The Structure of Intelligence

"Human Nature: a hopeful polemic"
Introduction
1. Chaos and Complexity
2. The Structure of Intelligence
____________________________________


One primary attribute that clearly distinguishes humanity from other species, partisan snarking aside, is our intelligence. As such, an understanding of its structure and function is likely to be central to any description of human nature. Decades of research have been invested in the relationship between brain and intelligence, but I wish to call your attention to one recent book (courtesy Instapundit, around two years ago) which I consider to be incredibly significant. "On Intelligence" (2004), written by Jeff Hawkins with Sandra Blakeslee, introduces a powerful theory of the neurological structure of intelligence, and discusses the impact such a structure will have in the quest for artificial intelligence.

I find this book incredibly significant because I have come to believe that the memory-prediction framework at the heart of Hawkins' theory of intelligence is the primary piece in a set of rules which convincingly account for the entire spectrum of human thought, capability, and behavior. I STRONGLY recommend reading his very accessibly written book, but in order to provide necessary context I will attempt to briefly summarize the parts of his argument that are essential to understanding my own. (for the record, we have no affiliation, although I did send him an email alerting him to an early version of this post! Therefore, any and all mistakes are entirely my fault.)

In a nutshell, here's how it works.

Basically, your brain creates and then auditions pattern-recognizing models for their ability to sieve useful information from the torrent of incoming sensory data. This is what gives you the quick reactions you need to function in real time, in spite of the sluggishness of neurobiological processing speeds and the overwhelming volume of data.

The most successful models are reinforced and optimized, becoming dominant. Why? Because a model accurate enough to dependably recognize a repeating pattern essentially allows prediction of the completion of the pattern before it happens, which allows the brain to anticipate events, thus enhancing the chances of survival.

These predictive models are then layered to achieve more sophisticated predictions. Hierarchical layers are created when existing models are dropped below the threshold of consciousness, a process which automatically happens once they reach a sufficient level of predictive precision to become usable. Processing power then becomes available to construct new models. These learn to recognize patterns in the outputs of completed models, which have already been pushed down to the next level, and the process repeats. The deeper the stack, the more sophisticated its predictive capabilities are. Events not predicted by a given model are passed upward through the stack until they are accounted for, or until they reach your awareness.

You can't prevent this process any more than you can command your heart to stop beating. It's what your neocortex does. Hawkins maps this activity to specific neurological structure of the brain, and does a much more thorough job of developing this memory-prediction framework in his book than I can here. The gross oversimplification I am inflicting on his material is necessary in order to hold down the length of this post!

By way of a clarifying illustration, your ability to read this essay offers a good example of how this hierarchical layering works:

When you learned to read, you started by learning to recognize letters. Once your letter recognition models got good enough to become dependable, your brain started to use their output as the basis for constructing new predictive models that could recognize entire words. This is why it isn't necessary for you to consciously process every single letter when you read, although you can, if you focus your attention appropriately.

Extending the example, we can say that further modeling of phrase units, sentence structure, grammar, and composition are the subjects of yet higher layers, and your ability to coordinate the motion and focus of your eyes make up lower, older, more basic layers. Your knowledge of vocabulary and grammar actually allow you to predict what you will read before you finish each word or senten (see?) Hawkins argues convincingly that this predictive element is pervasive and central to the operating structure of intelligence.

The process works in reverse as well; if I wish to run up a flight of stairs, I start with a high-level prediction (running up the stairs) that propagates backward down the hierarchy of models that describe the correct sequence of muscle firings until it fires the appropriate individual motor neurons in the appropriate complex pattern, which is why I don't have to micromanage individual muscle movements while I am on the stairs. The better the models, the more natural and efficient the movement, and the deeper in the stack they are, the more unconscious your movements will be.

The expression "muscle memory" is often used to describe deeply learned complex movements, especially those of musicians and other performing athletes. The speed and relaxed precision of their movements are due to the layers of highly accurate models of body structure and function, integrated with equally well developed models of the execution of the task at hand, whether it's meeting a ball with a diving catch or playing a cello concerto from memory. The unconscious ease of a physical talent at work bespeaks entire subsets of accurate models pushed below the threshhold of consciousness, whether they were discovered quickly and "naturally" (a great definition of intuition!) or learned and refined more slowly and painfully through extended trial and error.

Interestingly, this same structure also provides an explanation for the phantom limbs sensed by so many amputees; the predictive models remain even when the limb is gone.

I'll leave you with one more tidbit: Hawkins also points out that because the patterns being recognized by these models are composed purely of neurological activity (sound, light, and touch already having been converted to nerve impulses before reaching the cortex), there is absolutely no difference between abstract and concrete concepts at this level. While "abstract" patterns can be harder to recognize because we have had far less experience in recognizing them than we have had in learning to operate our bodies, they are not handled any differently within the brain. This is an extremely important insight, and we'll come back to it later.

Now the above summary is merely a taste--there is much more to the book, including a much greater emphasis on mapping these functions to brain structure--but it should give the flavor of Hawkins' theory.

What can we do with it?

That will be the subject of the next chapter.



While I construct the next part of my explanation, I am posting what I have just written. Check back soon for more!

Chaos and Complexity

"Human Nature: a hopeful polemic"
Introduction
1. Chaos and Complexity
2. The Structure of Intelligence
____________________________________


Books can and have been written about this subject alone (my library has James Gleick's Chaos (1987) and M. Mitchell Waldrop's Complexity (1992), which are great places to start if you're interested), so I won't spend too much time on it. Suffice to say, for the moment, that the advent of computer modeling offered a new path to realizing that the emergence of complex phenomena from the interaction of simple building blocks was not just possible, it was ubiquitous, and it could be seen to follow undiscovered sets of rules. Thus were born Chaos Theory and Complexity Theory, which are rapidly spinning off areas of specialty.

They can describe things like the unpredictability of weather, the shapes of trees and clouds, and the flocking behaviors of birds--all things other branches of mathematics have great difficulty with--in terms of bottom-up organization, where iteration of the right simple procedural rules that govern the interaction of individual agents (like birds) allows the correct modeling of the complex behavior of the entire flock.

The nonlinear dynamics of many complex systems prevent exact determination of the future positions of individual agents, because the iterative nature of the rules (the results are immediately fed right back into the system) induces great sensitivity to initial conditions. The tiniest errors in measurement are quickly magnified, utterly changing the outcome.

However, many such nondeterministic systems possess a different sort of order. The Lorenz attractor is one of the first examples found of an odd class of objects now known as strange attractors. For a certain range of values plugged into the three simple equations describing this nonlinear system, the solution will oscillate in three dimensions around and between two fixed points (the attractors), but (strangely) without ever repeating itself, crossing its path, or becoming deterministic.

The path it takes is smooth and continuous, lending it an almost hypnotic motion which makes it fairly predictable in the very short term, and its three-dimensional boundaries become evident over time, lending the structure an easily recognizable and beautiful shape, but long-term prediction of the moving point's specific whereabouts is impossible because of the magnifying effects nonlinear systems have on measurement errors. If you want to develop a better feel for how this works, you can play with an interactive version of the Lorenz attractor here, or a more complete one here if you like.

As my first Lorenz link eventually indicates (and you can test this yourself using my last link), not all possible values for the three constants in the equations trigger this complicated behavior sustainably. For some values, the system spirals down to a single attractor, becoming stable. It takes the right range of parameters to create its trademark chaotic oscillation between two strange attractors.

Why am I presenting all this?

Because as I see it, the intertwining of simplicity and complexity exhibited by these types of systems offer a compelling metaphor for the relationship between human nature, societies, and individuals. Equations whose simplicity is apparent (even to those of us who aren't mathematically literate enough to fully understand them) can describe systems which possess both broadly identifiable tendencies and unpredictable individual behavior, although neither attribute need be laid out explicitly in the equations themselves. It is thus entirely plausible that the complexity of human nature may be explained in terms of the interaction of a few simple rules.

Furthermore, the sustainability of the complex life such a system exhibits is impacted by changing the constants in its underlying equations. As a societal metaphor, the attractors might represent the stable anchors which not only prevent a dynamic free society from dissolving into anarchy, but also preserve its capacity to adapt and avoid stagnation. Changing the constants might change the nature of the attractors, possibly causing society to spiral into one, becoming locked in and losing too much of its adaptability to be sustainable over the long term, or go the other way and fragment entirely into anarchy. Conversely, given a universal human nature, a failed society might be brought into full flower by the right adjustments to those constants.

The more we know about the underlying equations that govern human nature, whether they are mathematically stated or not, the more hope we have of deciphering the effects changes in constant values will have on societal attractors.

As you can see, the vocabulary that has grown up around the Lorenz attractor and its relatives offers us an intriguing language with which to describe these relationships. You might say that through experience we have identified some of the attractors for human society (democracy, socialism and tribalism, for example) and human thought (liberal, conservative, monotheistic, etc.) Can we discover the governing equations of human nature which shape both the orbits of individuals and societies and their strange attractors? Could a better understanding of them allow us to safely apply new values as constants, opening up access to currently unsuspected capabilities?

At the moment, my answers are yes, and I don’t know.

The next chapter, "The Structure of Intelligence," can be found here.

Human Nature: a hopeful polemic

-Introduction-
1. Chaos and Complexity
2. The Structure of Intelligence
____________________________________


If we could all agree on the rules by which the game of being human is played, we might stand a better collective chance of winning.

Of course, in order for such a set of rules to be universally accepted, it would have to convincingly account for the entirety of human philosophy, behavior, and capability, with absolutely no exceptions. Republicans, pacifists, teenagers, suicide bombers, and concert pianists would have to co-exist within the same completely self-consistent framework, together with everyone else in human history. While reconciling the entire range of conflicts generated within and between individuals and cultures under one hopeful human banner remains a noble goal--one that has inspired the searching of countless theologians, philosophers, scientists, and college freshmen--it has also proven to be an elusive quarry.

I would like to offer my two cents on the subject.

These perennial human conflicts are quite clearly fed by potent springs. But if their power can be traced to the fundamental hydrological dynamics of human nature, and those dynamics can be spelled out, then perhaps their energy can be drawn to more productive outlets in the surface topology of human society.

An accurate mapping of the proximity of the varied human terrain above to the unified dynamic forces beneath might even reveal why different societal structures use different methods to tap that energy source. If we could identify the underlying rules which completely describe human nature, perhaps we could then identify the topological conditions under which various incomplete notions of human nature might steer a culture towards democracy, tribalism, or socialism.

Instead of imposing a utopian management of imperfect human nature through government, I propose the unleashing of human potential, including that of government, through a more complete understanding of human nature itself.

Any comprehensive theory of human nature must account for both the existence of many differing religions and the passions aroused by comparing them, but it does not need to endorse religion any more than a theory of digestion or blood circulation would, as long as it can focus on the mechanics of being human rather than the purpose. It is not necessary to discuss the nature of God in order to describe the structural function of my knee, as long as I don't get into how exactly my knee came to be engineered the way it is. As long as a theory doesn't preclude any particular religion, there remains room for pragmatic consensus on structural grounds.

My purpose is to advance our understanding of the human condition, and thus our ability to improve it, by articulating the laws which govern the deep and truly universal structure of human nature. I believe that humanity and all its works, for good or ill, can be reconciled within the framework I wish to present, and that understanding its structure will allow us to unleash the incredible potential inherent (and largely latent) in each human being.

Describing human beings as imperfect only masks our lack of knowledge, by measuring what we aren't; that measures the shortfall of humanity against our own expectations, rather than human nature itself. Negative definitions can help to narrow the focus of a difficult question, but they cannot efficiently describe the actual properties of anything complicated. Simply declaring human nature unknowable is no longer sufficient, as technology has amplified the destructive reach of the determined individual to the point where entire societies are at risk. We need something more precise.

I submit to you that it is our understanding of human nature that is imperfect, rather than human nature itself. I wish to introduce a powerful set of unifying rules and observations, most of them originally articulated by other people, which when taken together appear fully capable of accounting for human nature as it really is. Ideally, this description will provide a lever to move our political discourse out of the rhetorical firebombing stage and into a realm where solutions are actually possible, for I do believe that they are within our reach, if not yet firmly in our grasp. Firmer knowledge of the ground rules of human nature will allow us to transform the architecture of our human institutions where necessary, and it will go a long way toward showing us how.

The underlying rules which govern human nature are far simpler and far more elegant than what you might expect, given such a vast and untidy subject. This is because they are rules of process. They function much like the simple physical laws which constrain the intricate growth of crystals, but result in the endless individuality of snowflakes. Simple components, when they interact according to the right sorts of simple rules, can produce bewilderingly complex results. Various attempts to describe human nature have succeeded in finding the measure of some of its facets; my purpose is to reveal the hidden order which governs the formation of the entire crystal. As I hope to demonstrate, a complete structural understanding of the full scope of human nature will show us how to use both our differences and our similarities to develop the incredible potential waiting to be tapped in each individual human being.

I believe we haven't found these structural laws up to now because of several factors, including these: 1) human beings have only recently become computationally aware of the order that lurks in chaotic and complex systems, and the existence of that counterintuitive order is not yet common knowledge; 2) because of the complexity of the brain and the difficulties involved in studying it, we haven't got a universally accepted theoretical model of the relationship between intelligence, emotion, the brain, and consciousness; in fact, we haven't even had ANY overarching theory of how intelligence functions until very recently; and 3) the human beings trying to puzzle out human nature are not themselves immune to the cognitive and emotional effects of being human.



Because of the burgeoning length of this post, I am now experimenting with breaking its content essentially into chapters, each receiving its own linkable post. The next of this series of chapters (at present!) deals briefly with the promise of Chaotic and Complex systems, and can be found here.