January 19, 2004

Den Beste's "SuperHuman Intelligence"

First thing this morning, I got engrossed in Steven Den Beste's bogpost entitled "Superhuman intelligence".

He starts with serial versus parallel computer processing, then onto the human brain , and then this section about visual processing in the brain:

The vision center has been studied pretty intensively, however, and is moderately well understood now. In humans it occupies a section of the brain in the rear which is perhaps six centimeters wide and maybe 12 centimeters tall, and its job is to take a huge stream of data from the eyes and to process it to remove redundant information, to compensate for distortion (such as due to effects of lighting), to try to identify which sections of the field of view represent "objects", to evaluate differences in the images captured by both eyes and to calculate depth information from that, and in the end to create an abstracted symbolic representation of what is being looked at to send elsewhere in the brain for further processing at a higher level. The division of labor between the vision center and more general computing elsewhere is not known, nor how that abstract description is encoded, nor how we remember objects we've seen and can recognize them when we see them again, or identify the function of objects we've never seen before (like knowing that something is a chair the first time we see it).

At least in the early stages, visual processing is highly pipelined, probably with several parallel pipelines, and with groups of neurons at each stage of the process using the output of previous stages to create greater and greater degrees of abstraction about the image. So, for instance, in an early stage of the process in one pipeline there are neurons which are wired to look at a small number of pixels (i.e. the outputs of a small number of specialized light-detection nerves in the retina) all of which are physically very near one another, and to decide if there is a transition in brightness or color or texture between them. In a later stage, there are nerves which look at the output of transition-detectors which are evaluating contiguous regions, so as to try to determine if there is an "edge", and if so to determine its orientation and curvature.

By its nature, the most important aspect of all this is the details of interconnections between the neurons. If a transition-detector cell isn't actually watching contiguous pixels, then it won't be possible for it to make any deduction that is useful. But physical structure at that level of detail cannot be specified using the kind of encoding which DNA uses in the human genome. The gross structure of the brain is controlled, but wiring at the level of exactly which neurons hook to which other ones can't be.

Nonetheless, it works, because it is "self organizing". In the visual cortex of a newborn baby, the neurons are actually far more cross-connected than in adults. And in the first few months of life, one of the things a baby has to do is to literally learn how to see. Individual neurons begin to learn which of their inputs are actually important and which are not, and to pay attention to the important ones and ignore the unimportant ones. (Whether the unimportant connections are physically eliminated or remain connected but are ignored is not known. It may be a combination of both.)

It got my attentioin because he's writing about how we see with our brains and in this excerpt, he ends with the idea of "self organization"... a topic which has a political dimension which is extremely timely.

Later, he writes about hive minds and the internet:

But hierarchization can be used in internet-based hive minds just as it is in existing organizations, and most members of such hive-minds won't need to communicate directly with one another.

We know this can happen, because it's already happening. "Open Source Software" is one example and result of formation of that kind of hive-mind. Political blogging is another.

When it comes to blogs, power-law distribution tends to spontaneously induce creation of hierarchical organization, with contributors to high-traffic sites becoming de-facto leaders.

At any given instant, there will be a myriad of such hive-minds exhibiting a broad range of behaviors and capabilities. New ones would form all the time and old ones would falter and die. Some will only "think" in very restricted realms, while others may range broadly. In terms of observable intelligence level they would fall on a continuum. Some might be "geniuses", some "morons", some might operate at the level of animals or even plants. There will be the equivalent of diseases (i.e. chain letters). There will be the equivalent of parasites (file sharing networks pirating music and movies). Some will be valuable and constructive, some will be trivial and useless, and some will be insane. Some of those will be dangerously insane.

Sometimes hive-minds will form in response to others in order to oppose them. Hive-minds will compete and contend. Some will cooperate, forming coalitions. Sometimes that will cause them to merge. Some hive-minds will break into pieces, yielding children whose contributing members sort themselves based on their disagreements. And generally they'll be self-organizing, and many will be able to adapt to changing circumstances.

Memes are the thoughts of hive minds.

Posted by Dennis at January 19, 2004 10:06 AM

Leave a comment