the reader to either memorize the concepts or see how they fit together at a higher level. In the end I don't think one can be classifies as more effective than the other since they correspond to different needs. Consider two AI books, the "Modern Approach" textbook by Norvig and Russell, and "Human Compatible" by the same Russell. A student who wants to do AI research should read the textbook, and spend a lot of time studying it. A layperson who wants to get an overview of the AI field as it stands today, and see what are the possible problems that interest a layperson like himself, should read the latter book, spending much less time than the student.

ege_erdogan 6y ago

I think the textbook vs. Pinker/Taleb/Dawkins books can be coarsely classified as "fact books" and "idea books". Pretty much every textbook (especially introductory ones) is a fact book. They are structured summaries of a field of knowledge, and the goal is to convey as much information as possible. There is no room for extended personal commentary. An idea book on the other hand does not simply present facts, but fits them in a narrative designed by the author. You don't necessarily obtain the overview of a body of knowledge. It is more like having a long chat with the author, who is preferably an expert in said body of knowledge. For the learning part, idea books are more effective at getting the "big picture" across, while textbooks require more effort by the reader to either memorize the concepts or see how they fit together at a higher level. In the end I don't think one can be classifies as more effective than the other since they correspond to different needs. Consider two AI books, the "Modern Approach" textbook by Norvig and Russell, and "Human Compatible" by the same Russell. A student who wants to do AI research should read the textbook, and spend a lot of time studying it. A layperson who wants to get an overview of the AI field as it stands today, and see what are the possible problems that interest a layperson like himself, should read the latter book, spending much less time than the student.