You are on page 1of 9

##Humpty Dumpty

Many years ago, closer to the beginning of the information an MIT professor aske
d a talented undergraduate student to design a functioning computer vision syste
m over the summer. Computer vision wasn't seen as a problem worth the time of th
e professor or a PhD student, unlike chess playing computers. Unfortunately, tha
t student wasn't able to solve the problem over the summer. Not that he wasn't s
mart; he was *very smart* and is now a full professor at MIT himself. But Lewis
Carroll was more prescient than the MIT prof: once you break Humpty Dumpty, it's
hard to put him back together again. Or in the case of computer vision, if we d
on't know how Humpty Dumpty was put together in the first place, we will have an
*even harder* problem assembling him from scratch.
The reason is that computer vision is a much harder problem than chess playing.
For that matter, human vision is a much harder problem than chess playing. We k
now nature isn't as smart as MIT undergraduates, but it did have a few billion y
ears to figure out human vision, so you would think that a single student would
take longer than a summer to solve the vision problem. Forty years and thousands
of researchers later, we still haven't solved vision. We *do* have good solutio
ns to particular versions of the vision problem, but general purpose vision is f
ar from being understood and it's quite likely that there isn't such a thing as
general purpose vision to start with. Not only is the solution to computer visio
n much harder than originally envisioned, the original problem wasn't the proble
m worth solving in the first place.
In fact, the original statement of the computer vision problem abstracted away e
verything that was hard: vision isn't a problem in geometry, it's a problem in b
iology, which means it's all about context, adaption to the environment and impr
ovement in stages.
> There's no single problem to be solved. Instead, there's a network of coupled
problems that defies a magic bullet solution.
## What does this have to do with Knowledge Networks?
Why am I bringing up this story in a book on designing knowledge networks? For a
simple reason: while human knowledge networks have evolved over millennia rathe
r than billions of years, they are enormously complex networks that have evolved
organically; reverse engineering such networks is a challenging task, but assum
ing that introducing technology in higher education will "solve" the problem is
a foolish hope, especially when the highest profile version of the solution is t
o beam it directly on to billions of computer screens in the hope that an abstra
ct transfer of information is the same as creating knowledge. MOOCs are no more
the solution to the problem of higher education than AI systems in 1968 were sol
utions to the problem of computer vision.
I am sure the proponents of MOOCs are aware of these issues. After all, many of
them are computer scientists who are aware of the history of AI and the challeng
es we have faced in building artificial systems that mimic human intelligence. I
understand that silicon valley thrives on hype and the constant reiteration of
words like disruption. I also understand that venture capitalists want a return
on investment today rather than a decade in the future and politicians need to b
e seen doing something, *anything*, about the rising cost of higher education an
d the mountains of student debt.
Therefore, it's not surprising that short-term thinking is driving the push for
technology in higher education. Part of the problem is that the people driving t
his push have never thought about the nature of knowledge itself; neither are th
ey philosophers, nor are they real thinkers. They are computer scientists who ar
e riding a wave of silicon valley supremacy. I am not saying this in a snooty wa

y, of humanistic inquiry trumping grubby capitalism. I am saying it as an engine


er, a knowledge engineering professional who builds networks that create and com
municate knowledge. Like vision, human knowledge is hard; we need to understand
it's inner workings before we foist it on others.
Suppose that one morning, we woke up to read the headlines of the New York Times
where a history professor was touting his ability to build the worlds largest c
loud computing company. Wouldn't we take his pronouncement with a pinch of salt?
But building a cloud computing company is a lot easier than changing education
and knowledge. Why are we willing to take computer scientists seriously when the
y prognosticate about these matters? The obvious reasons are:
1. They are well-funded, powerful and riding a hype machine that gets air time e
verywhere in the media.
2. They are the good cops of a much more insidious agenda.
Let me state my gut feelings about the conspiracy up front: throughout the weste
rn world, people tacitly understand that the good times are ending. As a species
, we cannot sustain ourselves in the way we did before. Resources are getting sc
arcer every years. As purses tighten public funding and support for education wi
ll decrease. The increasing push for technology in education is nothing but a re
flection of the writing on the wall:
> The powerful are building a two tier system: a hands on, enriching education s
ystem for the elite with all the social networks that come with physical proximi
ty and a technologically mediated edutainment and instrumental "skill driven" sy
stem for the masses.
The new vision of higher education is an industrial model that worked for school
s, but with college being the new high school, it's being transplanted on to hig
her education. Fortunately for these powers, there's a crisis in higher educatio
n of their own creation that they are taking advantage in order to set an even m
ore exploitative system in motion.
## The Crisis in Higher Education
There's no doubt about a major crisis in higher education. Here's a simple, even
simplistic way to characterize the problem, though somewhat different from the
usual characterization:
1. Higher education is, at it's core, about knowledge production and the trainin
g of knowledge creators.
2. Knowledge production requires research: open-ended, prone to failure research
. What's called blue-sky research.
3. Blue sky research is mostly public funded, but channelled through the governm
ent.
4. Public funding for research is dependent upon public support for higher educa
tion.
5. Public support for higher education is conditional upon the benefits of highe
r education being available to people who want it at a cost they can afford.
6. The cost of higher education is rising rapidly. Public support is eroding rap
idly.
7. Therefore: *Rising costs imperil the entire project*.
Let me summarize it in once again: genuine, exploratory, open-ended inquiry that
creates public goods is the one thing we can all respect as a truly noble goal
of research. However, it's the province of a vanishingly small group of people w
ho depend on public support. The foundation of that public support isn't just th
e nobility of the cause, but the fact that the benefits of higher education are
distributed widely in society.

> In one line: your child could improve her life because of higher education.
If that bottom line is unsustainable, either because higher education is too exp
ensive or it's an unsure path to livelihood and dignity, the entire project of k
nowledge production is jeopardized.
## The
Let me
ing as
to the

Knovigator
start with the end goal: the knovitagor. The knovigator should make learn
simple as walking or running, so that education gets integrated deeply in
way we live "naturally."

## Oral Knowledge
The traditional Indian education system was primarily oral. Texts were written d
own on palm leaves and re-transcribed, but for the most part learning involved m
emorization of a vast array of texts, so that a substantial portion of the learn
er's tradition was available instantly. This is not unique to the Indian traditi
on, the Greek tradition was also similar. Socrates famously distrusted the writi
ng down of ideas. While we think of these oral practices as quaint, or even wors
e, emphasizing rote learning over conceptual understanding, there are good cogni
tive reasons underlying these practices.
For one, the emphasis on explicit and error free memorization means that the tra
ditions' knowledge is explicitly, consciously shared among the entire community
of learners. That's always useful and leads to exceptionally good training. In m
odern learning, the only place I know where similar practices are followed is in
theoretical physics and in mathematics, where you could argue that all research
ers share an identical (OK, maybe that's too strong) understanding of the basic
principles of the subject. It's a lot easier to settle disputes and arguments if
we start from the same assumptions. A house built on a solid foundation is much
more stable -- while we might laugh at the longevity of these ancient tradition
s and think of them as outmoded relics, that stable foundation leads to exceptio
nally robust knowledge. It's not surprising that physics and mathematics are the
subtlest of the sciences (am I betraying my biases here?).
Second, it means that by the time the novice becomes an expert (s)he has an expl
icit cognitive map of the tradition as a whole. Knowledge and skill merge into a
n extended mind that makes learning *natural*; because these traditions didn't h
ave access to print technology, they had to rely on extending the capacities of
their own minds and brains. That's a good thing. In many ways print technology t
ook us away from our biological inheritance. More on cognitive extension later.
Third, and most importantly, these traditions emphasize the person over the text
. Socrates is world famous without ever putting a word down on paper; the reason
is because Socrates the man is more important to us than Socrates the writer. K
nowledge resides in people rather than in books. It's very hard for us to apprec
iate that difference - we have an exceptionally abstracted view of knowledge. Ho
wever, in the times we live in, where information and knowledge are changing dai
ly, texts cannot adapt, but people can and will. If we shift our educational goa
ls from making texts to **making people**, we will have accomplished an immensel
y important and necessary shift for the complex world of the 21st century. How c
an we do so and what role can technology play in that shift?
## "Artificial Intelligence"
Let me put it another way: what we mean by intelligence is almost entirely "arti
ficial intelligence," because almost everything we take to be a demonstration of
intelligence comes from culture rather than nature.

OK, I am not saying that nature and culture are two different things altogether,
but I hope you get my point: whether it is learning mathematics or reading or w
riting, almost everything we consider to be a core skill or a universal educatio
nal outcome is "unnatural" in the sense that we don't develop those skills on ou
r own in the natural course of human development.
## Natural and Artificial
Of course, that's the whole point of education isn't it? To go beyond what natur
e and biology has given us? If the only knowledge we know is what we get from ou
r natural developmental course, we won't have rockets and transistors or, opera'
s for that matter.
So there's a tension: on the one hand, we want to make education and learning as
natural as possible, but at the same time, we want our education to go beyond (
while building upon) our biological endowment.
That's the challenge:
* How
* How
* How
ledge

to
to
to
of

build upon our biology?


integrate our understanding of biology into our learning engines?
redesign education from the ground up using our rapidly increasing know
our biological capacities for learning?

# The core logic of cogit


The core of cogit's architecture is *people*. What does that mean? Well, here's
the idea: cogit is a language (that will be embedded in a platform) for thinking
with others. While designing such a language, we can adopt two attitudes:
1. Design for content. In which case, we will center the language around what we
believe to be the content units of thought: ideas, articles, documents, manuscr
ipts etc.
2. Design for people. In which case we will center the language around the socia
l units of thought: mentor-student, peer-peer etc.
Designing for content is the norm. Almost all the publishing platforms in the wo
rld are for creating content, i.e., whether blogs or traditional magazines, jour
nals and books are for highlighting the content, not the person who created the
content. Knowledge and Creative professionals - academics, writers, photographer
s etc - get rewarded and paid for their content. Their output forms the core bus
iness for a range of organizations: universities, opera houses, publishing compa
nies and so on. Not at all surprising, since content production is the industria
l eras way of managing creative output, creating economies of scale around it an
d selling it as a commodity. After all, newspapers and books are objects, they c
an be bought, they can be copied and they can be revised and reorganized. Great!
Unfortunately, we now live in an era where the price of almost every commodity i
s going to zero. If content is a commodity, it too will drop in price dramatical
ly. Actually, let me modify that a bit: there will always be a market (and fame)
for exceptionally high quality content in traditional as well as new media: mov
ies, books, shows and nature/science papers will always have a market and so wil
l the institutions that house the creators of that content: the MIT's and Harvar
d's, Dreamworks etc. However, like any economy driven by commodification, it wil
l soon be a winner take all economy: there will be enormous fame and money for a
few people at the top, but not much for the vast majority of people.
That's already happened in many creative fields: how many musicians, artists and
writers are able to make a reasonable living? While Salman Rushdie (not that he
is egregious, but I can't resist the dig) cavorts with models? Or Steven Pinker

for that matter? The current knowledge system is one of the most hierarchical s
ystems in society - supply of labor far outstrips demand, there's a glut of cont
ent out there and on top, it's one of the few places in society where feudal val
ues are combined with capitalist desires. That's a recipe for an exploitative sy
stem. No surprises when adjuncts have to live on food stamps.
Consider this: in illiterate societies, there was an entire class of workers who
se only job was to write letters to others. I am sure it was a respected, reason
ably well paid profession, just like stenography and simple book keeping. Those
professions don't exist anymore. We have copy editors and accountants, but there
are far fewer of those. Documents are the stuff of everyday white-collar life,
but they are not created by others; we create them for ourselves.
What was true of the *form* of knowledge yesterday is true of the *medium* of kn
owledge today (which is why newspapers and magazines are dying) and will be true
of the *content* of knowledge tomorrow.
> What's still unique and irreplaceable? The creator of the content.
> In the age of almost friction free replica's, the only thing that has real val
ue: the individual.
We cannot replicate them yet. Therefore, it makes sense to build a media platfor
m around people. A platform that focuses on people can create value for creators
by giving them a venue to earn respect for their expertise as well as revenue f
rom the sales of commodity items (books, for example) as well as unique interact
ions (small courses, one off lectures, conferences).
> Texts don't teach students, teachers teach students.
A people focused platform also creates value for people who are learners and afi
cionados for the human bond between mentor and learner is the original position
of learning. Further, it creates a clear ladder on which people can rise: from n
ovice to apprentice to teacher-in-training to teacher to mentor. We are now an e
conomy of free agents. The old human relations have been completely atomized and
commodified. Loyalty to company or institution doesn't really exist, from the C
EO on down. There's a good reason: companies treat people like widgets - they ca
n be fired any day, there's constant downward pressure on salaries, outsourcing
and automation are always threats. In such a world, people need two things:
1. A personal platform (my.brand.com if you will) which you carry wherever you g
o, so that you aren't defined by the places you study or work for, and that acts
as a lifelong dairy, learn/work log, report book and resume.
2. A network of trust, consisting of both peers and mentors. That network is you
r home away from home, your succor in a cruel world.
Human beings cannot survive without virtues: trust, honesty, passion, values. Si
nce our institutions will not provide them, we should self-organize to do so our
selves.
## The Free Agent Economy
There are about seven "social needs" that all of us desire and deserve
1. Family - our immediate and extended relatives and the human bonds that connec
t them. Is often geographically distributed these days.
2. Friends - obvious; often geographically distributed.
3. Community, including religion - people who live close by, people who share re
ligious, artistic or moral sensibilities. Includes mentors who coach you through
life.

4. Work - both livelihood and professional identity. The ability to be productiv


e. Includes mentors that help you master a skill.
5. Interests - personal interests that connect you to a network of people who sh
are those interests. Includes hobbies and sports and mentors or role models who
you admire.
In traditional societies, all of these happened in one delimited geographical lo
cation. In fact, it's quite likely that your social roles in all the five bucket
s above were pre-ordained because of your race, gender, caste etc. The industria
l economy replaced that traditional hierarchy with a new one - bonds of class, l
oyalty to employers, union membership etc. That too is gone. We are all free age
nts now, which, while it could be freeing and lead to autonomy, is actually lead
ing us towards atomization and exploitation. How do we start countering that mov
e?
The first thing is to build a platform around you. Why should your life be distr
ibuted across different institutions that control you? While being in charge of
your information isn't the solution to these problems, it's a beginning. Right n
ow, the data about your life is scattered - your workplace controls some of it,
your church controls others aspects and your family and personal interests are o
ften too scattered to be of any aggregated value. Instead, we should build our l
ife around data that flows through us and where these social fields are API's th
at connect to our personal data store.
Once that's done, then the *flow of data* combined with *real human interactions
* can bootstrap into something that we all wish we had: your own personal wisdom
teacher. Think of it as a technology enabled Socrates, a personalized cultivato
r of wisdom enabled by technology. What we can do now, however is different from
what we could have ever done before. The personal Socrates seems like a one-way
technology: AI mediated wisdom that comes to you. Instead, we should be thinkin
g about two-way technology: AI mediated wisdom that you both *share* and *receiv
e*. Cogit is the grammar of that collective Socratic inquiry.
What's interesting about this model is that it makes sense: moral sense, pragmat
ic sense. It's the right thing to do: hybrid intelligence doesn't make human bei
ngs go away, it doesn't replace them with automation and machines. Instead, it g
ives people the tools to expand their capabilities. By doing so and keeping that
expansion cheap and by building positive feedback loops (via sharing) so that m
y expansion leads to your expansion, we create better capabilities for all of us
.
At the same time, it makes pragmatic sense - instead of driving expertise out of
business, it gives a venue for expertise to find new learners and collaborators
. In other words, by hitching it's star to the flourishing of it's client base,
a cogit based platform creates a trust economy that helps bind the free agent ec
onomy together. That trust means that you are both customer and provider and tha
t increases your commitment to the system.
## The Architecture of Cogit
Now that we have agreed that we want to build a platform around people - with th
e eventual goal of building a collective Socrates, what kind of tools do you bui
ld into such a platform? Well, whatever we do, we must keep one principle in min
d:
> The Y/O Principle: Everything in the platform is either something you *make* *
*yourself** or *make* with **others**. Making oneself can either mean self-trans
formation or something that you produce. Making with others is something that yo
u do in collaboration or in training.

Y/O should remind you of I/O, i.e., the input-output system and indeed, the Y/O
has an I/O embedded in it. When you make yourself or you make with others, we ca
n always look at the needs of the users of that making platform as an I/O model:
> Input --> Processing --> Output
The input piece is *learning*, that which you get from others; it's fundamentall
y embedded in a mentor-student relationship. The processing piece is *research*;
it can be mentor-driven or peer-driven. The output piece is *media*, which can
either be teaching-driven or publishing-driven. Each one of these is a vertical
, i.e., a separable component of the system that can be (and should be!) a *life
long commitment*. Let's look at these three verticals in some more detail:
1. Learning
2. Research
3. Media
### Learning
There are tons and tons of ways to approach learning, but we have to remember to
design our system around people. Once that principle is kept in mind, designing
a learning system is simple; we build communities around teachers. Our goal is
to make it as easy as possible for potential teachers to create and share learni
ng content and build a community around the content they share. It's exceptional
ly important that these teachers be urged to create a voice of their own. They a
re not widgets teaching the same material. Instead, they are masters who have a
point of view about the material they're sharing. Establishing a point of view n
eeds four *outputs* that I am listing below in order of commitment:
1.
2.
3.
4.

Curated sequences
Mentoring
Courseware
Publications

#### Curation
The easiest way to establish your own voice is through curation and the easiest
way to curate knowledge material is via learning sequence. The curated sequence
is the easiest thing to explain: It basically says:
> If you want to learn X, start at X_0 and end at X_N == X.
Of course, as part of the cogit language, it must have some design principles bu
ilt into it. For example:
> If X_N == X, then N <=7. Perhaps N is somewhere between 4 and 7.
The idea is this: people don't have the mind space for more than 7 chunks unless
they are really well trained. So keep your sequences to a length that people ca
n cognize.
#### Mentoring
Mentoring can mean various things, but here, I mean the smallest additional ste
p beyond curation. If in curation you gather resources and publish them, then in
mentoring you add a question and answer layer on top of these curated resources
. That's all. A simple way of thinking about it is:
> When you curate you add an intro and summary to a series of resources and tell

a story by doing so. Mentoring adds a Q&A to that so that your readers can ask
you questions about the story you are telling and you can respond.
#### Courseware
Courseware is a much more serious commitment on part of the mentor - they have t
o not just curate and answer questions, but also put forward a teaching program
on their own (or with others). It means, that you have to replace the curated co
ntent with your own content or at least layer your content on top of the curated
content and be available consistently while the teaching is happening.
#### Publication
Publications are the highest level of commitment: they are a layer on top of you
r teaching. The idea is this: by working on a topic with others, and then teachi
ng it, you will create enough material for a publication - either individual or
joint. That publication should be pushed out into the world.
## It's all about space
Note how space is so central to this conception of cogit. Everything in cogit is
either a locus where people get together to learn, create and teach knowledge o
r a flow of information and knowledge between loci. This is of course the standa
rd stock and flow model that's well known in systems dynamics. The point is that
knowledge as a system is no different. It should also be modeled as stocks and
flows, except that we are adding spatiality to the internal structure of stocks
and flows themselves. The spatiotemporality (in the abstract sense) of stocks an
d flows gives us a much more powerful language for talking about them than just
a notation does (which is what the traditional stock-flow model gives us).
If this reminds you of category theory, you would be on the right path. Except t
hat we don't want to be tied to particular mathematical ideas - we want to addre
ss the *underlying problem* that category theory solves for mathematical purpose
s, not the actual formal structure of category theory itself. So what is that pr
oblem?
> The basic problem is to understand the construction of a commons where a) agen
ts share information and knowledge b) build that commons with layered building b
locks that can add or subtract structure c) enable flow of knowledge from one co
mmon space to another.
Category theory solves these problems for mathematics, but most other discipline
s don't have the same rigor that mathematics has. For example, consider the prin
ciple of naturality in category theory. It's exceptionally unlikely that real wo
rld systems obey naturality. In general, real world systems are only approximate
ly symmetric. Therefore, we don't want to impose the mathematicians mindset on t
he problem without understanding it's design characteristics in detail.
Here's another way of putting it: the architecture of cogit should lead to the a
rchitecture of category theory if the design specifications become those of math
ematical structures, but not otherwise. We are aiming for a more general theory
of world-making.
Ultimately, what I believe is that we need a general study of form and organizat
ion, that takes seriously the claim that all form is designed without having a d
esigner. Cogit is a language of the design of form itself, not any particular th
eory of form, whether that be mathematics, computer science or architecture. Fro
m that perspective, it makes sense to uncover the design principles underlying o
ur various knowledge systems such mathematics, physics and history. Another way
of putting that uncovering is: if we think of a world as a cluster of conditione

d existence, then the purpose of the design principles approach is to reveal the
conditions that constitute that world's skeleton.

You might also like