This article first defines calculation, then introduces its research significance and development, discusses the differences and connections between calculation and calculation, and proposes a calculation-calculation model. Finally, it describes its application and summarizes it.
Nowadays, humans have not yet given a precise definition of artificial intelligence, but a definition that is in line with most people’s views refers to Artificial intelligence is about letting computers complete various things that the human mind can do. From Turing's prediction of artificial intelligence in the 1940s, to the development of expert systems in the 1980s, and now the application of artificial intelligence in the three major branches of thought has penetrated into all aspects of human life. Among them, connectionism emphasizes imitating the cerebral cortex neural network and the connection mechanism between neural networks, that is, using a multi-hidden layer processing structure to process various big data; behaviorism mainly imitates human or biological individuals and group control behavioral functions. It is manifested as a reinforcement learning method with a reward and punishment control mechanism; symbolism emphasizes the use of physical symbol systems to generate intelligent behaviors, and is mainly used in the application system of knowledge graphs.
The application of artificial intelligence has achieved certain results, but it has flaws that cannot be ignored. The deep learning algorithm under connectionism is not differentiable, has weak computational convergence, and has poor results in an open dynamic environment. The model itself is a "black box". Behavioral reinforcement learning oversimplifies the human behavior process, and only measures simple reward and punishment feedback processes in experiments. Secondly, behaviorism studies observable behaviors, but often ignores the internal activities of the mind, denies the importance of consciousness, and opposes consciousness to behavior. Symbolism and its knowledge graph encounter the problem of how to define "common sense" and the problem of knowledge representation and problem solving of uncertain things.
To make artificial intelligence close to the human mind, we need to explore what intelligence is. Most definitions of intelligence have one thing in common: Intelligence is the ability to solve problems, and more complex problems require higher levels of intelligence. Compared with addition and subtraction, solving differential equations requires a higher level of intelligence; compared with tic-tac-toe, playing Go well requires a higher level of intelligence. However, just because a machine can solve a specific problem does not mean that the machine has a high level of intelligence, even if the problem is very complex. Turning our attention to human neural networks. Neural networks are often simplified into system units composed of perceptrons, centers, and effectors, and each of the three systems is composed of many neurons with feedback from each other.
With the current neuroscientific analysis methods, this model is correct, but people often ignore the ratio of external information to internal information. People often simplify this system into an input-output system. In fact, the nervous system has 100,000 times more receptors for receiving internal information than for external information. In other words, the nervous system as a whole is more like a self-enclosed system. Moreover, the closed nature of formal neural networks allows constructivism to deduce the philosophical theory that objective existence cannot leave the constructor. Machines exist objectively and cannot solve problems without humans. Therefore, it is unrealistic to realize a general artificial intelligence system that is completely independent of humans. This article first defines calculation, then introduces its research significance and development, discusses the differences and connections between calculation and calculation, and proposes a calculation-calculation model. Finally, it describes its application and summarizes it.
Calculation is a method that uses a mixture of perceptual and rational means to deal with the mixed relationship between various facts and values. Even without data, deep situational awareness can be clearly understood at a glance . Calculation can be manifested in many aspects and processes: calculation is the behavior before calculation. Calculation is a non-spontaneous order, which is produced by human design in advance. It is an order produced by human design rather than human behavior. Calculation is a calculation without numbers and diagrams, that is, there is no mathematics. Calculation is a mixed calculation of rationality and sensibility of human beings with motivation, and it is an existing logic. Integration planning of formal and unknown logical forms.
If the future of “computing” lies in utilizing the complex physical behaviors of the universe, then the future of “computing” lies in utilizing the complex physical and non-physical behaviors of the universe. Different from machine calculations, human calculations are complex, with both a rational part that embodies facts and a perceptual part that embodies values. Moreover, the perceptual part can be inconsistent, contradictory (such as love and hate), or even dialectical. (Mutual transformation), therefore, perceptual value is one of the most important differences in intelligence between humans and machines. Of course, the rational facts and rational values between humans and machines are not completely equivalent. Specifically, the one-many relationship of humans and the one-many structure of machines are often not the same thing.
Cognition can also be called cognition, which refers to the process of people understanding external things, or the effect on people The process of information processing of external things by the sensory organs. It includes feeling, perception, memory, thinking, imagination, and speech. It refers to the process of people's cognitive activities, that is, the individual's reception, detection, conversion, reduction, synthesis, encoding, storage, extraction, reconstruction, concept formation, judgment and Information processing for problem solving. In psychology, it refers to the process of acquiring knowledge through mental activities such as forming concepts, perceptions, judgments, or imaginations, that is, the psychological function of individual thinking for information processing.
The cognitive model is a computational model that simulates human cognitive processes based on the understanding of human cognitive abilities. Cognitive abilities in the cognitive model here usually include aspects such as perception, representation, memory and learning, language, problem solving and reasoning. In order to build more intelligent machines, we hope to find inspiration from people, and at the same time better explore and study the human thinking mechanism, especially the human perception and processing mechanism of surrounding information, so as to create real machines. Artificial intelligence systems offer new architectural and technical approaches. Duch divides existing cognitive models into three categories based on differences in memory and learning: symbolic cognitive models, emergent cognitive models, and hybrid cognitive models.
People often feel "state" and perceive "potential". However, the state and potential of the world are often mixed, so people's understanding of the world is often Perception, so situation awareness became a fixed combination. Later, it was discovered that the corresponding "situation" is generally short-term. In order to deal with medium and long-term trends, some deeper insights are formed, that is, the ability to understand, judge, and predict beyond the facts themselves. , situational cognition is ready to come out. It can also be said that cognition is the perception of perception. Later, it was discovered that human cognition has limitations such as preferences, habits, priors, fuzziness, and memory, and the mechanisms of machines and collaboration can complement each other and learn from each other's strengths. Naturally, deep situational awareness of human-machine integration has been derived. knowledge concept.
The meaning of deep situational awareness is "the perception of situational awareness, which is a kind of human-machine intelligence that includes both human intelligence and machine intelligence (artificial intelligence)". It is the signifier. It involves both the attributes of things (signifiers, feelings) and the relationship between them (referent, perception). It can not only understand the original meaning of things, but also understand the implication. It is based on the situational awareness (including information input, processing, and output links) with Endsley as the main body, plus the overall system trend analysis of people, machines (things), environment (nature, society) and their interrelationships, and has " "Soft/hard" two regulatory feedback mechanisms; including self-organization and self-adaptation, as well as other organization and mutual adaptation; including both local quantitative calculation predictions and global qualitative calculation evaluations. It is an autonomous and automatic Information correction and compensation expectation-selection-prediction-control system for convergence effect.
In a sense, deep situational awareness is to organize the system in a specific environment to fully utilize various human cognitive activities (such as purpose, feeling, attention, motivation, prediction, automaticity) to complete the subject task. , motor skills, planning, pattern recognition, decision-making, motivation, experience and knowledge extraction, storage, execution, feedback, etc.). It can operate both in situations of insufficient information and resources and in situations of overload of information and resources.
Through experimental simulations and on-site investigation and analysis, we believe that there is a "leaping frog" phenomenon (automatic reaction) in the deep situational awareness system, that is, from the information input stage directly to the output control stage (skipping information processing integration stage), which is mainly caused by the clarity of the task theme, the concentration of organizational/individual attention and the conditioned habit reflex of long-term targeted training. Just like a person chewing gum, chatting, holding an umbrella and walking, he can unconsciously coordinate various tasks. A natural order of activity, the system performs near-perfect automatic control rather than conscious rule-conditioned responses. Deep situational awareness is essentially the process of creating and resolving many paradoxes such as change and immutability, one and many, autonomy and passivity. Therefore, the system should not be a simple human-computer interaction but an autonomous cognitive process (including expectations, choices, control, and even emotional fields) throughout the entire human-computer environment system. In view of the wide scope of research on deep situational awareness systems, system characteristics such as nonlinearity, randomness, and uncertainty are easily produced, so system modeling research often faces greater difficulties.
Human symbols, connections, behaviors, and mechanismism are multi-level and multi-angle, and even change levels and angles. In contrast, machine symbols, connections, behaviors, and mechanismism are Single layer, single angle and fixed layer and fixed angle. The essence of human thinking is a program that adapts to changes and can be created in real time. It can explain the connections between symbolism, connectionism, behaviorism, and mechanismism, and can open up these connections to achieve comprehensive processing. Davenport believes that once a certain kind of human intelligent behavior is broken down into clear steps, rules and algorithms, it is no longer exclusive to humans. How scientific discovery becomes a question that can be studied. The problem of human-machine hybrid intelligence is that the higher the degree of autonomy of the machine, the lower the human's awareness of the situation, and the greater the difficulty in smoothly taking over tasks between humans and machines. It can be called "reduced physiological load and increased psychological cognitive load" "Phenomenon.
Computing is a complex "computing" process of cross-domain multi-source heterogeneous systems that humans do without the help of machines. In a certain sense or degree, calculation is the interactive balance of the two "neuromorphic" processes of integration of viewing and performance and "integration of storage and calculation". Viewing (existence) is a non-real-time TOP-DOWN process that increases the scale or granularity. Calculation is a real-time bottom-up process of small-scale fine particles. Beyond leaping thinking, the human mind is not symbolic in nature and therefore cannot be calculated. The human brain is not a computer and has both physical and non-physical physiological and psychological properties. Meaningful values can be hatched from meaningless facts, and meaningless facts can be produced from meaningful values. This mixture of subjectivity and objectivity determines the computational characteristics of the mind, that is, limited rational calculation and infinite perceptual calculation coexist. For example, human innovation's "leap-forward" thinking is not based on calculation, that is, thinking that often does not follow language and logic. Therefore, artificial intelligence that is completely based on machines may not be able to have "leap-forward" thinking, so it is unlikely to have Creativity in a truly non-closed and open environment.
Various concepts and propositions in the real world have various combinations of fluidity and flexibility. Calculation is not symbolic, but procedural, and is also the manifestation process of consciousness. Consciousness may be the "explicitness" of many "hidden" things, the manifestation of implicit states and hidden potentials, the manifestation of implicit feelings or implicit knowledge, the manifestation of implicit attention and implicit memory, the manifestation of implicit judgment and implicit reasoning. , concealing analysis but concealing the manifestation of decision-making, concealing the manifestation of facts and implicit values, concealing human feelings and concealing the manifestation of physics. Eastern calculations used to be mainly based on calculations based on human relations and sophistication and management, but now they are incorporating new calculations based on physics, mathematics, legal principles, etc.
Now, more and more people believe that various algorithms must transcend superficial correlations and reach the level of true understanding, thereby achieving a higher level of human-machine fusion intelligence. Situation and potential are two connected entities. Situation and potential reflect the changing direction of two different points. It is more accurate to use calculation than calculation.
On this basis, comparing the calculation differences between the East and the West is like what Jacque Derrida, the father of deconstruction, said: the presence or absence of logical rationality. Derived from the civilizations of North Africa and Central Asia, the West discovered the power of science and technology and invented a series of related subject areas, forming a worldview and values with "calculation" as the core. Based on it and using logic and rationality as a tool, it has restored many laws of the material world, economic phenomena, and natural selection to human society, and has made great contributions. However, in recent times, many insightful people in the West, while fully exploiting the advantages of logic and rationality, have increasingly felt the limitations and deficiencies of logic and rationality, and have consciously or unconsciously turned their attention to the East, which they consider to be "mysterious" Wisdom, from physics to psychology to management, etc., from early Leibniz to Joseph Needham to Hofstadter and others, the "calculation" of Eastern thought and the "calculation" of the West are a good pair, and they are also qualitative and The perfect combination of quantification, subjectivity and objectivity, value and fact, system and reduction. It can be said: "There is no higher level than people, and there is no better calculation than calculation."
The ontology of calculation is a factual concept, and the ontology of calculation is value preference. The subject of calculation is people, and the subject of calculation is the system containing people. The subject of calculation is changeable, but the ontology remains unchanged; the subject of calculation remains unchanged, but the ontology is constantly changing. Computation uses parametric modeling, and computation creates parametric modeling. Calculation is often carried out in the order of sensing-storing-calculating-transmitting-using-feeding-evaluating, but calculation often breaks the order combination of sensing-storing-calculating-transmitting-using-feeding-evaluating according to the specific situation. -Calculate, you can also feel-calculate-evaluate for a while. For calculations, if objective facts are input, then deterministic objective facts will be output. It can be said to be based on real evidence, seeking truth from facts, and is a logical reasoning of rational being. This is not the case for calculations. Even if objective facts are input, it is not necessarily true. It will output deterministic objective facts, that is, the real input can be subjectively changed to choose and output value. Seeking truth from facts is a non-logical realization of the perceptual should. For example, enter 23, which can be Jordan, James, etc.
In the real game process, on the surface it is a rational process of mathematical calculation. In fact, there is also a perceptual process of calculation. To be more precise, it is a process of calculation. That is, the two sides are not only fighting rationally, but also having conflicts. The turmoil of a large number of perceptual factors is the battle of arts and sciences in Huashan where facts and values are mixed and hinged.
The basis of calculation is limited closure, and the characteristic of calculation is conditional openness. Calculation is to deduce infinite subjective value Should (possibility) from the limited objective fact Being (reality). Calculation is reasoning with certainty, calculation is reasoning with uncertainty, and calculation is a mixture of certainty and uncertainty. Machines only have local factual logic and do not have the overall value logic of humans. Therefore, humans and machines are combined to complement each other in functions and abilities, and use the sharp edge of human calculations to penetrate various "walls" that machine calculations encounter from time to time. . The valuable thing in human-machine integration is that the fact process becomes something meaningful to intelligent logic through the dynamic environment. Facts are not what they are because of what they are, but what they are in the integration with value. , which requires the establishment of a new logic system to support it, that is, the computing-computing logic system of human-computer integration.
So, what is calculation? What is calculation? Calculation is logic that starts from known conditions and solves "complexity". Calculation is intuition starting from unknown premises and handles "complexity". There are two words at the core of the calculation: "different" and "yi".
In calculation, the change and balanced processing of things in different fields is the core of calculation, and calculation requires the same structure, the same data, and the same properties to be carried out. The calculated results are often Unchanging and certain.
In contrast, people focus on value logic, machines tend to factual logic, people side on dialectical logic, and machines tend to formal logic. Different from machine calculations, human calculations are complex, with both a rational part that embodies facts and a perceptual part that embodies values. Moreover, the perceptual part can be inconsistent, contradictory, and even dialectical and mutually transformable, so the perceptual part Value is one of the most important differences in intelligence between humans and machines. Of course, the rational facts and rational values between humans and machines are not completely equivalent.
Specifically, the one-many relationship of humans and the one-many structure of machines are often not the same thing. Then human-machine integration is dialectical formal logic or formal dialectical logic, which involves a difficult problem of logical transformation, that is, how to transform factual formal logic into value dialectical logic, or how to transform value dialectical logic into factual formal logic. . On the surface, human dialectical logic is used to think about problems rather than solve problems. Solving problems depends on formal logic. In fact, this is due to a unclear understanding of formal computational logic and dialectical computational logic. Different from computational thinking, computational thinking is contrary to computational logic in many aspects.
The relationship between calculation and calculation is also inseparable. In the process of calculation, calculation is needed to guide the direction. In the process of calculation, calculation may be used as a basis to complete basic work. Both are indispensable. The ontology of calculation is a factual concept, and the ontology of calculation is a value preference. The main object of calculation, the subject of calculation, is the system that includes humans. Calculation cannot change factual concepts, but it can change the person who operates it; the system of people in calculation cannot be changed, but value preferences often change. Therefore, only the combination of the two can achieve better intelligence.
Artificial intelligence has achieved remarkable results, but the current artificial intelligence body is still far from reaching a level close to the human mind. In the face of complex environments, the level of artificial intelligence in the computing system is limited and cannot give full play to its characteristics. Intelligence is a complex system. In the era of pursuing computing power and algorithms to realize artificial intelligence applications, the role of humans in cooperation with intelligent agents cannot be ignored. The collaborative system between the machine's ability value (calculation) and the human ability value (calculation) still needs to be studied. This article analyzes the computing logic of machines and the cognitive ability of human "computing" from different angles, explores its capabilities and shortcomings, and proposes a computing-computing model to provide a feasible architecture for human-machine hybrid intelligence.
This article proposes a computing-computing model based on existing results in the computing and cognitive fields. The model includes the situation awareness layer, the cognitive decision-making layer, and the target behavior layer. The computational-computational model is constructed in three levels.
The situational awareness layer contains environmental information. The environment includes the environment within natural time and space and social time and space. The source of all problems comes from nature and society. It can also be said that the source of human knowledge is also the same. This environment contains certain and uncertain components. In addition, the situation awareness layer is also responsible for the collection and perception processing of situation information. Calculus in mathematics is similar to this process, where known data is manipulated to get closer to the answer to a problem. The field of intelligence collection and analysis in the field of command and control is also about the perception and processing of information. Compared with numbers and calculus symbols, the quantification of intelligence is more complicated and is more often handled by experienced commanders.
Traditional automation methods and machine learning algorithms may lead to "out-of-the-loop" errors because humans have low situational awareness of tasks and because humans' perception of tasks and environments are largely based on experience, making it easy to become complacent. Mood or lack of vigilance. Uncertain feedback from the environment can also have an impact on problems outside the loop, highlighting the importance of achieving a balance between tightly coupled and loosely coupled human-machine-loop interactions. Nowadays, the interpretability of data processed by algorithms has declined, leaving humans trapped "outside the loop". It has also caused the problem of declining trust in intelligent agents.
The cognitive decision-making layer is similar to the deep processing of situational awareness information. It not only depends on human cognition in the traditional sense, but also requires the ability of machine reasoning. Human reasoning is based on cognitive abilities such as intuition, logic, and correlation. The idea of calculation is included in it. Expert systems were popular in the 1980s, and expert systems based on artificial intelligence were all the rage. Machines could perform simple questions and answers, but the content of the questions and answers was highly limited. In addition to computing power, the main reason is the reasoning ability granted to machines based on one-to-one, one-to-many, and many-to-one knowledge mapping relationships. How to achieve elastic reasoning under dynamic representation is an issue worthy of consideration if sufficient hardware computing power is available. And scattering, diffusing, and insinuating the knowledge under dynamic representation to achieve multi-hop reasoning is the key to solving this problem.
Fuzzy logic provides a framework for mapping one or more continuous state variables to corresponding categories for reasoning and decision-making; the neural network uses programs to perform variable interconnection weights learned during the training process of a large database of typical cases. An arithmetic framework for knowledge expression; genetic and evolutionary algorithms, inspired by evolutionary genetics, use repeated simulation methods to narrow the range of potential options and select the optimal solution. Humans interact with machines through human-computer interaction interfaces, and machines assist humans in decision-making. Realizing the humanization of machine parts is a big step in the integration of man and machine.
The target behavior layer is mainly reflected in human-machine hybrid decision-making. When higher-level intelligence appears, humans will always be at the top of decision-making. This is a consensus reached in many fields (except under conditions of low human cost in goal pursuit). Sensemaking is critical to survival in the real world, and many studies show that humans strive to construct meaning from objects, events, and situations in the world. Humans are better at interpreting signs. For humans, construction is a psychological activity. In order for machines to effectively cooperate with humans, the reactions and decisions of machines and humans should interpret the same objects, events, or situations as the same signs, or obtain the same meaning. Machine-assisted decision-making also needs to provide a more complete meaning construction when necessary to achieve the goal of human-machine hybrid decision-making.
One of the most cutting-edge issues in current human-computer interaction is the interaction between humans and artificial intelligence, but we still don’t have Seeing the light before dawn, the reason is that there is only "calculation" between man and machine, but no "calculation". To simplify it a bit, there is only "calculation" and no "calculation".
The "interaction" in human-computer interaction can be divided into: state-based interaction, potential-based interaction, sense-based interaction, and knowledge-based interaction. Another classification is: fact-based (data ) and interactions based on (subjective) value, as well as interactions based on fact-value hybridity. Calculations in human-computer systems are calculations that humans do without mathematical models. Everyone is familiar with the "AND or NOT" logic in scientific and technological calculations, so I won't go into details here. The logic in science fiction calculations might as well be called "right and wrong". Among them, "yes" means assimilation, "non" means accommodation, and "should" means balance. When encountering future science fiction problems, use "yes" first, then "non", and then "should". When there is a big right or a big wrong, if the big right does not move, try the small right first, and then try the middle right. If it doesn't work, don't move the big wrong, try the small right first, and then try the middle right. The process of these trials is the balance of "middle". "Should" means constantly trying, adjusting, and balancing. The above is a new logic system that combines scientific and technological computing with human-computer computing. Human-computer computing logic grasps the direction of value and emotion, while scientific and technological computing logic refines the fact-based rational process.
Machines are the crystallization of physics and mathematics, the environment is the product of geography and history, and the human-machine system represented by the interaction of human-machine environment systems is a representation of complex forms and simple laws. The mission of science fiction lies in the future, so it must shoulder the important responsibility and responsibility of "the Tao is not Tao, and the name is not name". The existing scientific and technological logic system is gradually showing signs of fatigue with the in-depth development of various disciplines, such as Gödel's incompleteness law in mathematics, Heisenberg's principle of unmeasurability in physics, and Arrow's impossibility theorem in economics. The times are calling. New scientific principles and new technical means all require the emergence of a new logical system, a non-existent existence that is different from the traditional way of thinking and more consistent with objective facts and subjective values.
Philosophically speaking, the objective world exists completely independent of the subjective world, but this is a false proposition and does not really exist. What can be truly observed is the combination of objectivity and subjectivity. Due to the interaction between the observer and the observed world, we cannot understand the objective world with infinite accuracy. For example, when hardware capabilities reach a certain level, people will pay attention to the improvement of software capabilities. When software capabilities reach a certain level, people will pay attention to the improvement of human software capabilities. To measure the level of a technological product, you can try to make a preliminary judgment based on the speed and accuracy of its ability to "cross" and "collaborate" in different fields. In the same way, to measure the rationality level of a human-computer system, you can try to judge it based on its "crossing" capabilities. , "associate" the speed and accuracy of abilities in different fields to make preliminary judgments.
Simply put, the way science and technology handles problems is generally a production-based "if-then" causal relationship, while human-computer systems should be more heuristic and do not seek optimality but seek satisfaction. And heuristics can often handle nonlinear problems. When humans and machines are powerless in an extremely complex environment, it may be the time when logic collapses. Perhaps, heuristic calculations (calculations) in the human-computer field are precisely the key to the success of new logic compression.
One of the bottlenecks and difficulties in the human-machine field is the problem of multi-domain imbalance in the human-machine environment system, which is embodied in the problem of how to effectively realize "cross-over" and "collaboration" in different subject areas. This is not only related to Solving the shortcomings of "state but no potential" or even "no state but no potential" in various human-machine construction systems will also involve the source of "inefficiency and incompetence" of many human-machine environmental systems. Try to organically combine the basic theoretical domain composed of the humanities domain, art domain, and social domain with the scientific and technological domain composed of the physical domain, information domain, and cognitive domain, in order to achieve true "cross-over" and effective human-machine cross-domain collaboration. "Xie" lays the foundation.
At present, in terms of objective reality and practical applications, we must clearly realize that the current wave of human-computer environment systems is not based on a major breakthrough in the understanding of the mechanism of scientific and technological development, but only on finding a better way. The powerful method that can take advantage of the current imperfect technology and computer expertise - old logic and old technology, may not be able to pry open the door to create real illusions. Most of the "human machines" in people's minds are inseparable from experiences in understanding, intentionality, will, emotion, self-awareness and spirit. There have been no breakthroughs in research in these areas so far, and there are no signs of major breakthroughs in the near future. Moreover, relying on technology is poor in interpretability, common sense, learning and visualization, and has limited help in deepening understanding, raising awareness, and improving human-machine structure and expression.
Therefore, the traditional logical approach based on predetermined strategies and empirical judgments is no longer feasible. How to ensure that in a highly complex and high-load human-machine environment, when faced with insufficient key information, people must accurately obtain useful information from a large amount of situational information, form correct cognitions, and quickly and actively immerse themselves in becoming future people. problems that need to be solved urgently in the machine field. Perhaps the real mechanism capable of controlling human thought is fundamentally different from any conventional logic created so far, as Einstein said: "When mathematics speaks of reality, it is indeterminate; when mathematics is certain, it has nothing to do with reality", All human experience and information are contained in the future human-computer environment system interaction relationships and new logical practices.
Although artificial intelligence is more and more widely used in specific fields, people’s call for general artificial intelligence is becoming more and more obvious. No longer satisfied with the limited capabilities of weak artificial intelligence. With the development of computers today, computing power has been greatly improved, and quantum computers will increase computing power by several orders of magnitude. So we turned to thinking about the nature and origin of human intelligence, trying to find inspiration from cognitive neuroscience and other aspects to understand consciousness, hoping to give artificial intelligence autonomous consciousness, but with little success. A more practical and feasible solution is the integration of human-computer interaction, which fully integrates people's conscious thinking, that is, the logical calculation of computing and machines, and allows the machines to learn and understand human computing capabilities during the integration. Human-machine fusion intelligence will open a new era of intelligence.
What is human-machine fusion intelligence? Simply put, it focuses on describing a new form of intelligence produced by the interaction of human, machine, and environmental systems and making full use of the strengths of humans and machines. It is neither human intelligence nor artificial intelligence. Human-machine fusion intelligence is not a simple combination of man and machine, but allows the machine to gradually understand human decision-making, and allows the machine to gradually understand the difference in value weights based on human decision-making under different conditions. People form cognition through their perception of the surrounding environment and their own desires and impulses, while machines can only obtain data and information about the surrounding environment, trigger specific execution processes through specific data, and combine human cognitive abilities with the computing power of machines. Integrate together to establish new ways of understanding, and then make purposeful and regular decisions, producing an effect where human-machine integration is greater than human-machine integration.
Artificial intelligence is only a describable and programmable part of human intelligence, and human intelligence is the product of the interaction between humans, machines (things), and environmental systems. The matching classification of human-machine functional capabilities can be classified according to all artificial intelligence, human-dominated machine data assistance, human-dominated machine regularized computational reasoning assistance, human-dominated machine probabilistic computational reasoning assistance, human-dominated machine weak judgmental decision-making assistance, human-hosted machine Machines have strong judgment to assist decision-making division, but in any case, humans are always in a dominant position in human-machine systems, that is, humans are assisted by the host throughout the entire process to avoid causing the system to get out of control.
How to realize the organic integration of man and machine? The current human-machine integrated intelligence is still in its infancy, and there are still some key issues that need to be solved. The most critical of these is how to organically integrate the cognitive capabilities of machines with the computing capabilities of machines. Human-machine systems currently in the practical application stage have a clear division of labor between humans and machines. The main task is that machines display data to humans and humans operate machines through interfaces, which does not produce effective integration. The human thinking and decision-making process is based on constantly receiving external information. Through perception, understanding, association and imagination, the external information is internalized into one's own knowledge or experiential common sense. This is also a learning process, and then one can use one's own experience when encountering problems. common sense, and analyze and understand problems with the help of inspiration and intuition. Machines do not have this ability to abstract and irrationally think about external information.
Therefore, the key to integration is to improve the machine’s knowledge understanding and learning capabilities at the cognitive level, so that humans and machines can communicate and integrate at the cognitive level. Human connection and imagination are represented by the abstraction and transfer of a problem or knowledge, so that problems in different fields can be solved cross-domain. The more abstract the representation, the more universal the transfer ability can be, and it can break through the limitations of thinking. Humans can achieve non-axiomatic intuitive awareness through the internalization of prior knowledge, while machines only process objective data under rational logic. Allowing machines to improve their cognitive abilities is the key to smooth human-machine integration.
Another key issue that needs to be considered in human-machine fusion intelligence is intervention, that is, the timing and method of human-machine fusion. How should the system make a decision when a person suddenly intervenes, or there is an asymmetry in the perception and processing of information about the surrounding environment between humans and machines, which leads to conflicts between the decisions made by humans and machines. When the decisions of humans and machines conflict, there will also be interpretable issues, that is, how one party can explain its decision and persuade the other party. In addition, in the fusion of humans and machines, the historical accuracy of machine decisions is not high due to insufficient external information or some of its own reasons. At this time, there will be problems with people's trust in machines. In some special scenarios such as military applications, decision-making errors may occur. Tolerance is very low. If in a general scenario, the machine's decision-making accuracy is very high, in the long run, it will lead to excessive dependence on people, which can easily lead to the loss of self-confidence, courage, courage and other excellent characteristics of human nature.
Difficulties in the mechanism of human-machine hybrid intelligence: the mechanism of calculation, the mechanism of intelligent calculation (calculation), the definition of basic concepts such as trust, understanding, intention, and adaptation in human-machine hybrid, the mechanism of human-machine hybrid intelligence The key to the demarcation, adaptation, and intelligent calculation of functions and capabilities lies in how to achieve the dialectical unity of calculation and calculation. As for counter-calculation, counter-calculation can lead to paradoxes and contradictions, and counter-calculation can endlessly loop. How do situations interact with each other, and how can perception be integrated? , how to learn to think like your enemy and make objective judgments and inferences.
The manifestation of human-computer integration is the way of human-computer communication, human-computer interaction interface, auxiliary decision-making and human-computer function allocation, etc. A high-speed and effective two-way information interaction relationship should be established between humans and machines to avoid problems such as information asymmetry. Machines are good at situationally sensing more quantitative and specific information from the surrounding environment. After processing, it should be presented to humans in a concise and intuitive way. The thinking and decisions made by humans should also be explainable and logically applied to machines. The existing human-machine function allocation used in aviation, nuclear power, air traffic control and other fields is mostly static allocation based on some automation level scales. Different functions are relatively reasonably allocated according to human and machine capabilities through a certain system. The reasonable allocation of human-machine functions can make full use of and combine the respective advantages of humans and machines to reflect the intelligence of human-machine fusion systems.
The fundamental reason why today’s artificial intelligence systems are far from meeting people’s expectations is that the basis for constructing artificial intelligence is contemporary mathematics rather than real intelligent logic. First of all, mathematics is not logic, from numbers to graphs and then to sets. , from arithmetic to calculus to category theory, everything is a mathematical logic system based on axioms, and real intelligent logic includes both mathematical logic and dialectical logic, as well as many undiscovered logical laws, which have not yet been discovered The logical laws of the future are both the source of future mathematics and the emergence of real-life logic. Real intelligence is never the product of a simple brain, but the product of the interaction and mutual stimulation and awakening of people, objects, and the environment, such as an intelligent system planned by a designer. It also requires the manufacturer to carefully understand the processing and implementation, and it also requires the user to adapt to local conditions, targeted and flexible applications, etc. Therefore, a good human-computer integrated intelligence involves the effective unity of opposites between three or even more, and has an objective fact state. The calculation, as well as the calculation of subjective value trends, is a deep situational awareness system of people, objects, and the environment. However, the current artificial intelligence, whether it is based on rule-based mathematical models or statistical probability, is mostly based on calculations, but lacks the combination and embedding of human calculations, and thus is far away from the reality and flexibility of intelligence.
This article first defines calculation, then introduces its research significance and development, and discusses the difference between calculation and calculation. Contact and propose the computing-computing model, and introduce the application of computing in human-computer interaction and human-computer integrated intelligence.
Human-computer interaction is an important part of the development of artificial intelligence, which requires not only new theoretical methods, but also new exploration of the relationship between humans, machines, and the environment. The popularity of artificial intelligence continues to increase, and more and more products are entering people's lives. However, strong artificial intelligence has still not been realized. How to transfer human computational intelligence to machines is a problem that must be solved. We have built cognitive models from the perspective of cognition or constructed conscious Turing machines from the perspective of consciousness. These are attempts to understand and simulate human cognitive thinking, hoping to realize human computing capabilities. Computational - The study of computational models not only needs to take into account the rapid development of machine technology, but also the way of thinking and cognition of human beings, the interacting subjects, so that machines and humans can perform their respective duties and integrate and promote each other. This is the prospect of human-computer interaction. and trends.
Human-machine hybrid intelligence, just like intelligence, is neither a product of the human brain or brain-like, nor a product of humans themselves, but a product of the interaction of people, objects, and environmental systems, as Marx said: " The essence of a person is not an abstraction inherent in a single person. In its reality, it is the sum of all social relations." For example, although the wolf boy has all the structures and components of the human brain, it does not communicate with the human social environment system. Or interact, it is impossible to have human intelligence and wisdom. In fact, the future man-machine also contains three components: people, objects, and the environment. With the rapid development of science and technology, the objects are gradually replaced by man-made machines, which are referred to as human-machine environment systems. To be fair, , It is basically impossible for man-machine to surpass the current level of science and technology based on the existing mathematical system and thinking model, but it is possible in the past, present and future man-machine environment systems. Technology is logical, but man-machine is not necessarily logical. Man-machine is a very vast space, which can open heterogeneous collections at any time and combine objective logic with subjective hyper-logic.
The above is the detailed content of A brief analysis of calculation and computation. For more information, please follow other related articles on the PHP Chinese website!
Application of artificial intelligence in life
What is the basic concept of artificial intelligence
Which is more difficult, c language or python?
Huobi.com
webstorm adjust font size
Where is the flashlight of OnePlus phone?
The relationship between bandwidth and network speed
Solution to gmail being blocked
![[Web front-end] Node.js quick start](https://img.php.cn/upload/course/000/000/067/662b5d34ba7c0227.png)
