In the natural world, adaptation by any organism is a process whereby selected information about its environment is absorbed into the organism, allowing it to develop in a direction, seemingly in defiance of all laws of probability, from the less orderly and more probable, towards a more orderly and improbable degree of harmony with that environment.
In other words, entropy must be moved from within the organism out to the environment, as opposed to the opposite process, which statistically prevails in an inorganic ecology. Any organism that fails in this fundamental process will simply become inorganic.
The disagreements regarding the ontology of knowledge seem to blossom at the very philosophical characterization of the information being absorbed in this adaptive process. Various philosophical schools of thought have come to differentiate information from the concept of knowledge by introducing the intellectual demand that knowledge be both understanding and commonly understood. This philosophical argument muddies the water when one is considering the fundamental nature of knowledge, as this demand denies the logical prerequisite of obtaining knowledge by the individual organism in the first place.
In the design of the Organon Sutra, an effort was made to circumvent the academic battle of circular definitions inherent in any dialog based on these intellectual qualifications by returning to fundamental definitions, by asserting that all epistemic activities swell from the context of the biological adaptation just defined.
This effort attempted to avoid the philosophical disagreements and further the design goal of an intelligent agent by being very specific in the characterization of the information being absorbed in the adaptive process just described.
This specificity begins with a detailed conversation on the phrase “selected information” as used in the description of adaptation, with the focus of the conversation being an emphasis on the word ‘selected’, without which the dialog unavoidably degenerates into an intellectual clash of words.
And yet, even before the conversation begins, the question of what mechanism is performing the selection of information in the adaption process seems to leap out of the narrative.
At first glance, the interaction of an E. Coli bacterium with its environment seems fairly simple. But if we study this single-celled organism closely, we find that E. Coli can sense changes in temperature, acidity, certain chemicals, osmolarity, and most importantly, its main food source of glucose molecules.
Using flagella for movement, the organism is capable of a variety of motile responses to these environmental signals. It can move toward greater gradients of glucose solutes, and it has specific mobility strategies to move out of, and even maneuver around, areas that it senses to be toxic. And in addition to adapting to the current environment it finds itself in, it must compete with the immense amount of other bacteria that are present in its habitat.
This surprisingly wide range of behaviors for a single-celled organism demonstrates a “knowledge” of its environment just as sophisticated as multi-cellular organisms having a fully developed nervous array. Now certainly, the relatively brief lifetime of a bacterium does not afford the opportunity to learn these behaviors. So where does this sophistication come from?
The term for these cellular smarts is ‘phylogenetically acquired knowledge’ (PAK), which refers to all of the survival wisdom accumulated into the genome of an organism over the billions of years of its evolution.
And referring to species in general, the genome can only orchestrate adaptation when the environment is stable across generations. As genetic knowledge is accumulated, the stability of this genetic retention within a species is clearly correlated with that of the constancy or degree of change in the environment of that species. We see this in the sea, the most unchangeable habitat on earth, where organisms are subject to little environmental selective pressure in their genetic makeup. This is also the case with our E. Coli bacterium, whose genetic makeup has varied little within a relatively unchanging environment.
Contrast this with Early Man, descending from the steady environment of arboreal life with the primates, by adapting to the grassy savannah and developing true bipedal capabilities. This newly adapted form of locomotion (along with the concomitant erect posture, freeing the hands to do more than support mobility) allowed him to migrate with more promising food sources and flee from undesirable and threatening locations, which presented him with a changing environment potentially on a daily basis.
This environmental instability created an unprecedented demand for novel survival responses to that variable environment, which drove the evolution of an unprecedented central nervous system.
In a way, our ubiquitous bacterium is born far smarter than every modern day human infant. Evolutions’ solution to the bacterium PAK and the human PAK differ because over a few billion years, the environment of the bacterium has been far more predictable than the environment for the human. The bacterium (and most other earthly species) can survive with a “pre-programmed” PAK for survival, but the unpredictability of the human environment demanded an entirely different form of knowledge.
For sure, man has a far more biologically complex and sophisticated organic body than does the bacterium, which requires a far more sophisticated CNS to maintain basic organic regulation and metabolism. But there are many mammals with comparable organic development and sophistication that have not been endowed with intelligence.
The development of these novel survival responses in Early Man created what is known as evolutionary feedback cycles, which created nonlinearities in conventional Darwinian evolution, accelerating the natural genetic adaptation of his central nervous system on two different levels.
On the first level, the dynamic environment of Early Man forced Nature to develop human organisms that were born with less and less phylogenetically acquired knowledge. For the first time in biological history, organisms were being born with less and less ability to survive in their environment, reversing 3 billion years of evolutionary precedent.
This created an extended altricial period for prehistoric human newborns that was made possible by the second level of the evolutionary feedback cycle. The transition from an arboreal environment to life on the savannah also altered the social structures of prehistoric man, and especially prehistoric woman. Because of this transition, the estrus cycle in prehistoric human females changed from that of their primate ancestors, which prompted a societal shift from the dominant-male-centered troop organization of the primates to a more communal, quasi-tribal culture, which allowed an even higher degree of adaptability to that more variable environment.
This new communal structure freed the females to develop more after-birth nurturing skills, thereby further extending the postpartum development period of their newborns. Male-to-male competition became less important as the competition for mates was refocused to a need for the males to develop, and display skills, both protective and provisional, to assist with these ever demanding nurturing requirements.
We can understand the importance of this extended nurturing by making a few comparisons.
The chimpanzee, our nearest relative as a species, has a gestation period of about seven and a half months, and reaches adulthood essentially in nine years.
At the time of birth, a chimpanzee’s brain occupies a volume of about 350 cubic centimeters, and is already well developed, growing to only about 450 cubic centimeters by adulthood.
The human brain is (not coincidentally) also about 350 cubic centimeters in volume at birth, on average. But unlike the chimpanzee, during development it explodes in size, growing to about 1400 cubic centimeters on average by adulthood.
After birth, a little more than 20 percent of a chimpanzee’s brain is developed while exposed to the environment it will grow up in. By contrast, 75 percent of the brain in humans is developed while the human is exposed to the environment it will be adapting to.
This is three times the development percentage of our nearest relative on the evolutionary tree, and perhaps our closest neighbor on the intelligence scale. Only 25 percent of the brain in humans is developed before the environment begins to soften its edges, totally opposite of the progression of the chimpanzee, and far more than every other organism. (Perhaps this is why teenagers of every new generation tend to revolt against the old norms, they are actually trying to reconcile the adaptation of old and new environments. And this surely must have what the Austrian ethologist Konrad Lorentz was referring to in his paper “The Enmity Between Generations and Its Probable Cause”. )
Certainly, this percentage is the forge of intelligence. The true machines of Nature are those organisms that come into being with only a repertoire of phylogenetically acquired knowledge, and do not develop any adaptive knowledge, such as the E. Coli bacterium. It is only as the percentages of adaptive knowledge increases in a species do we see the progression to intelligence.
And to create artificial intelligence, we must develop a similar forge, an analogous furnace from which we can heat and shape the intelligent behaviors we desire to see in our artificial agent. It is this artificial forge that forms the concepts which are the vision for the Organon Sutra.
And this reinforces the assertion that was made at the end of Discussion 5, where it was declared that anticipatory schematas could not be built in to an artificial agent until that agent was exposed to the environment it will be interacting with. Here, we find that Nature has come to the same conclusion, and reversing 3 billion years of evolutionary precedent, demonstrates that adaptive knowledge cannot be phylogenetically engineered. Adaptive behavior can only be developed while the organism is exposed to the environment it will be adapting to.
And this is why genetic processes cannot be used to engender intelligence in any environment, because as was said, phylogenetic knowledge can only be accumulated into the genome while an organism’s environment is stable across generations. Using a genetic vehicle to develop and subsequently store knowledge is not a workable mechanism in a highly variable environment.
And as is the case with the progression of almost all AI research to date, this confirmation only adds to the mounting questions we must answer before proceeding with the design for our artificial agent, because the astute reader will recognize that we have just defined two additional types of knowledge, further complicating an already complex discussion. Our threads of discussion are becoming an increasingly tangled knot.
GO TO NEXT DISCUSSION
GO TO TOP OF DISCUSSION
Copyright © 2019 All rights reserved