Flying Scroll News
Artificial Intelligence From the Bible
Latest News and Issues (June 2005 to May 2007)
Attention: The Flying Scroll News has been discontinued. For the latest news, go to Rebel Science News.
Attention: The Flying Scroll News has been discontinued. For the latest news, go to Rebel Science News.
May 5, 2007 3:50 PM EST
I am seriously considering going back to my AI roots, so to speak, and revive my old Animal program. My rationale for doing so is that I need some serious funding to continue my work. I am thinking that a chess program that gets better as it plays would be attractive to lenders and investors alike, especially if it can go from a rank beginner to expert or even master level without the usual alpha-beta search algorithm. A user could raise his or her own chess brain and pit it against others. I would like to rewrite Animal almost from scratch but this time I would use the C# language and Microsoft's XNA Game Studio Express, thereby killing two birds in one stone since the product would be available for both Windows and Xbox. My plan is to eventually move to the game of Go, a beautiful board game that has so far resisted all traditional approaches commonly used by chess or checkers programmers. Go is the most popular board game in the world, being played religiously by millions of people in Japan, Korea, China and elsewhere. Current computer Go programs are famously weak. A Go program that can play a decent game is sure to be an instant hit. My only problem is time. Oh well.
April 2, 2007 9:40 AM EST
I have seen all sorts of interpretations of the book of Revelation. Some have maintained that John's prophecy referred to the bloody reign of Roman Emperor Nero during which Christians were severely persecuted. Others have claimed that Revelation is just a warning to the "Church" through various periods up to the present time. Although I have no direct pipeline to God and I don't get visions in the middle of the night telling me to say "thus says the Lord", I can confidently reject all the interpretations I have seen so far, either wholly or partially. At the risk of sounding like I am gloating (I am not), I think that I have been given a gift of sorts. I can sometimes be pretty good at seeing things that others miss. Consider, for example, what John wrote regarding his vision (Rev 1:3):
What does this mean? What did John mean by "for the time is near"? I have noticed something peculiar about the use of the preposition 'for' in the book of revelation. It should be translated 'because'. John uses it repeatedly to establish a very strong causal or correlational link between two clauses. In this instance, I think John is saying that "reading" and "hearing" the words of the prophecy will occur right before the time of the end. People have been reading and hearing the words of Revelation for close to two millennia and the end did not come. So obviously, 'reads' and 'hear' have special meaning, otherwise the sentence makes no sense. I think that John meant 'read' and 'hear' in the sense of understanding or decoding the actual meaning of the metaphors. I think that I am beginning to understand a good part of the metaphorical texts. Although I haven't yet heard from anyone or read anything interesting in this regard, I suspect that others may be doing the same thing and coming to the same conclusions that I have.
My question to my readers is this: can you read or hear the words of John's prophecy? If so, you should heed them, i.e., believe them and act upon your convictions. And get ready to be blessed in ways that you cannot imagine, for the time is near!
March 21, 2007 6:00 PM EST
The Lord works in mysterious ways but the devil is more or less predictable. He hits you where it hurts the most, at your weakest points. Ever since I came to the realization that the Bible contains revolutionary scientific secrets and decided to publicize my findings, my life has been hell. My health, my finances, my relationships with family and friends, all rapidly deteriorated. I have been physically assaulted by individuals I would never have suspected of harboring criminal intentions. I have seen some of my own relatives become so threatening and unreasonable that I can only conclude that they were under demonic control. I don't want to bore my readers with my problems but suffice it to say that the last six months have not been easy. I had very little time to work on my research and my access to the internet was severely crippled. The good news is that I'm still alive, as determined as ever. I have added a new section to the site, to deal with Bible physics (see news item). Even though I have not had a chance to update all the pages in the Artificial Intelligence From the Bible section, rest assured that I have not been idle, at least not mentally. I am continually meditating on the meaning of one Biblical metaphor or another. I am now confident that, given the necessary resources and enough time, I can implement a truly intelligent machine based on my current understanding.
The temple in Zechariah's vision symbolizes the seat of the brain's memory and its motor learning mechanism while the golden lampstand or menorah represents a single seven-node memory sequence. There are a great many coexisting behaviors in memory but only a few are active at any one time. The others are dormant and remain so unless awakened by sensory activity. This is symbolized by Zechariah being awakened by the angel. It is also the reason that Sardis (the sleeping or dead church) is told to wake up in order to remember (i.e., record) how it received and heard. It is important to remember that memory nodes can only be awakened in groups of seven. This is symbolized by Zechariah waking up to see the seven-lamp golden lampstand. Every node in a seven-node sequence is part of the same behavior group. Two memory groups (behaviors) cannot be awakened at the same time if they are in conflict. In case of conflict, the stronger group wins as seen in Rev 2:23: "And I will give to everyone of you according to your works".
I used to think that Philadelphia represented the motor layer (basal ganglia) but I recently realized that the motor layer is not dealt separately in the Biblical text. Both Pergamum and Thyatira perform motor behavior (symbolized by walking). Lately, I came to understand that Philadelphia is but a behavior that has matured to the point where it can be permanently locked in place. This is symbolized in verse 3:12 by "He who overcomes, I will make him a pillar in the temple of My God, and he will not go out from it anymore". Note that only a mature group can be used as a trainer for the church of Laodicea (a.k.a. the synagogue of Satan or the cerebellum). In addition, a mature behavior group is no longer subject to tribulation (testing). This is seen in verse 3:10.
So again, in the light of the above, I have revised my original interpretation of the church of Philadelphia. I no longer think that it represents the motor conflict layer (basal ganglia). I now understand that the basal ganglia (motor conflict detection) is already part of Pergamum and Thyatira. More to come.
September 20, 2006 1:10 PM EST
The temple in Zechariah's vision symbolizes the seat of the brain's memory mechanism while the golden lampstand or menorah represents a single seven-node sequence within memory. But the metaphor goes deeper than that. Allegorically speaking, there is another temple that must be built in these last days. It is not a physical temple built of stones but a symbolic temple, one which will reside in the hearts of men and women of faith. How can this temple be built and by whom? What is needed is a powerful revelation that will make people stand up and take notice. The prophet Jeremiah (29:30) wrote,
What I am asking here is this: what word of God is hot enough to spread around the earth and set hearts on fire and break their rock foundations into pieces? Where is the heat going to come from? I argue that the revelation that the secret of artificial intelligence was hidden in the Bible for thousands of years will be a very hot fire indeed. What is even more surprising is that AI is only the tip of the iceberg, so to speak. There are many more powerful secrets hidden in God's word. Soon the mystery of God as he spoke to his servants the prophets will be finished and the fire will go forth and show a lot of learned and not so learned folks for what they are, charlatans and liars. In the light of this fire, is it not fitting that Jeremiah continues in verses 30 to 32:
Yes. There are a lot of preachers and Bible interpreters out there who claim to understand the true meaning of God's symbolic word (e.g., the seven churches of Asia and Zechariah's vision of the golden menorah). They have their own interpretations which they copy from one another and preach to the unwary. I have news for them. The true meaning of God's word is much more glorious than their puny minds can dream of. He that has an ear, let him hear what the spirit says to the churches. Consider that this temple which must be built before the return of our Master is not meant for Christians only. It will be a bridge of friendship between the house of Judah (the Jews or the bride of God) and the lost sheep of the house of Israel (Christians) in these tumultuous last days.
There is a lot of work left to be done but I can't do it alone. I can only lay a few stones. Just as Joshua and Zerubbabel received the help of all the Jews returning from far away places, this temple will need the help of many people from all corners of the globe. It's the kind of help that will require faith in the glory of God and the power of his word to spread like a fire and shatter solid rock into pieces. But when the son of man returns, shall he find faith on the earth? (Luke 18:8)
September 10, 2006 9:30 PM EST
In the message to the church of Smyrna we are told of "the blasphemy by those who say they are Jews and are not, but are of the synagogue of Satan". And again, in the message to the church of Philadelphia (verse 3:9), we read "Behold, I will cause those of the synagogue of Satan, who say that they are Jews and are not, but lie--I will make them come and bow down at your feet, and make them know that I have loved you". As can be seen in the book of Zechariah, only the Jews were sent to Jerusalem to build the temple. We now know that these Jews symbolize signals that are fed to Smyrna (the temporal association cortex). But who do the non-Jews represent? In the past, I have identified them as sensory signals that go directly to the cerebellum (Laodicea) without passing through either sensory and association cortices (Ephesus and Smyrna) or sequence memory (Sardis). The difference between Jews and non-Jews is obvious in the text. The Jews arriving at Smyrna are poor although they ultimately become rich (Rev. 2.9) through trials and tribulation. By contrast, the non-Jews appear to be rich on the surface even though they are actually poor. Thus interpreting the symbolic difference between Jews and non-Jews comes down to understanding the difference between those who are poor and those who are rich.
We already learned from reading Zechariah 6:10 that the Jews arrived from Babylon in multiple concurrent groups. Thus we know that signals arriving from Ephesus (or Babylon, as symbolized in the book of Zechariah) must be tested for concurrency. This implies that all non-Jews will fail the test of concurrency a priori and thus are not allowed in Smyrna. It just so happens that the only type of signals in the nervous system that will always fail the concurrency test are proprioceptive signals. These are generated by special receptors located throughout the body. These proprioceptors detect such phenomena as joint angles and acceleration, muscle activations, etc... What distinguishes these receptors from other sensors (i.e., light sensitive cells in the retina) is that they do not emit transitory spikes. They start firing as soon as they detect a given condition and they continue to fire for as long as the condition persists.
All proprioceptors ultimately send their signals to the cerebellum (Laodicea) where they are used for motor control. Their "richness" lies in the fact that they have precise meaning and need not go through the initial testing and filtering as would visual or auditory signals. The problem is that, unlike the signals that reach the basal ganglia (Philadelphia) via the motor cortex (Thyatira), cerebellar signals have not undergone any sort of conditioning (classical and operant) and have not been tested for motor coordination. This is the reason that they must come and bow down at the feet of the church of Philadelphia in order to learn how to behave. As mentioned in the message to Laodicea, they are said to be blind, poor and naked. In addition, they are not given control of speech generation which requires attentional processing of the entire sensory space and of short-term sequence memory.
September 6, 2006 8:50 PM EST
In my research of the symbolic meaning of the seven churches, I have grown to appreciate the fact that every word in the text has a meaning and a purpose. I am always conscious of the fact that the book of Revelation is the work of a superior mind whose intelligence must not be underestimated. I used to think that Smyrna was part of the sensory cortex and, as a result, I wrongly assumed that the ten days of tribulation stood for the correlation factor (10 to 1) used for synaptic learning in the sensory cortex (Ephesus). After I revised my interpretation of Smyrna, I knew that I needed a correlation factor for Ephesus. Then it occurred to me that the first verse in the message repeats two metaphors that were already given elsewhere (Rev 1:13 and 1:16) in the first chapter of Revelation:
It is obvious that neither metaphor has anything to do with Ephesus in particular. The first one pertains to the seven churches as a whole while the second one is relevant only to the middle or fourth church (Thyatira). So why are they included in the message addressed to Ephesus? In my opinion, the idea that is being conveyed here is not the metaphors (since they don't apply to Ephesus) but the number seven. I am almost tempted to infer that the sensory cortex uses a 7 to 1 correlation factor but I must admit that I don't really know at this time. All I know is that I have used many ratios in my own experiments and they all seem to work. The difference is that the higher the ratio, the greater the number of correlations.
I had previously interpreted the flying scroll chapter in chapter 5 of Zechariah to represent two correlation ratios, one for sequence memory (Joshua and his friends) and one for association memory (concurrent inputs). Now that I have revised my understanding of Smyrna, it seems that the ten cubits of Zechariah's scroll symbolize the same thing as the ten days of tribulation found in the message to Smyrna. So it follows that the 20 cubits is a metaphor for the correlation ratio used in sequence memory. Again, for those who have been trying to make sense of it all, the seven-node sequences comprise the church of Sardis. This is where both short and long-term memory resides.
September 2, 2006 12:20 PM EST
A few months ago, I revised my understanding of the churches of Smyrna, Pergamum and Laodicea. I now interpret Smyrna to represent the part of the brain which is commonly referred to as the association cortex. This is where input signals from the sensory cortex (Ephesus) are combined to form concurrent groups. Each group consists of a number of input fibers connected to a single association neuron. I arrived at my new interpretation of Smyrna after noticing the correlations between the following verses in the books of Revelation and Zechariah:
In retrospect, the link between the above passages now seems obvious. Only the Jews arriving from Babylon were given the task of building the temple. Non-Jews were forbidden. Of course, the returning Jews, having been exiles in Babylon, had very little in terms of possessions. In other words, they were very poor. In my next article on Smyrna, I will write about the differences between Jews and non-Jews. I will also identify a link between Smyrna and the flying scroll mentioned in the book of Zechariah. More to come, soon.
August 26, 2006 11:50 AM ES
My understanding is that this area is merely a part of Smyrna (signal fusion) which receives signals directly from the auditory sensory cortex (part of Ephesus). I believe that damage to Wernicke's area will result in more than just impaired speech comprehension but in all sorts of auditory perceptual problems as well, including impaired music perception. This area sends its output signals directly to Broca's area. There is a strong possibility that Wernicke's area is also responsible for learning and recognizing sound sequences. Not that it matters much from the standpoint of this presentation.
August 23, 2006 9:30 AM EST
OK, I've gotten over my disappointment with the so-called Christian community. It is time that I get back to posting some news. In my research, I make progress not only by discovering new interpretations of Biblical metaphors but also by retracing my steps and correcting previous wrong assumptions on my part. As I have explained in the past, the seven churches of the book of Revelation symbolize the organization and operating principles of the left hemisphere of the brain which controls the right side of the body (hence the metaphor of the stars being held in the right hand of God). Certainly, any description of the left hemisphere would be incomplete if it did not include a mention of the so-called Broca's area, the part of the brain that is involved in speech processing, generation and language comprehension. A few months ago, I realized that my initial understanding of the message to the church of Pergamum was flawed. I had wrongly assumed that the metaphor of the sword protruding from the mouth of the son of God pertained to motor learning as a whole. It dawned on me that the mouth symbol has to do with a specific type of motor behavior (speech) and that Pergamum is a metaphor for Broca's area. I don't know how I missed it initially, seeing that it is such an obvious explanation.
My new understanding of Pergamum, in turn, gave birth to even more realizations regarding Pergamum's relationship with Laodicea (cerebellum). I now know that the cerebellum can be trained to do all sorts of motor tasks except produce speech. This is not something I learned from any existing literature on the brain. I did a quick search on the web to see if anybody else was aware of this but I could not find anything. So count this as one more testable prediction for the skeptics. I have also changed my interpretation of the message to the church of Smyrna but I will leave it for an upcoming news item. In the next few days, I will edit the text of the relevant pages to reflect my new understanding. Stay tuned.
May 15, 2006 10:55 AM EST
In the last few months, I have made a tremendous amount of progress in my understanding of the Biblical metaphors pertaining to the organization of the brain. My excitement and my enthusiasm are at an all-time high. As usual, my intention is to post all my discoveries here as I have done in the past. However, this has become too much of a one-way enterprise. I feel like a worker who is constantly laboring without ever getting any wages. I cannot fathom how something as momentous and as important as Biblical AI has garnered so very little support. Where are the true Christians? Is there no one left or are you so scared of the world that you are all cowering in fear? Please don't talk to me about your righteousness and my lack thereof. It's time we all face it. You and I are not righteous, period. There is only one righteous being and we all know who he is. We have all fallen short and a little leaven is enough to leaven the whole bread. So forget about your righteousness delusion: it is not going to save you.
A little less than two thousand years ago when our master was departing from the earth at the end of his mission of redemption, he asked, "When the son of man returns, shall he find faith on the earth?" (Luke 18:8). I know that there are some real Christians out there who understand that these are the end days when the deep mysteries of God, which he revealed to his faithful prophets of old, would be unsealed. What I want to know from you is this: where is your faith, goddamnit? If you really want to see the hand of the creator in your lives in these troubled last days, you had better put on a cloak of faith. This is getting ridiculous from my vantage point. As of today, I will no longer post anything new to the Flying Scroll News until I detect some signs of life from you. I am not asking for handouts anymore. As it is written, a worker is worthy of his wages. If I have taught you anything of value and you have benefited from my writings, then you owe me and I expect to get paid. Don't give me a dime otherwise; and please, don't visit the site and don't read any of it as it is not meant for you. Let me add that what I do with the money I receive for my work is nobody's business.
January 5, 2006 11:50 AM EST
I know. It has been more than a month since I posted my last article in the Flying Scroll news. I apologize but I have my usual excuses, the lack of time and resources. I have also been promoting my work in software reliability in the hope that it may generate funds that I can subsequently use in my AI research. In my last article, I promised to explain why Jezebel is a woman and the significance of this symbolism with regard to behavior formation. However, there is something rather important I would like to bring back to the forefront of this discussion before I return to Jezebel. It has to do with motor command timing. How does the brain decide how to time muscle movements?
At this point, I am pretty confident that the brain relies very little on sensory feedback while performing motor actions. Sensory input is used mostly for the creation of memory sequences and for correcting intervals within sequences. During playback, motor commands are generated automatically, whether or not they coincide with the expected sensory signals (symbolized by Joshua and his friends in the diagram below). This is probably the reason that neuroscientists have observed spike doublets in the hippocampus. I hypothesize that a doublet consists of the learned spike followed in close succession by the actual sensory signal. This is an important capability of memory organization and operation. The reason is that objects are more often than not partially occluded from view. So the brain must rely on its memory of previous experiences to compensate for incomplete or unavailable information.
At this point, the mechanism of memory formation is pretty much clear in my mind. My understanding is that the brain uses a huge number of seven-node memory sequences which are organized according to the Biblical metaphor of the golden lampstand described in the fourth chapter of the book of Zechariah. As I have mentioned elsewhere on this site, it is not so much the intervals between input signals in a sequence that are recorded, but their temporal relationships. What this means is that all intervals in the sequence are variable and proportional to each other such that, if one interval is halved, the others are halved as well. This mechanism is important in more ways than one. I already mentioned pattern completion, i.e., the ability to fill in the missing parts of an incomplete sensory pattern of signals. There is also a problem that psychologists have identified and it has to do with the generation of very fast movements such as throwing a rock or catching a ball. It is impossible for the brain to rely on proprioceptive feedback in such circumstances because it takes too long for the signals to travel from the muscles to the memory cortex. The solution to this problem is rather straightforward. The memory interval mechanism is such that it can extrapolate any interval, however short-lasting, from the learned relationships.
The first question that comes to my mind is, how does the brain decide how fast to playback a sequence? Is there a separate control mechanism that sits on top of memory and decides playback speeds? After all, we can intentionally move our limbs faster or slower, at will. I haven't found any metaphor in the Biblical text that suggests the existence of such a control mechanism. I am forced to conclude that it does not exist. So what triggers a change in playback speed? In my opinion, intervals are updated according to sensory circumstances. This is implied in the message to the church of Sardis in Revelation 3:3:
I interpret this to mean that sensory signals are used to record (remember and keep it) a new interval and this in turn is used to update or change (repent) all the other existing intervals. The new intervals are kept indefinitely until new sensory circumstances arise which will trigger a new update. For example, simply by observing the velocity of approaching traffic, a pedestrian crossing a street can adjust the speed of his or her gait so as to avoid getting hit. More to come soon. For more on motor learning, see the Pergamum page.
November 15, 2005 11:10 AM EST
In the letter to the church of Thyatira, we read the following:
Why is Jezebel given time to repent? And why must she repent only of her immorality (fornication)? Why not also idolatry? After all, Jezebel teaches and seduces God's servants to commit fornication and to eat things sacrificed unto idols (Rev 2:20). Is it a mere coincidence that Jezebel is a woman? In my opinion, these are important metaphorical clues which reveal the precise role of Jezebel. I will try to answer them in an upcoming article. As always, I am pressed for time but I am not about to give up. I've only just begun. Hang in there.
November 15, 2005 11:15 AM EST
Sir Isaac Newton, the founder of modern mathematical physics and the co-inventor (with Gottfried Leibniz) of calculus, is renown for having formulated the laws of motion and gravity. What most people do not know is that Newton had a lifelong passion for alchemy. Most of us associate alchemy (or chymistry as Newton called it) with black magic, mysticism and superstition, but Newton, one of the most rational scientists and thinkers of all times, had a different opinion. He regarded it as a serious science, on a par with his published works on physics. What gave alchemy its bad reputation was the custom of its practitioners of concealing their knowledge with the use of mythical-sounding allegories. In an upcoming PBS/NOVA show (scheduled for today) titled "Newton's Dark Secrets", we learn of Newton’s fascination with the Bible and alchemy. What grabbed my attention was the fact that Newton (and many other philosophers of his time, including Leibniz) believed that ancient Greek and Roman myths were not, as is commonly believed, pagan religious texts, but serious scientific knowledge written using clever and, at times, misleading metaphors.
Apparently, ancient philosophers (the so-called wizards and sorcerers) needed an effective way to document their science but considered their knowledge to be so disruptive that they felt compelled to use a coded symbolic language to hide its secrets from the public at large. This is a topic that I touched upon in my original article on Biblical AI. I will revisit this interesting subject from time to time. In the meantime, take a look at Newton's alchemy at NOVA's site.
November 9, 2005 3:50 PM EST
As many of you know, I live in Florida which was hit hard by hurricane Wilma about two weeks ago. There was widespread damage. Luckily for us, Wilma was a fast moving category 3 storm. Had it moved a little more slowly, or had it been a category 4, there is no doubt in my mind that there would have been a lot of casualties. The house in which I live was hit by a falling mango tree and the ceiling caved in in my bedroom. Most of southern Florida lost electric power. My area which was among those that suffered the hardest damage just got power back yesterday. The interesting thing is that twice in the same season, my area was hit by the eye of a hurricane. Hurricane Katrina passed through here a couple of months ago on its way to Louisiana and the gulf states. Florida was lucky that it was only a category 2 at the time. To cut a long story short, I am now back on the internet and I will be posting a new article on Thyatira and goal-seeking behavior in the next few days.
October 20, 2005 6:50 PM EST
OK. I am almost there and, in my opinion, the message to the church of Thyatira is the key. The importance of Thyatira cannot be overstated. In Revelation 2:1, John writes that the Son of man walks in the middle of the seven churches. Thyatira is the fourth candlestick and thus occupies the center position among the seven golden candlesticks. Unlike the other cell assemblies in the brain, Thyatira does not process sensory signals but, rather, provides an essential service for the other churches. It organizes individual memory sequences into distinct sensory-motor behaviors (nations). It discourages the activation of behaviors that lead to pain (or away from pleasure) while encouraging those that lead to pleasure and away from pain (Thyatira's love).
Goal-seeking behavior is an essential capability of intelligent adaptive systems. A goal is a desired sensory outcome. This is the reason for the metaphors of eyes like a flame of fire and feet like fine brass at the beginning of the Thyatira message. Sensory signals are indirectly coupled to motor signals via memory sequences. It is important to keep in mind that incoming sensory signals continually update intervals in memory. They are not sent immediately to the motor layer. Thyatira provides the mechanism needed for actually initiating motor output.
It goes without saying that, in order to achieve a goal, the goal must be known a priori, even before the actions that may lead to it. For example, getting to the other side of a river is a goal even if the means of doing so are not yet available or known to the system. A goal, in my opinion, is a future attractor (successor) that seduces one or more predecessor actions that may lead to the goal. Which brings me once more back to Jezebel:
Jezebel, in my opinion, represents a goal to be achieved but it is not necessarily a good goal. After all, she only calls herself a prophetess, not knowing the consequences of the goal. This means that, contrary to what some behaviorists have claimed in the past, a goal can exist without prior reinforcement. The initial goodness of a new goal is assumed unless a subsequent painful stimulus proves otherwise. Likewise, a pleasurable stimulus may further reinforce the initial goodness. Indeed, how can an intelligent agent test a goal unless it first tries to achieve it? When I watch young children at play, I am struck by the quasi-total randomness of their behavior. They do the darnedest things, things that do not necessarily lead to either reward or punishment. Curiosity is the tendency to seek goals, in my opinion.
In the old testament, Jezebel was known to put on face makeup to attract and seduce potential lovers. Analogically, a goal must have something that will attract the motor commands that are required to achieve the goal. I hypothesize that this something is a learned interval. The problem is that several preceding actions may lead directly to a goal but they may not be compatible. In other words, they may lead to motor conflicts (fornication and idolatry). A bad behavior is not disconnected unless it introduces one or more motor conflicts. Behaviors that are found to lead to pain or discomfort must be kept for future reference. In my next article, I will talk about how behaviors are selected and how they are initiated. Something to do with a morning star...
More to come...
October 4, 2005 11:30 AM EST
Over the last couple of weeks, I have struggled with the action timing problem. At first, I thought I had figured it out and that it was just a matter of conflict detection and reinforcement learning. But the more I thought about it, the more I kept running into the hard problem of timing. As always, when I am faced with seemingly intractable problems, my recourse is to read the relevant Biblical passages over and over in the hope of finding new nuggets that I might have missed. A couple of days ago, I got a powerful new insight while re-reading Thyatira for the umpteenth time. It has to do with Jezebel's teaching and seducing God's bond servants so as to lead them astray. It may be because the full meaning of a passage is sometimes lost in the translation (how I wish I was fluent in Greek and Hebrew!), but the subtle power of the metaphor had previously escaped me. It has to do with such things as prediction, goal generation and learning. I am giving myself a little time to allow this new idea to percolate in my brain while I'm preparing a new article. Stay tuned.
September 19, 2005 3:58 PM EST
One of the functions of memory is to enable thought. I define a thought as an internal behavior whereby individual sequences are activated and processed by a special thought mechanism that inhibits all motor commands generated by the processing sequences. While the system is thinking, output motor control (the outer courts) is delegated to the cerebellum (the gentiles), a supervised automaton that uses direct sensor-effector connections for motor control.
A train of thought is started whenever a neural sequence is executed. The neural mechanism is such that triggering the primary node (Zerubbabel) in the sequence causes the other nodes to fire at specific prerecorded intervals programmed by experience. In addition, the execution of one sequence may trigger the execution of one or more sequences. Consider the sequence diagram below:
My current understanding of the Biblical metaphors is that all inputs to a sequence originate from coincidence neurons (not shown) which receive their signals directly from the sensory cortices (Ephesus and Smyrna). Coincidence neurons can also sprout branches to more than one sequence. During sequence playback, the nodes fire in succession. What is not shown in the diagram is that the input coincidence neurons fire at the same time. As a result, all the nodes that are targeted by a coincidence neuron also fire and this activity, in turn, triggers the execution of other sequences resulting in a sequential train of thought. This way, the system can follow the chain and see where it leads, a must for reasoning and planning.
As I wrote previously, high-level behaviors consist of a number of small primitive actions. The problem is, how does the system determine the proper timing of various actions. A grasping behavior, for example, is useless unless it is triggered at the right time. I'll have more to say about this in a coming article.
September 19, 2005 11:58 AM EST
Behaviorists often claim that all actions and behaviors are created via reinforcement (conditioning). The problem with this stance is that, in order to reinforce a behavior, it must already exist. My current understanding, based on the Biblical symbology, especially the letter to the church of Thyatira (Revelation 2:18-29), is that conditioning is not about behavior generation but about behavior selection and composition. A different principle other than conditioning is necessary for the creation of behaviors. The usual approach to reinforcement learning is to reinforce all recently active behaviors and let them compete? The problem with this approach is that a low-level behavior such as walking straight ahead is not necessarily either good or bad. In other words, walking straight toward a hungry lion is bad but walking straight toward food is good. One runs the risk of reinforcing the wrong actions. This is the credit assignment problem.
Adaptive behavior is thus the result of the application of two learning principles: motor coordination and conditioning. This is what is meant by the passage in Revelation 2:23: and all the churches will know that I am He who searches the reins and hearts; and I will give to each one of you according to your deeds. My interpretation is that reins symbolizes motor coordination/reasoning while hearts is a metaphor for motivation/conditioning. During adaptation, an intelligent system combines low-level primitive behaviors (actions or groups, if you will) into high-level appetitive and aversive behaviors. Each action is a group of multiple coherent (non-conflicting) sequences. A high-level behavior or concept consists of multiple coherent actions. Behaviors may share actions but only one behavior can be active at any one time. During adaptation, the behavior may receive pain and pleasure stimuli. These are used to reinforce actions. Ultimately, only the good actions within a given behavior survive. This way, the context (the high-level composite behavior) determines whether or not a given action should be reinforced. So, an action may be good in one behavior and bad in another. This, in my opinion, solves the credit/blame assignment problem rather elegantly, without having to use some awkward temporal difference reinforcement mechanism.
In my next article, I will talk about thinking/reasoning. During thinking, the brain is able to anticipate future events and correct its behavior without having to actually perform the actions.
September 13, 2005 3:40 PM EST
It occurred to me recently that we are frequently able to figure out what someone had said earlier in a sentence even though we may have missed some of the initial words. We have the ability to use subsequent information to guess previous unintelligible information. It is easy to see how retro prediction can work using the memory mechanism I have proposed, based on the golden lampstand (seven-lamp menorah) symbology found in the book of Zechariah.
Interval prediction works because every interval in a sequence is a fixed percentage of any other interval in the same sequence. This means that, given any interval, the value of the others can be easily inferred and instantly calculated. So, if after receiving a Zerubbabel signal, a sequence fails to receive a Joshua signal as expected, all is not lost. If any of the friend signals arrives afterwards, all the other intervals can be instantly updated to their correct values. It suffices to make the Branch pathway bi-directional (not yet shown in the diagram). Alternatively, there could be two branches, which would be more in keeping with Zechariah's vision in verse 4:12: What are the two olive branches which are beside the two golden pipes, which empty the golden oil from themselves?
Now that I think about it, retro prediction must be what Zechariah 3:8-10 is alluding to:
In my opinion, the repeated use of the word day indicates that interval updating is a quick process that lasts only one neural cycle, or about ten milliseconds. I think this is a very powerful capability for causal/temporal reasoning because it enables pattern completion in both directions.
I am preparing an article on thinking and reasoning. It has to do with the two olive trees mentioned in the books of Zechariah and Revelation (the two witnesses). More to come...
September 13, 2005 12:20 PM EST
In the previous article (see below), I wrote that immature memory sequences lose their recorded intervals (memory traces) as soon as the sequences go back to sleep. Why? Is there something to be gained by forgetting intervals? After all, the neural mechanism that stores the intervals remains operational. Why not retain the intervals and use them again the next time when the sequence is re-awakened? In my opinion, the reason is that a great many intervals are useless over the long run because they are not predictable. That is to say, they are useful only while the system is paying attention to a given context. In the message to the church of Sardis we read (Revelation 3:2): Wake up, and strengthen the things that remain, which are about to die; for I have not found your deeds completed in the sight of My God. This is obviously referring to the fact that memory sequences retain their traces (the things that remain) only for a short period of time while they are awake. If the traces are not strengthened, they simply die. The implication is that a sequence is strengthened every time it wakes up and, if it is awakened often enough and the intervals keep the previous temporal correlations, the sequence will eventually retain its stored intervals permanently. So forgetting seems to be a way to keep the system from holding on to infrequent and uncorrelated intervals.
As I mention elsewhere, there is a direct link between memory sequences and motor output. This is alluded to in the message to the church of Thyatira (Revelation 2:2): The Son of God, who has eyes like a flame of fire, and His feet are like burnished bronze. Eyes refer to perceptive inputs while feet symbolize motor control. Placing both eye and feet in the same sentence is a way of showing that they are directly connected. The reason is simple: memory is useless unless it actually does something. It follows that the mechanism that creates non-conflicting motor connections between sequences and effectors must be very quick. Otherwise, the intelligent system would never have time to use newly formed memory traces before they die. In my opinion, these motor connections remain attached to their target effectors even after the traces disappear. This way, they can be reused immediately the next time the sequences wake up and record fresh new intervals. But what happens to the old connections if newly recorded intervals give rise to motor conflicts? I think that motor connections that worked once are never completely severed. They are temporarily disabled, when necessary, but are kept around to be reused again, if the opportunity presents itself.
September 1, 2005 11:55 AM EST
Short-term memory is pretty much instantaneous. That is to say, STM traces are created almost instantly, not only from newly sensed phenomena, but also from previously recorded long-term memory. It follows that the underlying storage mechanism of STM must already be in place in order for a new memory trace to be formed. In the model that I have proposed, memory can be viewed as a huge sheet of sequences aligned perpendicular to the sheet. Each sequence (see diagram below) is a neural circuit consisting of seven nodes. Sequences are formed automatically during perceptual learning. They are organized into conceptual groups under the control of single cells that I call concept cells. A group (or concept) can be either awake (active) or asleep (inactive) depending on the intelligent system's current focus.
The primary function of a sequence is to anticipate the arrival of sensory signals and to generate properly timed motor commands. This means that a node will fire at the right time even if the anticipated input signal does not arrive. This gives the intelligent organism the ability to complete a sequence even when parts of the information is missing. To do so, each node stores a memory trace, that is to say, the interval between two signals. There really is no distinction between STM and LTM. The same mechanism is used for both. STM is just an awakened LTM. However, unless a sequence is mature, the intervals are lost when the sequence goes to sleep. This means that reawakening and immature sequence has no effect other than making it available for recording a new set of intervals. Several questions immediately come to mind is: a) What is the point of having an underlying structure for recording memory traces if the traces can be quickly forgotten? b) Since memories are used for motor output, what happens to the motor connections after a sequence goes to sleep? and c) What happens to concept/context formation in short-lived memories?
More to come...
August 22, 2005 1:05 PM EST
There seems to be a general consensus among AI researchers that perceptual memory is organized as a probabilistic hierarchical tree. Sensory signals are supposed to travel up the nodes of the tree and a probabilistic associative value is encoded into each node. The accepted wisdom is that this is the most practical way to predict the probability that a node will fire. Needless to say, I disagree with this approach. Not because I think it is dumb (I, too, used to think it was the way to go), but because I have found that the Biblical metaphors suggest a completely different model. A major problem with the probabilistic node model is that it does not take the causal or common sense logic of sequences into consideration. For examples, we know that a door cannot be opened unless it is first closed and vice versa; and we know that we cannot empty a cup unless we first fill it. This is part of what I have been calling the Principle of Complementarity. Complementary logic is not only imposed by the environment, it is expected by the innate functional organization of the brain.
But there is more to common sense logic than just knowing the order of a simple sequence of two given stimuli. The logic demands an inherent ability to predict intervals or durations. Without this ability, an organism cannot anticipate the time of arrival of multiple stimuli and their correct sequential order. This is absolutely crucial to survival and to symbolic reasoning. A further problem with the probabilistic node approach is that it does not provide an effective mechanism for attention. The brain is able to focus on one concept at a time while ignoring all others.
These are some of the main reasons, in my opinion, that Bayesian networks have not lived up to their promise. There is a better way, of course. The solution (explained below) is somewhat counterintuitive. But it is as clever as it is simple and powerful. Now that I know and understand it, I realize I never would have guessed it on my own, not in a thousand years. It is like looking for a needle in the proverbial haystack.
It is beneficial to think of each hemisphere of the brain as a tightly integrated network consisting of two major subnetworks: a perceptual network and a conceptual network. The former deals with the timing of sensory signals and their correlations while the latter is concerned with motor coordination and command timing. Both subnetworks complement each other. An important consideration, one which is not lost on some AI researchers (e.g. Dr. Rodney Brooks of MIT) is that there is little time to waste as signals travel from the sensory layer to the motor layer. Response time must be kept to an absolute minimum so as to ensure the intelligent organism's survival in a fast changing environment.
The Biblical model does not assign a probabilistic value to every node in sequence memory. Certainly, it uses a probabilistic correlation factor to find the initial connections (see the Flying Scroll), but once the connections are found and once they reach a certain maturity, they are locked into place forever. From then on, every sequence is treated as if it had a 100% probability of occurrence. That is to say, every sequence is played back to completion every time its primary node fires. An important thing to realize is that once a sequential correlation is ruled out, it is never tried again. This is symbolized in Zechariah 5:4 thus:
In other words, bad connections are permanently destroyed. In my opinion, this is the reason that children find it easier than adults to learn new languages. As we grow older, we simply can no longer make the proper correlations because previously rejected connections cannot be retried. There is much truth to the old proverb that one cannot teach new tricks to an old dog.
So, how does the brain determine which nodes in a sequence should be allowed to fire and which ones should be inhibited? Obviously, it cannot base its determination by just examining the nodes since no node maintains information about its probability of firing. It must use a different set of criteria. The answer has to do with what I call groupthink. In the model that I am proposing (based on my interpretation of the Biblical metaphors), the brain uses complementary logic to weed out incompatible nodes within a group (concept). As it turns out, compatible (non-conflicting) nodes will automatically cluster into individual coherent groups under the control of special cells that I call invariant concept cells. The reason is that, due to the limited number of motor resources, the brain cannot think of everything at once. It can only concentrate on one group at a time. So whenever a group is activated, only the nodes that are part of the group are allowed to fire. But this does not tell us why or how one group is selected for activation as opposed to another.
Groups compete with each other for activation. Two specific criteria are used. In Revelation 2:23, we read:
As I wrote previously, reins (kidneys) symbolizes reason (memory sequences) whereas hearts represents the seat of emotions or motivation. I interpret this to mean that the probability that a group will be activated depends on: a) the number of nodes within the group that are receiving sensory stimuli and b) the motivational strength of the group. So, the probability that a node will activate depends on the activation of its entire group, not on a learned probability specific to the node. My understanding is that concept cells receive afferent connections from the nodes under their control (reins) and from visceral cells in the Amygdala (hearts). I am not entirely sure how the two criteria are combined (the Biblical text does not seem to think it's a big problem, otherwise it would have mentioned it) but it would seem, at this point, that a simple multiplication would work.
August 15, 2005 5:35 PM EST
I had time to meditate further on the intriguing and powerful metaphors used in the message to the church of Thyatira in the book of Revelation. I realized soon after I wrote my sequence sharing hypothesis that something was missing. I could not come up with a consistent and robust mechanism for group formation. As I usually do when I am faced with a sticky problem, I went back to the biblical text and read the relevant passages over and over in the hope that a new idea might pop up into my mind. It occurred to me that Jezebel was a powerful national leader in ancient Israel, who dealt directly with generals, captains and other leaders. She would never interface directly with the common folk. Then I got to thinking again about verses 2:26 and 2:27. We learn that overcomers in Thyatira will rule over the nations with a rod of iron, and that the nations will be broken into pieces like the vessels of the potter. I now interpret nations to mean conceptual agglomerations and that these are broken into small coherent behavior groups (pieces) governed by single control cells (the rest who are in Thyatira who have not known the deep things of Satan (2:24)). As seen in the figure below, I now see two levels in this hierarchy. The top level consists of high-level concepts (Jezebels). These control multiple small groups and can share groups among themselves.
Note that group B is shared between high-level concepts A and B. To form a concept, it suffices to find non-conflicting (or non-fornicating, as Revelation puts it) groups. Unlike a bottom-up pyramidal hierarchy, the concept cells do not receive recognition signals percolating up the tree. Their primary function is to activate and deactivate entire groups of related sequences (small rectangles). These receive signals from the sensory cortices and send output signals directly to the motor cortex. The nice thing about this arrangement is that the hierarchy does not cause any delay in signal propagation. That is to say, unlike a bottom-up perceptual pyramid, it does not add to the connectivity diameter of the brain. Sensory signals have to travel only about five or six neurons, or less, before reaching the motor layer. Fast sensorimotor response time is absolutely crucial to every organism's survival. The importance of having a narrow connectivity diameter cannot be overstated, especially when one considers the slow nature of biological neurons.
August 11, 2005 8:05 PM EST
There is no longer any doubt that there are such things as "grandmother" cells in the brain. A couple of months ago, Rodrigo Quiroga and his colleagues at the University of Leicester, UK, located and identified grandmother-like cells (source New Scientist) in the hippocampus of their patients. In my opinion, these neurons should be called invariant concept cells because they fire continually while a subject is focusing on a particular concept. At the very least, a grandmother cell is used as part of the attention control mechanism of the brain. That is, it has activation and deactivation control over a group of neurons in memory which respond to a specific object or concept.
Most AI researchers assume that a concept cell is the end point of a multi-level perceptual pyramid that is rooted in sensory data. There are several problems with this model.
First, if the goal of perception is to activate a single cell at the top of a hierarchical tree, what does a brain do with a single firing cell? In other words, what complex motor task can it perform with it? Not much. Certainly, if it could be built, one could use a perceptual pyramid to create an invariant recognition engine but there is a wide gulf between simple recognition and complex motor behavior.
Second, sensory signals must travel many levels up the tree in order to reach the top. Given the slow performance of biological neurons, stimulus response times would be unacceptably slow. As Dr. Rodney Brooks (director of the MIT AI Lab) points out, the connectivity diameter of the brain is only about five of six neurons, not nearly enough to account for the many levels that would be required to generate a grandmother cell. A diameter of six neurons is on a par with the brain's fast response time as evidenced by the psychological record.
Finally, there is the problem of attention. The pyramid hypothesis provides no mechanism for attentional focus. If a subject is presented with a picture with several objects side by side, the model cannot account for the subject's ability to concentrate on one concept/object at a time while ignoring the others.
In the model that I propose in these pages (based on my interpretation of certain Biblical metaphors), a concept cell is an attention cell. That is, as long as it continues to fire, it keeps the system focused on a particular concept and the behavioral and sensory channels associated with it. This ability is very important to intelligence. It is corroborated by psychological research [pdf] that shows the existence of a canalization of sensory and motor activity while performing a particular task. Canalization is only possible if the intelligent system has the ability to activate a group of related memory cells while deactivating all others. Two questions arise: a) what are memory cells and b) how does an intelligent system determine which memory cells are related so as to group them as a single coherent concept?
Memory (short and long-term) consists of a huge number of seven-node sequences (see Memory Model). These store intervals in such a way that the recording of the first interval is enough to predict the other six (dynamic pattern completion). The purpose of sequences is to anticipate the arrival of sensory signals so as to compensate for missing information. That is to say, if node B is expected to fire at a specific time after node A, the anticipatory mechanism will cause node B to fire even if it does not receive an incoming signal at the expected time. The nodes receive signals from the sensory cortices and send output signals directly to effectors to execute motor actions. Sensory learning and memory sequence formation is part of what I call bottom-up perceptual learning.
It is virtually impossible to generate an invariant concept cell from perceptual learning alone. Consider the grandmother cell example. We are able to recognize grandma even if she is facing away from us, by observing her gait or silhouette or by hearing her voice. We are somehow able to associate a myriad disparate patterns with a single concept, even in the presence of unrelated distracting patterns or incomplete information. How is that possible? In my opinion, we form concepts by finding and grouping related memory sequences using a trial and error process based on motor learning. This is because only related signals can be used to generate coordinated or non-conflicting motor commands. Motor behavior runs the gamut from speech generation to walking, grasping, turning one's head, moving one's eyes, etc... But behavior is not always observable, at least in humans. Thinking is a form of behavior that is not accompanied by external movements. We can all speak to ourselves or play back a melody internally without actually making any sound. The point I am getting at is that we can form a new concept or awaken a previously learned concept without having to express it through external behavior.
In order to understand what a motor conflict is, we must first specify what is meant by motor action and motor command. A motor action is a physical effect on the system's environment which consists of the system's own actuators, i.e., muscles, motors, solenoids, valves, etc... Every action has at least one associated effector which is a motor command neuron that services a given actuator. Each effector has a unique intensity level and may receive multiple command connections from multiple sources. Likewise, an actuator may be controlled by multiple effectors. Every action has duration and thus a beginning and an end. As a result, every effector is controlled by two complementary types of motor commands: start and stop. That is, a command can be used to either start a motor action or stop it. There are two types of motor conflicts: out-of-step and multiple-activation.
This is all there is to motor conflicts, actions and commands. It is all about timing. Conflict resolution simply consists of detecting conflicts as they happen and disconnecting bad command connections (synapses). These connections all have an initial strength which is increased by a small amount every time a command signal is received. However, whenever a conflict is detected, guilty connections are strongly weakened. Eventually, only non-conflicting connections survive.
Group formation is based on the premise that related sequences produce signals that can be used to generate coordinated (non-conflicting) motor commands. Every group is under the control of a single control cell which I earlier called an invariant concept cell. Every control cell tries to harness as many non-conflicting sequences as possible and may have sequences in common with other control cells. Obviously, given the finite number of motor resources and the impossibility of doing more than one task at a time, not all memory sequences can be active concurrently. Therefore, only one control cell and its group can be active at a time. The beauty of having control cells and memory groups is that it solves both the behavior selection problem and the credit assignment problem.
August 6, 2005 9:15 PM EST
I am getting close to a final breakthrough, real close. My current understanding of the principles and organization of the brain is enough for me to start writing code in earnest. In fact, I believe that the only things stopping me from building a truly adaptive and intelligent machine or computer program are time and money. My plan is to build a single hemisphere system at first, even though I think I have an excellent understanding of how a two-hemisphere system should work. For my purposes, one hemisphere will do just fine. My platform of choice is a four or six-legged, crab-like robot with multiple degrees of freedom in each leg, one or two manipulator arms for grasping, and stereoscopic visual and auditory sensors. It would have a number of other types of sensors as well, such as touch, torque, acceleration, inclination, etc... Failing that, I would love to work on a speech or handwriting recognition system. Unless I can obtain the necessary funding, I plan to modify my existing Animal program to incorporate my new understanding. Progress will be slow because I don't have much spare time.
My first intelligent system would not include a cerebellum because I don't think it is necessary in order to demonstrate true general intelligence. However, it is indispensable if one's goal is to build a machine with full human reasoning capabilities. The cerebellum is needed to assist the brain in doing non-volitional automated tasks while the system is thinking about something else. The reason is that thinking is a type of motor behavior and it requires the full use of the basal ganglia, the seat of motor control.
Those of you who have followed my work over the years know that I have stuck to it even in the face of ridicule from some in the scientific community. I have been called a crackpot and a crank by many. I am already on several crackpot lists on the internet. Some university professor even wrote a full article on his site just to mock me. Paul Z. Myers is proud of his credentials and his understanding of neurobiology. He loves to tell others that he is an associate professor of biology at the University of Minnesota, Morris. But I am not easily phased by misguided or mean-spirited criticism and I was never one to be intimidated by academia. I hope for his sake that Myers has a taste for crow because I now have orders of magnitude more confidence in the validity of my Biblical AI work than I did in the beginning. I can say in all honesty that I have made a number of significant breakthroughs over the last two years, the most recent ones occurring just in the last two weeks. None of these breakthroughs would have been possible had I not been given a glimpse of the significance and true meaning of the Biblical metaphors and had I not been willing to go out on a limb and take a risk. Faith is mostly courage.
A behavior is a group of motor sequences that performs a given task. Every behavior is controlled by a single parent cell which is known as a grandmother cell in the artificial intelligence and neuroscience communities. Most researchers assume that sensory signals travel up a hierarchical tree until they reach a grandmother cell. In their model, the grandmother cell fires after every relevant neuron at every level in the hierarchy has fired. My model, which is based on my understanding of the Biblical metaphors in the book of Revelation and Zechariah, posits that it is the grandmother cell that controls and creates the hierarchy, not the other way around. What this means is that a number of existing memory sequences is harnessed by a control cell so as to form a coherent group. Once the group has been assembled, it is up to the control cell to activate (awaken) or deactivate it. Note that this is a flat hierarchy, one which is formed in a top-down manner, as opposed to the bottom-up pyramidal hierarchy proposed by most researchers. I'll have more to say on this in a future article.
There are several factors that determine when and how a behavior is awakened. Every control cell receives feedback connections from the sequences it controls. I have reasons based on the Biblical metaphors (more on this in a future article) to suspect that a feedback connection is made with the primary cell of every sequence in the group. Since I believe that the primary cells are located in the hippocampus, it seems natural that the control cells should also reside in the hippocampus. My current thinking is that a control cell will wake up its entire group if it receives enough feedback signals from the group to warrant it. A problem arises when multiple groups are awakened concurrently because motor conflicts are bound to occur. Conflict resolution between groups is straightforward in my opinion. Every control cell has a motivational strength which is determined by past experience and the number of wake-up signals (a measure of salience) it receives from its group. Whenever two or more groups are in conflict, the group with the strongest control cell wins.
An organism is said to be adaptive if it refrains from behaviors that lead toward pain or discomfort and selects the ones that lead toward pleasure (appetitive behavior) and away from pain (aversive behavior). Every control cell receives visceral signals originating from pain and pleasure sensors. The important thing to keep in mind about pain and pleasure sensors is that the more recent the signal, the more influential it is. This is indicated in the letter to the church of Thyatira (Revelation 2:19) thus: And your works, the last more than the first. I hypothesize that, when awakened, control cells immediately send request signals (is my recent behavior good or bad?) to the Amygdala which, in turn, sends relevant data back to the control cells in the hippocampus. Regardless of the actual biological mechanism, the goal is compare behaviors based on a scale of goodness and saliency: And all the churches shall know that I am he who searches the reins and hearts and I will give to everyone according to his works (see Revelation 2:23). Reins (kidneys) symbolizes reason (memory sequences) whereas hearts represents the seat of emotions.
Whenever an intelligent organism has a choice between two or more behaviors, only the one that is deemed to be the most beneficial is activated and given motor access. The others must be deactivated (cast unto a bed. Revelation 2:22). If the behaviors have equal strength, then one must be selected at random.
August 4, 2005 12:05 AM EST
It seems that a large number of AI researchers have concluded that memory is organized as a feed-forward hierarchy. I disagree with their stance on the basis of strong psychological evidence and on my understanding of certain Biblical metaphors, but I do not dismiss their reasoning out of hand. It is certain that low-level behaviors are reused by higher-level goals. That is, we do not have to relearn how to walk or talk every time we need to accomplish a new goal (behavior). So, how can the flat memory hierarchy that I proposed (see below) give rise to the nested behaviors that we observe in humans and animals? In my opinion, this is accomplished through a simple mechanism that I call sequence sharing.
Memory is divided into a huge number of groups ( as the vessels of the potter are broken into pieces). A memory group is a collection of coherent motor sequences. Two or more sequences are said to be coherent if they don't generate motor conflicts while running concurrently. Several sequences can be combined into high-level groups under the control of a parent cell. Parent cells may share one or more sequences. As seen in the figure above, groups A and B have three sequences (red rectangles) in common.
Note: It is important to understand that this model is not based on either a feed-forward or a feedback hierarchy. That is, a group does not feed output signals so as to form a higher level node in a hierarchical tree. The curved lines in the picture are not output lines. They are control lines used by the parent cell to activate or deactivate an entire group at once. As I write elsewhere, each sequence consists of seven nodes. It receives input signals (not shown above) from the sensory cortex and sends command signals to the motor layer.
It turns out that the group and sequence sharing model explained above affords us with an easy solution to a problem in reinforcement learning known as the credit assignment problem. In the past, I often wondered how sequences could be assigned credit or blame since so many of them can be active simultaneously. Besides, a given sequence may be reused for different purposes, good and bad. Obviously, individual sequences cannot bear responsibility for their actions. It is easy to see how the problem can be solved by assigning credit, not to individual sequences, but to sequence groups. A group should be seen as an agent which performs one or more tasks which can have consequences, either good or bad.
In my next article, I will explain how memory groups can be used to generate both appetitive and aversive behavior.
August 1, 2005 14:10 PM EST
Recently, I have become convinced that
the sort of multi-level, feed-forward memory hierarchy proposed by Jeff Hawkins and
others is fatally flawed. For one, ample psychological evidence on
human and animal response timing clearly demonstrates that sensory
signals could not possibly travel more than six of seven layers within
the brain in order to reach the motor layer. A feed-forward hierarchical tree of the
sort needed to generate a grandmother recognition cell would need so
many layers that response timing would be much too slow to match the experimental
record or ensure the organism's survival.
August 1, 2005 12:40 AM EST
RebelScience.org now has its own discussion forums. All the forums are housed under one roof. Please feel free to register and let me know what you think.
July 25, 2005, 12:25 AM EST
In his book, On Intelligence, Jeff Hawkins mentions the brain's ability to recognize moving objects from various angles and under changing lighting conditions. He writes:
I think Mr. Hawkins is right about the crucial importance of invariant pattern recognition. No moderately complex intelligent organism could function without this ability. In my opinion, one of the reasons for the scientific community's failure to crack the problem is that neuroscientists and AI researchers have made a number of wrong assumptions regarding the organization of memory, not the least of which is the common belief that memory hierarchy is a result of perceptual learning, as opposed to motor learning.
Lately, as a result of my continuing research in decoding certain Biblical metaphors (see AI From the Bible), my understanding of memory organization (both STM and LTM) has risen to a point where I can confidently claim that the solution to the problem is finally here. I believe that invariance is the result of two memory mechanisms. First, memory is organized into a huge number of individual sequences, each consisting of seven temporally correlated nodes. This means that the timing of a sequence is not static but dynamic. Hence, a given sequence can be used to encode a sizeable range of temporal signatures associated with the same phenomenon. An example of this is the ability to recognize a learned melody regardless of tempo. Second, an entire group of non-conflicting or related sequences representing a complex and dynamic sensory object can be activated (awakened) by a single parent neuron. Once activated, the group will stay awake for as long as the object remains within the sensory space. The parent control neuron is what is known in the AI community as a grandmother cell. See Jezebel and Memory Hierarchy below and the Memory Model for more info.
Note that, in this scheme, there is no strict, unyielding hierarchy wired into the connectionist structure of memory during perceptual learning. More importantly, hierarchy is not encoded with the help of feedback loops or a feed-forward hierarchical tree, as Mr. Hawkins and others suppose. Contrary to conventional wisdom, a grandmother cell does not emerge from a group of perceptual neurons. Rather, it uses principles of motor coordination to assemble the group. Memory sequences are useless unless they are tied to motor behavior. Behavior groups continually wake up and go to sleep depending on sensory circumstances and conflict resolution. The result is that memory is divided into a huge number of small and self-coherent behavioral clusters. This is symbolized in the message to the church of Thyatira thus: "And he shall rule them [the nations] with a rod of iron, as the vessels of the potter are broken into pieces, [...]".
More to come...
July 24, 2005, 12:15 AM EST
First the bad news. The day after I posted my July 10 news item below, my computer suffered a power surge and blew up, spewing smoke and electrical sparks. At this point, I am not sure I can recover any of the stuff I was working on. I found myself stuck without a computer for two difficult weeks. Fortunately, thanks to the generosity of a family member, I now have a brand new Dell Inspiron 6000 laptop and I love it. It is much more powerful than my antique and now useless desktop machine. Problem is, I still have to find a way to recover my lost files from my old hard drives.
The good news is that I am making continued and rapid progress in my understanding of brain memory organization and the Biblical metaphors that pertain to it. I now know that I was mistaken about the role played by the black and white chariots in the hierarchical organization of memory. Direct memory feedback pathways, as symbolized by the two complementary chariots, have nothing to do with memory hierarchy. These chariots only symbolize crossed communication pathways between the two hemispheres of the brain. My understanding is that these pathways link the two hippocampal formations of the left and right hemispheres of the brain. I believe that the master or predictor nodes (see figure below) of the seven-node sequences reside in the hippocampal formation. The other six nodes reside in the various lobes of the cerebral cortex.
I now understand that memory hierarchy is not a product of perceptual learning but of motor learning and attention! This is counterintuitive, I know, but I see no other possible interpretation. Note that motor learning does not necessarily imply actual motor output to the muscles. The brain is able to rehearse volitional (conscious) motor behavior while inhibiting output motor commands. This way, conflicts can be weeded out before committing memory sequences to motor actions. The key to memory organization can be found in the message to the church of Thyatira in the book of Revelation. In it, we read an admonition against Jezebel's seducing the people into fornication and idolatry (motor conflicts). I interpret this to mean that multiple non-conflicting sequences are herded together to form a coherent behavior group under a single authority. The group can be either awake or asleep depending on circumstances. i.e., on whether or not it is conflicting with another awakened group. This is the mechanism of attention. Normally, a group will automatically wake up whenever it receives input stimuli from the sensory layers. However, it can do so only if it does not conflict with another group. If there is a conflict, the strongest group wins and the other goes back to sleep. The strength of a group is determined by past pain and pleasure stimuli. In a coming article, I'll explain how this mechanism gives rise to invariant pattern recognition and the "grandmother cell" phenomenon.
At this point, I am still struggling with a few issues such as group formation, whether or not groups can overlap or be used by another group, and the correct metaphorical meaning of Jezebel's children. More to come.
July 10, 2005, 10:50 PM EST
What follows is intended to give readers an idea of my current thinking on hierarchical memory within the context of Zechariah's vision and the seven churches in Revelation. In my opinion, Zechariah's vision of the four chariots is a metaphorical description of various efferent pathways coming out of cortical memory. Please read what Zechariah wrote in the beginning of chapter 6. My interpretation follows.
In my opinion, the bronze mountains are a metaphor for the source of the burnished bronze feet of the Son of God as written in the message to the church of Thyatira. Feet are made for walking (motor output) and the Son of God is said to walk in the middle or center of the seven churches, which is Thyatira. But why two mountains? This seems to be in keeping with the same duality that we have already seen in Zechariah's vision, e.g., the two olive trees, the two olive branches and the two golden pipes. My current interpretation is that it is a metaphor for the left/right hemispheres of the brain. The entire brain is symbolized by the whole earth metaphor. Thus the two bronze mountains represent the left and right cerebral (memory) cortices of the brain. This is important in order to understand the meaning of the four chariots.
The strange thing about the four chariots is that the destination of the red horses is not mentioned even though it is clear from the text that all four chariots are going somewhere. My usual intuition in cases where information is missing in one place is that it is given elsewhere. In this case, I think the destination of the red horses is given in the metaphor of the woman in the ephah. It seems that certain memory nodes generate signals that are altogether useless whether for motor or feedback purposes. My understanding is that these signals are sent to a special location (a signal garbage dump, so to speak) in the brain where signals go in but nothing comes out (symbolized by the lead cover on the ephah). I am not sure what the utility of a signal dump is but it is an interesting revelation about the human brain that should be testable. It may also apply to animal brains. Not being an expert in neuro-anatomy, I have no idea whether or not this is already something known by neuroscience. I would appreciate any feedback on this matter.
The white horses follow the black horses to the north country. What does this mean? In my opinion, it means two complementary types of feedback pathways, one going to Zerubbabel-type connections and the other to the coincidence detectors that generate friends and Joshua-type signals. The coincidence detectors are very important in this feedback scheme because they act as temporal filters which automatically block unrelated signals. I'll have more to say on this in a future page. The question is, why separate the signals into two distinct types? To answer this question it is necessary to remember that the left hemisphere is subservient to the right brain. This means that there must be communication between the two hemispheres. Obviously, the master does not communicate to the slave the same way the slave communicates to the master. How is it done? The clue comes from the message to the Church of Sardis where we learn the following in verse 3:5:
This, in my opinion, is how the confession (or communication) is accomplished: by presenting the signals to the master's (the Father's) angels the way Joshua, in Zechariah's vision, is presented before the angel. In other words, feedback from the left or subservient hemisphere generates Joshua or Friend-type signals in the right hemisphere. On the other hand, feedback fibers from the right or master hemisphere make Zerubbabel-type connections in the left hemisphere. This makes sense because Zerubbabel-type signals are what initiates motor sequences.
Note: An important aspect of feedback has to do with attention or focus. Only a relatively small number of node sequences (lampstands) are selected (i.e., are awake) at any one time. This limits the manner in which feedback is organized and used in memory. In a future update to the Thyatira page, I will provide more details into how this is accomplished.
These are mixed colors (black and white) horses and they go to the south country which I interpret to mean the motor layer. These horses are said to be strong which I interpret to mean that no motor connection is initiated unless the node has reached a certain strength which is an indication of its certainty. But why are the colors mixed? The answer is that, unlike the black and white horses, the dappled signals can be used for complementary motor commands. That is to say, they can be used either as start or stop commands. See the interpretation of the message to the church of Pergamum for more details on motor control.
More to come. Hang in there.
July 5, 2005, 11:45 AM EST
In the past, I have often wondered how we are able to go through our memories at a faster or slower rate than they were recorded from sensory experience. For examples, we can speak at different rates and we can recall and play back a learned melody at a fast or slow pace with relative ease. In view of the latest memory model based on the metaphors of Zechariah, it now seems obvious that what is important is not the absolute values of the recorded intervals but the relative timing or order of events. This is the "how" of memory mentioned in the message to the church of Sardis. The brain can scan memory traces at different rates without destroying their original temporal order. This is highly significant because future intelligent machines, by virtue of their fast electronics, will be able to reason/think at amazing speeds compared to humans and/or animals.
July 1, 2005, 1:40 PM EST
Ok. It's been three weeks since my last news update. I apologize for the delay. Again, I find it hard to allocate as much time into this project as I would like. Also I've been having difficulty interpreting the text of Zechariah 6:9-15 regarding the exiles and Joshua's crown. This morning, I had a sudden insight when I read verse 14 for the umpteenth time:
Compare the above to verse 10:
My interpretation is that, unlike Zerubbabel, Joshua's signal is not a single signal arriving from the sensory layers (Babylon). It is generated from the simultaneous arrivals of multiple signals (exiles) originating from the sensory layers. As indicated, the arrival of the concurrent signals is not deterministic. As soon as the signals (exiles) arrive, a memory trace (a reminder, symbolized by the crown) is created in the main node. As we have seen in the text of Sardis, what is memorized is the interval between two signals, in this case, between Joshua and Zerubbabel. This is where the Branch comes in:
The Branch is an output signal coming out of Joshua's node (for he will branch out from where he is). It's obvious that the second pronoun "he" in the verse is referring to Joshua. It's also evident that the Branch has something to do with the reminder (Joshua's crown). In my opinion, the Branch is a special anticipatory signal. As soon as a new memory trace is created in Joshua's node, a special signal is sent out to the other nodes to reset their intervals. There must be a correlation between the primary node's interval and the others. That is to say, Joshua's interval predicts the others.
This is useful in several ways. Some time ago, I gave the example of a bloodhound chasing a hare. The anticipatory mechanism of the brain needs to instantly calculate the future evolution of events in the environment. The brain also needs to fill in signal gaps in situations where part of the sensory information is missing or is not totally intelligible. An example of this is following a conversation in a noisy room.
At this point in the project, I feel that I am almost there. My only problem right now is time and money. More to come. Hang in there.
Addendum, 2:35 PM
I wrote above that there is a correlation between the primary node's interval and the other six. What is the correlation? In my opinion, it is a simple percentage. In other words, each friend node's interval is a percentage of the interval recorded at Joshua's node. When I update the Memory Model page, I'll explain how correlations are learned. On another note, I believe that the incoming friend signals are, like Joshua's signal, the result of the concurrent arrivals of multiple signals from the sensory layers (Babylon). Of course, there are also feedback signals but I'll leave this subject for a future page on the metaphor of the four charriots.
June 08, 2005, 2:05 PM EST
I now have a new memory model based on Zechariah's vision. I have not yet updated the relevant pages. I will upload a new memory model page soon with the accompanying explanation. In the mean time, here's the newest memory sequence diagram:
As you can see, things have changed quite a bit from the last model. It's amazing what a sudden new insight can do. More to come soon.
June 07, 2005, 2:10 PM EST
Once again, I find myself on the verge of changing my mind about some things that I had taken for granted. I now realize that my interpretation of Zechariah's vision is way too complicated. In a couple of days, I will upload a new memory model, one which is much simpler and much more consistent with the metaphors used in the vision.
©2004-2007 Louis Savain
Copy and distribute freely