Lei Feng network press: The author of this article Liu Wei, from Beiyou human and cognitive laboratory.
In a sense, human civilization is a process in which mankind continuously recognizes the world and himself. Cognition refers to the whole process of collecting, filtering, processing, predicting output, and adjusting feedback of useful data—information. , Looking at the earliest Mesopotamian civilization of mankind (more than 6,000 years ago), ancient Egyptian civilization (6,000 years ago) and the ancient Greek culture derived from it (the origin of modern Western civilization, dating back 3,000 years) Its essence reflects the relationship between people and objects (objective objects) . This is also the basis for the rapid development of science and technology.
The civilizations represented by ancient India often contained beliefs between people and gods. The ancient Chinese civilization ranked later was the only cultural pulse in the four major ancient civilizations that has been relatively complete. Its core principle reflects people. Communication with people, people and the environment (this may be the reason why the Chinese civilization continues to be an important reason).
Throughout the process of interaction between these people, machines, and the environment, the generation, circulation, processing, mutation, curling, amplification, decay, and disappearance of cognitive data are ongoing all the time... How to be full of this? In the process of variables, to maintain the stability and continuity of various possibilities? For this reason, people invented various theories and models, used many tools and methods, tried to find effective answers and universal laws in the order of nature and society. Until modern times, in the 16th century, a Catholic priest, Copernicus’s "Heart of the Sun", gradually transferred the authority of religion to science. Since then, experiments and logic have reconstructed a completely different space-time world over the centuries, reducing it again and again. People's physiological load, mental workload, and even mental workload...
With the continuous evolution of scientific thinking, there has been considerable technological advancement. The “Old Third Theory†(system theory, cybernetics, and information theory) has not faded, and the “new three theories†of dissipative structure theory, coordination theory, and catastrophe theory. It will come out on stage, electron tubes, transistors, and integrated circuits have not disappeared, and nanometer, supercomputers, and quantum communication technologies are eager to try. The artificial intelligence thoughts and technologies born in the 4th and 50th centuries were an important frontier that emerged in these basic fields.
However, due to the fuzzy of cognitive mechanism, the lack of mathematical modeling, the limitations of computing hardware and other reasons, artificial intelligence has not been able to quickly grow from small to large and from weak to strong. Judging from the current research progress in mathematics and hardware, it will be difficult to make breakthroughs in the short term. Therefore, how to open the breakthrough point from the cognitive mechanism has become the choice of many scientists. The purpose of this paper is to provide a preliminary introduction and review of in-depth situational awareness in order to promote the research and application of this field in China.
In June 2013, the U.S. Air Force Command officially appointed Mica R. Endsley, a woman scientist known for the Situation Awareness (SA), as the new chief scientist of the U.S. Air Force. This 1990 U.S. University of Industrial and Systems Engineering The professional graduated woman doctor Dr. and her previous appointment Mark T. Maybury (term 2010.10-2013.5, Ph.D. in 1991 graduated from Cambridge University Department of artificial intelligence of computer science department) all take the cognitive engineering in human-computer interaction as the research direction. Prior to September 2010, the chief scientist of the US Air Force was mainly based on the practice of aerospace or mechanical and electrical engineering.
This situation of appointing chief scientists in the context of cognitive science as a professional background is also quite popular in other branches of the U.S. military. This may mean that in the future trend of military-civilian technology development, the hardware manufacturing-oriented manufacturing and processing field is quietly giving way. The command and control system is based on the wisdom of software .
Coincidentally, just as professionals around the world, such as artificial intelligence and automation, have seriously studied Situational Awareness (SA) technology, the global computer community is working hard to analyze contextual awareness (CA) algorithms in the field of linguistics for natural language processing. The grammar, semantics, pragmatics, and other areas are also very enthusiastic. The discussion of situational awareness in the psychology department is also a lively place of the moment. The mainstream of Western philosophy is actually a philosophy of analysis (it is a philosophical school, its methods can be roughly divided into two Type: One is an analysis method of an artificial language, and the other is an analysis method of a daily language.) Of course, the main branch of cognitive science such as neuroscience is currently focused on brain consciousness.
All of us now live in an increasingly human-machine-environment (natural and social) system with increasingly active information. The command and control system is naturally regulated by the interaction of human and machine environments and the input, processing, output, and feedback of information. Ongoing topical activities, thereby reducing or eliminating the process of uncertainty in the results. Focusing on the core of the command and control system, Mica R. Endsley put forward a consensus concept about Situation Awareness (SA) at the 1988 International Human Factors Annual Meeting: “...theperception of the elements in the environment Within a volume of time and space, the comprehension of their meaning, and the projection of their statusin the near future.†Then change the situation ").
Specifically as shown below:
Figure 1 Dynamic Decision Situation Awareness (SA) Model (Endsley, 2000)
The model is divided into three levels. Each stage is prior to the next stage (necessary but not sufficient). The model follows an information processing chain, from perceptual interpretation to predictive planning, from low to high, specifically:
The first level is the perception of the components in the environment (input of information), the second level is a comprehensive understanding of the current situation (information processing), and the third level is the prediction and planning of the subsequent situation (information output) .
In general, people, machines, environment (nature, society) and other components that make up a particular situation often undergo rapid changes. In this fast-paced situation, there is not enough time and enough information to form a The overall perception of the situation, understanding, so accurate quantitative prediction of the future situation may be greatly reduced (but should not affect the qualitative analysis of the future situation).
In the era of big data, for artificial intelligence systems, based on the logical relationship between exclusion, attraction, competition, and risk-taking between various components and their interference components, we established a rule based on discrete rules and continuous probability (even The qualitative and quantitative integrated decision-making models that reflect the objective situation, including emotion-based and epiphany, have become increasingly important .
In short, big data mining that does not understand data representation relationships (especially heterogeneous variation data) is not reliable, and intelligent prediction systems based on such data mining cannot be reliable.
In addition, in the intelligent prediction system, it is also often faced with some problems that are difficult to distinguish between management defects and technical failures. How to conceptualize non-concept problems? How to homogenize heterogeneous issues? How to build unreliable components into a reliable system? How can the mistakes/errors of human beings be minimized by forming the front/back (rigid, flexible) feedback system in the intelligent prediction system, while maximizing the effectiveness of the machine and the environment?
In response to this, the 1975 Computer Turing Award and the 1978 Nobel Prize Winner Simon HASimon proposed a clever countermeasure: the limited rationality that the non-conceptual and unstructured elements in the infinite range can be extended into finite space-time. The concept of flexible, structural components that can be manipulated, so that a linear, indefinite system can be linearized and satisfactorily handled (not pursuing a needle in the sea but satisfied with a bowl of water ), and then related to the irrelevant things on the surface, intelligent predictions become more intelligent. However, in practical engineering applications, due to various imperfect factors (subjective and objective) and processing methods, there are many flaws in situational awareness theory and technology throwing. In view of this, we have tried to propose the concept of in-depth situational awareness. as follows.
So far, the storage of machines is still implemented formally, and people's wisdom is often realized visually. The calculation of artificial intelligence is the reality of formalization, and people's calculations are often objective logic plus subjective intuition fusion. The result. The calculated predictions do not affect the results. The calculated expectations often change the future. In a sense, the in-depth situation is not perceived by the calculations, but it is perceived as a result. Autonomy has both advantages and disadvantages. The attempted correction from the inside out is the verification of experience—analogue migration of experience. If the calculations are brain machines, then the calculations are the brains, and how big the world is.
Some people think: artificial intelligence is that human beings know themselves and know themselves. In fact, artificial intelligence is just human beings trying to understand themselves, because the origin of the “I am who†is far from being determined...
The question “Who Am I?†is the initial problem of autonomy. It is also the origin of the coordinates of all intelligent coordinate system frameworks. Memory is the directional vector (intentionality) in this coordinate system, and the storage of the von Neumann computer system. Differently, the procedural rules and data information here are not static, but are rather random in the interaction of the human-environment system (so the individual brain-like meaning is not significant). The flexibility of this change often reflects the independence The size of sex.
For example, language communication is a model of autonomy. It is based on interaction scenarios (not scenarios). No matter how you test it, it is a reaction between a script and a non-script. The accuracy of the test can be used to determine whether or not...
Some people divide the language into three fingers, namely, name, referent, and referent, and point out that the study of these three parties and the relationship between them has always been the problem and challenge faced by artificial intelligence. Coincidentally, in the 19th century, British scholars put forward the concept of signifiers and referents. Thinking about them, these are probably nothing more than the attributes (signifiers, feelings) involved and the relationship between them (meaning, perception). Question now! In fact, a word, a sentence, or a paragraph is inseparable from the autonomous situational limit. We know that there is much more (significant) than we can say (signifying). If you don’t believe it, think about the people you’ve met and talk! Tracing the root cause, the reason is generally based on the rational transformation mechanism: Sensibility is the wormhole of rationality, and it traverses the restraint and restraint of rationality. Reason is the black hole of sensibility and limits the willfulness and willfulness of sentiment. It can be said that the sense of self-government is driving the sensibility and being enslaved by reason...
The essence of intelligence lies in self-determination and "similar" judgment, and it is due to grasp the "similarity benchmark" measure properly. One of the advantages of people over machines is that they can find patterns of things earlier from less data. One of the reasons for this is that the machine has no coordinate origin, that is, the question of who "I" is.
For humans, whether things exist, their existence is not objective, but the result of our observation with subjective purposes, and this mixture of subjective and objective is often the product of the context of the context. For example, the construction and deconstruction of processes such as Being, Should, Want, Can, and Change are often carried out simultaneously. In addition, even the same sensation (such as vision) has specific orientation and abstract meaning. In addition to physical contact, the handshake can also be accompanied by psychological suggestion. The human brain can generate "the mapping from Euclidean space to topological space" when carrying out autonomous activities, that is to say, when doing choices and control, people can use the similarity benchmark (not The proximity in the European space, but the rational contact network, is changing, and it is decided to carry out the classification of the situation.
The free-adjusted environmental system triggered the reverse movement of the autonomous system, and thus formed a multi-directional movement or multiple movements between the human-machine environment and the resulting contradictions and conflicts . The solution to this inconsistency and even the opposite problem is often not a matter of mathematical knowledge alone. A problem with boundary, conditional, and constrained solutions is a mathematical exploration. When the same problem is unbounded, unconditional, and unconstrained, the solution often becomes philosophy. the study. For example, how does fiction correct the truth, how does the truth feedback and fiction? This will be a very stimulating question.
The difference between human learning and machine learning lies in: Human learning is a mixture of fragmentation and integrity, so it is more adaptable and has been performing stable predictions and failures in the context of insufficient information (resources, such as time and space). Stable control, mis-premises, and out-of-control scenes happen from time to time, so how to do two or three times...multiple timely and timely multi-level feedback adjustments become more and more necessary. In this regard, people deal with non-structured and non-standard situations. To be better than the machine, but in the standardization of the standard scene, the machine is relatively better than others. And this kind of adaptability is cumulative, and it will gradually form a kind of personalized rational expectation. At this point, the autonomous (expectancy + forecast + control) mechanism has begun to take shape and grow up... "The true sign of intelligence is not knowledge. But imagine."
Einstein said: "Imagination is more important than knowledge, because knowledge is limited, and imagination sums up everything in the world, promotes progress, and is the source of knowledge evolution."
Fiction is a physical representation of intelligence, and it is evident from the fact that familiarity, paradoxes, paradoxes, and other facts that can be emphatically preposterous.
The mainstream machine learning approach is to first generate a “model†from the sample using a “learning algorithmâ€, and then use this model to solve practical problems for the algorithm. Actual problems often do not strictly distinguish between the learning process and the problem-solving process, and the entire system operation is decomposed into a large number of "basic steps", each step being implemented by a simple algorithm. The convergence of these steps is determined in real time, and there is generally no strict repeatability (since the internal and external environments are not repeatable). Therefore, a general intelligent system should have no fixed learning algorithm, and there should be no invariable problem-solving algorithm, and “learning†and “reasoning†should be the same process. In addition, human learning is a fusion of causation, correlation, and even customs. Some of these can be programmed. Many are difficult to describe clearly (such as some subjective feelings, tacit knowledge, etc.), while machine learning is dominant. The connotation of knowledge is far greater than the extension of an implicit concept.
In fact, for the cognitive process of human beings, the relationship between rules and probabilities is agglomerative, rules are the existence of large probabilities, and the probabilistic nature is a state without rules. The habit is the unconscious behavior of the rules, and learning is the cumulative process of probabilities, including the familiarity and familiarity corrections. In general, the former is unconscious, the latter is conscious, and it is a compound process. In addition, the process of information processing by people is variable, sometimes the release of automated habits, sometimes semi-autonomous conscious and unconscious balance, and sometimes purely artificial slow, but this process is not a simple transmission of information, but also includes How to construct the organization in the knowledge vector space from the corresponding grammatical state, and reconstruct a variety of semantic and pragmatic systems .
Basic point of view
The meaning of in-depth situational awareness is “the perception of situational awareness, which is a human-machine intelligence that includes both human intelligence and machine intelligence (artificial intelligence)â€, refers to the signifier + signifier, and refers to both Attributes (signifiers, sensations) are also related to the relationships between them (referring to the senses). They can both understand the sounds outside the strings and understand the meaning of the words. It is based on Endsley's perception of the subject situation (including information input, processing, and output links). It is an overall system trend analysis that includes people, machines (objects), environment (nature, society), and their relationships. "hard" two kinds of regulation feedback mechanism; including both self-organization, self-adaptation, but also include his organization, mutual adaptation; both including the local quantitative calculation and prediction, as well as the global qualitative calculation assessment, is an autonomous, automatic convergence The effect of information modification, compensation expectation-selection-prediction-control system.
In a sense, in-depth situational awareness is to complete thematic tasks and organize the system to fully use various types of human cognitive activities (such as purpose, feeling, attention, motivation, prediction, autonomy, motor skills, plans, etc.) under specific circumstances. Comprehensive expression of pattern recognition, decision making, motivation, experience and knowledge extraction, storage, execution, feedback, etc. It can not only operate under the conditions of insufficient information and resources, but also can act under the conditions of information and resource overload.
Through experimental simulation and on-site investigation and analysis, we believe that there is a phenomenon of “jumping frogs†in the deep situational awareness system (automatic response), that is, directly entering the output control stage from the information input stage (skipping the information processing and integration stage). This is mainly Due to the clear theme of the task, the concentration of organizational/individual attention, and the habitual reflection of long-term targeted training, it is possible to uncoordinately coordinate the order of various natural activities as if a person were chewing gum and chatting while walking with an umbrella. The system performs near-perfect automatic control, rather than consciously conditioned conditional responses. Compared with the general situational awareness system, the sampling of their information will be more discrete, especially in information filtering after perceiving various stimuli (the basic function of the “filter†of information is to allow the specified signal to pass relatively smoothly. The attenuation of other signals, use it to highlight useful signals, suppress/attenuate interference, noise signals, improve the signal-to-noise ratio or the purpose of selection), demonstrated a strong ability to "get false, save, and get fine." For each stimulus object, it includes both useful information features and redundant other features, and the in-depth situational awareness system has the ability to accurately grasp the key information features of the stimulus object (which can be understood as The sneak peeks the ability to know leopards, so it is possible to form a quick search of stepped AI rather than refinement and operational research. Optimizing the pruning plan predicts cognitive abilities and performs thematic tasks automatically and quickly. For the general situational awareness system, because there is no cognitive response capability of the depth situational awareness system, the perceived stimuli object includes not only useful information features but also other redundant features, so the information sampling volume is large. Slow integration of information, delay in predictive planning, and weak implementation.
In the context of time and task pressure, “experienced†in-depth situational awareness systems are often based on discrete empirical thinking schemas/scripts for cognitive decision making activities (rather than assessment based). These schema/scriptive cognitive activities It is the basis for the formation of an automatic model (ie, it does not require analysis at each step). In fact, they are based on the accumulation of previous experience in response and action, rather than through conventional methods of statistical probabilistic decision-making choices (basic cognitive decision-making context assessment is based on schema and script. Schema is a type of concept or The description of the event is the basis for the formation of long-term memory organization.In the “Top-Down†information control process, the information of the perceived event can be mapped according to the most matching existential thinking schema, and the “Bottom-Up†information is processed automatically. In the process, inconsistent matching is adjusted according to the thought pattern aroused by the perceived event, or the active thinking search pattern matches the latest changing thinking schema.
On the other hand, in-depth situational awareness systems are sometimes forced to make conscious analysis decisions on some changing task scenarios (automatic mode can no longer guarantee the accuracy requirements for accurate operation), but the in-depth situational awareness system rarely shifts attention to Non-theme or background factors, this will make it "distracted." This phenomenon may be related to complex training rules, because in the rules, the general situational awareness system is required to be executed according to the program, and the rule program sets a threshold (ie, the information is stimulated to meet the specified information) that triggers its situation recognition. In fact, the dynamic situation often causes the threshold to change; for this reason, the deep situational awareness system, through a great deal of practice and training experience, forms an implicit dynamic triggering situational awareness threshold (ie, it is useful for itself. The key information features are activated, not specified.
A “Top-Down†process extracts information that relies on (at least by its influence) the previous knowledge of the characteristics of the thing; a “Bottom-Up†process extracts information only related to the current stimulus. Therefore, any process involving the identification of a thing is a "Top-Down" process, that is, an organization process of known information about the thing. The "Top-Down" process has been proven to have an effect on depth perception and visual illusion. The "Top-Down" and "Bottom-Up" processes can be processed in parallel.
In most normal situations, the situational awareness system is based on the "Top-Down" process to achieve the goal; in an abnormal or emergency situation, the situational awareness system may achieve a new goal according to the "Bottom-Up" process. In any case, the in-depth situational awareness system should remain proactive (proactive) in context (eg, using feedforward control strategies to stay ahead of context changes) rather than reactive (eg, using feedback control strategies to keep up with situational changes). ) This is very important. This proactive (proactive) strategy can be obtained through training in response to abnormal or emergency situations.
Under the real complex background, a comprehensive and comprehensive study of depth situational awareness systems and technologies, according to the information transfer mechanism in the process of man-machine-environment systems, the construction of accurate and reliable mathematical models has become a goal pursued by researchers. . The experience of human cognition shows that people have the ability to search for specific targets from complex environments and have selective treatment of target information. This search and selection process is called focus attention. In the case of multi-batch, multi-target, and multi-task situations, it is a big problem for people to quickly and efficiently obtain the required information. How to apply the environment focus and self focus mechanism of human cognitive system to the learning of multi-module deep situational awareness technology system, determine the input of attention mechanism according to the processing task, and make the whole depth situation aware The system effectively completes information processing tasks and forms efficient and accurate information output under the control of the attention mechanism, which may provide a new way for solving the above problems. How to establish a moderate-scale multi-module deep situational awareness technology system is the first problem to be solved. In addition, how to control the integration and coordination among various functional modules of the system is an important issue that needs to be solved.
Through research, we look at the situational awareness cognitive technology problem in this way: First, the deep situational awareness process is not a passive response to the environment, but an active behavior. The deep situational awareness system is acquired through the stimulation of environmental information. Filtering, changing situational analysis strategies, extracting invariance from the dynamic information flow, producing near-perceived operations or controls under human-machine environment interactions; secondly, the calculations in depth situational awareness technology are dynamic and non-linear (same Cognitive computing is similar in calculation. Usually, it is not necessary to calculate all the problems at one time, but to calculate the required information. Furthermore, the calculations in the in-depth situational awareness technology should be self-adaptive, and the characteristics of the command and control system should be Should change with the interaction with the outside world. Therefore, the calculation in the depth situational awareness technology should be the result of the interaction of the external environment, equipment, and human cognitive sensors. The three are indispensable.
Studying deep situational awareness system technology based on human behavioral characteristics, that is, researching the perception and response capabilities of organizations organized in an uncertain dynamic environment, emergency command and organization systems for major events (wars, natural disasters, financial crisis, etc.) in social systems. The rapid processing of faults in complex industrial systems, system reconstruction and repair, and the design and management of humanoid robots in complex environments all have important reference values.
The construction of meaning
In depth situational awareness, instead of constructing a situation, it constructs a meaning framework for the situation, and then realizes deep-level prediction and planning in a number of uncertain situations.
In general, the sensation is often a fragmented property, and the knowledge is the simultaneous establishment of the relationship (relationship). The person's sense and knowledge process is often performed at the same time (the machine is not), and people can simultaneously perform physics, Psychology, physiology and other attributes, sense of relationship, and knowledge can also be mixed with cross-sensing and perception. It will produce some intuition or emotion over time, from irrelevant to weak, weak to related, relevant to strong, and even The formation of a "jumping frog phenomenon": irrelevant related dominance, that is, intuition, analogy plays a very important role in this process, and is a bridge to convert implicit tacit knowledge into explicit rules/probability.
According to phenomenology, perception is the key to consciousness, that is, to be aware of the surrounding objects and the world they constitute. The perception of an object is its own action on the object obtained by integrating the interaction experience of the object and the object.
For example, the perception of an apple on a nearby table can be eaten, walked over to get it in hand, and throw it up. It is generally believed that the perception is signal input, but in reality, the computer accepts video signal input but has no vision because the computer is incapable of moving. Perception needs to be combined with its own actions, which gives input signal semantics, although the input signal does not necessarily result in certain actions. The generation of perception begins with the input signal, self-motion, and the coordination and integration of environmental objects, and the formation of an empirical memory is integrated. When the relevant signal is encountered again, a perception of the object (actions that can be performed on the object) is generated. Of course, only perception may not be enough. Intelligent systems also need the ability to reason, think, and plan. But these capabilities can be built on a perceptual platform.
Compared with humans, human beings have strong language or information chunking capabilities, limited memory and rationality; machines have weak language or information blocks, infinite memory and rationality, and simultaneous implementation of language (program) operations and self-monitoring mechanisms. It is the basic principle that guarantees the reliability of the machine. People can communicate in a grammatical manner when using their mother tongue, and in many situations they can perceive the ambiguity of language, pictures, and music. For example, human hearing, vision, and touch are distinguishable and emotional. It is often possible to perceive information or concepts that are only intended to be unspoken (such as philosophy, which is difficult to learn through learning). Although the machine can play chess and answer questions, the ability to respond to cross-field situations is weak, and it cannot respond to contradictory or ambiguous information (the lack of necessary competitive risk selection mechanisms). Insufficient, will not use inductive reasoning and other methods to form concepts, put forward new concepts, but also extravagance of the theoretical form of metaphysics.
Differences between humans and machines in the processing of language and information are mainly reflected in the ability to relate things that are irrelevant on the surface. Although the age of big data may change, for machines, the abstraction of abstraction is also the gap between the decision-making method based on the conditions of rules and probability statistics and the mechanism of judgment based on emotions and meditation (human-specific). Still exist.
A great man once described the difference between logic and imagination: "Logic will get you from A to B, Imaginationwill take you everywhere". In fact, one of the greatest characteristics of human beings is the ability to fuse the logic with imagination, figuration, and abstraction in a specific context. This flexible and flexible dispersion aggregation mechanism is often closely related to the mission context.
As with the concept of words, some philosophers insist that the meaning of words is inherent in the physical objects that exist in the world, while Wittgenstein believes that the meaning of words is determined by the context in which people use words. . The reason is probably due to the competition and risk-taking phenomenon in the diode-like mechanism. There are also people's consciousness. If you want to talk and stop, you will be in a difficult situation. The roots of ideological struggle are related to uncertainty and related to the uncertainty of people, things, and situations. Limited reason may have some connection with it. The key is how to balance and find a satisfactory solution (a bowl and a needle). Not looking for an optimal solution (a needle in the sea).
In contrast, the Alpha dog parameters of the machine program that recently defeated Go world champion Li Shishi were adjusted very well (compared to the number of parameters of some advanced artificial intelligence software programs is several orders of magnitude lower than that of Alpha dog, but the current debugging differences. It is mainly the qualitative comparison of different resource allocation schemes that have not yet reached the phase of adjusting the value. This balance of parameters is just the critical line of the competitive adventure mechanism, just like the demarcation line of yin and yang fish in the Tai Chi diagram. There is always a contradiction between qualitative and quantitative adjustment parameters in competitive risk-taking behavior. Qualitative is a directional problem, while quantification is a precise problem. How to be red and specialized is often a bit to be or not to be.
For human beings, how does the most mysterious consciousness arise? There are two problems in it: one is the basic structure of consciousness, and the other is the experience of mutual accumulation. The former can be either physiological or abstract. It is the difference between humans and machines. The latter is necessary for humans or machines. Consciousness is the product of human-machine environmental interactions. The current machine theory is not theoretically based on the human environment. (active) interactions, so there is no reference coordinate system between you and me.
Some people say that "there is no intelligence in the current artificial intelligence, there is no knowledge in the current knowledge system, everything is played by human beings, and it seems to be logical, natural, convenient and easy to memorize and maintain." This is certainly biased. But it also reflects a certain truth: consciousness is the product of human-machine environmental system interaction. The current machine theoretically does not have (active) interaction with the human-machine environment system, so there is no reference coordinate system between you and him, so it is difficult to reflect All sorts of orders imply a stable and continuous meaning .
A well-known photographer once said something to the photographer: Ten words:
Photographs are not good enough because you are not close enough to life.
The lens captured with the eyes can only be called a photo, and the lens captured with the heart can be called art.
What I express is true self. It is out of my heart.
Sometimes the simplest photos are the most difficult.
Only good photos, no guidelines for good photos.
The photographer must be part of the photo.
I think the shadow is more attractive to me than the object itself.
Famous books, music, and painting all give me inspiration and inspiration.
I do not like to use photography as a mirror to reflect only the facts, so I have room for imagination in my expression.
I spent all my life waiting for the interweaving of light and scenery, and then letting magic come into my camera.
These ten sentences seem to make sense in the construction of meaning in depth situational awareness.
Sometimes the data can be understood (defined) as the representation of human stimuli should be correct (not necessarily a symbol), even if you see a word, hear a sound, ..., without a variety of stimuli, intelligence may not develop, Grow (not assemble). Einstein once said: "Words and languages ​​seem to have no effect in my thinking project. The physical entities I think about are symbols and images. They can be reborn and combined as I wish."
Language is the linearization of symbols, and language also limits thinking . These differences are like the difference between human-machine intelligence: a memory type (machine type) and a fuzzy type (human type). People have the advantage that they can be larger and larger. The irrelevant correlations of scales (even beyond the language), the limitations of the machine are discussed here (including size data): Limited correlation. For example, to describe a system that can track and locate objects in three-dimensional space, the system can infer the relationships among these three-dimensional objects by incorporating the position and orientation into the properties of a target. Although large data redundancy may also result in precision interference or cognitive overload (information redundancy is a self-preservation strategy in the era of big data), in many applications, small data should also be of great help, because small data is more dependent on analysis after all. The accuracy (its short board is no information redundancy as big data compensation).
In short, in- depth situational awareness is essentially a process in which many public opinions, such as change and invariance, one and many, autonomy and passiveness, are generated and resolved. Therefore, the system should not be a simple human-computer interaction but should be an autonomous (including expectation, choice, prediction, control, and even affective areas) cognitive process throughout the human-machine environment system.
In view of the wide range of researches on deep situational awareness systems, it is easy to generate nonlinear, random, and uncertain system characteristics, making the system modeling study often face greater difficulties. In the previous studies, a variety of valuable theoretical models were proposed and used to describe the situational awareness system behavior. However, these models are not comprehensive enough to consider the actual and influencing factors of the actual engineering application system, and they lack the experiment of model usability.验è¯ï¼Œæ‰€ä»¥æœ¬æ–‡é‡ç‚¹å°±æ˜¯é’ˆå¯¹æ·±åº¦æ€åŠ¿æ„ŸçŸ¥æ¦‚念的实质åŠå½±å“å› ç´ è¿™ä¸¤ä¸ªå…³é”®é—®é¢˜è¿›è¡Œäº†è¾ƒæ·±å…¥æŽ¢è®¨ï¼Œè¿½æ ¹æº¯æºï¼Œä»¥æœŸæ—©æ—¥å®žçŽ°é«˜æ•ˆå®‰å…¨å¯é 之深度æ€åŠ¿æ„ŸçŸ¥ç³»ç»Ÿï¼Œå¹¶åº”用于相应的人机智慧产å“或系统ä¸ã€‚
雷锋网(æœç´¢â€œé›·é”‹ç½‘â€å…¬ä¼—å·å…³æ³¨)注：转载请è”系授æƒï¼Œå¹¶ä¿ç•™ä½œè€…和出处，ä¸å¾—åˆ å‡å†…容。