A Metamodel and Framework for Artificial General Intelligence From Theory to Practice
Hugo Latapie, Ozkan Kilic, Gaowen Liu, Yan Yan, Ramana Kompella, Pei Wang, Kristinn R. Thorisson, Adam Lawrence, Yuhong Sun, Jayanth Srinivasa
FFebruary 12, 2021 1:30 MetamodelAGI
Journal of Artificial Intelligence and Consciousness © World Scientific Publishing Company
A Metamodel and Framework for Artificial General IntelligenceFrom Theory to Practice
Hugo Latapie, Ozkan Kilic, Gaowen Liu, Yan Yan, Ramana Kompella, Pei Wang, Kristinn R.Th´orisson, Adam Lawrence, Yuhong Sun, Jayanth Srinivasa [email protected], [email protected], [email protected], [email protected], [email protected],[email protected], [email protected], [email protected], [email protected],[email protected]
Received 15th December 2020Revised ... December 2020This paper introduces a new metamodel-based knowledge representation that signifi-cantly improves autonomous learning and adaptation. While interest in hybrid machinelearning / symbolic AI systems leveraging, for example, reasoning and knowledge graphs,is gaining popularity, we find there remains a need for both a clear definition of knowl-edge and a metamodel to guide the creation and manipulation of knowledge. Someof the benefits of the metamodel we introduce in this paper include a solution to thesymbol grounding problem, cumulative learning, and federated learning. We have ap-plied the metamodel to problems ranging from time series analysis, computer vision,and natural language understanding and have found that the metamodel enables a widevariety of learning mechanisms ranging from machine learning, to graph network anal-ysis and learning by reasoning engines to interoperate in a highly synergistic way. Ourmetamodel-based projects have consistently exhibited unprecedented accuracy, perfor-mance, and ability to generalize. This paper is inspired by the state-of-the-art approachesto AGI, recent AGI-aspiring work, the granular computing community, as well as AlfredKorzybski’s general semantics. One surprising consequence of the metamodel is that itnot only enables a new level of autonomous learning and optimal functioning for machineintelligences, but may also shed light on a path to better understanding how to improvehuman cognition.
Keywords : Artificial Intelligence and AI and AGI and General Semantics and Levels ofAbstraction and Neurosymbolic and Cognitive Architecture
1. Introduction
The field of artificial intelligence has advanced considerably since its inception in1956 at the Dartmouth Conference organized by Marvin Minsky, John McCarthy,Claude Shannon, and Nathan Rochester. The exponential growth in compute powerand data, along with advances in machine learning and in particular, deep learning, Preprint of an article submitted for consideration in [Journal of Artificial Intelli-gence and Consciousness] © a r X i v : . [ c s . A I] F e b ebruary 12, 2021 1:30 MetamodelAGI Latapie et al., 2020 have resulted in remarkable pattern recognition capabilities. For example, real-timedetection of pedestrians has recently achieved an average precision of over 55%[Tan et al., 2019]. Natural language understanding systems are achieving superhu-man performance on some tasks such as yes/no question answering [Wang et al.,2019]. However, as Yan [LeCun, 2020] wrote,“trying to build intelligent machines byscaling up language models is like building high-altitude planes to go to the moon.You might beat altitude records, but going to the moon will require a completelydifferent approach.” Beyond the need to improve the accuracy of pattern recognitionbeyond current levels, current deep learning approaches suffer from susceptibilityto adversarial attacks, a need for copious amounts of labeled training data, and aninability to meaningfully generalize.After over 30 years of intense effort, the AGI community has developed the theo-retical underpinnings for AGI and affiliated working software systems [Wang, 2005,2006]. While achieving human-level AGI is arguably years to decades away, some ofthe currently available AGI subsystems are ready to be incorporated into non-profitand for-profit products and services. Some of the most promising AGI systems wehave encountered are OpenNARS [Wang, 2006, 2010], OpenCog [Goertzel, 2009;Goertzel et al., 2013], and AERA [Th´orisson, 2012]. We are collaborating with allthree teams and have developed video analytics and Smart City applications thatleverage both OpenCog and OpenNARS [Hammer et al., 2019].After several years of applying these AGI technologies to complex, real-worldproblems in IoT, networking, and security at scale, we have encountered a fewstumbling blocks largely related to real-time performance on large datasets andcumulative learning [Th´orisson et al., 2019]. In order to progress from successfulproofs of concept and demos to scalable products, we have developed the DeepFusion Reasoning Engine (DFRE) metamodel and associated DFRE framework,which is the focus of this paper. We have used this metamodel and framework tobring together a wide array of technologies ranging from machine learning, deeplearning, and probabilistic programming to the reasoning engines operating underthe assumption of insufficient knowledge and resources (AIKR) [Wang, 2005]. Asdiscussed below, we believe the initial results are promising: The data show a dra-matic increase in system accuracy, ability to generalize, resource utilization, andreal-time performance when compared to state-of-the-art AI systems.The following sections will cover related theories and technologies, the meta-model itself, empirical results, and discussions as well as future work. Appendix Acontains some background information that may help readers gain a deeper under-standing of this material.
2. Related Theories and Technologies2.1.
Korzybski
After living through WWI, Korzybki, was concerned about the future trajectory ofmankind. He focused his research on the creation of a non-metaphysical definitionebruary 12, 2021 1:30 MetamodelAGI
A Metamodel and Framework for Artificial General Intelligence of man that was both descriptive and predictive from a scientific and engineeringperspective. He focused on what he termed the “time-binding property” that en-abled human societies to advance exponentially from a technological perspective.As his objective was to discover the source of humanity’s self-destructive tendencies,he created a model of the human nervous system he called “the structural differen-tial”, which is the primary inspiration for our metamodel. Korzybski developed atheory that explains the power of the human nervous system, the weaknesses thatcause many of humanity’s major problems such as world wars, and a path to op-timal/correct functioning of the human nervous system. The Institute of GeneralSemantics, which Korzybski founded, continues to train of educators around theworld.Korzybski focused on helping people better utilize the considerable power of thehuman nervous system in part because the combination of exponential advancementof technology and a primitive way of using it could lead to large-scale destruction.Given that we have far more powerful compute capability and weapons, the saneoperation of all autonomous learning systems, human or machine, is of even greaterimportance. OpenNARS
OpenNARS (see [Hammer et al., 2019]) is a Java implementation of a Non-Axiomatic Reasoning System (NARS). NARS is a general-purpose reasoning systemthat works under the Assumption of Insufficient Knowledge and Resources (AIKR).As described in Wang [2009], this means the system works in Real-Time, is alwaysopen to new input, and operates with a constant information processing ability andstorage space. An important part is the Non-Axiomatic Logic (see Wang [2010] andWang [2006]) which allows the system to deal with uncertainty. To our knowledge,our solution is the first to apply NARS to a real-time visual reasoning task.
Embeddings
Graph embedding [Cui et al., 2018; Hamilton et al., 2017] is a technique used torepresent graph nodes, edges, and sub-graphs in vector space that other machinelearning algorithms can use. Graph neural networks use graph embeddings to ag-gregate information from graph structures in non-Euclidean ways. This allows theDFRE Framework to use the embeddings to learn from different data sources thatare in the form of graphs, such as Concept Net [Speer et al., 2017]. Despite its per-formance across different domains, the graph neural networks suffer from scalabilityissues [Ying et al., 2018; Zhou et al., 2018] because calculating the Laplacian matrixfor all nodes in a large network may not be feasible. The levels of abstraction andthe focus of attention mechanisms used by an Agent resolve these scalability issuesin a systematic way.ebruary 12, 2021 1:30 MetamodelAGI Latapie et al., 2020
3. The DFRE Metamodel
The DFRE metamodel and framework are based on the idea that knowledge is ahierarchical structure, where the levels in the hierarchy correspond to levels of ab-straction. The
DFRE metamodel refers to the way that knowledge is hierarchicallystructured while a model refers to knowledge stored in a manner that complies tothe DFRE metamodel. It is based on non-Aristotelian, non-elementalistic systemsof thinking. The backbone of its hierarchical structure is based on difference , a.k.a.antisymmetric relations, while the offshoots of such relations are based on symmet-ric relations. As in figure 1, even a simple amoeba has to differentiate distinctionsand similarities because preserving symmetric and antisymmetric relations is fatallyimportant.
Fig. 1. Amoeba distinguishing between distinctions and similarities.
Korzybski [Korzybski, 1994] dedicated the majority of his professional life toanalyzing and studying the nature of this hierarchical structure. While it is wellbeyond the scope of this paper to discuss the details of his analysis, our initial focuswas to incorporate these fundamental principles. • K1 – the core framework of knowledge is based on anti-symmetric relations – Spatial understanding: right/left/top/bottom – Temporal understanding: before/after – Corporal understanding: pain/satiation – Emotional understanding: happy/sad – Social understanding: friend/foe – Causal understanding: X causes Y • K2 – symmetric relations add further structure – A and B are friendsebruary 12, 2021 1:30 MetamodelAGI
A Metamodel and Framework for Artificial General Intelligence – A is like B • K3 – knowledge is layered – Sensor data is on a different layer than high level symbolic information – Symbolic information B, which expands or provides context to sym-bolic information A, is at a higher layer / level of abstraction – In the symbolic space, there are theoretically an infinite number oflayers, i.e., it is always possible to refer to a symbol and expand uponit, thus creating yet another level of abstraction • K4 – since knowledge is structure, any structure destroying operationssuch as confusing levels of abstraction, treating an anti-symmetric relationas symmetric, or vice-versa, can, if inadvertently applied, be knowledge-corrupting and/or a knowledge-destroying operation. However it shouldbe noted that creative problem solving and other adaptive behaviors mayrequire mixing levels of abstraction. The key is to ensure that the long-term structure of the metamodel is meticulously maintained and that theseoperations occur by design and not by accident.The DFRE Knowledge Graph (DFRE KG) groups information into four levels asshown in Figure 2. These are labeled L0, L1, L2, and L* and represent different levelsof abstraction with L0 being closest to the raw data collected from various sensorsand external systems, and L2 representing the highest levels of abstraction, typicallyobtained via mathematical methods, i.e. statistical learning and reasoning. Thelayer L2 can theoretically have infinitely many sub-layers. L* represents the layerwhere the high-level goals and motivations, such as self-monitoring, self-adjustingand self-repair, are stored. There is no global absolute level for a concept andall sub-levels in L2 are relative. However, L0, L1, L2 and L* are global conceptsthemselves. For example, an Agent, which is basically a computer program thatperforms various tasks autonomously, can be instantiated to troubleshoot a problem,such as one related to object recognition or computer networking. The frameworkpromotes cognitive synergy and metalearning, which refer to the use of differentcomputational techniques (e.g., probabilistic programming, Machine Learning/DeepLearning, and such) to enrich its knowledge and address combinatorial explosionissues.One advantage of the DFRE Framework is its integration of human domainexpertise, ontologies, prior learnings by the current DFRE KG-based system andother similar systems, and additional sources of prior knowledge through the mid-dleware services. It provides a set of services that an Agent can utilize as shown in Korzybski argues that what is currently limiting humanity’s advancement is the general lackof understanding of how our own abstracting mechanisms work [Korzybski, 1949]. He considersmankind to currently be in the childhood of humanity and the day, if it should come, that humanitybecomes generally aware of the metamodel, is the day humanity enters into the “manhood ofhumanity” [Korzybski, 1921] ebruary 12, 2021 1:30 MetamodelAGI Latapie et al., 2020
Fig. 2. DFRE Framework with four levels of abstraction.
Figure 3.
Fig. 3. DFRE Framework.
The Sensor Data Services are used to digitize any real world data, such asvideo recordings. Similarly, the Data Structuring Services restructure data ,e.g.,ebruary 12, 2021 1:30 MetamodelAGI
A Metamodel and Framework for Artificial General Intelligence rectifying an image, if needed. These two services are the basis for Image ProcessingServices which provide a set of supervised and unsupervised algorithms to detectobjects, colors, lines, and other visual criteria. The Sensor Data Analytic Servicesanalyze objects and create object boundaries enriched with local properties, suchas an object’s size and coordinates, which create a 2D symbolic representationof the world. Spatial Semantic Services then uses this representation to constructthe initial knowledge graph that captures the spatial relations of the object as arelational graph. Any L2- or high-level reasoning is performed on this knowledgegraph.Graph-based knowledge representation provides a system with the ability to: • Effectively capture the relations in the sub-symbolic world in a world ofsymbols, • Keep a fluid data structure independent of programming language, in whichAgents running on different platforms can share and contribute, • Use algorithms based on the graph neural networks to allow preservationof topological dependency of information [Scarselli et al., 2009] on nodes.All processes are fully orchestrated by the Agent that catalogues knowledge bystrictly preserving the structure while evolving new structures and levels of abstrac-tion in its knowledge graph because, for DFRE KG, knowledge is structure. MultipleAgents can have not only individual knowledge graphs but also a single knowledgegraph on which all can cooperate and contribute. In other words, multiple Agentscan work toward the same goal by sharing the same knowledge graph synchronouslyor asynchronously. Different Agents can have partially or fully different knowledgegraphs depending on their experience, and share those entire graphs or their frag-ments through the communication channel provided by the DFRE Framework. Notethat although the framework can provide supervised machine learning algorithmsif needed, the current IoT use case is based on a retail store which requires unsu-pervised methods as explained in the next section.
4. Experimental Results
The DFRE Framework was previously tested in the Smart City domain [Hammer etal., 2019], in which the system learns directly from experience with no initial trainingrequired (one-shot), based on a fusion of sub-symbolic (object tracker) and symbolic(ontology and reasoning) information. The current use case is based on object-class recognition in a retail store. Shelf space monitoring, inventory managementand alerts for potential stock shortages are crucial tasks for retailers who want tomaintain an effective supply chain management. In order to expedite and automatethese processes, and reduce both the requisite for human labor and the risk ofhuman error, several machine learning and deep learning-based techniques havebeen utilized [Baz et al., 2016; Franco et al., 2017; George and Floerkemeier, 2014;Tonioni and Stefano, 2017]. Despite the high success rates, the main problems forebruary 12, 2021 1:30 MetamodelAGI Latapie et al., 2020 such systems are the requirements for a broad training set, including compilingimages of the same product with different lighting and from different angles, andretraining when a new product is introduced or an existing product is visuallyupdated. The current use case does not demonstrate an artificial neural network-based learning. The DFRE Framework has an artificial general intelligence-basedapproach to these problems.
Fig. 4. Retail use case for DFRE Framework.
Before a reasoning engine operates on symbolic data within the context of theDFRE Framework, several services must be run, as shown in Figure 4. The flowstarts with a still image captured from a video camera that constantly records theretail shelves by the Sensor Data Services as in Figure 4.a, which corresponds toL0 in Figure 2. Next, the image is rectified by the Data Structuring Services inFigure 4.b for better line detection by the Image Processing Services, as displayedin Figure 4.c. The Image Processing Services in the retail case are unsupervisedalgorithms used for color-based pixel clustering and line detection, such as proba-bilistic Hough transform [Kiryati et al., 1991]. The Sensor Data Analytics Servicesin Figure 4.d create the bounding boxes which represent the input in a 2D world ofrectangles, as shown in Figure 4.e. The sole aim of all these services is to providethe DFRE KG with the best symbolic representation of the sub-symbolic worldin rectangles. Finally, the Spatial Semantics Services operate on the rectangles toconstruct a knowledge graph, which preserves not only the symbolic representationof the world, but also the structures within it in terms of relations, as shown in Fig-ure 4.f. This constitutes the L1 level abstraction in the DFRE KG. L1 knowledgeebruary 12, 2021 1:30 MetamodelAGI
A Metamodel and Framework for Artificial General Intelligence graph representation also recognizes and preserves the attributes of each boundingbox, such as the top-left x and y coordinates, and the center ’s coordinates: height , width , area and circumference . The relations used for the current use case are in-side , aligned , contains , above , below , on left of , on right of , on top of , under and floating . Since the relations in the DFRE KG are by default antisymmetrical, thesystem does not know that aligned(a,b) means aligned(b,a) , or on left of and onright of are inverse relations unless such terms are input as expert knowledge or arelearned by the system through experience or simulations. The only innate relationsin the DFRE metamodel are distinctions , which are anti-symmetric and similarity relations; and the rest is learned by experience.The system’s ultimate aim is to dynamically determine shelves , products and unknown/others , as illustrated in Figure 5, and to monitor the results with times-tamps.While L2 identifies only the concepts of shelf , product and unknown , and thepossible relations among them, the reasoning engine, NARS [Wang, 2006, 2010;Wang et al., 2018], creates their L1 intensions as an evidence-based truth systemin which there is no absolute knowledge. This is useful in the retail use case sce-nario because the noise in L0 data causes both overlapping regions and conflictingpremises at L1. This noise results from not only the projection of the 3D worldinput data into a 2D framework, but also the unsupervised algorithms used by L1services. The system has only four rules for L2 level reasoning: • If a rectangle contains another rectangle that is not floating, the outer rect-angle can be a shelf while the inner one can be a product. • If a rectangle is aligned with a shelf, it can be a shelf too. • If a rectangle is aligned with a product horizontally, it can be a product too. • If a floating rectangle is stacked on a product, it can be a product too.
Note that applying levels of abstraction gives the DFRE Framework the power toperform reasoning based on the expert knowledge in L2 level mostly independentof L1 level knowledge. In other words, the system does not need to be trained fordifferent input; it is unsupervised in that sense. The system has a metalearning ob-jective which continuously attempts to improve its knowledge representation. Thecurrent use case had 152 rectangles of various shapes and locations, of which 107were products, 16 were shelves, and the remaining 29 were other objects. Whenthe knowledge graph in L1 is converted into Narsese, 1,478 lines of premises thatrepresent both the relations and attributes were obtained and sent to the reasoner.Such a large amount of input with the conflicting evidence caused the reasoningengine to perform poorly. Furthermore, the symmetry and transitivity propertiesassociated with the reasoner resulted in the scrambling of the existing structure inthe knowledge graph. Therefore, the DFRE Framework employed a Focus of At-tention (FoA) mechanism. The FoA creates overlapping covers of knowledge graphsfor the reasoner to work on this limited context. Later, the framework combines theresults from the covers to finally determine the intensional category. For example,ebruary 12, 2021 1:30 MetamodelAGI Latapie et al., 2020
Fig. 5. LoA for retail use case. when the FoA utilizes the reasoner on a region, a rectangle can be recognized as ashelf. However, when the same rectangle is processed in another cover, it may beclassified as a product. The result with higher frequency and confidence wins. TheFoA mechanism is inspired by the human visual attention system, which managesinput flow and recollects evidence as needed in case of a conflicting reasoning re-sult. A FoA mechanism can be based on objects’ attributes, such as color or size,with awareness of proximity. For this use case, the FoA determined the contexts bypicking the largest non-empty rectangle, and traversing its neighbors based on theirsizes in decreasing order.ebruary 12, 2021 1:30 MetamodelAGI
A Metamodel and Framework for Artificial General Intelligence The framework is tested in various settings with different camera angles andproducts placements as shown in Figure 6.
Fig. 6. DFRE Framework qualitative result in retails use cases with different environments
In Figure 6, each row represents samples from different settings: rectified frames,bounding boxes, and instantaneous output of the reasoner. We would like to em-phasize that the system does not require any retraining or any change in order toadapt to the new setting. It requires only a camera to be pointed to the scene; thenit automatically generalizes.DFRE Framework was tested 10 times in 4 different settings with and withoutthe FoA. The precision, recall and f-1 scores are exhibited in Table 1.
Table 1. DFRE Framework experimental results.
Category without FoA (%) with FoA (%)precision recall f1-score precision recall f1-score product shelf other overall accuracy 46.30 (min/max: 30.13/84.65) (min/max: 88.10/100.00)
The results indicate that the FoA mechanism improves the success of our AGI-based framework significantly by allowing the reasoner to utilize all of its computingresources in a limited but controlled context. The results are accumulated by theframework, and the reasoner makes a final decision. This approach not only allowsus to perform reasoning on the intension sets of L1 knowledge, which are retrievedthrough unsupervised methods, but also resolves the combinatorial explosion prob-lem whose threshold depends on the limits of available resources. In addition, oneebruary 12, 2021 1:30 MetamodelAGI Latapie et al., 2020 can easily extend this retail use case to include prior knowledge of product typesand other visual objects, such as tables, chairs, people and shelves, as allowed bythe DFRE KG.
Graph Embedding for Link Predictions
Recall, a graph G ( V, E ), where V is the set of all vertices, or nodes, in G , and E , isthe set of paired nodes, called edges. | V | ∈ Z is the order of the graph, or the totalnumber of nodes.As mentioned before, DFRE takes advantage of graph embedding, a transfor-mation that constructs a non-dimensional knowledge graph G into a d -dimensionalvectors space S ∈ R | V |× d . Among the many benefits, such as creating a Euclideandistance measurement for G , link predictions can be established between node vec-tors in S . Preliminary experimental results have given a great deal of insight intothe relationships between nodes that might otherwise not be present from the graphspace.The main algorithm used by DFRE to transform our knowledge graph G to a2-dimensional vector space is Node2Vec. [Grover and Leskovec, 2016]. This frame-work is a representation learning based approach that learns continuous featurerepresentations for all nodes in a given knowledge graph G . The benefits from thisalgorithm, and the motivation for use in DFRE, construct the graph embeddingspace S where link prediction, and other methods of measurement, can be usedwhile preserving relevant network properties from the original knowledge graph.The functionality behind Node2Vec is similar to most other embedding pro-cesses, by use of the Skip-Gram model, and a sampling-strategy. Four argumentsare input into the framework: the number of walks, the length of the walks, p , and q , where p is referred to as the return hyper-parameter, and q is the I/O hyper-parameter.Once a 2-dimensional vector representation has been assigned to every node n ∈ V , our embedding vectors space S ∈ R | V |× can provide additional metrics usedfor machine learning and prediction measures. One such measure is link predictionused to understand the relationship between nodes in a graph that might not beobvious from the graph space.Consider the nodes n , n ∈ G ( V, E ) such that n and n are not similar ideasin the graph (e.g. the probability ( n , n ) ∈ E ( G ) is low). Once the nodes arerepresented in vector form ˆ n , ˆ n ∈ R ⊂ S , we establish a linear relationshipbetween the two such that a line y = ax + b is satisfied, where a = ˆ n − ˆ n ˆ n − ˆ n and b = ˆ n − a ( ˆ n ).Let (cid:15) > ∀ ˆ n k that lies within the range of y ± (cid:15) , we consider these nodevectors to be associated hidden links between two the two ideas ˆ n and ˆ n .Additionally, if the line y is divited into four evenly distributed quadrants y ..y and grown by small perturbations where y (cid:48) = y ± (cid:15) (cid:48) such that (cid:15) (cid:48) = (cid:15) + γ and γ ∈ (0 ,
1] until there exists at least one ˆ n k in every quadrant. We call this set ofebruary 12, 2021 1:30 MetamodelAGI A Metamodel and Framework for Artificial General Intelligence node vectors S n .This set of node vectors S n gathered within range y (cid:48) provide DFRE a rela-tionship that might not be immediately obvious from the graph space alone. Thedistribution of the vectors along the quadrants is revealing in away such that, forexample, consider the two disjoint subsets ( ˆ n k ) i and ( ˆ n l ) j of S n . Without loss ofgenerality, if ( ˆ n k ) i ∈ y (cid:48) and ( ˆ n k ) j ∈ y (cid:48) , where i << j , we see that the relationshipskews towards the set of node vectors that lie within the range of y (cid:48) .Additionally, within the embedding space, consider a finite set of clusters { C , C , . . . } , each corresponding to its own central idea. For any arbitrary cluster C i , if a new node vector ˆ n (cid:48) is introduced in S such that ˆ n (cid:48) ∈ C i , then we can easilyleverage this proximity into our sub-symbolic space to identify any additional nodevectors.We find the main benefit to graph embedding is that we now have an unsuper-vised method for correlating the graph embedding space with additional embeddingspaces that are generated using unsupervised machine learning techniques.
5. Philosophical Implications
The nativism-versus-empiricism debate, which posits that some knowledge is in-nate and some is learned through experience, was ascribed in the ancient world bythe Greek philosophers, including Plato and Epicurus. Today, Descartes is widelyaccepted as a pioneering philosopher working on the mind as he furthered and refor-mulated the debate in the 17 th century with new arguments. Perception, memory,and reasoning are three fundamental cognitive faculties that enhance this debate byexplicating the building blocks of natural intelligence. We perceive the sub-symbolicworld, and abstract it in memory, and reason on this symbolic world representa-tion. All three place concept learning and categorization at the center of the humanmind.The process of concept learning and categorization continues to be an activeresearch topic related to the human mind since it is essential to natural intelligence[Lakoff, 1984] and cognitively inspired robotics research [Chella et al., 2006; Lietoet al., 2017]. It is widely accepted that this process is based more on interactionalproperties and relationships among Agents, as well as between an Agent and itsenvironment, than objective features such as color, shape and size [Johnson, 1987;Lakoff, 1984]. This makes the distinction of anti-symmetric and symmetric relationscrucial in the DFRE Framework, which assumes that the levels of abstraction arepart of innate knowledge. In other words, an Agent has L0, L1 and pre-existingL2 by default. This constitutes a common a priori metamodel shared by all DFREAgents. Each Agent instantiated from the framework has the abstraction skill basedon interactional features and relationships. If the concepts in real life exist in inter-actional systems, natural intelligence needs to capture these systems of interactionswith its own tools, such as abstraction. These tools should also be based on in-teractional features by strictly preserving the distinction between symmetry andebruary 12, 2021 1:30 MetamodelAGI Latapie et al., 2020 anti-symmetry.The mind is a system as well. Modern cognitive psychologists agree that con-cepts and their relations in memory function as the fundamental data structuresto higher level system operations, such as problem solving, planning, reasoning andlanguage. Concepts are abstractions that have evolved from a conceptual primitive.An ideal candidate for a conceptual primitive would be something that is a stepaway from a sensorimotor experience [G¨ardenfors, 2000], but is still an abstractionof experience [Cohen, 1997]. For example, a dog fails the mirror test but exhibitsintelligence when olfactory skills are needed to complete a task [Horowitz, 2017]. Ababy’s mouthing behavior is not only a requisite for developing oral skills but alsofor discovering the surrounding environment through one of its expert sensorimotorskills related to its survival. The baby is probably abstracting many objects into ed-ible versus inedible higher categories given its insufficient knowledge and resources.What is astonishing about a natural intelligence system is that it does not need aplethora of training input and experiments to learn the abstraction. It quickly andautomatically fits new information into an existing abstraction or evolves it into anew one, if needed. This is nature’s way of managing combinatorial explosion. Ob-jects and their interconnected relationships within the world can be chaotic. Naturalintelligence’s solution to this problem becomes its strength: context. The concept of‘sand’ has different abstractions depending on whether it is on a beach, on a cam-era, or on leaves. An Agent in these three different contexts must abstract the sandin relation to its interaction with world in its short-term memory. This cumulativeset of experiences can later become part of long-term memory, more specifically,episodic memory. The DFRE Framework uses a Focus of Attention (FoA) mecha-nism that provides the context while addressing the combinatorial explosion prob-lem. The DFRE metamodel’s new way of representing practically all knowledge astemporally evolving (i.e. time series) can be viewed as the metamodel’s conceptualspace. For example, the retail use case given in Section 2 starts with a 3D worldof pixels that is abstracted as lines and rectangles in 2D. The framework producesspatial semantics using the rectangles in the 2D world. Based on this situation, afew hundred rectangles produce thousands of semantic relations, which present acombinatorial explosion for most AGI reasoning engines. For each scene, the DFREKG creates contexts, such candidate shelves, runs reasoners for each context, andmerges knowledge in an incremental way. This not only addresses the combinatorialexplosion issue, but also increases the success rate of reasoning, provided that thelevels of abstraction are computed properly [Gorban and Tyukin, 2018].Abstracting concepts in relation to their contexts also allows a natural intel-ligence to perform mental experiments, which is a crucial part of planning andproblem solving. The DFRE Framework can integrate with various simulators, re-run a previous example together with its context, and alter what is known forthe purposes of experimentation to gain new knowledge, which is relationships andinteractions of concepts.Having granular structures provides structured thinking, structured problemebruary 12, 2021 1:30 MetamodelAGI
A Metamodel and Framework for Artificial General Intelligence solving, and structured information processing [Yao and Deng, 2012]. The DFREFramework has granular structures but emphasizes the preservation of structures inknowledge. When a genuine problem that cannot be solved by the current knowledgearises, it requires scrambling the structures and running simulations on the newstructures in order to provide an Agent with creativity. Note that this knowledgescrambling is performed in a separate sandbox. The DFRE Framework ensures thatthe primary DFRE KG is not corrupted by these creative, synthetic, “knowledge-scrambling” activities.
6. Discussion on Metamodel and Consciousness
For the purposes of this discussion, we will define consciousness as autonomous self-aware adaptation to the environment. This means that an abstraction of the self aswell as the environment of the self is learned autonomously. Human consciousnessbuilds on the prior capabilities of chemistry-binders (plants) and space-binders (an-imals) with the unique ability of infinite levels of abstraction [Korzybski, 1994]. Forany concept, one can envision creating a higher-level meta-concept. We are able toformulate symbolic representations that can be externalized and shared. The hu-man model of the self evolves not only via direct interaction with the environmentand cogitation, but also by watching other humans and modeling them. Human con-sciousness as an implementation of the metamodel appears to be dynamic in nature.The concept of self can grow to encompass family, friends, social/work groups, andbeyond. Understanding our nature as time-binders that form a collective conscious-ness of ever-increasing power (cognitive and physical) over time, civilizations, andgenerations, appears to lead to higher cognitive functioning of the individual.Human societies consisting of billions of people networked together in real-time,with petabytes of shared storage and petaflops of compute, may see the evolution ofexo-cortical consciousness. In fact, many argue that this exo-cortical consciousnessalready exists with the growing number of autonomous self-healing systems deployedand connected throughout the world.Since the metamodel hypothesizes the existence of an exo-cortical consciousness,it consequently yields to the possibility of implementing artificial consciousness, e.g.,in robots. Artificial consciousness, which is also known as machine consciousness,is a field designed to mimic the aspects of human cognition that are related tohuman consciousness [Aleksander, 2008; Chella and Manzotti, 2009]. In the 1950s,consciousness was seen as a vague term, and inseparable from intelligence. [Searle,1992; Chalmers, 1996]. Fortunately, the improvements in technology, and compu-tational and cognitive sciences have created new interest in the field. [Chella andManzotti, 2011] reviews that the most important gap between artificial and biolog-ical consciousness studies is engineering autonomy, semantic capabilities, intention-ality, self-motivation, resilience and information integration. The Agents based onthe metamodel have autonomy. They also have semantic capabilities and intentionsto seek solutions or communicate with other Agents for knowledge sharing, whichebruary 12, 2021 1:30 MetamodelAGI Latapie et al., 2020 are set by self-motivation.Chella and Manzotti [2011] emphasize that consciousness is a real physical phe-nomenon, can be artificially replicated, is either a computational phenomenon ormore. We bring forth the metamodel as an enabler for the achievement of artificialconsciousness. The abstraction mechanism constantly and automatically creates theabstractions of the sensor data and the system’s own experience. The metamodelis based on generation of new knowledge using self-perception and experience, andshares knowledge among the Agents of similar nature to support collective con-sciousness and resilience. The creation of self through experience gives the meta-model the ability of enhanced generalization and autonomy. Similarly, focus of theattention mechanism to segment complex problems semantically is also related toconsciousness because attention and consciousness are interrelated [Taylor, 2007,2009]. Implementation of attention is important because control theory is relatedto consciousness and plays a leading role in the intentional mechanism of an Agent.
7. Conclusion
Several mathematical models and formal semantics [D¨untsch and Gediga, 2002; Be-lohl´avek, 2004; Wang and Liu, 2008; Wille, 1980; Ma et al., 2007] are proposed tospecify the meanings of real world objects as concept structures and lattices. How-ever, they are computationally expensive [Jinhai et al., 2015]. One way to overcomethis issue is with granular computing [Yao and Deng, 2012]. The extension of a con-cept can be considered a granule, and the intension of the concept is the descriptionof the granule. Assuming that concepts share granular common parts with varyingderivational and compositional stages, categorization, abstraction and approxima-tion occur at multiple levels of granularity which plays an important role in humanperception [Hobbs, 1985; Yao, 2001, 2009]. The DFRE Framework has granularstructures but emphasizes the preservation of structures in knowledge. Being in theextension of a concept does not necessarily grant the granular concept the rightto have similar relations and interactions of its intensional concept up to a certaindegree or probability. Each level must preserve its inter-concept relationships andits symmetry or anti-symmetry in a hierarchically structured way.We have outlined the fundamental principles of the DFRE Framework. DFREtakes a neurosymbolic approach leveraging state-of-the-art subsymbolic algorithms(e.g. ML/DL/Matrix Profile) and state-of-the-art symbolic processing (e.g. reason-ing, probabilistic programming, and graph analysis) in a synergistic way. The DFREmetamodel can be thought of as a knowledge graph with some additional structure,which includes both a formalized means of handling anti-symmetric and symmet-ric relations, as well as a model of abstraction. This additional structure enablesDFRE-based systems to maintain the structure of knowledge and seamlessly sup-port cumulative and distributed learning. Although this paper provides highlightsof one experiment in the visual domain employing an unsupervised approach, wehave also run similar experiments on time series and natural language data withebruary 12, 2021 1:30 MetamodelAGI
A Metamodel and Framework for Artificial General Intelligence similar promising results.
8. Future work
Nowadays there is a rapid transition in the AI research field from single modalitytasks, such as image classification and machine translation, to more challengingtasks that involve multiple modalities of data and subtle reasoning, such as visualquestion answering (VQA) [Agrawal et al., 2015; Andersonet al., 2018; Zhuet al.,2020] and visual dialog [Vishvak et al., 2019]. A meaningful and informative conver-sation, either between human-computer or computer-computer, is an appropriatetask to demonstrate such a reasoning process given the complex information ex-change mechanism during the dialog. However, most existing research focuses onthe dialog itself and involves only a single Agent. We plan to design a more reliableDFRE system with implicit information sources. To this end, we propose a novelnatural and challenging task with implicit information sources: describe an unseenvideo mainly based on the dialog between two cooperative Agents.The entire process can be described in three phases: In the preparation phase,two Agents are provided with different information. Agent A1 is able to see the com-plete information from different modalities (i.e., video, audio, text), while Agent A2is only given limited information. In the second phase, A2 has several opportuni-ties to ask A1 relevant questions about the video, such as the person involved, theevent happened, etc.
A2 is encouraged to ask questions that help to accomplish theultimate video description objective, and A1 is expected to give informative andconstructive answers that not only provide the needed information but also moti-vate A2 to ask additional useful questions in the next conversation round. Afterseveral rounds of question-answer interactions, A2 is asked to describe the unseenvideo based on the limited information and the dialog history with A1. In this tasksetup, our DFRE system accomplishes a multi-modal task even without direct ac-cess to the original information, but learns to filter and extract useful informationfrom a less sensitive information source, i.e. , the dialog. It is highly difficult for AIsystems to identify people based on the natural language descriptions. Therefore,such task settings and reasoning ability based on implicit information sources havegreat potential to be applied in a wide practical context, such as the smart hospitalsystems, improving current systems.The key aspect to consider in this future work is the effective knowledge transferfrom A1 to A2. A1 plays the role of humans, with full access to all the information,while A2 has only an ambiguous understanding of the surrounding environmentfrom two static video frames after the first phase. In order to describe the videowith details that are not included in the initial input, A2 needs to extract usefulinformation from the dialog interactions with A1. Therefore, we will propose a QA-Cooperative network that involves two agents with the ability to process multiplemodalities of data. We further propose a cooperative learning method that enablesus to jointly train the network with a dynamic dialog history update mechanism.ebruary 12, 2021 1:30 MetamodelAGI Latapie et al., 2020
The knowledge gap and transfer process are both experimentally demonstrated.The novelties of the proposed future work can be summarized as follows: (i) Wepropose a novel and challenging video description task via two multi-modal dialogagents, whose ultimate goal is for one Agent to describe an unseen video based on theinteractive dialog history. This task establishes a more reliable setting by providingimplicit information sources to the metamodel. (ii) We propose a QA-Cooperativenetwork and the goal-driven learning method with a dynamic dialog history updatemechanism, which helps to effectively transfer knowledge between two agents. (iii)With the proposed network and cooperative learning method, our A2 Agent withlimited information can be expected to achieve promising performance, comparableto the strong baseline situation where full ground truth dialog is provided.
Appendix A. Background Material
This paper presents a few simple, yet potentially revolutionary ideas regarding thenature of knowledge. Unfortunately, we find that current habits of cognition canmake understanding this paper difficult for some people irrespective of level offormal education, intelligence, and so forth. This preface is intended to highlightsome of the potential blockages to understanding, along with causes, in an attemptto assist readers to potentially benefit from this paper.We begin by first diving into what may for many appear to be an oxymoronicconcept: “simple yet revolutionary.” In general, although this is largely an unstatedbelief/feeling, it would make much more sense that the term revolutionary wouldbe associated with a certain amount of complexity and/ or dramatic, immediateand obvious impact, e.g., a meteor killing the dinosaurs, Nikola Tesla inventingthe induction motor, or Einstein formulating a new theory of space-time. Peoplegenerally have difficulty gaining an intuitive understanding for small, incremental,simple changes that still have tremendous world changing impact over longer periodsof time.This particular cognitive blockage, as pointed out by Roberto Unger [Unger andSmolin, 2015], is probably at the heart of why society has not yet learned how tocontinually transform and adapt to changing technological, political, and economiccircumstances. Societal changes that are incremental are discounted as insignificant,and those that are dramatic are considered too dangerous, thus locking society intoa mode where tectonic political pressures build to monumental levels leading inex-orably to world wars and the like. However, for the purpose of understanding thispaper, it may help to keep in mind the rice and the chessboard problem [Wikipedia,2019] which demonstrates how, in an exponential growth situation, a single grainof rice quickly grows to several times the world production of rice [Unger, 2014].In order to understand how exponential growth laws apply to systems, such asthe AI systems which are the focus of this paper, or societies which we mentiononly in passing, we remind the reader that technological and knowledge progress atan exponential growth as highlighted in the references (pg. 73 of [Korzybski, 1921],ebruary 12, 2021 1:30 MetamodelAGI
A Metamodel and Framework for Artificial General Intelligence and [Swanson, 2020]).Another major impediment to understanding this paper is what Korzybskitermed elementalistic thinking. Elementalistic thinking is the mental habit of at-tempting to understand systems by focusing primarily on the elements of the sys-tems rather than the interactions between the elements. In complex systems consist-ing of a large number of elements, understanding individual elements does little tohelp understand the overall system. For example, in software consisting of millionsof lines of code, understanding the individual bits and bytes is of little to no benefitin understanding the nature of the entire software/hardware system.Korzybski identified many fallacies in our thought processes largely stemmingfrom cultures and educational systems based on modes of thinking originating inAristotle’s time circa 300 BC. While our scientific understanding has progressed byleaps and bounds, we remain largely bound by incorrect legacy beliefs. The meta-model we present in this paper is intended to create intelligent systems that exhibitnew levels of autonomous learning and operation. The irony is that to truly under-stand these concepts, one must be at least partially unhindered by the problemsthe metamodel solves.The metamodel is based on the addition of very simple structure to a basicknowledge graph consisting of nodes and links. We add structure to represent dis-tinction relations (i.e. this is not that). These distinction relations we represent asanti-symmetric and form the backbone of our knowledge structure. We then addstructure to represent similarity relations. Finally, we formalize Korzybski’s struc-tural differential to encode different levels of abstraction [Korzybski, 1949].While at the atomic level, these changes are trivial, at the system level, theeffects we have seen are revolutionary. The metamodel enables systems to learn andreason without inadvertently mixing levels of abstraction. This actually resolves allof the issues mentioned in this section as “cognitive blockages.” It should be notedthat Korzybski created the field of General Semantics to help humanity evolve to ahigher level of cognitive functioning. He spent many years creating a foundation andworking with students and educators. He found that younger children were able tofully absorb and adapt to these ideas extremely quickly. Unfortunately for adults,it is a longer process but always successful when there is real interest.We recommend keeping Baudrillard’s idea from Simulacra and Simulationin mind [Baudrillard, 1994]:“ The simulacrum is never that which conceals thetruth—it is the truth which conceals that there is none. The simulacrum is true.”In the sense that the mental models we humans create define our individual worldsand what is possible within those worlds. When these mental models are incorrect,as they are bound to be at times (e.g. the belief that something cannot be bothsimple/incremental and revolutionary), it has far reaching implications. At best,our mental models are useful abstractions, e.g. our simplification of the quantumscale world. If we keep this in mind, perhaps we will not attempt to touch theseemingly solid disk mental model we create for a dangerous spinning object likeebruary 12, 2021 1:30 MetamodelAGI Latapie et al., 2020 a spinning metal object [Korzybski, 1949]. More importantly, we will continuallyquestion whether our views are based on valid abstractions for the current context.As has been pointed out in the machine learning community: “all models are wronghowever some are useful.” [Box, 1976].
References
Agrawal, A., Lu, J., Antol, S., Mitchell, M., Zitnick, C. L., Batra, D., Parikh,D. VQA: Visual Question Answering. Proceedings of the IEEE InternationalConference on Computer Vision (ICCV).Aleksander, I. 2008. Machine consciousness. In
Scholarpedia . (2):4162.Anderson, P. and He, X. and Buehler, C. and Teney, D. and Johnson, M. and Gould,S. and Zhang, L. Bottom-Up and Top-Down Attention for Image Captioningand Visual Question Answering. Conference on Computer Vision and PatternRecognition (CVPR).Baudrillard, J. 1994. Simulacra & Simulation . University of Michigan Press.Baz, I., Yoruk, E., Cetin, M. 2016. Context-aware hybrid classification system forfine-grained retail product recognition. In
Image, Video, and MultidimensionalSignal Processing Workshop (IVMSP), 2016 IEEE 12th IEEE . 1–5.Belohl´avek, R. 2004. Concept lattice and order in fuzzy logic.
Annals of PureApplied Logic . 128(1-3):277–298.Bono, E. D. 1969.
The Mechanism of Mind . Jonathan Cape.Box, G. E. P. 1976. Science and Statistics. In
Journal of the American StatisticalAssociation . (356):791-–799.Chella, A., Dindo, H., Infantino, I.. 2006. A cognitive framework for imitationlearning. In The Social Mechanisms of Robot Programming by Demonstration,Robotics and Autonomous Systems . 5:403—408.Chella, A., Manzotti, R. 2009. Machine Consciousness: A Manifesto for Robotics.In
International Journal of Machine Consciousness . (1):33—51.Chella, A., Manzotti, R. 2011. Artificial Consciousness. In Perception-Action Cycle .637—671.Chalmers, D. J. 1996. The Conscious Mind: In Search of a Fundamental Theory.New York, Oxford University Press.Cohen, P. 1997. Projections as Concepts. Computer Science Department FacultyPublication Series (194), https://scholarworks.umass.edu/cs/_faculty/_pubs/194 . Last accessed 18 May 2020.Cui, P., Wang, X., Pei, J., Zhu, W. 2018. A survey on network embedding.
IEEETransactions on Knowledge and Data Engineering .D¨untsch, I., Gediga, G. 2002. Modal-style operators in qualitative data analy-sis. In
Proceedings of the 2002 IEEE International Conference on Data Mining ,Washington, D.C. 155—162. Springer.Franco, A., Maltoni, D., Papi, S. 2017. Grocery product detection and recognition.
Expert Systems with Applications . 81:163—176.ebruary 12, 2021 1:30 MetamodelAGI
A Metamodel and Framework for Artificial General Intelligence G¨ardenfors, P. 2000. Conceptual Spaces: the Geometry of Thought. MITPress.Cambridge, Massachusetts.George, M., Floerkemeier, C. 2014. Recognizing products: A perexemplar multi-label image classification approach. In
Proceedings of the European Conferenceon Computer Vision . 440–455. Springer.Goertzel, B. 2009. OpenCogPrime: A Cognitive Synergy Based Architecture forGeneral Intelligence. In
Proceedings of the International Conference on CognitiveInformatics , Hong King.Goertzel, B., et al. 2013. The CogPrime Architecture for Embodied ArtificialGeneral Intelligence. In
Proceedings of the IEEE Symposium on Human-LevelAI , Singapore.Gorban, A. N., and Tyukin, I. Y. 2018. Blessing of dimensionality: mathematicalfoundations of the statistical physics of data.
Philosophical Transactions of theRoyal Society A: Mathematical, Physical and Engineering Sciences , 376.2118.Grover, A., Leskovec, J. node2vec: Scalable Feature Learning for Networks. ACMSIGKDD Internation Conference on Knowledge Discovery and Data Mining(KDD).Hamilton, W. L., Ying, R., Leskovec, J. 2017. Representation learning on graphs:Methods and applications.
IEEE Data(base) Engineering Bulletin . 40:52-–74.Hammer, P., Lofthouse, T., Fenoglio, E., Latapie, H. 2019. A reasoning based modelfor anomaly detection in the SmartCity domain. In
NARS Workshop at AGI-19 , Shenzhen, China. https://cis.temple.edu/tagit/events/papers/Hammer.pdf.Last accessed 19 May 2020.Hobbs, J. R. 1985. Granularity. In
Proceedings of the 9th International JointConference on Artificial Intelligence . 432–435.Horowitz, A. 2017. Smelling themselves: dogs investigate their own odours longerwhen modified in an olfactory mirror test.
Behav. Proc. . 143:17–24.Johnson, M. 1987.
The Body in the Mind . University of Chicago Press.Jinhai, L., Mei, C., Xu, W., Qian, Y. 2015. Concept learning via granular comput-ing: A cognitive viewpoint.
Information Sciences . 298:447—467.Kiryati, N., Eldar, Y., Bruckstein, A. M. 1991. A probabilistic Hough transform.
Pattern Recognition . 24(4):303–316.Korzybski, A. 1921.
Manhood Of Humanity, The Science and Art of Human Engi-neering . NY: E. P. Dutton & Company.Korzybski, A. 1949. Videoswith Alfred Korzybski. . Lastaccessed 14 Dec 2020.Korzybski, A. 1994.
Science and Sanity: An Introduction to Non-Aristotelian Sys-tems, 5th Edition . NY: Institute of General Semantics.Lakoff, G. 1984.
Women, Fire, and Dangerous Things . University of Chicago Press.LeCun, Y. 2020. Facebook’s chief AI scientist says GPT-3 is ‘not very good’as a dialog system. https://thenextweb.com/neural/2020/10/28/facebooks-yann-lecun-says-gpt-3-is-not-very-good-as-a-qa-or-dialog-system/ . Last accessed 13 Decebruary 12, 2021 1:30 MetamodelAGI Latapie et al., 2020
Biologically, InspiredCognitive Architectures . :1-–9.Ma, J. M., Zhang, W. X., Leung, Y., Song, X. X. 2007. Granular computing anddual Galois connection. Information Sciences . 177(23):5365-–5377.Murahari, V. and Chattopadhyay, P. and Batra, D. and Parikh, D. and Das, A..Improving Generative Visual Dialog by Answering Diverse Questions. Proceed-ings of the Conference on Empirical Methods in Natural Language Processing(EMNLP).Pauli, W. 1935. Part I. General: (A) theory. Some relations between electro-chemical behaviour and the structure of colloids.
Transactions of the FaradaySociety . :11-–27.Scarselli, F., Gori, M., Tsoi, A. C., Hagenbuchner, M., Monfardini, G. 2009. Thegraph neural network model. IEEE Transactions on Neural Networks. IEEETransactions on Neural Networks . 20(23):61—80.Searle, J. R. 1992. The Rediscovery of the Mind Cambridge (Mass), MIT.Speer, R., Chin, J., Havasi, C. 2017. ConceptNet 5.5: An Open Multilingual Graphof General Knowledge. In
Proceedings of the AAAI 31 .Swainson, B. 2020. The ExponentialInternet. . Lastaccessed 13 Dec 2020.Tan, M., Pang, R., Le, Q. V. 2019. EfficientDet: Scalable and Efficient ObjectDetection. https://arxiv.org/pdf/1911.09070v7.pdf . Last accessed 27 Jul 2020.Taylor, J. G. 2007. CODAM: A neural network model of consciousness. In
NeuralNetworks . 20:983–992.Taylor, J. G. 2009. Beyond Consciousness?. In
International Journal of MachineConsciousness . (1):11-–22.Th´orisson, K. R. 2007. Integrated A.I. Systems. Systems. Minds & Machines .17:11–25.Th´orisson, K. R. 2012. A New Constructivist AI: From Manual Construction to Self-Constructive Systems. In: P. Wang & B. Goertzel (Eds.)
Theoretical Foundationsof Artificial General Intelligence . 145–171. New York: Springer.Th´orisson, K. R., Bieger, J., Li, X., Wang, P. 2019. Cumulative Learning. In
Proceedings of 12th International Conference on Artificial General Intelligence(AGI-19) , Shenzhen, China, 198–209.Tonioni, A., Di Stefano, L. 2017. Product recognition in store shelves as a sub-graph isomorphism problem. In
International Conference on Image Analysis andProcessing , 682-–693.Unger, R. M., Smolin, L. 2015.
The singular universe and the reality of time .Cambridge University Press.Unger, R. M. 2014. Roberto Unger: Free Classical Social Theory from Illu-sions of False Necessity.
Online Lecture . Retrieved on 13 December 2020 fromebruary 12, 2021 1:30 MetamodelAGI
A Metamodel and Framework for Artificial General Intelligence Wang, L. D., Liu, X. D. 2008. Concept analysis via rough set and AFS algebra.
Information Sciences . 178(21):4125-–4137.Wang, P. 2005. Experience-grounded semantics: a theory for intelligent systems.
Cognitive System Research . 6:282—302. doi:10.1016/j.cogsys.2004.08.003Wang, P. 2006.
Rigid Flexibility: The Logic of Intelligence . Dordrecht: Springer.Wang, P. 2010.
Non-axiomatic logic (nal) specification . Retrieved on October 22,2020 from: https://cis.temple.edu/ pwang/Writing/NAL-Specification.pdfWang, P. Insufficient Knowledge and Resources—A Biological Constraint and ItsFunctional Implications In
Proceedings of 2009 AAAI Fall Symposium Series .Wang, P. 2019. On Defining Artificial Intelligence.
Journal of Artificial GeneralIntelligence . 10(2):1–37.Wang, P., Li, X., Hammer, P. 2018. Self in NARS, an AGI System.
FrontiersRobotics and AI . 5(20). doi: 10.3389/frobt.2018.00020Wang, A., Pruksachatkun, Y., Nangia, N., Singh, A., Michael, J., Hill, F., Levy, O.,Bowman, S. R. 2019. SuperGLUE: A Stickier Benchmark for General-PurposeLanguage Understanding Systems. https://arxiv.org/pdf/1905.00537.pdf . Lastaccessed 13 Feb 2020.Wikipedia. 2019. Wheat and chessboard problem. https://en.wikipedia.org/wiki/Wheat and chessboard problem . Last accessed 14Nov 2020.Wille, R. 1980. Restructuring lattice theory: an approach based on hierarchies ofconcepts. In: I. Rival (Ed.),
Ordered Sets , 445–470. Dordrecht-Boston: ReidelYao, Y.Y. 2001. Information granulation and rough set approximation.
Interna-tional Journal of Intelligent Systems . 16:87–104.Yao, Y.Y. 2009. Integrative levels of granularity. In Bargiela A and Pedrycz W(Eds.),
HumanCentric Information Processing Through Granular Modelling . 31–47. Berlin: Springer.Yao, Y.Y., Deng, X.F. 2012. A Granular Computing Paradigm for Concept Learn-ing. In Ramanna, S., Jain, L., Howlett, R.J. (Eds.),
Emerging Paradigms in Ma-chine Learning . 307–326. London: Springer.Ying, R., He, R., Chen, K., Eksombatchai, P., Hamilton, W. L., Leskovec, J. 2018.Graph convolutional neural networks for web-scale recommender systems. In