Sunday, March 09, 2014

orderly abstract intelligence is a highly principled remix

When arguing against the natural evolutionary origin of humanity, flattery doesn't hurt. Consider the entire class of arguments which praise the uniqueness of humans in comparison to other biological organisms. For if humans are so incredibly special, then how could they be products of natural processes? How could their distinctiveness arise from an evolutionary continuum?   

Most often, the topic of human uniqueness focuses on intelligence and behavior. Does any other animal do _____? Is any other animal motivated by ______? Can any other animal easily solve puzzles like _______? And if not, then how are humans different? Expressed in a single debate-ready question: what is the mechanism behind the human flair for abstract intelligence, as opposed to the earthy intelligence demonstrated by "lower" animals?

Of course, in the context of evolution denial, mentioning this mysterious question probably isn't intended to lead to physical answers; the questioner is probably insinuating that the question has an equally mysterious metaphysical answer. However, regardless of the intent behind it, the question is still intriguing for anyone who accepts materialistic naturalism. Clearly there isn't yet a proven definitive answer to it—everyone will notice when there is one, because it will lead to unmistakable artificial intelligence. Meanwhile, informed speculation is necessary to fill the gap.

So speculate I shall. I concede that humans' abstract intelligence must operate somehow. My current thinking is that the question is framed to be subtly misleading. It deliberately hints that abstract intelligence is an additional centralized semi-independent ability which only human brains possess. In this view, the non-abstract, i.e. concrete and animal-like, portions of the brain receive and refine sense data and primal drives. Then those results flow into the supplementary "humans are awesome" brain part (e.g. the antiquated hypothesis of the pineal gland's function), where abstract intelligence performs miracles on the information.

This simplistic paradigm is unlikely. My current impression is that abstract intelligence acts more like an energetic remix of concrete intelligence. It's a complex set of coordinated interactions with all the types of concrete intelligence; it reuses, repurposes, redirects, even reprograms. It achieves unprecedented combinations. Although it's certainly a novel innovation in brain design that obviously produces drastically exceptional output, it's not built as a completely distinct type of intelligence. This more complicated paradigm is compatible with the known structure of the human brain, which indeed includes lengthy and numerous interconnections among the various parts that are devoted to identifiable types of concrete intelligence. The prefrontal cortex is a famous site for the junctions of these prominent interconnections, thereby creating excess routes.

It's not difficult to imagine the routes guiding the flow of data. But what if abstract intelligence also consists of routes for the flow of delegating reasoning? What if it's partly a metaphorical engine for translating data from one type of concrete intelligence to a second, and enabling the second's specialized processing on the initial data? What if the primary strategy is to "summon help" from diverse intricate types of concrete intelligence that have evolved for eons? For example, invoking the visual cortex to examine an abstract image as if it had originated from the eyes. If this were accurate, then it's an amusing instance of inversion of control, like the tail wagging the dog. When the triggered "lower" intelligence sends the solution back to the dispatching junction, the solution itself triggers yet another flow. I suppose this abstract give-and-take could last for a while, as intermediate solutions provoke further intermediate solutions.

On the other hand, unhampered conscious remixing of both data and reasoning would be chaotic. Under normal conditions the operation of abstract intelligence isn't a subjective experience of total bedlam, so presumably the remix obeys any number of organizing principles. The first could be the presence of fresh salient external data. Perhaps such data interrupts and overpowers most abstract loops of brain activity. Evolution surely has rewarded brains which prioritize sudden pertinent events in the organism's surroundings.

The second organizing principle could be the bonds supplied by narratives. Narratives furnish a strongly coherent context for abstract intelligence, and the importance of context shouldn't be underestimated. In fact, I suspect that narratives are extremely influential and versatile tools for steering human thought at all levels of abstractness. Narratives derive strength from three features. First is language. The decoding of language systematically and rapidly activates swarms of ideas. So any organizing principle that harnesses language already has extensive access to the brain. But note that strong narratives can also be nonverbal. The second strong feature of narratives is episode. Like episodes in reality, episodes in narratives can have wide-ranging effects simultaneously. An episode isn't one thing, sensation, or concept. It's a blend of many in a momentary span of time, and the blend can cause any number of cascading reactions in the brain all at once. And in effect the blend reinforces the mutual association of all the data in the specific episode, according to the proverb for behavioral conditioning: "neurons that fire together wire together." The third strong feature of narratives is sequence. Narratives present the episodes in a particular sequence, and thus sequencing is an indispensable clue for comprehending and extracting relationships among episodes. The sequence is raw material for countless examples of abstract intelligence. It may tie together otherwise disparate data items: the sequence of merging any collection with a count of two and any collection with a count of three yields a collection with a count of five. It facilitates replaying or reciting: even a familiar telephone number demands more effort to state in reverse. I'm deeply interested in the precise representation of sequences in the "massively parallel" brain, because generally parallelism and sequencing are hard to reconcile.    

The third organizing principle could be timing. This includes forms of external synchronization such as tempo, rhythm, and repetition. Perhaps insistent timing effectively filters brain activation patterns, through direct disruption of patterns that don't match the timing. Moreover, timing seems like an unavoidable factor in the brain's coding of information. Nerve cells need to activate in timed groups—each nerve cell is too simple to encode information by itself. Each cell's contribution to the "meaning" is interdependent on other cells activated at that time, like a single minuscule display-dot that at one time is part of a "B" and at a later time is part of a "6". Timing shows up significantly in more ways, too. Overall brain activity has a customary timing, such as alpha waves or beta waves. During conversation, unexpected timing of phrases is considered highly distracting...and a sign of non-native speakers. During abstract data handling, timing is related to the boundaries between separate items: the start of the following item shouldn't be timed before the end of the preceding item. I wonder if it might encode data containment as well, via several coinciding timings of differing lengths. The larger enclosing unit could have a long timing, and the smaller enclosed units could have shorter timings that begin and conclude within the longer timing of the larger unit. Needless to say, music of all kinds is a prime example of the operation of this organizing principle.
          
The fourth organizing principle could be remodeling: restructuring the brain in order to act as a better-fitting model of specific information. As previously mentioned, remodeling ultimately occurs at cell level, when each cell's selective activation has lasting (or should I say memorable?) side effects by activating another cell. At coarser levels of detail, remodeling is merely a cumbersome synonym for learning—not just learning greater motor-nerve control over the body but also over inaccurate abstractions. Brain remodeling is like continuously redrawing a map by trial and error, whether that entails erasing mistakes or adding more information. Drastic changes are less possible at older ages, of course. Meanwhile, at a still-coarser level of detail, remodeling is societal and cultural. Explicitly and implicitly, culture introduces fundamental abstractions and then introduces more and more elaborate abstractions by remixing its fundamentals. Naturally, the preceding three organizing principles are effective tools for brain remodeling. For instance, a ritualistic chant may be a sensory experience that expresses a narrative and is always repeated according to predetermined timing. Remodeling nurtures and vitalizes abstract intelligence's potential. From generation to generation, remodeling ensures that the total cultural products of abstract intelligence are communal. The teaching, modification, and preservation of culture comes from a long-lived population of humans who remodel each other's brains.

After accounting for the constraining influence of organizing principles, it's more plausible that orderly abstract intelligence could occur through remixes of brain impulses. In this way, the mass of unintelligent pieces is neither a disjointed mob nor a rigid clockwork. If it were the disjointed mob, then it couldn't make any coherent cumulative progress. If it were the predictable clockwork, then it couldn't adapt to imperfect or surprising data. The inner workings of a principled remix lies in the middle between these extremes. That mode in the middle really isn't any more obscure than the two extremes, so perhaps it's unhelpful to use the magical-sounding term "emergent" just to describe its large-scale pattern effects. Something similar occurs all the time in other huge dynamically-structured groupings which incorporate small interactions among individual items or subgroups: colonies of eusocial insects, Internet routers, political movements in democracies, buyers and sellers in economies, cells in multicellular organisms, air masses in weather. For these common groupings, usually nobody objects to the simultaneous existence of the whole and also every item. Strictly speaking, the whole itself is an extant grouping and therefore it and its effects aren't emerging. The appearance of emergence is a side-effect of enlarging the viewing scale. The contents of abstract intelligence "emerge" from a large-scale view on innumerable synaptic teams.

Then the relevant follow-up question is how this applies to the experience of emergence in the brain. The other cases of large-scale observation have obvious observers and methods, such as an economist collecting many surveys or a weather satellite tracking many clouds. The counterpart in the brain could be one more organizing principle: resistance. If a cell at a somewhat central position resists activation until it receives a multitude of synchronized signals from other cells, then it distinguishes between small-scale and large-scale activation. Its activation "represents" a trending pattern. The resistant cell "ignores" a tiny impulse sent by lone cells but it "observes" a relatively wide signal sent by cooperative cell groups. And this setup could be repeated again, with entire groups of somewhat central resistant cells chained in turn to other groups of resistant cells. In effect, the cumulative resistance builds and builds, until the cells near the end of the chain only activate whenever the sending group of cells is especially plentiful. The active cell groups overcome resistance by forming the necessary alliance, acquiring the necessary key, lowering the necessary drawbridge, switching on the necessary transistor gate voltage. Some of the most influential coalition cells could be the ones which are tied to pain and pleasure, so that relatively neutral sensations or ideas are less likely to dominate overall brain activity. Resistance produces a selective result. Chaotic cell activations normally lose and are filtered out. Normal consciousness is infamously narrow because there can only be few popular victors. On the other hand, the competition is held a multitude of times, so it's also possible for past losers to overcome resistance later. At first, half-baked thoughts might have difficulty leaping the hurdles to consciousness, but once those thoughts have developed further, the resurfacing i.e. "emergence" could be sudden—the half-baked subconscious thought grew into an unconventional intuitive insight which broke into consciousness.

However, for emergence in particular, I realize that resistance and my other organizing principles are more complicated and/or unpoetic in comparison to metaphysical explanations. It's more comforting to propose that orderly abstract thinking is imposed reliably on the disarray from "above", rather than confronting the proposal that it's constructed from out of the "below" disarray using inherently fallible strategies. Unfortunately, reality on its own continues to refuse to conform to human preferences. 

No comments:

Post a Comment