Table of Contents
Here's are some axioms I posit:
- Many smart people read a lot. Some of the smartest and/or most successful people (Warren Buffet, Elon Musk, Benjamin Franklin, Bill Gates) are un-coincidentally some of the best read.
- 2) Scientific text-books, academic papers, instructional videos / lectures (MOOC, youtube), and scholarly essays are some of the most rich and valuable types of resources for facilitating learning. EDIT: There's also a lot of crap among these formats, which is a problem worth addressing re: deduping, saving edit history.
- 3) Life is short. There's too much to read; there are too many of these resources in existence to reasonable vet in our lifetime. e.g. Amazon, alone, knows of 30,000+ texts on Calculus. How many webpages do you imagine there are which you will never even see in your life time?
- 4) We duplicate each others' efforts a lot. And we do little to prevent duplication. And it costs everyone. We all independently waste inordinate amounts of time searching for and vetting for many of the same books. When authors write new books, they haven't read a fraction of the books on the topic. The next book on calculus will be largely duplicative of other resources in the field, only contributing to the problem of information overload.
- 5) We suck at rating books / heuristics. We compensate for information overload and ignore duplication by using heuristics to surface the best stuff. It's very hard to determine which works are best, and in which order they should be read. Algorithms for ranking books are still primitive (number of reads, citations, star ratings, fulltext search match). It's even harder to know which high quality resources we're missing.
- 6) We suck at showing our work. Partly because many of us don't feel accountable to others unless there's financial incentive. It's easy enough to leave ourselves notes, it's difficult to describe our method and leave a trail which others can understand, follow, reproduce, and further.
- 7) Context is key. Most experts have institutional knowledge about which books and sections to read and in which order content should be read. This is often encoded in collegiate curricula. Most colleges and departments have very different curricula.
- 8) Curricula are imperfect. They suffer from many of the same things as books, web pages, and many of the other mediums described in #2. There are many duplicates, they are not responsive to rapid paced feedback, and they are not interoperable (i.e. There's no easy way to share, browse, edit, or publish these curricula. There's no way to easily cross-reference, compare, and deduplicate these curricula). The open syllabus project is a good step towards addressing this.
- 9) A curricula is basically a mashup of different types of content. The unit is usually something which needs to be purchased or is under copyright. It's not universally accessible.
- 10) Copyright policy makes it hard for us to collaborate by enhancing or mashing up book content.
In many ways, what we are able to accomplish in our lifetimes is limited by the efficiency of our tools and processes for communicating and processing information. Before email, snail-mail wasted days of our life. Before cars and planes, horses wasted weeks of our lives. Technologies like Google and Wikipedia not only help us answer questions more quickly, but perhaps more importantly enable us to address a class of problem we'd have had to give up on entirely 20 years ago. But many processes of our society are just as inefficient as they were 20 years ago. If not worse. Including Books.
One of the world's most paralyzing and important problems is knowing what resources are worth reading / investing time on. Especially given we live during an age where evaluating a work can cost money and thus may not be achievable at scale. Where distributing a work for discussion may violate copyright fair use. And where publishing new works can be as easy as clicking an upload button.
Life is short, there are too many books to vet, there are tons of duplicate works, our vetting/rating heuristics are bad, it's too hard and expensive to access books to evaluate them at scale, and there's presently no effective way to harness the community to answer this question in a reliable way. There's currently no way (i.e. no platform, framework, or protocol, e.g. wikipedia) where by the community can answer this question in an objective way.
There are interesting community efforts like the Less Wrong "Best Textbooks on Every Subject". I purchased the domain https://thebestbookon.com (temporarily defunct -- ssl warning) and spent months trying to learn how one could create a system for diplomatically and objectively determining the best book on a topic. I also crowd sourced a list of great textbooks from people I respect: Awesome textbooks. I've also worked with the community to create a fulltext search interface which leverages the existing curatory efforts of Mortimer Atler who create a 52 volume collection of what he thought were the most important, "Great Works". By and large, it's still a very much open problem. I now spend much of my time thinking about this problem at the Internet Archive through the Open Library project.
Once we have the best book on a subject, then what? The further I explore this problem, the more something becomes obvious to me. Knowing the best book on a topic is not enough*. We need to harness the communities intelligence to figure out what sequences of books should be read. And not just books, but sections.
* this said, finding the "best" 1M scholarly books/texts/papers really is a really good place to start. Books as units are fairly manageable. The fact that they evolve slowly is actually of benefit given technology can hopefully catalog them faster than they are published. We should strive to figure out what the best 1M works are, find a way to legally make them all publicly available, and work together to create curricula around them
We need to be able to ask a question or input a topical query and be presented with 3 potential canonical entry points. The tradeoffs between these starting points need to be evaluated and made visually clear to the user; the user should be able to, by inspection, intuit how these entry points into the conversation/dialog eventually connect (assuming they do, also useful to know) -- i.e. the characteristics of each path through knowledge. We need to be able to visualize (as a graph) what the path looks like from this starting point to intermediary learnings. That is, physically see (at a bird's eye view) how a chapter from one book, leads to a youtube video, leads to a paragraph or figure from another book, to an academic paper, to a fundamental conclusion or learning.
People need to be able to create or propose various paths (a la arguman -- a platform for mapping arguments and voting on good points). We need to create a framework which is capable of honoring and respecting both our and others' personal values of what is "good" content.
This means we need better tools for adapting and expressing media which as been designed for physical interactions within digital contexts. One should be able to seamlessly reference a chapter, paragraph, sentence, or word and link it to other such units. And these connections/links should be annotatable (like the body of an html anchor tag, as opposed to its href)
We need to overcome the limitations of the book and create a community ecosystem which is safe for the rapid development (connection/curation and creation) of knowledge (at the speed of thought rather than the speed of publishing a new edition), an ecosystem which prevents duplication, allows merging and flagging of content (e.g. wikidata), preserves provenance/history (e.g. git), and which enables works and all their sections, components, and figures to be connected in a map, which may be navigated and annotated in whichever way makes sense to the viewer. In many ways, these concepts are not far removed from the notion of Vannevar Bush's "memex" (discussed in the essay, "As We May Think"). More thoughts about what this experience looks like[12,13]
-  http://lesswrong.com/lw/3gu/the_best_textbooks_on_every_subject
-  http://tr.arguman.org -- map arguments
-  https://www.khanacademy.org + metacademy.org + arbital.com -- structured sequences for learning
-  http://explorer.opensyllabusproject.org -- Mapping the college curriculum across 1M+ syllabi
-  michaelkarpeles.com/explanations -- to become whyloop.org: recursive explanations
-  dissertate.org -- a wiki for publishing and exploring curricula: curated directed sequences of textbooks and academic papers on specialized topics.
-  ideas on deduping books, starting with tables of contents
-  there's too many explanations
-  a curated list of other similar posts
-  on curricula
-  on what such a graph system / navigable curricula might look like for math (e.g. math.mx -- visualizing what maths you know)
-  How books should work
-  Complaints about books
cc: Drew Winget, Joytika Jit, Jan Paul Posma, Juan Batiz-Benet, Jacob Cole, Andrey Fedorov, Adrian Perez, Gordon Mohr, Richard Caceres, Lachlan Ford