Inevitability of incremental reading

From supermemo.guru
Jump to navigation Jump to search

This text is part of: "History of spaced repetition" by Piotr Wozniak (June 2018)

Computational inevitability

Incremental reading makes it possible to dramatically increase the speed of reading, while increasing comprehension, applicability and retention of knowledge.

Incremental reading was cosmically inevitable. Over the millennia, countless minds had the same intuitions and frustrations about the process of innovation and problem solving. We always tried to effectively augment human memory with well organized external sources of knowledge. At the end of the last millennium, with many of the knowledge preservation goals achieved, frustration shifted to augmenting the way we retain high quality knowledge in memory. The human brain has no match in its ability to creatively associate ideas. This is why we want to achieve smoother integration between the concept network of the brain with external knowledge networks that will also gradually conceptualize with the assistance of artificial intelligence. Soon, we will end up with one super-intelligent, super-concept network (see: Society as a concept network).

Incremental reading and neural creativity are just a bridge between the 1990s (emergence of Google, popularization of spaced repetition), and the immediate future in which global knowledge and individual brains will undergo ever closer integration. This text speaks of multiple threads of convergence between the thought processes of people involved in knowledge management in the past and in the present.

This convergence process is multi-threaded, continual, and involves independent ideas from people who heard little or nothing about incremental reading. Possibly, many never even will, unless the chaos of facts is clarified with the emergence of the semantic web.

Intelligence is a force that makes some ideas computationally inevitable

Convergence of ideas

The brain is a concept network. It illustrates how simple processors (neurons) organize into an intelligent network (brain). Similar principles of intelligent self-organization are pretty universal from microscopic to galactic levels (see: Complex adaptive systems). This self-organizing trend is now increasingly observable in all human activities based on knowledge, innovation and networking. The Internet became a form of a concept network. The semantic web will be the next conceptualizing stage. It will be a framework of super-intelligence where all human inventions of the past, present and future will get re-integrated and crystallized in the same way as neural networks and knowledge undergo conceptualization in the brain.

Krzysztof Biedalak has recently observed that our company is rather powerless in its attempts to bring forth the true history of spaced repetition. I agree. Misattribution and misappropriation and rampant. This is part of the darwinian process in which knowledge and innovation spread and crystallize. Our efforts to correct falsity meta-vectors that circulate in human meta-knowledge are comparable to the efforts of tiny concept neurons in the brain to dominate control over a set of input patterns. This is natural and this is healthy.

The first step towards the invention of incremental reading is the observation that knowledge is power. To access knowledge we preferably need semantic routes that make selection and application easy. In the case of the brain, it is relatively easy. The brain itself will not be interested in knowledge that is irrelevant. The relevance, by definition, is semantic. In other words, we remember effectively only things that find a well-attached place in the knowledge tree. The problem of forgetfulness can be resolved by review and this is why spaced repetition is precious in preserving access to semantic knowledge.

For millennia, we have used written records to preserve most valuable and most applicable knowledge. For written records, preservation is less of a problem. It is access to knowledge that becomes critical. Knowledge on paper is not associative. it can take part in innovation only if it is, at least temporarily, brought back to processing in the brain. Hundreds or thousands of scientists or writers worked on their own system of notes. They would often store it in notebooks in a linear fashion. However, any method of semantic organization would have tremendous benefits. For those purposes, little pieces of paper with notes, stored on special shelves, drawers, index cards, filing cabinets, or mini-notebooks, are immediately useful and necessarily obvious. The ideas in the vein of Bush's Memex or Zettelkasten have been invented over and over again in history (source).

Storing stuff in small-ish notes is the fundamental principle in creating a device called “Zettelkasten” (German for “slip box”, or “card index”). Vladimir Nabokov, Jean Paul and Arno Schmidt wrote their novels’ drafts on index cards. German sociologist Niklas Luhmann’s productivity was increased to epic proportions (70 books, 400 articles) with the help of his Zettelkasten

Nabokov's "Pale Fire" was deemed by Ted Nelson as a form of hypertext. The list of scientists with similar ideas and/or methodologies is endless. The earliest record I have in my notes is that of Carl Linnaeus - the tireless student of taxonomy, which is nothing else but a knowledge tree describing the evolutionary origin of species (long before Darwin). I often mention Niels Bohr, and a lovely person of Larry Sanger to who we are all grateful for the birth of Wikipedia. They all sought semantic structure in knowledge, ways to preserve it in memory, and ways to use it in creative thinking and writing.

History of memory notes

The evolutionary tree of note taking and memory augmentation is rich enough for an entire book. Here I summarize some influential milestones in the mainstream, i.e. the tree trunk. The lesser known branches are like a creeping ivy: they separate, return, inspire, transform or wilt.

Note taking

On my Genius checklist I put a point "write down your ideas". This is not just because note-taking was characteristic of many genius minds. It is because great ideas tend to be fleeting. They can easily get forgotten and may not reoccur. I have documented many cases of my own ideas that volatilized and were recovered years or decades later due to the recurring nature of incremental reading.

Charles Babbage said: "Write down the thoughts of the moment. Those that come unsought for are commonly the most valuable". Graham Bell kept scrupulous notes of his ideas that collectively could help Bell break Edison's patent record. Botvinnik kept meticulous notes of his chess game ideas and encouraged his students to follow the same practice. It might be easier to list a hundred more examples than to find true genius of science that was not meticulous with his notes.

Note-taking is as old as print. Note-taking technologies were implemented in countless forms, and usually independently. Note taking in creative work is inevitable.

Interleaving

Interleaving is a learning technique in which we switch from subject to subject while learning. Its efficiency has been confirmed by rich research. It has been employed for hundreds of years. It has been employed in a lame fashion, e.g. at school, when interleaving is forced from above. It has also been employed by individual readers or thinker along the creative demand (e.g. as in incremental reading). In "Antifragile", Nassim Taleb explained his natural approach to interleaving powered by the learn drive. It explains why schools inflict harm by destroying the learn drive, while trial-and-error approach to reading on demand underlies the emergence of new semantic quality:

The minute I was bored with a book or a subject I moved to another one, instead of giving up on reading altogether - when you are limited to the school material and you get bored, you have a tendency to give up and do nothing or play hooky out of discouragement. The trick is to be bored with a specific book, rather than with the act of reading. So the number of the pages absorbed could grow faster than otherwise. And you find gold, so to speak, effortlessly, just as in rational but undirected trial-and-error-based research. It is exactly like options, trial and error, not getting stuck, bifurcating when necessary but keeping a sense of broad freedom and opportunism. Trial and error is freedom

Ebbinghaus to Memex

Research by Hermann Ebbinghaus in the 1880s reflects the ageless human interest in the properties of memory. Not much later, in 1912, Edward Thorndike dreamed of "incremental reading on paper" that could dispense of a teacher. His reasoning was still tainted by the idea of extrinsic goal setting that is characteristic of schooling:

If, by a miracle of mechanical ingenuity, a book could be so arranged that only to him who had done what was directed on page one would page two become visible, and so on, much that now requires personal instruction could be managed by print

However, even today, most of the scientists are rarely bothered by the pain of forgetting. Making notes and Google can satisfy most of the needs for most of them. Zettelkasten is a product of this reasoning in where the access to knowledge takes precedence over storing it in the brain. When Vannevar Bush conceived Memex in the 1930s, he saw it as a memory augmentation device. Memex can be seen as a granddaddy of incremental reading, but it was to live outside the human brain and needed no spaced repetition. Today, the closes embodiment of Memex is Google. A predominant view today is that we need not burden memory because we can Google anything. Naturally, all prolific problem solvers know that creativity feeds on creative association, which feeds on memory, which has a limited capacity if only the working memory was to be used. Long-term problem solvers need vast resources of associative knowledge available right in their own memory.

Descendants of Memex

Descendants of Memex are too many to list (for a summary see: Godfather).

J.C.R. Licklider credited Bush's Memex as his main inspiration when writing about "thinking centers" in "Libraries of the Future" (1965), in which he arguably imagines semantic web accessed in natural language and capable of basic intelligence.

Doug Engelbart read "As we may think" while waiting for a ship in 1945. He started his "Augment" project in 1962 (the year I was born). The project evolved in a form of hypertext web serving researchers in storing their work. When the government stopped the funding (1975), some of the team went on to Xeroc PARC, which influenced Steve Job, which influenced Hypercard, which influenced the web.

Memex also had a powerful impact on Ted Nelson whose ideas matured over decades soaking in inspiration from multiple sources. In 1972, Nelson declared "Memex is here". Nelson was an avid reader. He was also a writer. He considered paper a form of prison. He conceived an idea of hypertext that was a bit too complex for his day (see Gary Wolf's gripping "Curse of Xanadu"). Nelson did not like the linearity of paper, but would still accept navigational linearity of hypertext. His ideas need a bit of semantic parallelism known from the brain (as in neural creativity), where navigational pathways compete for attention in a parallel process. Bush's "associative trails" could scan prospective routes before presenting them to human attention.

Nelson was too far ahead of his time. The tragedy of hyper-genius is that it will only be fully understood a few generations from now. Only when Tim Berners-Lee took a simpler route towards hypertext, did we see the explosion of the web. Seymour Papert would envisage Knowledge Machine. I had my own ideas of SuperMemo Knowledge Machine (1995). At first, SuperMemo World tried to implement the machine literally as described in "Where do we want to go?" (1996). When it appeared that venture capital was not interested, SuperMemo went on with a less systemic implementation: incremental reading (2000). Instead of molding Internet protocols to incorporate spaced repetition, we opted for molding user's experience.

Tim Berners-Lee

Using the inspiration from Bush's Memex, Engelbart's knowledge repository, and the hypertext ideas of Ted Nelson, Tim Berners-Lee envisioned a system that could improve information exchange in large teams. In March 1989, while employed at CERN, Tim wrote a proposal for improved information management. His main concern was to improve keeping track of large projects. As it was the case with Memex and Xanadu, Tim's prime concern was knowledge aggregation and access with learning being just a side effect that has always been taken for granted. Great minds tend to believe that good semantic anchors are all that is needed to remember. They are right in that semantic learning makes knowledge stick. But forgetting curves in spaced repetition show that without a degree of assistance, memory may be lost in the long term. Memex, Xanadu, Zettelkasten or Google are great for knowledge access, but they all need to be augmented with tools supporting human memory.

Today, Tim Berners-Lee works over the semantic web that will become the embodiment of all prior efforts in structuring knowledge for the use by human intelligence and artificial intelligence.

Recent parallel inventions

As I type these words, thousands of people wonder how to augment their cognition with smart note-taking, concept maps, smart review, etc. The richness of the web provides endless opportunities. This leads to parallel co-inventions and re-inventions.

Michael Nielsen

Not long ago, for the parallel nature of human invention in the area of hypermedia and learning, I chose a quantum physicist and a passionate student of neural networks, Michael Nielsen as an example. The choice came from the fact that Nielsen re-invented incremental reading using Anki as the underlying spaced repetition mechanism. Nielsen is now also involved in projects that seem a reincarnation of ideas that were conceived decades ago. In case of SuperMemo World, the ideas of Dr Antoni Szepieniec never materialized due to the failure to secure sufficient funding. Today, the same ideas may require a fraction of the originally conceived investment. Here is an example: Quantum computing for the very curious at Quantum Country.

For more about Michael Nielsen see: Michael Nielsen re-discovers incremental reading with Anki

Andy Matuschak

Andy Matuschak is another example of a mind who craves knowledge, appreciates the networked beauty of semantics, and wants to escape the linearity of books and the thought process (see: Books don't work).

I love Andy's approach to intellectual endeavors, which I see as the polar opposite of peer review in terms of its inspirational and curiosity value. For Andy, the openness of Twitter is not enough. He literally wants to connect his brain to the world and engage in a two way communication:

One of my favorite ways that creative people communicate is by “working with their garage door up,” to steal Robin Sloan’s phrase. This is the opposite of the Twitter account which mostly posts announcements of finished work: it’s Screenshot Saturday; it’s giving a lecture about the problems you’re pondering in the shower; it’s thinking out loud about the ways in which your project doesn’t work at all

His emphasis on weaving semantic networks, helps Matuschak provide a good prescription for creativity based on a creative association of two ideas, which happens in working memory, but relies on well-structured long term memory:

Consider some hypothetical leap of insight you’d like to be able to make. To make that leap, you’ll typically need to evolve many independent, partially-formed ideas simultaneously, until they suddenly converge in a flash of inspiration. If you need to iterate on more than a few pieces at once, you may struggle to keep them all in your head

Only Matuschak himself may try to figure out the origins of his inspirations. For one, he is a friend of Michael Nielsen. On the other, he admits having been inspired by Zettelkasten. His principles of semantic note taking seem to lead to an equivalent of well-finalized topics in incremental reading. He calls his notes "evergreen notes" for their perpetual applicability:

It’s hard to write notes that are worth developing over time. These principles help: (1) Evergreen notes should be atomic. (2) Evergreen notes should be concept-oriented. (3) Evergreen notes should be densely linked. (4) Prefer associative ontologies to hierarchical taxonomies

Matuschak and Nielsen work on an interactive semantically networked "book" on Quantum Computing that incorporates several of the ideas underlying an efficient reading system with support for long-term memory (see the previous section).

Robert Heaton

With omnipresent quick access to knowledge, it takes about 4-5 years to re-discover incremental reading. Interestingly, programmers and physicists seem to dominate the list of the discoverers.

Robert Heaton is a nice young programmer, a graduate of physics (Oxford), a voracious reader, and a prolific writer. He is busy with a million things, a baby and a blog, but he was still able to re-invent incremental reading in just about 5 years (2013-2018). His hunger for knowledge and frustration with forgetting made him arrive at a pretty efficient reading system. I can summarize his thinking in a few sentences to show how it recapitulates the reasoning process that lead to incremental reading. For the full text see Heaton's detailed account: How to read?.

Inventing incremental reading in the words of Robert Heaton:

  • People laugh about how they don’t remember anything they learned in school (see: Problem of schooling)
  • Five years ago I realized that I remembered almost nothing about most books that I read (see: Mechanism of forgetting)
  • It occurred to me that calling yourself an auto-didact doesn’t mean you actually know anything
  • I’ve evolved a system to help me remember more of what I read (see: SuperMemo Knowledge Machine)
  • Learning comes from repetition, but books are long and verbose and not designed with this in mind (see: Minimum information principle)
  • Once you have finished the book, make a “writeup”. This involves summarizing the book, doing further research and making flashcards (see: active recall)
  • Anki (see: spaced repetition)(discovering the concept by Googling is also a form of speed-innovation)
  • Write down any additional thoughts or connections that come to mind (see: incremental writing)

In addition to the above, Heaton developed a system of symbols equivalent to processing attributes, which in SuperMemo are often implicit or deferred (i.e. decisions are made upon re-encountering an extract).

Heaton's system is missing priorities and tools for handling knowledge overflow. Note-taking combined with spaced repetition sound great at the beginning, but after reading 30-40 books using his algorithm, Robert Heaton will certainly discover that he needs some extra guns to tackle the problem of knowledge volume.

Simon Hørup Eskildsen

In innovation, one of the most precious attributes of the brain is an incessant hunger for knowledge. This hunger is produced by the learn drive system. The power of the learn drive stems from the fact that it expands the semantic network of knowledge then feeds itself on its newly exposed branches. This lead to a popular saying: "the more you know the more you know you don't know". This precious mechanism for rich lifelong learning is systematically destroyed by the system of compulsory education, which we must end as soon as possible to boost the survival chances of mankind. Individuals who survive school unscathed, or who recover their learn drive after school, make the bulk of the innovating generation. Good survival at school is usually due to childhood freedoms or steering the course of learning through passions. A hungry brain is usually also a generalizing brain. This means that new knowledge undergoes generalization with an increase in applicability. This speeds up the acquisition of new knowledge, powers creativity, and innovation. Those hungry brains that delve into rich areas of abstract knowledge will show particularly rich output. We need many more similar hungry brains to power the new wave of innovation that will conceptualize human meta-intelligence. I mentioned Nielsen and Matuschak before.

Recently, someone called my attention to another prolific brain. This is Simon Hørup Eskildsen who, at the moment of writing (Apr 2020), is a Director of Infrastructure Engineering at Shopify. This text shows how Eskildsen evolved his own efficient system for long-term learning. He was driven by the same forces as myself, decades ago. The same forces that made hundreds or thousands of great minds follow decades or centuries before. However, Eskildsen operates in a new reality in which he can compose his system of various digital tools, components and sub-components. His is a digital native for who a modern software and networking toolset are as common as morning coffee. We could literally breed millions of digital natives if we only let kids enjoy their childhood to the best of their individual environments and passions. Instead, we scare them with Digital Dementia, and impose screen limits to make sure they engaged in letteracy verging of letter paperacy.

Eskildsen's girlfriend diagnosed the magic beind Eskildsen's learn drive pretty accurately, even if Eskildsen cannot see it clearly due to his own natural blind-spots (source):

I read 30-50 books each year on topics ranging from the history of transistors to ancient philosophy. My girlfriend jokes that "because Simon never went to university, he still enjoys studying". I’m not sure. I think I’d love university

Even though college provide the illusion of freedom by the fact that it is voluntary, the method of study and testing ruin the love of learning pretty efficiently even for those who chose their major out of sheer passion.

My evolution from spaced repetition to incremental reading took roughly 15 years, and matured over the next 20 years. Eskildsen rushed through his own at roughly four years. His superior speed comes from better access to rich knowledge, life in a digital world, and possibly less injury from schooling. My ten years of college were a drag, and a diversion mixed up with a bit of semi-relevant learning.

Most decisively, Eskildsen could gulp from the nutritious well of inspiration from sources described in this text. I could only dip into an ancient paper library dominated by pre-war German publications. For clarity, I read German only in Google Translator. This contrast alone is a great testimony to an ever increasing pace of innovation.

Extracts (highlights)

The use of a highlighter pen is as old as print. At Newton's time, it had a form of "notes on the margin". Cutting out pieces of print for a system of index cards has been employed by many scientists (incl. Niels Bohr described in the section: Creativity). In incremental reading, the role of a highlight is played by an extract. As Eskildsen likes to read on Kindle, he uses his own system for picking the cream of the crop (clippings):

Anything that I find that’s important, I’ll highlight on my Kindle. All of those highlights automatically go to my Readwise where I can add them to my learning system. Readwise is fantastic because it automatically scrapes all of my Kindle highlights and puts them into one place for me where I can search, tag, and review them

Naturally, the role of extracts is wider in SuperMemo. In addition to serving as highlights, they are a fodder for cloze deletion. Moreover, extracts may also contain pieces for later reading. In this process, they are recursively fed back into the learning process, in which attention is incrementally focused on increasingly narrow pieces of knowledge. In those processes, incremental reading moves from outlines and generality to details and quintessentiality.

Subset review

Eskildsen appreciates the value of subset review. He came up with his own version by reading books in a single area of interest in parallel. His prime rationale is the comparison of different perspectives from different authors. In SuperMemo, this approach is also very helpful to boost knowledge darwinism and creativity:

I go on streaks of reading a lot of books on a particular topic around the same time. Doing this is useful because it means I don’t have to just trust one author’s perspective on a particular topic — and helps me connect a lot of facts together, so I can understand things better

Knowledge tree and concept maps

In incremental reading, knowledge is organized semi-automatically into an knowledge tree. Connectivity between pieces of knowledge can then be expanded, on demand, with links, hyperlinks or concept links, which form meta-structures in the shape of "concept maps". Eskildsen uses a note repository for knowledge structure, and flashcards for memory support:

My learning system itself has two components: a flashcard system and a custom-built note repository inspired by the Zettelkasten — which is a note-taking system developed by the social scientist Niklas Luhmann. The Zettelkasten is where I spend time processing, categorizing, and connecting what I read

Spaced repetition

Eskildsen suffered from the same frustrations that I experienced in the years 1982-1983. He put a lot of effort into reading, only to discover the ravages of forgetting. What had to come next was, naturally, spaced repetition:

Once I started to read more books, I realized that I didn’t really remember much from them. That bothered me. So I went down the rabbit hole of building a bunch of systems to help me remember what I read

Cloze deletion

In the original description of his methodology, Eskildsen did not mention cloze deletion. In a good Zettelkasten tradition, he is crafting his items in Anki. However, with a rich flow of knowledge, the cloze step is also inevitable. At certain speeds of learning, generating items with a click is just too convenient.

I’ll then periodically open the list of “flash” tags in Readwise, and then translate them into flashcard form in Anki

The origins of cloze deletion in SuperMemo are simple and clearly inevitable:

  • 1995: introducing topics for the sake of course presentation (e.g. in language learning)
  • 1998: introducing richly formatted text that encourages copy and paste from the internet, and reading in SuperMemo
  • 1999: I personally increased my copy-and-paste rate in item formulations, and copying a keyword from the question to the answer turned out very convenient, obvious, and inevitable
  • 1999: SuperMemo 9.5 (aka SuperMemo 99) introduced first simple cloze deletion options (see: What's new in SuperMemo 99)

Some of my colleagues do not use cloze. They still stay in the era of hand-crafted items in the spirit of Zettelkasten. I insist that this keeps them blind to the possibility of cranking up the volume of learning. It is easy to predict that Eskildsen will move in the direction of cloze sooner or later. His knowledge hungry personality makes it inevitable.

Creativity

The way Eskildsen uses his system of notes is a clear indication that he loves the creative effect of richly interconnected web of personal knowledge. The power of the "conversation" with one's own notes have been discovered over and over again. It is the same force that drove Luhmann, Niels Bohr, or Georgios Zonnios. Zonnios case is particularly interesting because he seems to have come up with the idea of incremental writing at just about the same time when I started employing the concept. There were two factors that I had in common with Zonnios. Like Eskildsen, we are both pretty hungry for knowledge, and we both used incremental reading as a preceding inevitable step. The core of the "creativity invention" is a simple conclusion: if I read about other people's ideas, why not add my own conclusions to the process. The rest is just a series of inevitable conclusions.

Niels Bohr approach to a system of notes was also a good example to show inevitability of Zettelkasten or incremental reading. Bohr used the power of intermittent reading and intermittent processing to maximize his creative output. He would keep dozens of drawers with outlines of ideas on sheets of paper (source). He would return to individual drawers from time to time, over months or years, esp. if he was inspired by a conversation, publication, thought, or experiment. He would then keep reading a single stack, ponder, expand notes, add new notes, sort or duplicate, etc. Many of those drawers ended up as scientific publications. All his stacks of ideas needed was a bit of prose to glue them together. In that sense, Niels Bohr employed a rudimentary form of incremental writing in his scientific endeavors.

Speed-learning

Luhmann is said to have accumulated 20,000 notes in Zettelkasten. Digital learning is easier, and Eskieldsen arrived at 10,000 cards in Anki in four years:

I am creeping up on 10,000 cards in Anki, and I’ve been doing this for over 4 years. It’s probably the most impactful habit I have in terms of impact over time invested

In SuperMemo, 10,000 is easily achievable in a year, esp. when learning vocabulary. In incremental reading, the speed of generating items is stratospheric, and can go well beyond the brain's ability to remember. The degree of processing, however, is more important than the speed and numbers. Luhmann might have achieved with his 20,000 more than it might be achieved with automatically rushed 200,000 elements in incremental reading. In the end, the brain's capacity and the speed of learning seem pretty steady and pretty limited (see: How much knowledge can human brain hold?). This is why a smart selection of knowledge is essential. Its contemplative processing is also of value. The speed of incremental reading is probably most useful in sifting away tons of knowledge we do not actually want to learn. The mankind's tempo of generating new ideas keeps increasing. This is why this filtering power of incremental reading is increasing in relative importance.

Future

In the future, ideas of incremental reading will be extended and consumed by larger trends of integrating human and artificial intelligence into a global unified concept network. Vannevar Bush could see the future back in 1945 (see: As We May Think). Today, we have all tools we need. They just need a little bit of work and experimentation.

In incremental reading, in the first step, the evolution of knowledge priority in a knowledge collection will need to resemble analogous conceptualization processes in the brain with spreading activation and two-component model of memory as the underlying mechanisms for stabilization and forgetting. This is necessary to rationalize the management of an increasingly voluminous inflow of new knowledge. For example, when we retrieve information from the collection, execute subset review, or perform specific actions during a repetition, priorities will undergo modification via semantically driven spread of activation. The changes will undergo modification along the rules of two-component model of memory.

For example, if you execute a subset review on coronavirus, the priorities of knowledge associated with the subject will increase, while competing knowledge will lose on priority. If the virus gets semantically connected with newly emerging topics, e.g. ventilator or respirator, those topics will receive higher priority, and so will topics associated with ventilators in that particular user's collection (e.g. the life story of one's grandmother who had to receive treatment). The changes in priority will be more pronounced for newly important topics, but forgetting will be slower for topics that receive regular up-prioritization. This way the entire user's collection will undergo processes that are similar to those that occur during learning, except the entire process will initially occur outside user's brain and will only affect the review process in the future.

Knowledge personalization will make it possible to employ a degree of connectedness with global resources. For example, if I insist to disagree with the world on a specific subject, I may block human knowledge from affecting my collection, while still receiving hints on existing contradictions. In the future, such blocks can be implemented as an inhibition from own coherent models dominating the concept network. In contrast, when I accept global consensus, e.g. due to its lower priority, my collection knowledge would get updated in the process analogous to updating Wikipedia today. Update recommendations will follow our interests and priorities. This may look a bit like YouTube recommendations today. Global concept network will know our interests and where our knowledge shows inconsistencies with the global consensus, or with a local consensus (e.g. subsystems in a concept network associated with a particular author or a particular point of view).

The above example shows that knowledge will show clear separation between all human knowledge (e.g. as available via Google), personal knowledge (as available in SuperMemo), and actual knowledge stored in an individual's brain. Human knowledge will also form clusters typical for a concept network. There will be an ongoing process of integrating all those resources and the convergence of the mechanism in which knowledge is consolidated and forgotten in the brain, as well as in all human resources.

Tim Berners-Lee effort of creating standards for the semantic web will help human knowledge get organized in a way analogous to which it is stored in the brain. Artificial intelligence will then easily capitalize on simple mechanisms known from concept networks. The process requires no special technology, advanced math, or know-how. It will be incremental, evolutionary, accelerating, and inevitable. As not everyone is happy with the power of artificial reasoning, we need to observe that the benevolence of such a super-network will reflect human benevolence and will be an emergent force based on game theory, the quest for intelligence, and the quest for survival. The battle of good and evil will be no more fierce or significant than a competition of neural subsystems.

Human brains and artificial intelligences will incrementally self-organize into a global concept network



For more texts on memory, learning, sleep, creativity, and problem solving, see Super Memory Guru