JP7193870B2 - Artificial brain system and intelligence device used in it - Google Patents

Description

The present invention relates to artificial brain technology.

A neural network has been proposed as an arithmetic processing model that imitates the information processing style of the brain (see Patent Document 1). This neural network is a model that imitates neurons and synapses in the brain, and consists of two stages of processing: learning and discrimination. In the learning stage, the features are learned from many inputs to build a neural network for classification processing. In the identification stage, a neural network is used to identify what the new input is.

In recent years, techniques in the learning stage have made great progress. For example, deep learning is making it possible to build multilayer neural networks with high expressiveness. The knowledge structure of the neural network can be said to be the optimum knowledge structure because the processing speed is high when inference is performed as knowledge in which the structure of conceptual knowledge does not change. In addition, the more parent data is learned, the deeper the knowledge of feature concepts becomes, and it can be said that this is an excellent knowledge inference system when evaluating the entire concept knowledge.

JP 2010-146514 A

However, in a neural network, which can be said to be an artificial brain that imitates the information processing style of the brain, it is necessary to extract features from a large amount of input information for learning, and the learning takes a lot of time. In addition, it is not possible to perform identification operations (inference operations) during learning, which requires a great deal of time, and the neural network cannot be used as a brain during learning. In addition, conventional artificial intelligence systems using neural networks cannot autonomously determine and operate based on their own will.

SUMMARY OF THE INVENTION The present invention has been made in view of such circumstances, and provides an artificial brain system that can autonomously determine and operate according to its own will.

The present invention also provides intelligent devices for use in such artificial brain systems.

A system according to the present invention includes an information vessel that holds provided information as valid information in an outputtable state, and one or more knowledge units communicatively coupled to the information vessel, wherein Each of the one or more knowledge units includes a knowledge storage section for storing knowledge information including data values representing each of a plurality of concepts constituting one knowledge; A knowledge management unit that manages knowledge information so as to associate the concept with a data value, and activates the concept awakened in the knowledge information managed by the knowledge management unit. The concept is awakened when the information effectively held in the information vessel matches the data value of the concept in the knowledge information managed by the knowledge management unit based on the intention attribute that determines the concept. activation information determining means for determining a concept to be activated in the knowledge information managed by the knowledge management unit; and an information transmission means for providing the information vessel with the data value of the concept determined by the information determination means.

With such a configuration, when data values of one or a plurality of concepts included in a plurality of concepts constituting one knowledge (knowledge information) are provided to the information vessel, the knowledge unit corresponding to the knowledge is provided. Then, assuming that one or more concepts are awakened in the knowledge, a certain concept included in the one knowledge (knowledge information) is activated based on the intention attribute (will of the knowledge unit), Data values for the activated concepts are sent to the information vessel. That is, transmission of data values to the information vessel in response to data values of one or more concepts provided to the information vessel is performed as one knowledge-based artificial brain activity (knowledge activity).

In the artificial brain system according to the present invention, the information vessel may include a database in which provided information is registered as valid information.

According to such a configuration, when information is registered in the database, the information is held in the database as valid information that can be accessed by each knowledge unit.

In the artificial brain system according to the present invention, an information input unit for providing information as a data value of a concept constituting a certain knowledge from an external device to the information vessel; and an information output unit that outputs a conceptual data value provided to the external device to the external device.

According to such a configuration, when information as a data value of a concept constituting a certain knowledge is provided to the information vessel from an external device through the information input unit, the knowledge is provided in response to the provision of the information. The data values of the activated concepts that make up that knowledge are emitted from the corresponding knowledge units to the information vessel. Then, the data value is output from the information vessel to the external device through the information output section.

In the artificial brain system according to the present invention, at least one of the one or more knowledge units further includes a concept that the activation information determining means is awakened from information effectively held in the information vessel. If the concept to be activated cannot be determined based on the intention attribute due to the lack of knowledge, the knowledge managed by the knowledge management unit is based on the learning intention attribute that requests learning of the missing concept. a first learning request means for providing learning request information representing a learning request for a concept to be learned in identification information to the information vessel; and responding to the provision of the learning request information to the information vessel by the first learning request means. and when the information provided as educational information to the information vessel matches the data value of the concept to be learned, the learning control means provides the concept of the data value as an awakened concept to the activation information determining means. And, it can be configured to have.

According to such a configuration, in the knowledge unit, from the information effectively held in the information vessel, the concept to be activated cannot be determined based on the intention attribute due to the lack of concepts to be awakened. If so, learning request information for the missing concept is provided to the information vessel. When the information provided as educational information to the information vessel in response to the provision of the learning request information to the information vessel matches the data value of the missing concept (concept to be learned), the concept of the data value is As an awakened concept, it serves to determine the concept to be activated. As a result, the shortage of concepts to be awakened is eliminated, the concepts corresponding to the awakened concepts in the knowledge information are activated based on the intention attributes, and the data values of the activated concepts are stored in the information vessel. be sent.

In the artificial brain system according to the present invention, at least one of the one or more knowledge units further lacks a corresponding data value in the knowledge information to be managed by the knowledge management section. a second learning request means for providing the information vessel with learning request information representing a learning request for the concept; knowledge updating means for storing information provided to the information vessel as educational information as data values corresponding to the concept in the knowledge storage section, and updating the knowledge information managed by the knowledge management section; , can be configured.

According to such a configuration, in the knowledge unit, when a concept occurs in which the corresponding data value is missing in the knowledge information to be managed by the knowledge management unit, the learning request information about the concept is stored in the information vessel. provided. Information provided as educational information to the information vessel in response to the provision of the learning request information to the information vessel is stored in the knowledge storage unit as a data value corresponding to the concept, and is managed by the knowledge management unit. knowledge information is updated.

In the artificial brain system according to the present invention, when information representing a concept to be added to the one knowledge is effectively held in the information vessel, the knowledge is managed by the knowledge management unit . It can be configured to have a concept adding means for adding as a concept constituting the knowledge represented by the information.

According to such a configuration, in the knowledge unit, when information representing a concept to be added to one corresponding knowledge is provided to the information vessel, the knowledge information managed by the knowledge management unit The above concept is added to the knowledge represented by . In this state, the added concept learning request information is provided to the information vessel as described above. Information provided as educational information to the information vessel in response to the provision of the learning request information to the information vessel is stored in the knowledge storage unit as a data value corresponding to the concept, and the knowledge management unit stores The knowledge information to be managed is updated.

In the artificial brain system according to the present invention, a learning output unit that outputs the learning request information provided to the information vessel from any one of the plurality of knowledge units to an external device, and a learning output unit that outputs the learning request information from the external device as the educational information and an educational input for providing information to the information vessel.

With such a configuration, when learning request information is provided to the information vessel from a plurality of knowledge units, the learning request information is output to the external device through the learning output section. On the other hand, information as education information responding to the learning information is provided to the information vessel through the education input unit. The educational information is then taken as a response to the learning request information into the knowledge unit that issued the learning request information.

A knowledge device according to the present invention comprises a knowledge storage section for storing knowledge information including data values representing each of a plurality of concepts constituting one knowledge; A knowledge management unit that manages information so as to associate the concepts with data values, and a concept to be activated corresponding to the concept awakened in the knowledge information managed by the knowledge management unit. When the provided information matches the data value of the concept in the knowledge information managed by the knowledge management unit based on the determined intention attribute, the concept is regarded as awakened by the knowledge management unit. activation information determination means for determining a data value of a concept to be activated in the knowledge information managed by the activation information determination means among the knowledge information stored in the knowledge storage unit and information transmitting means for transmitting the data value of the determined concept.

According to such a configuration, when data values of one or a plurality of concepts included in a plurality of concepts that constitute one piece of knowledge (knowledge information) are provided, the knowledge device performs the above-mentioned Assuming that one or more concepts are awakened, a certain concept included in the one knowledge (knowledge information) is activated based on the intention attribute (will of the knowledge unit), and the activated concept is activated. Data values are emitted. That is, transmission of data values in response to data values of one or more provided concepts is performed as one knowledge-based artificial brain activity (knowledge activity).

In the knowledge device according to the present invention, when the activation information determination means cannot determine the concept to be activated based on the intention attribute due to lack of concepts to be awakened from the provided information. , first learning for outputting learning request information representing a learning request for the concept to be learned in the knowledge information managed by the knowledge management unit, based on the learning intention attribute for requesting learning of the lacking concept; requesting means; and awakening the concept of the data value when information provided as educational information in response to the output of the learning request information by the first learning requesting means matches the data value of the concept to be learned. and learning control means for serving the activation information determination means as a concept.

Further, in the knowledge device according to the present invention, when a concept occurs in which the corresponding data value is missing in the knowledge information to be managed by the knowledge management unit, learning request information representing a learning request for the concept is generated. a second learning request means for outputting; and information provided as educational information in response to the output of the learning request information by the second learning request means is stored in the knowledge storage unit as a data value corresponding to the concept. and knowledge update means for updating the knowledge information managed by the knowledge management unit.

Further, in the knowledge device according to the present invention, when information representing a concept to be added to the one piece of knowledge is provided, the concept is represented by the knowledge information managed by the knowledge management unit . It can be configured to have a concept adding means for adding as a concept constituting the knowledge to be obtained.

According to the artificial brain system of the present invention, each knowledge unit transmits data values to the information vessel in response to data values of one or more concepts provided to the information vessel based on the intention attribute. Since it is performed as an activity based on one piece of knowledge (knowledge activity), it is possible to perform an action that is autonomously determined by one's own intention (intention attribute).

FIG. 1 is a conceptual diagram conceptually showing a knowledge model. FIG. 2 is a diagram showing a model of knowledge. FIG. 10 is a diagram showing intelligence knowledge (knowledge A) in a model of the knowledge activation model; FIG. 10 is a diagram showing another intelligence (knowledge B) in the model of the intelligence activity model; FIG. 10 is a diagram showing the state of a cybervessel in a model of an intellectual activation model; FIG. 4 is a diagram showing a model of knowledge (description of conceptual bases and intention attributes). FIG. 1 shows a model of a GLIA cell model (No. 1). FIG. 2 is a diagram showing a model of the GLIA cell model (No. 2). FIG. 10 is a diagram showing a model of the GLIA cell model (No. 3). FIG. 10 is a diagram showing a model of the GLIA cell model (No. 4). FIG. 10 is a diagram showing a model of the GLIA cell model (No. 5). FIG. 10 is a diagram showing a model of the GLIA cell model (No. 6). FIG. 10 is a diagram showing a model of the development of a subjective learning action (task-dependent type); FIG. 10 is a diagram showing a model of development of subjective learning action (self-adjusting type); FIG. 4 is a diagram showing a task-dependent knowledge model; FIG. 10 is a diagram showing a task-dependent flow; FIG. 12 illustrates a self-adjusting knowledge model; FIG. 10 is a diagram showing the flow of a knowledge adjustment type; FIG. 4 is a diagram showing a task-dependent learning relationship; FIG. 3 illustrates a self-regulating learning relationship; 1 is a block diagram showing the configuration of an application system including an artificial brain system according to one embodiment of the present invention; FIG. 1 is a block diagram showing a configuration example of an artificial brain system according to an embodiment of the present invention; FIG. FIG. 4 is a diagram showing a first example of a knowledge management table for managing knowledge information; FIG. 10 is a diagram showing a second example of a knowledge management table that manages knowledge information; FIG. 10 is a diagram showing a third example of a knowledge management table that manages knowledge information; FIG. 13 is a diagram showing a fourth example of a knowledge management table that manages knowledge information; 4 is a flow chart showing the flow of operation of a processing section in a knowledge unit (knowledge device); 4 is a flowchart showing the flow of activation processing; FIG. 4 is a diagram specifically explaining the wisdom inference function (awakening and activation) in the knowledge unit (knowledge). FIG. 10 is a diagram specifically explaining a knowledge inference function (cooperative awakening) in a knowledge unit (knowledge); 10 is a flowchart showing the flow of task-dependent learning processing; It is a figure explaining a subjective learning function (task dependent type) concretely. 4 is a flowchart showing the flow of self-adjusting learning processing; It is a figure explaining a subjective learning function (self-adjustment type) concretely. FIG. 10 is a diagram showing an example of a knowledge management table (first stage) that changes due to self-adjusting learning; It is a figure which shows an example of the knowledge management table (2nd stage) which changes by self-adjustment learning.

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.
First, a theoretical model of an artificial brain system (brain-type artificial intelligence system) according to the present invention will be described.

Content structure Active intention model
§1 Intellectual reasoning:
An effect in which knowledge is evaluated by awakening and activating knowledge.
§2 Subjective learning:
An action in which a learning request is made by the learning motivation of intellectual intelligence.

1. Introduction In the conventional artificial intelligence technology "stochastic branching type", it is not possible to change the stochastic formula determined as intelligence before the operation of artificial intelligence during operation. This means that artificial intelligence cannot correct its "intelligence" while it is operating, and if we compare it to a human being, it is impossible to develop knowledge through learning and come up with new ideas in life. In addition, in order to change the "intelligence" of artificial intelligence, it is necessary to stop artificial intelligence, which is to put it in a state of brain arrest.

The artificial intelligence technology discussed below is not the conventional "probabilistic branching type" but the "wakeful intelligence type" technology. This artificial intelligence technology is capable of (1) correcting intelligence while the artificial intelligence is in operation, and it is possible to develop intelligence while moving the brain. In addition, (2) intelligence can be made to have a will and autonomously make judgments and act, allowing autonomous learning of artificial intelligence and new combinations (ideas) between intelligences. Furthermore, (3) the state of intelligence can be stored and combined in units of time and environment, and emotional actions such as emotional intelligence awakening and evaluation based on past memories can be performed. Artificial intelligence technology with the above characteristics will be described below.

§1. Intellectual reasoning 1. Active Intention Inference Model “Active Intention Inference Model” is a model in which the intelligence itself is given active action functions and intention attributes, and based on the information it perceives, the intelligence itself acts on intentional actions and subjective actions based on its will. It is an inference model that performs a learning action.

1.1 Conceptual diagram of active reasoning intelligence model The model of active intention reasoning consists of intelligence models shared by multiple knowledges. FIG. 1 is a conceptual diagram of an intelligence model that performs active intention inference. The intelligence model is described below.

1.2 Structure of intelligence model
(1) Artificial Brain (Cyber Vessel)
・Cybervessel is a virtual container of an artificial brain, where "intelligent organization (knowledge)" exists and knowledge evaluation is performed.
・The inside of the Cyber Vessel is filled with data transfer substances (wave-like cyber particles) that diffuse the intelligence waves sent and received by the "Intelligent Organization".
・"Intelligent Organizations" exist in an independent state without physical unity, but information is exchanged by intelligence waves.

(2) Intelligence Organization (Wisdom)
The evaluation organization/intelligent organization that exists within Cyber Vessel is an independent individual organization that performs knowledge evaluation, and consists of one "AI cell" and multiple "information cells".
・In order to communicate with each other, "AI cells" and "information cells" are linked by cyber synapses that can transmit information without physical bonds.
・When the "information cell" senses the intelligence wave and awakens, the "information cell" is activated by the action of the "AI cell" and transmits the evaluation result as an intelligence wave within the Cyber Vessel.

*Intelligent organization << {AI cells}
… {information cells}
… {information cells} 》

(3) AI cells (Active intention cells)
AI cells, one of the cells that make up the intelligent tissue, are the cells that govern the actions of the intelligent tissue, and are composed of an "intelligent body" that serves as a blueprint for the knowledge evaluation action and an "autonomous body" that implements the knowledge evaluation action. ing.

* AI cells { (intelligent body) (independent body) }

(4) Intelligent Unit
One of the cell bodies that make up the AI cells, the intelligent body, is the cell body that serves as the blueprint for the action of intellect evaluation. there is

*Intelligence ([Intent attribute] [Concept base])

(5) Autonomous Unit
Autonomous bodies, one of the cell bodies that make up AI cells, are cell bodies that carry out the action of evaluating knowledge, and execution factors are defined for each type of action.
・In version 1, factors are defined for ``intentional activation'' and ``subjective learning''.
・The "intentional activation" action is the action of the autonomic body [activation factor], and when it receives an awakening notification from the informational cell, it instructs the evaluated informational cell to activate data transmission based on the intentional attribute.
・"Subjective learning" action is to make a learning request to the external operator "life management cell (described later)" based on the intention attribute when receiving a learning motivation notification from the information cell due to the action of the autonomous body [learning factor].

* Autonomous body ( [active factor] [learning factor] )

(6) Intelligence Cell
Information cells, one of the cells that make up the intelligent system, define knowledge expression values, and are composed of "data values" that express knowledge and "concept values" that express the meaning of data.
・Information cells have a "receptor organ" that senses intelligence waves and a "donor organ" that transmits data values as intelligence waves.
・As incidental effects, information cells have an "inductive awakening" effect that notifies AI cells of awakening when they are sensitively awakened by intelligence waves, and a "learning motivation" effect in which information cells that do not have data values notify AI cells of data learning. have

* Information cell { (data value) [concept value] }

2. Functions of Wisdom Awakening and Cooperative Awakening Wisdom awakening is carried out by intelligence waves (information) within Cyber Vessel. In FIG. 1, the order of knowledge awakening will be described below.

2.1 Movement of knowledge awakening First, I will explain the knowledge awakening of “Knowledge A”.
(1) External stimulus information is converted into intelligence waves ([1] [2]) by life control cells (see below) and diffused within the cybervessel.
(2) The "receptor organ" of the information cell senses the intelligence wave, and the information cell (data [1] [2]) awakens.
(3) When the information cell awakens, it notifies the AI cell of the awakening.
(4) AI cells receive information cell awakening notifications, perform knowledge evaluation based on the intelligent body's [will attribute], activate the autonomic body [activation factor], and give the evaluated information cell (data [3]) " Data value” is activated.
(5) Information cells receive instructions from AI cells, and transmit data values as intelligence waves ([3]) from the donor organ into the vessel.
As mentioned above, the phenomenon in which "wisdom A" sends data values is called "wisdom activated".

When one knowledge is awakened and activated, and an intelligence wave is transmitted into the Cyber Vessel, other knowledge is awakened and activated sequentially by receiving the intelligence wave. This phenomenon is called coordinated awakening. With reference to FIG. 1, cooperative awakening will be described below.

2.2 Coordination Awakening Movement In the order described in the previous section, "Wisdom A" is activated, and an intelligence wave ([3]) is transmitted into Cyber Vessel.
"Wisdom B" is subsequently activated.
(1) Sensitive information cells (data [1] [3]) are awakened by the intelligence wave of external stimulation ([1]) and the intelligence wave activated by knowledge A ([3]).
(2) When the information cell awakens, it notifies the AI cell of the awakening.
(3) AI cells receive the information cell's awakening notification, perform intelligence evaluation based on the intelligent body's [will attribute], activate the autonomous body [activation factor], and tell the evaluated information cell (data [4]) " Data value” is activated.
(4) Information cells receive instructions from AI cells, and transmit data values as intelligence waves ([4]) from the donor organ into the vessel.
By the above action, after ``wisdom A'' is activated, ``wisdom B'' is also activated, and the two intelligence waves ([3] [4]) emitted from both knowledges become the input information ([1 ] [2]) and output to the outside. It is also possible to output intelligent waves ([3] [4]) by controlling life management cells (described later).

3. models of intelligence
3.1 Knowledge model A concrete model of the intelligence model is explained below. FIG. 2 shows a model of knowledge.

In the model of FIG. 2, the outer circle is the body of knowledge. The three pentagonal stars are the "information cells", and the central hexagon is the "AI cell". The thin dotted triangle extending from the "information cell" to the "AI cell" is a "cyber synapse" where information can be transmitted without physical cohesion. Only one exists. The semi-circle on the outer circle of knowledge is the receptor that senses the intelligence waves, and the triangle on the outer circle is the donor that emits the intelligence waves.
In Figure 2, there are three information cells {(pink) [color]} {(spring) [season]} {(cherry blossom) [flower]}, and the central AI cell is an autonomous body with [activation factor]. And there is an intelligent entity with [concept base] [will attribute].

3.2 Model of knowledge activation model The model of the knowledge activation state is explained below.
3A-3C are models of the intellectual activity model. FIG. 3A describes an intelligence organization (knowledge A) and FIG. 3B describes an intelligence organization (knowledge B). FIG. 3C is the state of activated knowledge and intelligence waves within the cybervessel. Using this concrete model, the activation of actual knowledge and the awakening of cooperation will be explained below.

3.3 Activation of knowledge A
[1] The external input "spring" and "pink" information is converted into intelligence waves by the life control cells (described later) and diffused within the cybervessel.
[2] The receptors of “Information Cell [Spring]” and “Information Cell [Pink]” of “Wisdom A” respond to external input intelligence waves (“Spring”, “Pink”) and “Information Cell [Spring]” and “Information Cell [Pink]” awaken.
[3] "Information Cell [Spring]" and "Information Cell [Pink]" notify the "AI Cell" of awakening.
[4] "AI cells" receive the awakening notification of the information cells, perform knowledge evaluation based on the intelligent body's [will attribute] "When [season] and [color] awaken, send [flower]" and autonomously. It activates the body [activation factor] and gives an activation instruction to transmit a [data value] to the [information cell [cherry blossom]], which has the concept of [flower].
[5] "Information cell [cherry blossom]", which has the concept of "flower", transmits the data value "cherry blossom" as an intelligence wave from the donor.

3.4 Activation of knowledge B
[1] Receptors of "Information Cell [Spring]" and "Information Cell [Sakura]" of "Knowledge B" are externally input intelligence waves ("Spring") and activated "Information Cell [Sakura]". In response to (the same information cell), "Information Cell [Spring]" and "Information Cell [Sakura]" awaken.
* The awakening of "Information Cell [Spring]" is awakened when the first external input intelligence wave ("Spring") spreads.
* "Information Cell [Sakura]" of "Wisdom A" and "Wisdom B" are the same.
[2] "Information Cell [Spring]" and "Information Cell [Sakura]" notify the "AI Cell" of awakening.
[3] "AI cells" receive the awakening notification of the information cells, perform knowledge evaluation based on the intelligent body's [will attribute] "When [seasons] and [flowers] awaken, send [events]" and autonomously. It activates the body [activation factor] and gives an activation instruction to transmit a "data value" to the "information cell" which has the concept of "event".
[4] "Information cell", which has the concept of "event", transmits the "admission" of the data value as an intelligence wave from the donor.

3.5 Output of Wisdom A and Wisdom B The intelligent waves of “spring”, “pink”, “cherry blossom”, and “entrance” that are diffused within the Cyber Vessel are converted into output signals by the life control cells (described later) and sent to the outside. output to
As mentioned above, ``Wisdom A'' is first awakened and activated, and then ``Wisdom B'' is awakened and activated. A collaborative awakening is taking place.
The spread of knowledge activation by coordinated awakening makes it possible to obtain the output values of many other knowledges related to the activated knowledge from the input value (intelligence wave) to one knowledge.
The data value (output value) transmitted from the intellect under the condition of the arousal state of the information cells (input value) of this intellect activation action is determined by the contents defined in the intelligence body's [will attribute]. Dependent.

4. Functions of Conceptual Base and Intentional Attribute Using a model of knowledge, we will explain the "conceptual base" and "intentional attribute" that make up the intelligent body of AI cells. FIG. 4 shows a model of knowledge (description of conceptual bases and intentional attributes).

4.1 Conceptual basis
The intelligence of AI cells has an information definition body called "concept base", which defines the conceptual structure of knowledge.
The conceptual basis of knowledge in FIG. 4 is composed of the concept of {representative [flower] [season] [color] [country]}.
A concept is a high-level expression representing the Gaien relationship of things, not a specific naming element. A conceptual group defines an inclusive expression such as a property or a component of an object.
Concepts have ``representative concepts'' that represent conceptual groups. In FIG. 4, [flower] is a representative concept, and means "the concept of [flower] is composed of the concepts of [season], [color], and [country]".
The conceptual information of the conceptual base is modified by life management cells (see below) through new knowledge learning.

4.2 Intention Attributes
AI cells have an information definition body called "will attribute", which defines the action instructions of information cells.
The intention attribute in FIG. 4 defines the following two action instructions.
(1) “When the concept values ``flower'', ``color'' and ``season'' are awakened, activate ``country'' and ``flower''.
(2) "When the concept value 'flower' and [season] awaken, learn and act on [color]"
Intention attributes determine the action of knowledge based on the arousal state of information cells.
(1) is an instruction to activate the information cells of [Country] and [Flower] by awakening the information cells of [Flower], [Color] and [Season], and (2) is an instruction to activate the information cells of [Flower] and [Season]. It is an instruction to send out the desire to learn of the information cells of [color] by awakening the information cells of [color].

The definition information of the intention attribute is not modified by intelligence action, but is modified by life management cells (described later) upon receiving an instruction from the outside (operator). Using the knowledge model shown in Fig. 4, the knowledge activity will be explained below, centering on the functions of the conceptual base and intention attributes.

4.3 Intellectual activity without learning activity (action of activator)
When the "activation effect" of the will attribute is evaluated by the awakened state of knowledge (information cells), the autonomous body (AI cell) [activation factor] is activated and the knowledge is activated.
[1] As input information, "What are the pink flowers in spring?" is input, the life control cells (described later) extract "spring", "pink" and "flower" as input information through morphological analysis and spread as intelligence waves.
[2] The information cells {information cells (pink)} {information cells (spring)} {information cells [flower]} that responded to the diffused intelligence waves "spring", "pink", and "flower" awakened and AI cells to notify. At this time, {information cells (pink)} and {information cells (spring)} are responsive to data values, and {information cells [flower]} are responsive to concept values.
[3] According to the definition of [will attribute] (1), AI cells activate the autonomic entity [activation factor] of the activation action, and activate the {information cell (Japan)} and {information cell (Sakura)}. to direct. {Information Cells (Japan)} and {Information Cells (Sakura)} receive activation instructions from AI cells and transmit "data values" as intelligence waves into Cyber Vessel.
[4] Output to the outside is performed by life control cells (see below).

4.4 Intellectual activity with learning action (action of learning factors)
When the "learning action" of the will attribute is evaluated by the awakened state of knowledge (information cells), the autonomous body (AI cell) [learning factor] is activated and the learning request of knowledge is performed.
[1] As input information, "What about spring flowers?" ] is input, the life control cells (described later) extract "spring" and "flower" as input information through morphological analysis, and diffuse them into the Cyber Vessel as intelligence waves.
[2] The information cells {information cells (spring)} {information cells [flowers]} that respond to the diffused intelligence waves "Spring" and "Flower" awaken and notify the AI cells. At this time, the {information cell (spring)} responds to the data value, and the {information cell [flower]} responds to the concept value.
[3] According to the definition of [will attribute] (2), AI cells activate the autonomy of learning action [learning factor] and instruct learning action to {information cells [color]}.
[4] {Information Cells [Color]} receive learning instructions from AI cells, and transmit "motivation to learn" as intelligence waves within Cyber Vessel.
[5] "Motivation to learn" asks the operator for the "data value" of [color] by the life control cell (described later).
[6] "Pink" information is input by the operator, and the intelligence wave "Pink" spreads within the Cyber Vessel.
[7] The intelligence wave "pink" awakens {information cells (pink)}, and the action of the previous section (4.3) [3] continues.
In the first awakening of knowledge (information cells), the "learning action" is activated, and the learning action that encourages education from the outside (operator) is performed. ” has been activated.

5. Knowledge management (genetic life management)
Knowledge changes in value and presence depending on the state of data values and instructions from an external operator. This change action is called "knowledge management action", and it may act on a cell basis, act on definition content, or act upon external input or output. Genetic life management cells (GLIA cells) are responsible for this knowledge management function.

5.1 Genetic life imperative administrator cells GLIA cells
GLIA cells are responsible for all knowledge, cells, and definitions within the cybervessel, and cells differ according to their content of action.

Figures 5A-5F are phantoms of the GLIA cell model. Six types of GLIA cells and autonomically defined agents are described.
[1] GLIA-A cells (AI cell life control factor: see Figure 5A)
Manage the life of AI cells.
[2] GLIA-D cells (information cell life control factor: see Figure 5B)
Manage the life of information cells.
[3] GLIA-C cells (conceptual life control factor: see Figure 5C))
Manages the life of conceptual groups.
[4] GLIA-I cells (will attribute life control factor: see Figure 5D)
Manages will-attribute lives.
[5] GLIA-Li cells (input intelligence life control factor: see Figure 5E)
Manage the life of input intelligence.
[6] GLIA-Lo cells (output intelligent life control factor: see Fig. 5F)
Manage the life of the output intelligence.

5.2 Functions of GLIA cells Specific functions of GLIA cells are explained below.
[1] GLIA-A cells (AI cell life control factor: see Figure 5A)
・[Generation factor] AI cells are created when new knowledge is registered.
・[Replication factor] If the concept is the same, the concept is duplicated from the original AI cell.
・[Extinguishment factor] When knowledge is erased, AI cells are extinguished.
[2] GLIA-D cells (information cell life control factor: see Figure 5B)
[Generation Factor] An information cell is created when new knowledge is registered.
・[Extinguishment factor] When the information of knowledge is erased, the information cell is extinguished or non-awakened.
[3] GLIA-C cells (conceptual life control factor: see Figure 5C)
• [Generation factor] Registered in the concept base when a new concept is registered.
・[Modifier] Modify the conceptual basis of AI cells when the conceptual structure is modified.
・[Disappearance factor] If the conceptual base of knowledge is erased, it will be extinguished.
[4] GLIA-I cells (will attribute life control factor: see Figure 5D)
・[Generation factor] Registered in the intention attribute when a new intention is registered.
・[Modifier] Modify the will attribute of AI cells when the will is modified.
・[Disappearance factor] If the will is erased, the will attribute will be extinguished.
[5] GLIA-Li cells (input intelligence life control factor: see Figure 5E)
・[Input Factor] Converts input information into intelligence waves and spreads them within Bessel.
・[Answer factor] Converts answer information by learning action into an intelligence wave and spreads it within Bessel.
[6] GLIA-Lo cells (output intelligent life control factor: see Fig. 5F)
・[Output factor] Converts the intelligence wave of the data value activated by intelligence to the outside and outputs it.
・[Question factor] Intelligence waves that ask questions through learning action are converted to the outside and output.
・[Response factor] Intelligence waves that respond by learning action are converted to the outside and output.

§2. Subjective learning6. Principle of subjective learning The subjective learning function in the active intentional reasoning model is constructed based on the subjective learning model of educational theory.

6.1 Self-directed learning In education theory, self-directed learning (self-directed learning) is defined as “having an interest in learning, relating it to the direction of one’s career development, looking back on one’s own learning activities with an outlook, and connecting to the next.” Learning”, and is defined as “There are 'task-dependent type' and 'self-regulating type' in subjective learning.” [1]
``Task-dependent'' subjective learning is ``taking on a task with interest,'' while ``Self-regulating'' subjective learning is ``directing oneself by using learning goals, learning methods, and metacognition. It continues to be that it is to adjust it by attaching it.

6.2 Knowledge and Intelligence Learning Intelligence is the state of knowledge expressed by AI cells and information cells possessed by knowledge. When intelligence demands learning action, each organization of knowledge activates its own intentional action to carry out learning action.

6.3 Subjective learning by intelligence Based on the definition of the above paper, subjective learning by intelligence of wisdom is defined as ``intelligence has its own purpose according to external stimuli and the intelligence state of the surroundings, and learns the task while being aware of the direction and outlook. It can be defined as “learning activities toward a solution”. This learning action is not a conventional learning action by instruction of education from the outside by an operator, but an action in which intelligence sends a learning request to the outside according to its own purpose, prompting education and learning intelligence. .
Intelligence's "self-purpose" is the purpose of intelligence itself to judge its own state of intelligence and achieve a stable intellectual state equilibrium (career formation), and to achieve self-balanced perfect intelligence. The goal is to perform the necessary information acquisition activities (learning activities) on their own as complementary actions to lead to the next intelligence evaluation.

6.4 Subjective learning action If we explain the subjective learning action of knowledge in line with the "task-dependent type" and "self-adjusting type" of education theory, the subjective learning action of intelligence also occurs when evaluating external questions. There are two types of learning: “learning by missing tasks (task-dependent type)”, and “learning by self-demand for unvalued values (self-adjusting type)” when the concept of intelligence changes.
First, "learning by insufficient tasks (task-dependent type)" is learning in which the intelligence itself requests learning of the missing tasks to the outside when the sufficient conditions for passing due to insufficient tasks are not satisfied when the intelligence evaluates the tasks. It is action.
Next, "learning by self-request for unvalued (self-adjusting type)" means that when the completeness of intelligence is not guaranteed by the conceptual structure, the intelligence itself makes a learning request to the outside for information on the complete formation of intelligence. It is a learning effect.
The subjective learning of intelligence will be explained in detail below.

7. Mechanism of subjective learning
7.1 Motivation to learn This section explains the motivation to learn that occurs in subjective learning of intelligence in the active intention reasoning model. FIG. 6A shows a developmental model of task-dependent subjective learning, and FIG. 6B shows a developmental model of self-adjusting subjective learning.

"Motivation to learn" is learning request information generated within intelligence when intelligence (AI cells) activates learning action (autonomous body [learning factor]). Intelligence learning requirements may arise from the evaluation results of AI cells or from information cells to which AI cells are subordinate.
For example, in the case of the "task-dependent type" in Fig. 6A, it is generated from the "learn and act on [color]" of the intelligence [intentional attribute] of the AI cell, and in the case of the "self-adjusting type" in Fig. 6B, the information cell arises from the learning action of the {info cell [color]} where there is no "data value" of .
As specific information about the willingness to learn, the concept value (which may include the modifier value) requested to learn and the learning instruction are handed over together.

7.2 Learning Factors [Learning Factors] are execution modules activated by AI cells, and are defined as "autonomous bodies" of AI cells.
[Learning factor] is activated when AI cells perceive the desire to learn, and transmits the information required for learning (motivation to learn) as an intelligence wave within Cyber Vessel through the information cells. The intelligence wave transmitted to the Cyber Vessel outputs learning motivation (concepts and instructions) to the outside (operator) through the GLIA cells (Lo).

7.3 Occurrence of subjective learning action As explained above, there are two types of subjective learning action of intelligence: task-dependent type and self-adjusting type.

(1) Task-dependent type FIG. 7A shows a task-dependent knowledge model, and FIG. 7B shows a task-dependent flow.
Task-dependent learning occurs from the intelligence [intentional attributes] of AI cells.
[Intention attribute] defines the learning action (generation of motivation for learning of information cells) by the awakening state of information cells so as to promote learning action. By awakening information cells, intelligence evaluation of AI cells' [will attribute] is performed and learning motivation is generated. Based on the learning action, AI cells activate "autonomous bodies" [learning factors] and instruct information cells to transmit their willingness to learn.

(2) Self-regulating FIG. 8A shows a self-regulating knowledge model, and FIG. 8B shows a self-regulating flow.
The self-regulating learning action is generated from the "learning motivation action" of information cells.
The structure of the information cell is a pair structure of "data value" and "concept value". When a state of only "concept value" occurs due to the transformation of information cells, information cells seek "data values", so they notify AI cells of their willingness to learn through the "learning motivation effect", which is an incidental action. AI cells receive notifications of learning motivation from information cells, activate "autonomous bodies" [learning factors], and instruct information cells to transmit learning motivation.

8. Teaching action and learning action
8.1 Structure of subjective learning Intelligent subjective learning consists of the interrelationship between ``education from the outside (operator)'' and ``learning action of artificial intelligence''.
(1) Task-dependent learning relationship Fig. 9 shows a task-dependent learning relationship.
In a task-dependent learning relationship, intelligence receives "evaluation instructions" from the outside (operator), but if the conditions for intelligence evaluation are insufficient, intelligence makes a "learning request" to the outside (operator) at its own discretion. to send. The intelligence receives an "answer" from the outside (operator) and returns an "evaluation proposal" of the final goal.
In this case, the action of intelligence ``receiving an answer from a learning request'' is called a ``learning action'', and the action of an external (operator) ``inputting an answer to a learning request'' is called an ``educational action''.
(2) Self-Regulating Learning Relationships FIG. 10 shows self-regulating learning relationships.
In the self-adjusting learning relationship, the intelligence receives "information registration" from the outside (operator), but if an incomplete state occurs in the information of the intelligence (information cell), the intelligence will make its own decision to register the information from the outside (operator). Send a "learning request" to Intelligence forms a complete intelligence state by receiving "answers" from the outside (operator).
In this case, the action of intelligence ``receiving an answer from a learning request'' is called a ``learning action'', and the action of an external (operator) ``inputting an answer to a learning request'' is called an ``educational action''.

8.2 Functions of educational action and learning action Intelligent subjective learning cannot be established only by the learning action of intelligence. It is realized by cooperation with an external (operator) educational action that assists the learning action of intelligence.
Different transition states are expressed depending on how to hold the main focus for both educational action and learning action in subjective learning. The educational action of intelligence differs according to the "primary focus of educational action", and the external (operator) learning action differs depending on the "primary focus of learning action".

8.3 Main focus of educational action There are two educational states in the educational action performed by the outside (operator). One is the action of the outside (operator) giving answer information about the willingness to learn of artificial intelligence, and the main point of education is the content of learning of intelligence. On the other hand, the outside (operator) gives new information, and the main purpose of education is to improve knowledge from the outside.
In the "task-dependent learning relationship" in FIG. 9, when the intelligence makes a learning request based on the willingness to learn, the outside (operator) gives the intelligence a reply to the learning request. Also, in the "self-regulating learning relationship" in FIG. 10, the outside (operator) gives knowledge information to the intelligence in order to improve the intelligence.

8.4 Focus of learning action There are two learning states in the learning action performed by intelligence. One is the action of transmitting the willingness to learn to the outside (operator), and the main point of learning is the transmission of the willingness to learn from intelligence. The other is the action of registering answers from the outside (operators) as knowledge, and the main point of learning is to understand the answers from the outside.
In FIGS. 9 and 10, the arrow from knowledge (left side) to the operator (right side) is the action of "transmission of willingness to learn", and the arrow from the operator to knowledge is the action of "understanding of external answers".

8.5 Transmission of motivation to learn by concept value Regarding “transmission of motivation to learn”, as explained in the previous section, the main focus is transmission of motivation to learn by information cells, but on the other hand, “transmission of confirmation for understanding external instructions” exists. When the object of awakening of information cells is not a "data value" but a "conceptual value", this effect is to transmit the willingness to learn about other concepts with [will attribute] in order to increase the confidence of awakening of the knowledge. It is action.
For example, when "flowers (conceptual value)" of "spring (data value)" are awakened, it is possible to transmit a willingness to learn about "color (conceptual value)."

authority [1]
What is proactive learning?-From theory to ``proactive, interactive and deep learning''- (2018)
Shinichi Mizokami (Doctor of Education, Kyoto University)

Next, an example of an artificial brain system (brain-type artificial intelligence system) constructed based on the active intention inference artificial intelligence model described above will be described.

An application system using this artificial brain system is shown in FIG. This application system is, for example, the "Japanese Flower Encyclopedia System" that performs actions based on knowledge about flowers.

In FIG. 11, an application system 100 includes an artificial brain system 110 and an interface device 120 (PC, smart phone, etc.) that interfaces (UI) between the artificial brain system 110 and a user (person). The interface device 120 and the artificial brain system 110 are connected directly via a wire, directly via a wireless connection, via a wired and wireless connection, or via a predetermined network (such as the Internet) so that information can be transmitted and received. . The interface device 120 has an input sentence analysis function that performs sentence analysis of text information input by the user and extracts words from the text information, and an output sentence that generates sentences from output information from the artificial brain system 110. It has a generation function. The interface device 120 displays the input information (text information) and the output information (text information) from the artificial brain system 110 in response to the input information in a format similar to the output format of the chatbot, for example. , audio output, etc.

The artificial brain system 110 has an information processing device 111 and an information storage device 112 . The information processing device 111 has the aforementioned functions of intelligent reasoning and subjective learning. The information processing device 111 includes, for example, hardware resources such as von Neumann type or non-von Neumann type computers (including CPUs), GPUs (General Purpose Graphic Processors), DSPs (Digital Signal Processors), and FPCAs (Field Programmable Gate Arrays). or software included in them. The information storage device 112 stores various kinds of knowledge information (knowledge DB), a knowledge management table used for managing the knowledge information, and other various kinds of information necessary for the processing of the information processing device 111, which will be described later.

The artificial brain system 110 is specifically configured as shown in FIG. 12 by the functions of the information processing device 111 and the information storage device 112 .

12, an artificial brain system 110 includes an information vessel 10 (corresponding to a "cybervessel" (see FIG. 1)) and a plurality of knowledge units (knowledge devices) 20(1), 20(2), 20(3). ), . . . 20(n) (corresponding to "knowledge" (see FIG. 1)). Each of the plurality of intelligent units 20(1)-20(n) is coupled to the information vessel 10 so as to be able to transmit and receive information. The information vessel 10 holds the provided information as valid information in an outputtable state. The information vessel 10 is connected to an interface device 120 (external device) via an information input section 11 and an information output section 12 . Information from the interface device 120 is provided to the information vessel 10 through the information input section 11 , and information effectively held in the information vessel 10 is output toward the interface device 120 through the information output section 12 . The information vessel 10 is also connected to an interface device 120 (external device) via the training input section 13 and the learning output section 14 . When learning request information representing a learning request for a concept that constitutes knowledge from any of the plurality of knowledge units 20(1)-20(n) is provided to the information vessel 10, the learning request information is: It is output toward the interface device 120 through the learning output unit 14 . Educational information can also be provided to the information vessel 10 from the interface device 120 through the education input unit 13 . The knowledge unit 20(i) that provided the learning requirement information to the information vessel 10 as described above can incorporate said educational information provided to the information vessel 10 . The information vessel 10 can be configured using, for example, a database, and information registered in this database (information vessel 10) is held as valid information.

Each of the plurality of knowledge units 20(1) to 20(n) is configured as follows.

A knowledge unit 20(i) (i=1 to n) corresponding to one knowledge about Japanese flowers comprises a processing section 21, a knowledge storage section 22 and a communication section 23. FIG. The processing unit 21 is configured as one function of the information processing device 111 (see FIG. 11), and executes processing related to activation/learning of concepts based on intention attributes as will be described later. The knowledge storage unit 22 stores knowledge information including data values representing each of a plurality of concepts forming one knowledge (information cells). Further, the knowledge storage unit 22 stores a knowledge management table (AI cells/intelligent bodies) that associates each concept that constitutes the knowledge with the data value thereof, and stores the knowledge management table (AI cells/intelligent bodies) necessary for processing in the information processing device 111 . Also used to store information.

For example, one piece of knowledge, ``Cherry blossoms are pink flowers in the spring season,'' is composed of three concepts [season], [color], and [flower]. , the data value "spring" for the concept [season], the data value "pink" for the concept [color], and the data value "cherry blossom" for the concept [flower]. For example, as shown in FIG. 13A, this knowledge information associates the concept [flower] with the data value "cherry blossom", associates the concept [season] with the data value "spring", and associates the concept [color ] and the data value [pink] are managed by a knowledge management table (knowledge management unit). Also, one piece of knowledge is managed as knowledge about a representative concept. The above-mentioned knowledge that "cherry blossoms are pink flowers in spring" is knowledge about the representative concept [flower]. managed by

For example, as shown in FIG. 13B, the knowledge information representing the knowledge about the representative concept "flower" that "a flower whose color is purple in spring is wisteria" is represented by the concept [flower] and the data value "wisteria" as shown in FIG. 13B. ”, the concept [season] and the data value “spring”, and the concept [color] and the data value [purple]. can. Further, for example, the knowledge information representing the knowledge about the representative concept "flower" that "the pink flower in the spring season is lotus flower" is represented by the concept [flower] and the data value, as shown in FIG. 13C. It can be managed by a knowledge management table (knowledge management unit) that associates "Rengeso", associates the concept [season] with "spring", and associates the concept [color] with the data value [pink]. can. Furthermore, the knowledge information representing the knowledge of the representative concept "event" that "the event in which the season is spring and the flowers are cherry blossoms is the entrance ceremony" is the concept [event] and the data value "entrance ceremony" as shown in FIG. 13D. expression", the concept [season] with the data value "spring", and the concept [flower] with the data value [cherry blossom]. can be done.

Returning to FIG. 12, under the control of the processing unit 21, the communication unit 23 inputs the information effectively held in the information vessel 10 (receiving function by the receptor (receptor)), and the information vessel 10 Output information to (transmitting function by donor machine (donor)).

Next, the operation of the application system 100 (Japanese Flower Encyclopedia System) including the artificial brain system 110 described above will be described.

When the user inputs information related to a question (for example, "What are the pink spring flowers?") to the interface device 120, the interface device 120 displays the question, outputs the question as a voice, and also displays the question. , and extract the wording used in the question. For example, the words "pink", "spring" and "flower" are extracted for the question. The interface device 120 then outputs information (text information) representing the extracted phrases (for example, the phrases “pink” and “spring” related to “flower”) to the artificial brain system 110 . Information (for example, “pink” and “spring”) output from the interface device 120 is provided to the information vessel 10 through the information input unit 11 in the artificial brain system 110, and is held (registered) in the information vessel 10 as valid information. ) is done.

In each knowledge unit 20(i) of the artificial brain system 110, the processing section 21 executes processing according to the procedure shown in FIG. The processing will be explained.

The processing unit 21 repeatedly accesses the information vessel 10 (database) through the communication unit 23 at predetermined intervals. When the information is effectively held (registered) in the information vessel 10 as described above, the processing unit 21 takes in the information held in the information vessel 10 via the communication unit 23 (receptor), and the information is intelligently stored. It is determined whether or not it matches any concept data value managed in the identification management table (see FIGS. 13A to 13D) (S11). If the information held in the information vessel 10 matches the data value of any of the concepts managed in the knowledge management table (YES in S11: YES), the processing unit 21 detects the information held in the information vessel 10 (S12), and it is determined whether other concepts should be activated due to the awakening of the concept (S13). Each knowledge unit 20(i) has an intentional attribute (intelligent body) that determines the concept to be activated when a certain concept is awakened (for example, the intentional attribute "concept [season] and concept [color] When Awakens, activate the concept [Flower]”). Even if this intention attribute is defined in the program executed by the processing unit 21, or defined in the form of a decision table in the information storage device 112, , may be defined in other formats.

If the information held in the information vessel 10 does not match any concept value managed by the knowledge management table (NO in S11), the processing unit 21 performs learning (self-adjusting learning) described later. In order to determine whether or not to execute such processing, the information held in the information vessel 10 is definition information representing a new concept in knowledge about a representative concept (for example, [flower]) managed in a knowledge management table. (S18).

Here, if it is determined that a certain concept contained in the knowledge information should be activated according to the activation condition (for example, if the concept [season] and the concept [color] are awakened) (YES in S13), the process The unit 21 executes activation processing (S14: activation information determination means: AI cell/autonomous body/activator). This activation process is specifically performed according to the procedure shown in FIG. In this activation process, a concept to be activated corresponding to the awakened concept in one knowledge is determined based on the intention attribute (S141). Then, the processing unit 21 refers to the knowledge management table, reads the data value corresponding to the concept to be activated from the knowledge storage unit 22, and transmits the data value via the communication unit 23 (providing device) to It is provided to the information vessel 10 (S142: information transmitting means). Returning to FIG. 14, after that, the processing unit 21 performs the above-described processing while confirming that a predetermined stop condition (for example, disappearance of knowledge, stop of knowledge function, etc.) is not satisfied (NO in S17). Execute repeatedly. In the process, if the above-described stop conditions such as knowledge disappearance or function stoppage are satisfied (YES in S17), the above-described processing is terminated (knowledge function stoppage, knowledge disappearance).

The data values of the activated concepts provided by the knowledge units 20(i) and operatively held in the information vessel 10 as described above are transmitted from the information vessel 10 through the information output 12 to the interface device 120. output. When the data value is input, the interface device 120 generates a sentence representing the answer to the question using the data value, and outputs the sentence representing the answer by display, voice, or the like. This allows the user to obtain an answer from the artificial brain system 110 to the question.

The operation of the artificial brain system 110 described above will be described more specifically with reference to FIG.

As described above, when the user asks the interface device 120, "What are the pink spring flowers?" ], the interface device 120 displays the question, outputs the voice, etc., analyzes the text of the question, and identifies the words "pink", "spring", and " Extract the word "flower". Then, the interface device 120 outputs information (text information) representing the words “pink” and “spring” related to the extracted “flower” to the artificial brain system 110 . The words “pink” and “spring” output from the interface device 120 are provided to the information vessel 10 through the information input unit 11 in the artificial brain system 110 and held (registered) in the information vessel 10 as valid information.

Knowledge information including the data value "cherry blossom" for the concept [flower], the data value "spring" for the concept [season], and the data value "pink" for the concept [color] is managed in the knowledge management table shown in FIG. 13A. On the other hand, the knowledge unit 20(i) having the intention attribute "Activate the concept [flower] when the concept [season] and the concept [color] are awakened" performs the following processing.

As described above, when the information of “spring” and “pink” is held in the information vessel 10, the knowledge information (see FIG. 13A) stores the concept [season] of the data value “spring” and the data value “pink”. ” is confirmed as an awakened concept (S11, S12 in FIG. 14). Then, the concept [flower] is determined as the concept to be activated based on the intention attribute "When the concept [season] and the concept [color] awaken, activate the concept [flower]" (S13, S14 in FIG. 14). (S141 in FIG. 15)), the data value "cherry blossom" of the activated concept [flower] in the knowledge information is provided from the knowledge unit 20(i) to the information vessel 10 (S142 in FIG. 15), That information vessel 10 is effectively held.

At this time, the knowledge management table shown in FIG. , and the same procedure ( The processing is performed in S11 to S14) of FIG. As a result, when the information "spring" and "pink" are effectively held in the information vessel 10, the concept [flower] is activated corresponding to the awakened concept [season] and [color]. The data value "lotus flower" of [flower] is provided to the information vessel 10 and is effectively held in the information vessel 10 .

As described above, the data value "cherry blossom" from knowledge unit 20(i) and the data value "lotus flower" from knowledge unit 20(j), which are effectively held in information vessel 10, are output by the information output section. 12 to the interface device 120 . When the data values "sakura" and "rengeso" are input, the interface device 120 generates sentences representing answers to the questions using the data values, for example, "It's cherry blossoms" and "It's lotus flowers." The sentences "It's cherry blossoms" and "It's lotus flowers" are displayed and output by voice or the like. This allows the user to ask, 'What are the pink spring flowers? can be obtained from the artificial brain system 110 as an answer to the question "Sakura" and "Rengegusa".

By the way, the knowledge management table shown in FIG. In yet another knowledge unit 20(k), which has the intention attribute "When the concept [season] and the concept [flower] are awakened, activate the concept [event]", a similar procedure is performed. (See FIG. 14).

The operation (coordinated arousal) of the artificial brain system 110 in which the processing in the knowledge unit 20(k) is added to the processing in the knowledge unit 20(i) described above will be described with reference to FIG.

As described above, the message from the interface device 120, "What are those pink spring flowers?" is effectively held in the information vessel 10, the knowledge unit 20(k) corresponding to the knowledge that does not have the concept of [color] is It does not respond to the "pink" information (NO in S11 of FIG. 14). In that knowledge unit 20(k), since the information held in the information vessel 10 does not contain the data value of the concept "flower" (insensitive), the activation condition (concept [season] and If the concept [flower] awakens) is not satisfied (NO in S13 of FIG. 14), the activation process based on the intention attribute (S14 of FIG. 14) is not performed.

In such a situation, as described above, an intelligence that responds to both "spring" and "pink" effectively held in the information vessel 10 and confirms the concept [season] and the concept [color] as arousal concepts. The awareness unit 20(i) provides the information vessel 10 with the data value "cherry blossom" of the concept [flower] activated corresponding to the awakening concept, and the data value "cherry blossom" is valid for the information vessel 10. is held to Then, the knowledge unit 20(k) corresponding to the knowledge that does not have the concept of [color] but has the concept of [flower], along with "spring" effectively held in the information vessel 10, the knowledge unit 20(k) i) (YES in S11 of FIG. 14), the concept [season] and the concept [flower] are confirmed as awakening concepts (S12 in FIG. 14). Here, since the activation condition (when the concept [season] and the concept [flower] are awakened) is satisfied (YES in S13 of FIG. 14), activation processing (S14 of FIG. 14, S141 and S142 of FIG. 15) ) is executed, and based on the intention attributes (“Activate the concept [event] when the concept [season] and the concept “flower” are awakened”), the concept [event] ] is activated to provide the corresponding data value "Entrance Ceremony" to the information vessel 10 where it is held in effect.

As described above, the data value "Sakura" from knowledge unit 20(i) and the data value "Entrance ceremony" from knowledge unit 20(k), which are effectively held in information vessel 10, are the information output It is output to the interface device 120 through the unit 12 . When the data values "Sakura" and "Entrance Ceremony" are input, the interface device 120 generates sentences representing answers to the questions using these data values, such as "It's Sakura" and "There is an entrance ceremony", The sentences "Sakura" and "There is an entrance ceremony" representing the answer are displayed and output by voice or the like. This allows the user to ask, 'What are the pink spring flowers? As an answer to the question "Sakura", the artificial brain system 110 can obtain information about an event called "entrance ceremony" related to spring flowers and cherry blossoms.

By the way, from the information of the wording (for example, "spring") extracted from the question (for example, "What are the spring flowers?") provided to the interface device 120 by the user and effectively held in the information vessel 10, , some knowledge units 20(i) (for example, knowledge units with the intention attribute "When concept [season] and concept [color] are awakened, activate concept [flower]"), the awakened concept Due to the lack of (e.g. concept [color]), it is not possible to determine which concept to activate based on the intent attribute. In the processing section 21 of such a knowledge unit 20(i), the information effectively held in the information vessel 10 does not satisfy the condition for activating the concept in the processing according to the procedure shown in FIG. NO), and further determines whether or not the learning conditions are satisfied (S15). Then, if the learning conditions are satisfied (YES in S15), the processing unit 21 executes processing (task-dependent learning processing described later) related to learning (subjective learning).

Learning (subjective learning) includes task-dependent learning and self-adjusting learning. Task-dependent learning is learning of deficient tasks. Specifically, in this system, task-dependent learning refers to learning about a lacking concept (missing task) when there is a lack of awakened concepts to activate one concept in knowledge information. is. In addition, self-regulated learning is learning about the changed (eg, increased) concept that is performed by self-requirement when the concept that constitutes one knowledge is changed (eg, increased). Specifically, in this system, self-adjusting learning is learning about a concept when knowledge information lacks a data value (when the number of concepts constituting knowledge increases).

Each knowledge unit 20(i) has a learning intent for instructing learning of the deficient concept when the deficient concept to be awakened makes it impossible to determine the concept to be activated based on the intent attribute. Attributes (intelligents) are defined. For example, in the knowledge unit 20(i) in which the above-mentioned intention attribute "Activate the concept [flower] when the concept [season] and the concept [color] awaken" is defined, "When the concept [season] awakens, If the concept [color] is not awakened, learn the concept [color]" is defined as the learning intention attribute. Even if this learning intention attribute is defined in the program executed by the processing unit 21, it is also defined in the form of a decision table in the information storage device 112 in the same way as the intention attribute described above. It may be one that is specified in another format, or it may be one that is specified in another format.

In such a knowledge unit 20(i), the information effectively held in the information vessel 10 (e.g., the data value "spring") indicates that the lack of an awakening concept (e.g., the concept [color]) causes activation. I can't decide which concept to transform. Then, when the task-dependent learning condition (for example, the condition that the concept [season] is awakened but the concept [color] is not awakened) is satisfied (YES in S15), the processing unit 21 sets the learning intention attribute to Based on this, task-dependent learning processing is executed (S16). This task-dependent learning process (S16) is executed according to the procedure shown in FIG.

In FIG. 18, the processing unit 21 learns the missing concept (for example, the concept [color]) is confirmed as a concept to be learned (S161), and learning request information representing a learning request for the concept is provided to the information vessel 10 via the communication unit 23 (donor) (S162: 1 learning request means: AI cell/autonomous body). After that, the processing unit 21 waits for the educational information to repeatedly confirm whether or not the educational information provided from the interface device 120 is effectively held in the information vessel 10 (S163).

On the other hand, the learning request information (e.g., representing a learning request for the concept “color”) provided to the information vessel 10 from the knowledge unit 20(i) is output through the learning request output unit 14 toward the interface device 120. be. When the learning request information is input, the interface device 120 generates a sentence (for example, "What is the color of the flower?") relating to the information request based on the learning request information, and displays the sentence relating to the information request. , output by voice, etc. In response to this information request, when the user inputs information responding to the information request (for example, "the color is pink") to the interface device 120, the interface device 120 outputs the information by display, voice, etc. , extracts words used in the information (eg, “color” and “pink”), and outputs the extracted words (eg, the word “pink” related to “color”) to the artificial brain system 110 . The information (for example, "pink") output to this artificial brain system 110 is provided to the information vessel section 10 through the educational information input section 13 in the artificial brain system 110, and the information vessel section 10 is provided with the wording (for example, , “pink”) is effectively held (registered) as educational information.

When the information as the educational information is effectively held in the information vessel 10 (YES in S163), the processing unit 21, which has been in the waiting state for the educational information as described above, Information (for example, "pink") is taken from the information vessel 10 as the educational information. Then, the processing unit 21 refers to the knowledge management table and determines whether or not the educational information matches the data value of the missing concept (for example, the data value "pink" of the missing concept [color]). (S164). When the information as the educational information matches the data value of the missing concept (YES in S164), the processing unit 21 activates the activation process (S14), and uses the missing concept as the awakening concept to activate the activation process (S14). ) (S165: learning control means). As a result, the processing unit 21 determines the concept to be activated based on the intention attribute in a state in which the shortage of awakened concepts is resolved, and transmits the data value of the activated concept (FIG. 14). , S141 and S142 in FIG. 15).

On the other hand, if the information as the educational information does not match the data value of the missing concept (for example, the data value "pink" of the missing concept [color]) (NO in S164), the processing unit 21 performs activation processing (S14 ) without activating the learning (task-dependent learning). In this case, this knowledge unit 20(i) is not activated and does not return information to the user regarding the answer to the user's question.

The operation of task-dependent learning in the artificial brain system 110 described above will now be described more specifically with reference to FIG.

When the user asks the interface device 120 the question ``What are the spring flowers? , the question is displayed, output by voice, etc., and information (text information) of the word "spring" related to the "flower" is provided from the interface device 120 to the artificial brain system 110. Information relating to the word “spring” provided to the artificial brain system 110 is effectively held (registered) in the information vessel 10 .

For example, the knowledge information including the data value [cherry blossom] of the concept [flower], the data value of "spring" of the concept [season], and the data value of "pink" of the concept [color] is stored by the knowledge management table shown in FIG. Management, intention attribute ``If concept [season] and concept [color] awaken, awaken concept ``flower'''', and learning intention attribute ``When concept [season] awakens and concept [color] does not awaken, concept In the knowledge unit 20(i) where "learn [color]" is defined, the following task-dependent learning is performed.

In Step 1 of FIG. 19, the knowledge unit 20(i) that has taken in the information of "spring" held (registered) in the information vessel 10 as described above converts the concept [season] of the data value "spring" into the concept of arousal. However, since the concept [color] of the data value "pink" is not awakened, the concept to be activated cannot be determined (NO in S13 of FIG. 14). Therefore, based on the learning intention attribute "learn the concept [color] when the concept [season] is awakened but the concept [color] is not awakened", the learning request information about the concept [color] is transferred to the knowledge unit 20 ( i) is transmitted to the information vessel 10 (S16 in FIG. 14, S161 to S163 in FIG. 18).

When learning request information about this concept [color] is provided from the information vessel 10 (artificial brain system 110) to the interface device 120, the interface device 120 generates the information request sentence "flower color" based on the learning request information. What is ...? ] is generated, and the text of the information request is displayed or output by voice or the like. In response to this information request, when the user inputs the answer "The color is pink" to the interface device 120, the interface device 120 outputs the answer information representing the answer by display, voice, etc., and also the concept [color]. Information (text information) of the wording “pink” is output to the artificial brain system 110 . In the artificial brain system 110, the "pink" information related to the answer from the interface device 120 is effectively held (registered) in the information vessel 10 as educational information.

Subsequently, in Step 2 of FIG. 19, when the information of the word "pink" of the concept [color] is held in the information vessel 10 as educational information, the information "pink" is taken into the knowledge unit 20(i). . Then, referring to the knowledge management table (see FIG. 13A), the information "pink" taken in as the educational information matches the data value "pink" of the missing concept [color] (in S164 of FIG. 18). YES), the missing concept [color] is awakened. While the concept [season] has already been awakened, if the lacking concept [color] is awakened in this way, the will attribute "concept [season] and [color] will activate the concept [flower]. Based on this, the concept [flower] is activated, and the data value "cherry blossom" of the concept [flower] is transmitted to the information vessel 10 (S14 in FIG. 14, S141 and S142 in FIG. 15).

When the data value “cherry blossom” of the concept [flower] is validly held in the information vessel 10 , the data value “cherry blossom” is provided to the interface device 120 . Then, the interface device 120 outputs the sentence "Sakura desu" representing the answer by display, voice, or the like. This allows the user to ask the question 'What are the spring flowers? by answering “It’s pink” (educational information) to the question “What is the color?” What is it? can be obtained from the artificial brain system 110 as an answer to "Sakura".

Note that, for example, the knowledge information including the data value [wisteria] of the concept [flower], the data value "spring" of the concept [season], and the data value "purple" of the concept [color] is managed as shown in FIG. 13B. Managed by the table, the intention attribute "When the concept [season] and the concept [color] awaken, awaken the concept "flower"", and the learning intention attribute "When the concept [season] awakens and the concept [color] does not awaken" Also in another knowledge unit 20(s) in which "learn the concept [color]" is defined, the task-dependent learning process as described above is performed. However, in the task-dependent learning process, the information "pink" as educational information for the learning request for the concept [color] does not match the data value "purple" for the missing concept [color] (S164 in FIG. 18). NO), the activation process (S14) is not started. Then, the processing related to learning (task-dependent learning) ends as it is. In this case, this knowledge unit 20(s) (artificial brain system 110) responds to the question from the user: "What are the spring flowers?" ], the learning request ``What is the color?'' is output, but information regarding the final answer is not returned to the user.

Next, a description will be given of the learning of a concept when there is a concept that does not have a data value in the knowledge information (when the number of concepts constituting knowledge increases), that is, the process of self-adjusting learning.

By the way, there is a case where the user wants to add a concept to one knowledge (for example, the knowledge about the representative concept [flower]). In this case, the user inputs into the interface device 120 information representing a concept to be added to the knowledge (eg, a concept [country] to be added to the flower knowledge). Then, the interface device 120 displays definition information representing the concept to be added to the knowledge (for example, the knowledge of the representative concept [flower] includes the concept [country]), and outputs it by voice or the like. The definition information is output to the system 110 . In the artificial brain system 110 , the definition information from the interface device 120 is effectively held (registered) in the information vessel 10 .

In the artificial brain system 110, as described above, the processing unit 21 in each knowledge unit 20(i) retrieves the definition information effectively held in the information vessel 10 in the process of processing according to the procedure shown in FIG. is definition information representing a new concept in knowledge about a representative concept (for example, [flower]) managed in its own knowledge management table (see FIGS. 13A to 13D) ( S18). Here, if the definition information held in the information vessel 10 is the definition information representing a new concept in its own knowledge (YES in S18), the processing unit 21 defines through the communication unit 23 (receiver) Information is taken in, and concepts (eg, concept [country]) included in the definition information are represented by knowledge information managed in the knowledge management table (see FIGS. 13A to 13D) (eg, , knowledge of the representative concept [flower]) is added as a concept composing (S19: concept adding means). Specifically, the concept represented by the definition information is added to the knowledge management table. As a result, a concept (for example, the concept [country]) lacks a corresponding data value in the knowledge information managed by the knowledge management table. In this way, when knowledge information lacks a data value, the processing unit 21 executes a self-adjusting learning process (S20). This self-adjusting learning process is executed according to the procedure shown in FIG.

In FIG. 20, the processing unit 21 refers to the knowledge management table and confirms a concept whose data value is missing in the knowledge information (for example, the concept [country]) as a concept to be learned (S201). Then, the processing unit 21 provides the information vessel 10 via the communication unit 23 (providing device) learning request information representing a learning request for the concept (for example, [country]) confirmed as the concept to be learned. (S202: second learning request means: AI cell/autonomous body/learning factor). The learning request information indicates a request for a data value of a concept (eg, [country]) added to knowledge (eg, knowledge about the representative concept [flower] of the data value "cherry blossom"). After that, the processing unit 21 waits for the educational information to repeatedly confirm whether or not the educational information is effectively held in the information vessel 10 in response to the provision of the learning request information (S203).

On the other hand, the learning request information provided from the knowledge unit 20(i) to the information vessel 10 is output to the interface device 120 through the learning request output unit 14. FIG. When the learning request information is input, the interface device 120 generates a sentence related to the information request (for example, "What is Sakura no Kuni?") based on the learning request information, displays the sentence related to the information request, Output by voice, etc. In response to this information request, when the user inputs information (for example, "My country is Japan") in response to the learning request to the interface device 120, the interface device 120 outputs the information by display, voice, etc. , extracts the words used in the information (eg, “country”, “Japan”), and outputs the extracted words (eg, the word “Japan” related to “country”) to the artificial brain system 110. This information (for example, "Japan") is provided to the information vessel 10 through the educational information input unit 13 in the artificial brain system 110, and the information of the wording (for example, "Japan") is provided to the information vessel 10 as effective educational information. is held (registered) in

When the information as educational information is effectively held in the information vessel 10 (YES in S203), the processing unit 21 waiting for the educational information as described above sends the information via the communication unit 23 (receiver). Information (for example, "Japan") is taken in from the vessel 10 as the educational information. Then, the processing unit 21 causes the knowledge storage unit 22 to store the taken-in information (eg, “Japan”) as the data value of the additional concept (eg, concept [country]), and stores the additional concept (eg, [ The knowledge management table is rewritten so that the imported information (for example, "Japan") is associated with the country]) as a data value (S204: knowledge updating means). This updates the knowledge information. After that, the processing unit 21 outputs knowledge update completion information to the information vessel 10 via the communication unit 23 (donor) (S205). This completion information is provided to the interface device 120, and the interface device 120 outputs the completion information (eg, "Registered"). This allows the user to know that a new concept (eg, the concept [country] with a data value of “Japan”) has been added to the specified knowledge (eg, flower knowledge related to cherry blossoms).

The self-adjusting learning described above will be described more specifically with reference to FIG.

When the user inputs into the interface device 120 the information "flowers have a country" indicating that a new concept [country] is to be added to the knowledge of flowers, the information is displayed, output by voice, etc. , the definition information about the knowledge of flowers "the knowledge of the representative concept [flower] includes the concept [country]" is provided to the artificial brain system 110, and the definition information is effectively held (registered) in the information vessel 10. be. At this time, for example, the knowledge information shown in FIG. In the knowledge unit 20(i) managed by the management table, the following self-adjusting learning takes place.

In Step 1 of FIG. 21, the knowledge unit 20(i) that incorporates the definition information "the knowledge of the representative concept [flower] includes the concept [country]" held (registered) in the information vessel 10 as described above 22A, the concept [country] included in the definition information is added to the knowledge management table as a concept to be managed (S18, S19 in FIG. 14). This causes knowledge unit 20(i) to generate the concept [Country] with missing data values (see FIG. 22A). Then, the knowledge unit 20(i) transmits to the information vessel 10 learning request information requesting the data value of the concept [country] added to the knowledge about the representative concept [flower] of the data value "cherry blossom". (S201, S202, S203 in FIG. 20).

When a learning request about this concept [country] is provided from the information vessel 10 (artificial brain system 110) to the interface device 120, the interface device 120, based on the learning request, sends the information request sentence "What is Sakura no Kuni? mosquito? ] is generated, and the text of the information request is displayed or output by voice or the like. In response to this information request, when the user inputs the answer "Country is Japan" to the interface device 120, the interface device 120 outputs the answer information indicating the answer by display, voice, etc. Information (text information) of the word “Japan” is output to the artificial brain system 110 . In the artificial brain system 110, the information "Japan" related to the answer from the interface device 120 is effectively held (registered) in the information vessel 10 as educational information.

Subsequently, in Step 2 of FIG. 21, when the information of the word "Japan" of the concept [country] is held in the information vessel 10 as educational information, the information "Japan" as the educational information is stored in the knowledge unit 20(i). be incorporated into Then, "Japan" is added to the knowledge information stored in the knowledge storage unit 22 as the data value of the additional concept [country], and as shown in FIG. [Country] is rewritten so as to be associated with the data value "Japan" (S204 in FIG. 20).

Thereafter, a completion report is provided from the knowledge unit 20(i) via the information vessel 10 to the interface device 120, which displays the information "registration complete" indicating that the concept addition process has been completed, It is output by voice or the like. Thereby, the user can know that the concept of [country] has been added to the knowledge of flowers in the artificial brain system 110 .

According to the artificial brain system 110 as described above, each knowledge unit 20(i) responds to one or more concept data values (concept arousals) provided to the information vessel 10 based on intentional attributes. Since the transmission (activation) of the data value to the information vessel 10 is performed as an activity (knowledge activity) based on one knowledge, the action (knowledge inference function) becomes possible.

Also, in each knowledge unit 20(i), if the data value of the concept provided to the information vessel 10 alone cannot transmit (activate) the data value, the missing concept data is determined based on the learning intention attribute. A learn value request is made to obtain the missing data value through educational information provided in response to the learn request. Then, taking into consideration the data value of the missing concept, the data value is transmitted (activated) to the information vessel 10 . Therefore, according to the artificial brain system 110 described above, autonomous task-dependent learning (subjective learning) is possible according to one's own intention (learning intention attribute).

Furthermore, in each knowledge unit 20(i), when a concept with missing data value occurs in the knowledge to be managed, the situation is determined, and a learning request is made for the concept with missing data value. The data value of the learned concept is supplemented by instructional information provided in response to the request. Therefore, according to the artificial brain system 110 described above, autonomous self-adjusting learning (subjective learning) becomes possible according to one's own intention.

In the artificial brain system 110 described above, self-adjusting learning is performed when a concept lacking data values occurs due to a request to add a new concept to knowledge, but the invention is not limited to this. For example, even when a concept whose data value is missing due to some other factor, such as when the data value of the managed concept is deleted due to some trouble, self-adjusting learning can be executed.

Further, according to the artificial brain system 110 described above, addition of the knowledge unit 20(i), deletion of the knowledge unit 20(i), knowledge information of the knowledge unit 20(i), intention attribute , by modifying the learning intent attribute (eg, self-regulating learning reference), the artificial brain's knowledge can be modified (learned). Therefore, it is possible to correct (learn) the artificial brain without stopping its function as the artificial brain.

In the above-described artificial brain system 110, the information handled (knowledge information, data values, input information, output information, learning request information, educational information) was text information, but is not limited to this, voice, image, etc. Other types of information such as (including still images and moving images) may be used.

In the artificial brain system 110 described above, all the knowledge units 20(1) to 20(n) have the function of subjective learning, but the present invention is not limited to this. Even if all the knowledge units 20(1) to 20(n) do not have the function of subjective learning, only some (one or more) of the knowledge units have the function of subjective learning. There may be. Also, the subjective learning function provided in the knowledge unit may be either a task-dependent learning function or a self-dependent learning function.

The information processing device 111 and the information storage device 112 in the artificial brain system 110 may be configured as one piece of equipment housed in one housing, but they may be distributed and connected to a network (for example, the Internet). may be configured to be coupled via. More specifically, even if the information vessel 10 and the plurality of knowledge units 20(1) to 20(n) are distributed and connected via a network (eg, the Internet), , a plurality of knowledge units 20(1) to 20(n) may be distributed for each group, and the plurality of groups may be connected via a network (for example, the Internet).

Although the embodiments of the present invention have been described above, the embodiments and modifications of each part are presented as examples, and are not intended to limit the scope of the invention. These novel embodiments described above can be implemented in various other forms, and various omissions, replacements, and modifications can be made without departing from the scope of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are included in the invention described in the claims.

INDUSTRIAL APPLICABILITY The artificial brain system according to the present invention has the effect of being able to autonomously judge and act according to one's own will, and is useful as an artificial brain technology.

10 information vessel 11 information input unit 12 information output unit 13 education information input unit 14 learning request output unit 20(1) to 20(n) knowledge unit 21 processing unit 22 knowledge database 23 communication unit 100 application system 110 artificial brain system 111 information processing device 112 information storage device 120 interface device

Claims (11)

an information vessel that holds provided information as valid information in a state that can be output;
one or more knowledge units communicatively coupled to the information vessel;
each of the one or more knowledge units,
a knowledge storage unit storing knowledge information including data values representing each of a plurality of concepts constituting one knowledge;
a knowledge management unit that manages the knowledge information stored in the knowledge storage unit so as to associate the concept with the data value;
Based on intention attributes that determine concepts to be activated corresponding to concepts awakened in the knowledge information managed by the knowledge management unit, the information effectively held in the information vessel is stored in the knowledge information. Determining a concept to be activated in the knowledge information managed by the knowledge management unit assuming that the concept has been awakened when it matches the data value of the concept in the knowledge information managed by the management unit activation information determining means for
an information transmitting means for providing the information vessel with data values of concepts determined by the activation information determining means among the knowledge information stored in the knowledge storage unit. 2. The artificial brain system according to claim 1, wherein said information vessel includes a database in which provided information is registered as valid information. an information input unit for providing the information vessel with information as a data value of a concept constituting a certain knowledge from an external device;
3. The artificial brain system according to claim 1, further comprising an information output section for outputting to an external device a data value of a certain concept provided to said information vessel from information transmitting means in said each knowledge unit. At least one of the one or more knowledge units further:
If the activation information determination means cannot determine the concept to be activated based on the intention attribute due to the lack of concepts to be awakened from the information effectively held in the information vessel, the shortage providing learning request information representing a learning request for a concept to be learned in the knowledge information managed by the knowledge management unit to the information vessel based on the learning intention attribute for requesting learning of the concept; a learning request means;
When the information provided as educational information to the information vessel in response to the provision of the learning request information to the information vessel by the first learning request means matches the data value of the concept to be learned, the data 4. The artificial brain system according to any one of claims 1 to 3, further comprising learning control means for providing the activation information determination means with the concept of value as an awakened concept. At least one of the one or more knowledge units further:
second learning request means for providing, to the information vessel, learning request information representing a learning request for a concept for which a corresponding data value is missing in the knowledge information to be managed by the knowledge management unit; When,
Information provided as educational information to the information vessel in response to the provision of the learning request information to the information vessel by the second learning request means is stored in the knowledge storage unit as a data value corresponding to the concept. 5. The artificial brain system according to claim 1, further comprising knowledge update means for updating knowledge information managed by said knowledge management unit. When information representing a concept to be added to the one piece of knowledge is effectively held in the information vessel, the knowledge represented by the knowledge information managed by the knowledge management unit is added to the concept. 6. The artificial brain system according to claim 5, comprising concept adding means for adding as a constructing concept. a learning output unit that outputs the learning request information provided to the information vessel from any one of the plurality of knowledge units to an external device;
7. The artificial brain system according to any one of claims 4 to 6, further comprising an educational input unit for providing information as said educational information from an external device to said information vessel. a knowledge storage unit storing knowledge information including data values representing each of a plurality of concepts constituting one knowledge;
a knowledge management unit that manages the knowledge information stored in the knowledge storage unit so as to associate the concept with the data value;
Information to be provided is managed by the knowledge management unit based on intention attributes that determine concepts to be activated corresponding to concepts awakened in the knowledge information managed by the knowledge management unit. Activation for determining the data value of a concept to be activated in the knowledge information managed by the knowledge management unit assuming that the concept has been awakened when the data value of the concept in the knowledge information matches the data value of the concept an information determining means;
and information transmission means for transmitting a data value of the concept determined by the activation information determination means among the knowledge information stored in the knowledge storage unit. If the activation information determination means cannot determine the concept to be activated based on the intention attribute due to the lack of the concept to be awakened from the provided information, the learning request for the lacking concept is determined. first learning request means for outputting learning request information representing a learning request for a concept to be learned in the knowledge information managed by the knowledge management unit, based on the learning intention attribute to be performed;
when the information provided as educational information in response to the output of the learning request information by the first learning request means matches the data value of the concept to be learned, the concept of the data value is regarded as the awakened concept 9. The intelligent device according to claim 8, further comprising learning control means for serving said activation information determination means. second learning request means for outputting learning request information representing a learning request for a concept when a concept lacks a corresponding data value in the knowledge information to be managed by the knowledge management unit;
Information provided as educational information in response to the output of the learning request information by the second learning request means is stored in the knowledge storage unit as a data value corresponding to the concept, and the knowledge management unit stores 10. A knowledge device according to claim 8 or 9, comprising knowledge update means for updating managed knowledge information. When information representing a concept to be added to the one piece of knowledge is provided, the concept is added as a concept constituting the knowledge represented by the knowledge information managed by the knowledge management unit . 11. A knowledge device according to claim 10, comprising concept adding means.

Images