Upload A Book To Your Brain: A Viable Theory for 2016

Yesterday, somewhere around 800 new books were published in the US. And there will be 800 more today, and 800 more tomorrow. The amount of written material that comes into existence everyday is magnificent, fully incomprehensible. Even if you read a book a minute for the next 100 years, you wouldn’t even get halfway through the books that already exist.

There has to be other ways to get the knowledge that’s trapped in ink and refurbished tree into our brains. The classic sci-fi throw back is to the metal spike in the matrix, the ability to electronically transfer information to your brain. The science is closer to computer sentience than it is to mind uploading, but I have a theory on a workaround. What if a software could understand what you already know in such a way that it could accurately predict how you would read and internalize a book? You would then, conceivably, be able to skip the middle man (the book), and have a computer read it on your behalf and give you a summary of how you would have read it. Here’s the theory:

Assumptions:

1. On average, when reading, my brain sees sentences as falling into one of three categories; known, “interesting”, or particularly interesting. Known statements are not be committed to memory, but only act to reinforce existing memories.
2. An “interesting” statement would be evaluated based on its relevance, first to major broad groups of knowledge, working towards narrower and narrower categories of knowledge until it fits comfortably into a sub-category or requires the formation of its own sub-category.
3. Particularly interesting information is that which fits into more than one category at one of the highest levels. This information creates a bridge between broad categories, fosters a deeper understanding, and inspires awe.
4. If I were reading, say, 3 books a day, my true internalization would be limited. It might include a brief summary of the complete mass of “interesting” statements as well as a specific quoting of particularly interesting statements. In this theory, these two inputs constitute the Read Report.
5. Interacting with indirect language is most closely connected and stored near other interactions with indirect language. Indirect language is that which was not meant for anyone directly (as direct conversations and interactions are), but for many people indirectly. Most forms of media including; books, movies, television, poetry, music, lectures, podcasts, and more, fall into this category. Thus I principally use my existing knowledge of indirect language to judge new information received by way of indirect language.

The AI

The challenging part, of course, would be the preparation. This software, given the assumptions, would need to access as much of my pre-existing indirect language as possible. That would mean a database of the scripts, lyrics, or transcripts of every meaningful (in this case, memorable) piece of indirect language I’ve encountered since the age of stable memory accumulation (say 5 or 6). This database would be weighted chronologically (oldest media receives a low ranking and the newest media having the highest ranking) so that the newest language I’ve come into contact has the strongest effect on new information.

A rudimentary categorization of phrases by key words would be carried out. The transcript of a conversation guided by fairly simplistic questions (What do you do for work? What are you interested in? What inspires you?) would provide linguistic data to further rank the rudimentary emerging categories, giving categories that are expressed to be more important a higher weight.

Once many generations of categories and sub-categories have been established, a second survey can be carried out, this one very specific. A list of words associated with each category will be made and assigned to that category as tags.

Once the database is complete, the first piece of text can be submitted. The program will sort through the text sentence by sentence, comparing each to the existing database. A certain similarity percentage (say 80%) will result in a known classification and discard the information. The assumption/hope would be that the vast majority of the text would fall into this category.

The remaining information will be classified as interesting, and its path to its most similar category will be recorded. In the Read Report, all the “interesting” paths will be combined and summarized en masse.

The final bits of information, those with a low enough similarity rating to be “interesting” but which can be sorted into one category but may also be described by the sub-category or tags of a second category, are deemed particularly interesting. These sentences will be printed in full in the Read Report.

After Thoughts

Though a software like this that could deliver a meaningful summary of a large set of information would require a huge amount of processing power, a tremendous amount of time creating the database, and a significantly better understanding of the computer science, the theory creates an interesting narrative.

If a technology existed and could be widely used that interpreted knowledge and condensed whole books into say a 1-page Read Report, how would that effect the way we write books? Would writers eventually lose the will for long form and decide to beat the computers to the punch by writing the Read Reports themselves?

But as the computer is interpreting the information and delivering it based on correlation to existing knowledge, maybe the opposite would occur, maybe books would get exponentially longer as writers attempt to create books that more and more people find relevant (or at least more and more computers think their people will find relevant).