A newborn’s brain contains the basic information and mechanisms that are needed for immediate survival and for the mental development of the individual. Those innate features are represented by the nodes of the model. The innate nodes can detect certain input patterns and activate appropriate responses, including body reflexes and learning mechanisms. Incorporating new information in a brain is accomplished by adding and modifying synapses between neurons. In the model this is represented by adding new nodes to the network, and by adjusting synaptic weights between nodes. All new additions to the innate network are linked to existing nodes and among themselves by the universal relationships: temporal, item-parts, and class exemplars.
The visual system of a baby contains innate circuitries that can detect a variety of light patterns. Innate detectors can detect light and dark pixels in the field of view. Other innate detectors detect groups of pixels that have the same illumination. Others can detect the boundary line between areas of different illuminations. All those are basic concepts, which are represented by nodes. They are exposed to the stream of visual stimuli that reach the brain, and extract from it combinations of visual features. The activators organize the extracted information as concepts in the information structures of the brain.
Quite early, the baby learns to recognize his mother's face. 'Mother's face' becomes a new item in the baby's memory. An entire information structure is recruited to represent this new concept. At the bottom of the structure are the innate concepts that represent individual pixels and their features. They serve as parts of recruited items. The item 'nostrils', for example, has the corresponding dark pixels as its parts. The item 'outer contour of the hair' has as its parts the boundary pixels between the hair and the background, and so on.
In addition to representing acquired information, some of the new concepts serve as detectors. They fire when their pattern recurs in the information stream. They trigger learning activators, which add records of the new event to the information structure.
With time, the baby learns to recognize faces of other people, such as his father. A similar information structure is constructed for the new concept 'father's face'. Then, other innate detectors and activators are triggered. They detect the similarities between 'father's face' and 'mother's face'. They recruit a new node 'face', which serves as a class node , and establish class-exemplars connections between 'face' and the nodes 'mother's face', and 'father's face'.
Then, the new concept 'face', acts as a detector, and together with the other existing detectors and activators define the concept 'person', which is a class node whose exemplars are all the items that have 'face' and 'move around', and so on.
New nodes act as detectors and activators, and add more nodes to the network
The innate brain has a small number of concepts, some of which are associated with one another. Functionally, they act as detectors and activators. In addition to executing innate life-sustaining activities, they detect and record certain patterns of signals that flow through the brain. The recording is realized by recruiting new nodes, which represent new concepts, and by forming new connections between nodes, which represent the basic associations: temporal, item-parts, and class-exemplars. At any given time, all the innate and acquired concepts act as detectors and activators. They extract new concepts from the signals that flow through the brain and add them to the existing information structures. This is how the brain expands its information structures.
Figure :The blocks of the innate brain
Basic and compound concepts
The innate detectors and activators of the brain underlie the basic concepts with which our entire perception of the world and of ourselves is built. Innate visual detectors define the colors that we can sense, they divide visual scenes into parts, distinguish moving objects from those at rest, and so on. Our taste buds and smell sensors determine the different flavors that we can detect. Our auditory detectors determine the sound frequencies that we can hear and divide the auditory scene into parts, manageable by the brain. The brain's innate activators determine our basic motor responses to the world and our basic feelings such as joy and fear. Innate temporal detectors enable us to detect and sort events according to the time that they happen. Innate detectors and activators determine how the brain records new information and how it retrieves and uses recorded information. All the acquired concepts that the brain uses are related, directly or indirectly, to the innate concepts of the brain.
Human commonsense is the outcome of the universal relationships between nodes
The innate brain has quite a few basic detectors and activators, but the number of concepts that it can build with them is enormous. Our perception of the world is made up of basic concepts that are combined into compound concepts. The logic relationships AND, OR, and NOT are the glue that holds together the compound concepts. Those logic relationships are represented in the model by the synaptic weights of the universal relationships: the temporal, the item-parts, and the class-exemplars relationships.
In an item-parts association, the logical relationships that holds the parts together is AND. If part 1, AND part 2, AND so on till the last part are present, then their item is either also present, or the brain may assume that it is present. In reverse, if the item is present, then part 1, AND Part 2, AND etc. are either present or the brain may assume that they are present. The item and the parts may be basic or compound concepts. For example: “If it looks like pizza, AND it smells like pizza, AND it tastes like pizza, then it is (or you may assume that it is) a pizza”. And in reverse, “If it is a pizza, then it looks like a pizza, AND it smells like a pizza, AND it tastes like a pizza (or you may assume so).
An item-parts information structure is used for two purposes. First, as a record of past experiences. Second, as a detector that associates ongoing experiences with past ones. The ongoing experience does not have to match exactly a past one. For example, if only the part 'it smells like a pizza' is happening, the brain may assume that a 'pizza' is happening. A sub group of the original parts may invoke the presence of the item. Such partial groups are cues of the item.
In a class-exemplars association, the logical relationship that holds the exemplars together is OR. If exemplar 1, OR exemplar 2, OR etc. is present, then their class is either also present, or the brain may assume that it is present. In reverse, if a class is present, then one or more of its exemplars are present, or the brain may assume so. For example, if it is a dove, or a finch, or a robin then it is a bird. And in reverse, if it is a bird then it is a dove or a finch or a robin.
Like items, class information-structures are also used for two purposes: as a record of past experiences and as a detector that associates items of ongoing experiences with recorded classes. A class has cues. An item that has such a cue as its part is qualified as an exemplar of the class. For example, an item that has a beak and feathers is probably an exemplar of 'bird'. A cue is related to its features by the item-parts (AND) association. The relationships between an item and its cues and between an exemplar and its cues is by the class-exemplars (OR) association.
The legacy of a class is a group of concepts associated with the class. They are held together as a group by the logical AND. When a concept inherits them, they become associated with it by the same associations that they have with the class. For example, the class 'bird' has a legacy of 'has two legs' AND 'hatched from an egg'. If you see for the first time a swan in a lake and, based on its beak and feathers, you conclude that this is a bird, you might also conclude that it has two legs and it came from as egg –the inheritance of the class ‘bird’. The ‘two legs’ is a part of 'bird', and 'hatched from an egg' is a compound relationship between 'egg' and 'bird'.
Both item-parts and class-exemplars associations are not exclusive. An item represents a group of parts, but it does not exclude other entities from being its parts. A class represents concepts that have certain common features, but it does not exclude concepts that has additional features.
The brain can explicitly exclude certain concepts from being parts of an item, and from being exemplars of a class. This is accomplished by adding the basic qualifier NOT to the logic AND and OR relationships. Thus, item-parts associations use the logic relationships AND and AND-NOT, and class-exemplars associations use the logic relationships OR and AND-NOT. For example, if the cue of the class ‘bird’ is ‘exemplar of animal’ AND (‘can fly’ OR ‘lays eggs’) all birds, including kiwi and ostrich will be included. To exclude flying fish from this class, AND NOT ‘exemplar of fish’ could be added to the cue.
Mutually exclusive classes and conflicting information
Mutually exclusive classes are classes that cannot have a common exemplar at the same time. For example, 'live cat' and 'dead cat' are two mutually exclusive classes. 'Mitzi' cannot be an exemplar of both classes at the same time.
The brain builds its information structures based on experiences and verbal instructions. It may happen, especially in verbal instruction, that an item is recorded as an exemplar of mutually exclusive classes. For example, if one is told and records in memory that 'Mitzi is dead' and then that 'Mitzi is alive', the information structure contains conflicting information.
The brain may sometimes realize that it contains conflicting information. It has several mechanisms to handle such situations. 1. The information may be tagged as conflicting so that the brain seeks more information about it, or uses it cautiously. For example, decide to tag the information about the disposition of 'Mitzi' as questionable. 2. The cause of the conflict may be erased, thus eliminating the conflict. For example, assuming that Mitzi is alive, and erasing the record that it is dead eliminates the conflict. 3. The cause of the conflict becomes a member of a new mutually exclusive class. For example, the class 'paradoxical cat' can be added to the classes 'live cat' and 'dead cat', and 'Mitzi' becomes an exemplar of the ‘paradoxical cat’ class.
Conflicting information may reside unnoticed in the brain. This is due to the way that the brain records new information. When new information is recorded, new nodes establish explicit associations with some of the existing concepts. That may be in conflict with implicit information somewhere else in the network. For example, the existing information is ‘Abe is taller than Bob’ and ' Bob is taller than Charlie'. If now 'Charlie is taller than Abe' is added, a conflict is created. The conflict is due to implicit information about the height relationship between Abe and Charlie (it is implied that Abe is taller than Charlie).
Handling conflicting information
Although conflicting information may reside in the memory, the brain has innate mechanisms that can detect it. We sense something 'wrong' when the two concepts 'Mitzi is dead' and 'Mitzy is alive' are activated at the same time. It is not known what is the biological underlying circuitry for such processes. The model can handle them in an ad hoc way.
The brain may employ various strategies to handle conflicting information once it encounters it. For example, the brain may direct the attention to the conflicting part, try to search for new information that clarifies it, and then correct it. (I see Mitzy running around, so she cannot be dead.) The brain may also leave it as is. (I will have to find later how Mitzy is doing.)