The state-sensing detection-data processing-correlation analysis of the machine is close to "walking with light". Therefore, in some scenarios, the electronic calculations in artificial intelligence are far beyond human beings. Coupled with the recent rapid progress of quantum computers, it is even more so. People are beyond reach, but the narrow bottleneck of the machine has not been resolved, that is, there are situations and weaknesses, feelings and weaknesses, and events that human situation perceives often "traverse" time, space, and logic, and subsequent events can be Exposure in advance. Using cognitive insight into the theory of relativity to correct the distorted time, space, and logic, we can regain the contextual elements of the normal sequence of time and space logic, and use sophisticated calculation methods under the guidance of macro-calculation thoughts. , Situation, situation, sense, and knowledge are rectified, compensated, and adjusted to each other, so as to obtain a good practice result of in-depth situational awareness of human-machine integration.
The in-depth situational awareness of human-computer fusion is a complex system in itself, "complex" means regularity (structure), "miscellaneous" means non-regularity (structure), and the complex things referred to by "complex and miscellaneous" should be, It has the regularity (repetitiveness) that we can recognize, and it cannot be completely attributed to a certain regularity, nor can it be completely trapped in irregularity. This is the "middle zone" proposed by Gellman to be in disorder and order. "The part and the whole of this kind of thing are similar, but not strictly similar, because random factors invade from time to time in the repeated iterations, that is, the generation and evolution process, but they are unpredictable, leading to the breach of strict self-similarity. Therefore, it cannot be described by deterministic methods. Such objects cannot be described by statistical methods, because after all, certain rules are constantly repeated in their generation and evolution process, and they have obvious invariance under scale (level) transformation, that is, regularity. "
Yesterday’s WeChat article: "Human Consciousness and Machine Models" amended as follows:
Consciousness is a kind of feeling. It is not produced by the eyes, ears, nose, tongue and body, but generated by various internal/external interactions. The "pattern" recognition is carried out by the mind, but the "pattern" here is composed of various "patterns". "Extraordinary pattern". The consciousness of humans and some spiritual animals is not only the recognition of facts, but more importantly, the recognition of value connotations, while the machine is only the recognition of factual patterns.
Artificial intelligence is a kind of factual intelligence, and human intelligence is a kind of value intelligence. Of course, facts have different degrees of facts, and values also have different degrees of value. Reality is often a mixture of facts of different degrees and values of different sizes. Most accidental facts often come from non-situational, non-scene and non-logical value crossing.
The reason why people can grasp the direction is the dynamic value after dialysis of facts, and the ability to determine the subjective value of a thing or fact under a specific situational task, not just the objective salience, so it is organic, and the machine is not subjective, otherwise, The essence of a machine is a calculation process without value and risk responsibility, and it is inorganic. In addition, the usual choice of two is often as heavy as Mount Tai at critical moments, so only value is not enough, but the courage and courage to bear the consequences are also needed. Subjective judgment and emotional value are crucial, so the final decision on fighting and firing cannot be left to the machine.
Any thing or fact has many facets. To simplify it, we might call it duality, right and wrong, or 1 and 0, but this thing/fact will change in value with the development of the relationship, or quickly Slow, short or long, the role of human beings is to appropriately embed the value of this thing/fact with the objective state of development, and to advance with the potential, instead of marking the label like a machine. What’s more interesting is that this duality (or multi-faceted) of things/facts is naturally reflexive, and will be triggered to be implemented in a specific task situation, and it will become non-zero, and 0 will give birth to 1, which is also One of the root causes of uncertainty.
Humans have an ego or superego and constantly strengthen their subjective preferences (ego), while machines do not have the distinction of ego or superego, and they always perceive the world relatively objectively. Naturally, humans cannot easily reach a consensus with machines. Similarly for a thing or fact, the label of the machine will not be personalized and flexible, and it will not extend the growth and change, and the human will, such as "holding your hand and growing old together" is already very different from the original intention.
Machine data information knowledge labels will not grow, but human data information knowledge concepts will change. For example, one inch of time and one inch of gold are variable, and holding hands and growing old together are also very different from the original intention.
People are very complicated, but machines make it simple. Once the label is fixed, there is no change to circumstances, no right and wrong.
The way to make data not boring is to give value to form information, such as 315 holidays, 23 Jordan, 24 Kobe, 658 community bus, 20200207 Li Wenliang; the way to make information not boring is to condense meaning to generate knowledge, such as 1+1=2 An inch of time and an inch of gold; the way to keep knowledge from sluggish is to evolve into an everlasting intelligence...
Learning can make people better and more calmly ascend and deal with the unknown, but not the machine. In addition to state (such as action sequence, number of characters and numbers, etc.) in human learning, there is momentum (development and change trend), which has a knock-on effect of overlooking; in machine learning, whether it is deep learning, reinforcement learning or other learning, there is no need to learn from The overriding of the momentum from the momentum to the momentum can only follow the trend from the momentum to the momentum.
In depth situational awareness, potential is direction and speed, and direction is more important; state is degree and size, and degree is slightly ahead; knowledge is essence and connection, and essence is particularly prominent; sense is phenomenon and attribute, and phenomenon is greater than attribute; depth is people The fusion and interaction of the computer environment, the fusion of interaction.
The difficulty of situational awareness lies in the hybridity of situation and situation and the hybridity of perception and knowledge. What is more difficult is the hybridity of situation, situation, perception and knowledge. There are true and false, there are true and false, there are false and true, and there are false and true...
One of the criteria for evaluating the quality of in-depth situational awareness is the ability to be a person, machine, environment, situation, situation, sense, and knowing auxiliary lines. If you do it well, you can solve it easily. If you don't do it well, it is totally different.
The fusion of human-computer groups is actually a process of convergence. On the one hand, the human-machine network makes its organization a process of continuous decentralization and decentralization. On the other hand, its organization makes the human-machine network a more convergent and integrated process. The realization of the two goals around the task is both contradictory and harmonious. There is a unity of opposition, just like when a young infantry-style volunteer army faces a heavily loaded US army, it is like a steel nail when scattered, and like an iron fist when gathered, dragging the World War II lion into a thin horse of the Korean War.
At present, there are two solutions for human-computer integration: the first is to let people participate in the training process of the system (this involves the question of when, where, and how people can effectively participate in the system), and the second is Assign as many decision-making tasks as possible to people to complete (this involves how to select suitable decision-making tasks for people and how to prevent the "projection effect"). Although some people have done experiments to prove that these two points can increase people's trust in the machine/system, is this conclusion broad? Need further in-depth analysis and research.
In the future, humans and machines will no longer have a master-slave relationship. Each is an independent agent. Through value confirmation, it records the contribution of each agent to the entire value, so as to carry out value distribution and industrial chain. Changing from the original "relationship chain" to a "value chain", the relationship between humans and machines seems to be looser, but the synergy is also stronger.
A group photo of the Earth and the Moon, taken by Mariner 10, which was launched on November 3, 1973, about 2.6 million kilometers away from the Earth...