Software intelligence: the "world" and "boundary" of AI sentient beings in AAAs system

One 、 nerve neural System

At present, we know very little about how the brain works . It is called neural network （neural networks） Our system is a statistical method , It has the same advantages and disadvantages as other statistical methods , But they have little resemblance to the way real neurons work .

Each of these methods is based on a specific technology ： mathematical statistics 、 Rule of grammar 、 Dictionary format or neural network . Each of them They all ignore the language aspect that technology does not adapt . however , For people , Language and all aspects of life are seamlessly integrated , They are not hindered by the boundaries between different technologies .

1、 Neuron computing similar to chip computers

The power of wisdom power From our great diversity .

aaas The system adopts a method called Language games The neuron computing network similar to chip computer is used to deal with the huge diversity of language mechanisms .

The goal is to use imperfections 、 Neurons that can go wrong （ Computing components ） Building a flexible 、 Fault tolerant systems . Different processes supported by different neurons can realize different language games , Even different aspects of the same language game .

  in other words , Language games It is different from any method based on a specific technology in the current neural network , According to the   In cognition One More basic , Implementation independent surface features . It's not It is not divided according to the kind of technology that produces the language compartmentalize Language , But through the subject matter （subject matter） And how to use it （mode） To divide languages . This method seems more natural . But it raises a question , That is, how the computer recognizes the game being played , Especially when all aspects of multiple games are combined in the same paragraph or even the same sentence .

The solution to this problem will be explained in the next article .

Two 、 A model of the nervous system

                  chart 18.1 The basic model of most neural networks

chart 18.2 Enhancement and extension of the basic model of neural network  

The cognitive objects are systematically divided into three kinds of complexity .  chart 18.1 The original image of is suitable for objects with intermediate complexity , It is the image with the help of media （ Large objects in computers , Such as CLOB and BLOB） Structural analysis of . chart 18.2 Suitable for high complexity objects （aaas New giant data objects in , Including intelligence quotient 、 EQ and equivalent ）, Need to use the complex and diverse language mechanisms of mesoscopic systems - Wave like objects , It is system analysis based on specific rules . For simple objects （ Numbers 、 A string of characters and numeric characters ） There may be no middle tier for feature analysis .

Corresponding to three kinds of neural systems with computational complexity , Only when the cognitive object is the most complex, the manipulation of neuron computing is required .

• In the simplest case , Only input layer and output layer , Only the central nervous system （ midbrain ） Conditioned reflex of . It is based on the initial weight , If the initial value condition is satisfied, the . Each is an independent one-way .【 inheritance 】
• Medium complexity , Need a brain （ the forebrain ) The participation of neural networks . By changing the weight , Fires when operand conditions are met . After the parallel input of the system, organize the calculation and then output .【 encapsulation 】
• High complexity requires the joint participation of the forebrain and hindbrain . Hindbrain neurons use temporarily established transmission channels and configured cache and management capabilities to schedule and measure Forebrain neural network , For the... Connected to it 4 Assign tasks to nodes and assign random number conditions  . This is a negotiation process , Until the random number conditions satisfying the respective operand conditions are obtained .【 polymorphism 】

3、 ... and 、 explain

most important of all   Three conditions are preset ： Operand condition, random number condition and machine number condition .

Three different computing tasks ：

• In industry , Used on site （ equipment ） level 、 Workshop level and factory level .
• logically , Corresponding to Well behaved data （ grouping DP equipment , Object technology - Controllable ）、  The nature of a well-organized transaction （ Divide the process section , information technology - Exchangeable ） Well defined parts （ Division , Operational technology - Scalable ）.

【------- To be continued ------------】