An Hierarchical Turing Test for General AI

By Herb Wiggins, M.D.; Clinical Neurosciences; Discoverer/Creator of the Comparison Process/COMP Theory/Model; 14 Mar. 2014
.
Have been following Jeffrey Hawkins’ work with some interest, and note his work in hierarchical  temporal memory models. It’s likely that Jeff Hawkins has the closest working model to creating a good simulation of AI.
 .
These insights which may help facilitate creating AI & can be explained in   detail using concepts of comparison process, structure/function relationships, least energy, complex systems thinking, and related methods.
 .
Using structure/function relationships (clinico-pathological correlation) traditionally, we can generate & create unlimited information about how the brain works, and its corresponding outputs, vision, language, motor/sensory thinking, memory, and so forth. These tools which have seemed to work for me to create higher brain functions right off the cortex, the mind brain interface, (cortical columns of Mountcastle), are S/F relationships, found by the millions in clinical comparisons between lesions in brain and functional deficits, & are self evident to us all. However, using a comparison process model, which largely operates in the CC’s, a very much better understanding can be found.
.
The model is both elegant, simplifying, as well as highly fruitful. Because a comparison can be made between or among most every event in existence both inside brain and outside (in the various permutations), it’s part and parcel of the structure/function method. And appears to be a universal tool. Least energy is also a comparison process, comparing energy, resources, time, and efficiencies of different methods, to find new insights. & likely universally applicable, too.
.
 I ask this deep question. How does it happen that the brain is organized upside down & reversed right for left in humans, and very likely most normal primates, as well?
Right hemisphere controls left body and vice versa, in both motor and sensory homunculi. This is well known from the time of Wilder Penfield and numerous published cases. The toes are at the top, & in the interhemispheric fissure. Then moving down the cortex of the motor strip comes the instep, the arch, the ankle, and the ankle is connected to the leg bones, & those to the knee, then the thighs, etc. The face is at the bottom, and the trunk and arms fit in between in a smooth transition. Why is this so? In addition, the visual system is also reversed right for left and upside down, with the sup. visual fields functions processed in the inf. occipital lobes and the inferior fields from 3 to 6 to 9 o’clock in the top parts of the visual cortex. & again, left field from 6 to 9, to 12 on the right visual cortex, & vice versa for the right visual fields.
.
And there is the optic chiasm which takes the right visual field to the left, and vice versa, and the inferior fields to the superior cortex in a smooth transition of reversed upside down & right for left. & the decussations of the pyramids of the inf. brain stem take right sensori-motor afferents from the limbs to left brain, the same way as do the cortical efferents are taken right hemisphere to left body, also. So wherever we find the decussations of the pyramids, we know the brain is most likely upside and right for left oriented. and not just in humans & primates, but likely for most mammals, including the ancient marsupials, and the platypus and echidna.
.
Why & how does this arrangement come about? We hear of the top/down models which were fashionable a few years back, but were somehow not clearly satisfying, either. Somehow incomplete.
.
The best, most efficient answer, & the key to understanding human/primate/mammalia brain neuroanatomy is quite simple. When we take a magnifying class, and hold it at arm’s length, comparing the images of the room around us, the image thru that double convex lens is reversed right for left and upside down. This is the visual comparison which shows us what’s going on. The eyes have it!  The brain is organized visually, and never in my years of study and work has this simple correlation/comparison been described nor discussed. It’s a deep and universal observation, spanning most higher animals. The ability to create new insights using visualization methods is highly human, and species specific. Altho some primates can copy our actions, the meaning of many actions, as in cargo cults, oftimes eludes them.
.
Quite simply, we are visual creatures, & our brains are organized largely upon the images of events upon our retinas, and very likely least energy, as well. We have stereoscopic vision, different from other species whose visual fields rarely overlap, but for our cousins, the primates. The evidence is also in the internal capsule’s organization, that of the massive connections of the white matter to the cortices, and the internal mirroring deep structures of the globus pallida and the thalami as well. A truly universal structure/function comparison.
.
But not only does this occur in humans and land mammals, but also in the dolphins, where the same decussation of pyramids are also seen. We don’t have to dissect these creature to see it, but simply image their nervous systems to find it. & thus we know their hypertrophied sound cortices also are organized in the same way, to highly and efficiently coordinate with the older visual systems. This is basic neuroanatomy, and although the top/down, old at bottom and younger at the top is interesting, it doesn’t have the depth, profundity & universality of the right for left, upside down features.  Even our cerebellum is organized the same, as well.
.
The comparison process when generally applied to data both within and outside of us, gives us the ability to see events in new ways and relationships which we have missed. For instance, colour vision. How does this come about, this model of events in existence? We can compare our brains’ color system with the EM spectrum for new knowledge. We know that orange is a frequency of light in the spectrum. But not in the visual cortex. It can be generated by mixing yellow & red pigments, but still we perceive orange! Blue green, by mixing blue and green, where the blue-green, real and existing frequencies are there, but not a mix. And then purples by mixing blue and red. And browns, a highly significant colour for some colour blind persons, are red/green in combined EM frequencies. Show us by comparison the “brown” frequency in the spectrum!! It’s not there. Thus our rhodopsins interacting with photons’ black/white continuum, create the colours simply and elegantly by mixing. But they cannot convey to us the existence of the spectrum based upon frequency, even tho, in fact, it may use the energies of colours to detect those.
.
Another insight is the rainbow, by which Newton was able to see frequencies of light by creating them from refraction. The rainbow is rare, or is it? In fact, it’s a daily phenomenon. Part of a Kuhnian revolution is “new seeing”. New comprehensions created by a new model. As Einstein once stated, “Every advance in physics is preceded by an epistemological advance.” This might also be true of neurosciences. Where IS the daily rainbow seen? Sunrise and sunset. First, it’s black, hardly any colour, then it’s IR with black/red. Then red, then orange, yellow, a bit of green, and then finally blue sky. We see the greens from the plants, and the browns as well, seeing bark and branches. & with the black and then shades of greys with clouds, the colour palette is set for us everyday. And at night once again with the sunset rainbow, too.
.
Consider that our eyes see colours at the brightest and most common frequencies which the sun creates, yellow/green, the center of our visual sensitivities. If an eye needed to detect light best, it would have highly receptive colour sensation for yellow/green. Which eyes do. Those frequencies, which the sun creates at its highest numbers of photons, thus brightest, give the greatest amount of information. It’s efficient, least energy. If we lived on a planet whose sun were yellow to slightly orange, our photopigments would be most sensitive at those frequencies, would they not? So our eyes are clearly attuned, evolutionarily & structurally to the sun. So is the rest of our nervous attuned closely to other common events in existence. As will be shown shortly. These are but a few of the  of the many kinds of insights which comparison process thinking can give.
.
These insights have been missed, for want of the comparison process and least energy tools. & perhaps, we can still see a lot, by just watching, to paraphrase Yogi Berra. Observation still trumps most all models, and creates, corrects, and extends them, too. & so the sciences are self-correcting by carefully observing repeating events in existence.
.
Thus do we find this comparison process which creates many methods to be useful. At your UC Berkeley campus, Dr. Paul J. Stark, PhD, Statistics chair, has a lovely video in which he discusses, “How do we know treatments have effects? The Methods of Comparison.”
.
“the effect is ubiquitous.” That is, universal in application. and he’s worked it out rather well, though he’s likely talking about methods, plural, rather than one. The comparison methods create information, and as Dr. Karl Friston states, regarding least energy (LE), it’s “consilient”. Crossing many fields in application, as Dr. Stark also writes. Pieces of this model are all over, but not yet fully integrated and developed. This is yet again more evidence of near universality.
.
Dr Friston’s works on connecting least energy to how brain functions is pretty well spot on and well worked out over the last 25 years. My work on evolution from the comparison, LE approach, pretty well confirms his, and that of others who’ve also seen those relationships when applying LE.
.
This is cutting edge evolutionary biology, but few have realized it. & at the heart of it lies LE and comparison processing.
.
When the comparison process and least energy are added in, we get a very great deal of information back out. The inputs and outputs of our CC’s can do this repeatedly, and even input outputs to create the hierarchies of our understandings. Essentially, our language and modeling systems are built upon this. Cognition psychology taught us a very great deal, but hasn’t progressed enough yet, tho intuitively  they realized that recognitions are basic & the case. But what is behind recognition? Clearly, it’s comparison process. When we perceive events in existence, we at once compare those to our long term memories, and if we find a match, we recognize it, that is we “re-know it”. The language, etymology reflect the structure which gave rise to it. and this is another key to understanding understanding and thinking about thinking.
.
The constant calls to LTM for recognition and the learning which establishes those memories are very clearly related. Events in our universe repeat themselves multiply. Those which are important to us, reinforce themselves into LTM. These are the CC’s modeling the repeating events by reinforcement. & then we create our landmarks & so forth upon those perceived events. This is essentially the behaviorist model, but it’s missing big pieces, neurochemical and neurophysiolgoical. It’s incomplete, which will be shown below. The highly repeating events in existence, are yet another basic clue to AI and how it can be created. Plus they weigh in heavily with respect to confirmation of events in existence by the sciences.
.
But when we want to understand description which is largely verbal, we also can find the quantitative version of description, measuring, most useful. When we take a ruler or tape to measure length, we compare the set, stable length standard to the event to be measured. This creates data, or information of length. & when we compare to a word standard, we describe relative to that standard, say a colour among the ROY G BIV standards. Is it red, or yellow, or green, or blue, etc.? So it’s likely that our descriptive measure of events, using words, which are primary, are intimately related by comparison processing to measuring standards as well. They both are comparison methods. Virtually all we do to create information & understanding is thus rooted in comparison processing in our cortices. Again, essential to creating AI.
.
Most words are likely descriptive standards which we compare to events in existence to understand those events. That is we relate them to events. and this is the point, it’s the comparison which creates the information and knowledge, as the measurement creates the data and information also. Ideas/Words are relatively arbitrary standards by which we measure events and describe them, qualitatively.
.
For instance, “son”. We know how he is related to his parents, mother and father, his siblings, and his parent’s sibling as his aunts and uncles and his parent’s sibling. The “relatives” of the son are clear. This builds up the hierarchies of genealogies and how we are related to others in our families. The son has a father, as his father was a son, and can have a grandfather, too, or grandsons and so forth. Each comparison creates an hierarchy. Each relative has set specific relationships created and read by comparison processing. In order for AI to be able to function, it MUST be able to understand, create, and navigate the hierarchies of our understandings with simple facility. This could be termed a “hierarchical Turing test” for AI.
.
And we see the hierarchies all around us. Comparison process both creates the hierarchies and navigates among them. The hierarchies of the dictionaries in alphabetic order, for instance.  Every word ordered by alphabetic comparison processes using trial and error, and dictionaries read by same process. It BOTH writes and reads, & is thus LE. It does lots with a little. The taxonomies of the species, in the millions; the taxonomies of the languages, both current and extinct, each finding its relationship to the others by hierarchic relationships among comparisons without limit of the words in each language (&/or dialects), and how those are related by “comparative linguistics” to the other. The Teutonic are clearly related by comparison, massively. The next hierarchy of the Indo-european. The Ungaro-altaic related to Suomi and Magyar, as well. Each, massively compared by words to the others. The Semitic languages, of ancient Egyptian, Arabic, Amharic, Coptic, Aramaic, etc., are all seen by massive comparison of words to be closely related languages. Each comparison combines to create a huge amount of data supporting the ubiquity and limitless use of comparison process, innate to our cortices. 34 millions of elements & compounds ordered in the IUPAC hierarchical listing, alone; all of them comparison process created, placed and read.
.
This is Einsteinian epistemology as well, because there are NO absolute measures or times or spaces, or much else. Most all that we measure is “relative”, that is, a comparison standard, which Einstein believed was arbitrary, but in fact is not likely that arbitrary. Thus does comparison process tie in neatly with the established facts of relativity, and explain it as well.
.
For instance, he stated that one could use any planet, or any position in our solar system as a fixed point to which we could relate everything else in the universe. But he missed this crucial point. Newtonian physics puts the sun at the center, because of least energy. Orbits about the sun are least energy. Orbits around the earth are not. Least energy rules, yet again. And this is the point. our linguistic standards, and the means by which are brains operate are NOT unlimitedly listings of complexities, but related to least energy rules as the standards. Thus our measuring systems, Hands for horse heights at the shoulder, and feet for length, are least energy. & based upon comparison to human anatomies, at first. Least energy rules are often the means by which we cut the Gordian knots of complexity.
.
For our senses we feel hot and cold. Hot, hotter, hottest, and cold, colder coldest. Hot, more hot, most hot, and cool, more cool and most cool. Notably, the central “hotter” words, etc., ARE the comparative forms!!! Again, language shows us the way, but we did not see those clear clues. Those scales are most all comparisons!! These are many instances of linear temperature scales.
.
But put a warm object on a cold hand, and it will feel hot. Put a cold object on a cold hand and it will seem normal. or a cold object on a very warm hand? Extreme temperatures again. Sensation is thus largely comparison process, too. And we compare the relatively fixed standard of our temperature scales, based upon the STP boiling point of water and its vaporizing points, do we not? Simple, ubiquitous water, becomes the basis of human temperature scales. An efficient, stable standard is thus created by simplicity and the commonality of water.
.
Hardness is the Moh’s relative scale to talc which is soft, limestone harder, corundum (sapphire) and diamond, hardest. Thus hardness description is also comparison process. Now we use GPA’s, but that is relative to a more standardized and thus more efficient hardness standard, is it not?
.
And for visual images? Why should we think that the visual system uses our geometries for modeling the universe of events? Non-euclidean models the real universe, and neither do we see perfect circles, right angles, squares and triangles very much in the natural world, either. Instead, more fractal types of features, which are clearly NOT Euclidean, even as our space/time maths use non-euclidean geometries. And so for shapes the visual system uses curves and roundness, because those are the commonest features the occipital lobes detect. The mother’s face is marked by roundnesses and curves. That likely sets the standard for shape processing in our visual cortex.
.
Thus, we see optical illusions, some of the commonest of which involve seeing straight lines as curved. Esp., the two straight lines drawn through a series of nested circles. This optical illusion is illustrative of all the others. and how to show those lines are NOT curved, which our visual systems insists they are? Take a clear plastic straight edge ruler and lay it down next to the straight line. and the illusion disappears at once. & generalizing, for nearly every optical illusion there is a comparison correction, or more than one, which will show the illusion. Again, recalling the colour generation of our visual systems, the comparison shows the illusion and the comparison correction(s) which fixes it. THAT is significant evidence again how the comparison process creates sensations of all sorts through central processing.
.
I referred this article to the Euro Radiological Society and am STILL getting hits after 2 years. & this is how we can further investigate how the visual system creates images, by the misfires of comparison processing in creating the whole, unlimited panoplies of optical illusions. Among those illusions lies the structure/function relationships about how our visual systems work. and not just ours, but most all animals with eyes, too.
.
Radiology is simply reading images and reporting the findings. But how is this “reading” done? Simple, just like description and measurements are done against fixed standards to create meaning and insights, and new data/information. Each medical specialist, & the more so radiologists, have relatively fixed, efficient standards by which “normal” is judged. By comparing those standards by which they have learned efficiently to read images, they know using a set routine, if AP and lateral chest X-rays, by massive comparisons, are normal or not. The same for all MRI, fMRI, EEG’s, MEG’s, ultrasounds, arteriograms, etc., massive, standardized comparison processing. The same for lab data of patients which is compared to normal standard. How many more instances of comparison processing do we need to realize it’s ubiquitous & universal, and thus proved to exist? It’s self evident.
.
What of the plate tectonics model in geology? It’s complex system, largely visual thinking, too. Setting up the standards of upwelling zones, or mid-oceanic rift zones, subduction zones, etc. If AI can learn to understand tectonics, and its hierarchies, and how those all interact and work, as well as recognize its features, WITHOUT previous specific training in new examples it hasn’t seen, then General AI is upon us. My work has shown some of this water. But it’s likely there are many more wellsprings of our humanity’s creativity, understanding, and much else.
.
& there it is. The mind/brain interface of the cortex, which in the CC’s of Mountcastle, a simple, single, repeating process creates the outputs of the mind, including languages, sensations, creativities, the many, multiplicit functions of consciousness, as well as the moral conscience. The latter of which can be imaged at this time.
.
Thus language is easy to generate, from the simple repeating “dada, mama” to the complex. And math as well. Take the linear, number line of counting, 1 + 1, 2 +1,. etc. We can count up and then down. From 2 to 8 there are six counts and vice versa. From that we know there are 4 two’s in 8, and vice versa. By subtraction the same, and by division the same. The we get the exponential both base 10, natural and logarithmic methods.
.
And the hierarchies are again there, the 1’s place, the 10’s, the 100’s, the 1000’s, 100’s of 1000’s, millions and so forth. And the counting, 1st hierarchy; addition/subtractions, second hierarchy; 3rd hierarchy, multiplication and division; 4th, exponentials, etc.
.
For geometries and algebra, we see that when we compare the circumference to the diameter, we derive Pi, as a ratio, a proportion. We get constants, and speed in the same ways. Thus do we get algebra, which is essentially ratios and comparisons, too. And we get trigonometry which are the precise ratios of idealized right triangles by comparing their 2 sides, and angles, to derive the other side, Most all comparison processing, is it not?
.
And yet it’s far, far deeper than that. When we value items against each other by barter, a bushel of grain is worth some silver, or some copper. Or we create money which measures values of items against a relatively stable, set constant called, for instance, a dollar or pound. Its costs and values, and each currency, each value can be compared to all others by “conversion factors”, or constants, derived by, you guessed it, comparison processing. And when we shop for the best items, for the best costs, we use both massive comparison processing as well as least energy to get the best for our bucks, too.
.
Most all of our brain higher cortical outputs are comparison processes. Again as some AI experts have stated, “a single repeating principle in the brain which creates predictive control.”
This is how predictive control is obtained:
 .
Comparison processing creates classifications, indices, and social classes among humans as well as the flocks, herds, schools, and so forth. “Like knows like” through universal recognitions created by comparison processing, that is thinking. IQ also becomes easy to define and measure, and not only in humans, either.
 .
Communication becomes translation & lots easier, because of
 .
Ich bin hier.
I am here.
Je suis ici.
Estoy ahi.
Sum hic.
 .
Most translation is comparison processing, tho context is also CP and is not easily recognizable to a computer. It cannot recognize as easily as we can and with high facility, that we talk differently to bosses, employees, lover, parents, and so forth. Thus context is comparison process, easily seen once we begin to apply CP.
 ,
“It’s beautiful.” he said to her later in the day.
“It surely is.” she replied. Computers will have a hard time recognizing the sunset context here, let alone the others which are social.
 .
& how to communicate with dolphins? What do we recognize in common with them? By the rule of commonality, fish, sharks, bubbles, colours, and so forth. Recently it’s been claimed dolphins have names for each other. because they observed that specific whistles often in a pod attracted the attentions of single members. & when those were played back, only the one whose “name” was broadcast showed interest. Trial and error, comparison process.
 .
& when we go into space and meet other species space faring or not? How did our trading ancestors do it?
 .
This is wood, this is water, this is meat, and what’s your name for those? Those events in existence which we have in common, those events in existence which create recognitions by continuing reinforcement because similar events repeat themselves. We show the aliens ice, liquid water, and water vapour. Give them our names for those, and at once we have 3 phases of matter and water, which is the commonest life giving element, & must be ubiquitous, as well. Commonest standards, commonest used, even for weight as the 1 cc. gram of water, too. and densities, yet another comparison of mass over volume? The rest flows easily from there. Universal understanding, at a stroke. Simple, easy, elegant, highly fruitful.
 .
 .
Have gone on way too long, but the model is highly applicable to most everything. Without limits. The universality of the CP and the LE models is notable. It can even, where Hawking’s “The Grand Design” lamented, we cannot integrate the classical models of thermodynamics, relativity and QM, while he missed biology, which also isn’t consistent with two of them, either. The commonalities of understanding can likely create a Unified model of most everything. Defragmenting the sciences, as well.
 .
ER equals EPR. Some of your physics colleagues will see this relationship as part of unifying physics of relativity and entanglement, as this model predicts. There are many, many other bridging concepts which can be used, of which LE is key, too. Esp. in solving the complexities of Quantum wave equations. There are approaches possible to achieve even that. Well past renormalization, which is a specific app of a more general solution kind.
 .
The implications for AI of where the hierarchies of our understanding come from, how to both create and read and extend those using creativity are thus almost at once apparent. The system can even model neurochemistry, using Dopamine and its ancient and central, 10+ receptor sites, which both create motion and Emotion (D1 and D2, largely). DA can be compared against most all the other neurochemicals such as serotonin, to create sleep/wake cycles, why we dream & even how to better treat migraine headaches, using complex systems thinking. Or how to make Viagra 50 mg. last about 3 days, where its normal duration is 6-8 hours. Then there’s Cialis.
 .
Bacterial resistances to antibiotics, which are very easily overcome, as well, using this new paradigm & epistemology of comparison process and least energy, structure/function, and complex systems thinking. Thus we do more with lots less. The secret of growth of all kinds.
 .
Hope you found this interesting. I can answer a LOT of questions, including how and why events go viral, fashions/fads & where humor comes from and how to create professional vs. amateur skills for improving education across the board, without limit. It’s behaviorism on super charge, largely extended, because the internal brain source of the classical & operant conditioning model has been found, too.
 .
These methods can answer many questions about AI, and how to get to the Promised Land of general AI. Half the problem of getting there is solved if we know WHERE we are going. And if some go by trial and error, in all its combinatorial complexity, compared to those of us who know WHERE to find our goals, that is we know better & more completely what we are trying to simulate, the win goes to the best model because it’s swiftest & most efficient.
Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s