Artificial intelligence

This is a topic too broad to encapsulate in an article, but here I just want to present some technical solution issues, hope everyone will give their opinions.

1. Ideas:

- In terms of application of artificial intelligence, it is not controversial, but in fact there are too many artificial intelligence operating in many different areas and really do not have proper integration and not yet achieved so-called wisdom (self-learning ability). So let us imagine an artificial intelligence that will become the center of knowledge of mankind, capable of learning, communicating and controlling connected peripheral systems.

- Currently we have many support tools such as voice recognition, face recognition, 3D technology . but all of them are too discrete, what if they combine all of them?

- Internet data can be extremely rich and confusing and storing raw form, there is no association, any correlation in these data, if we systematically store it under What about the definition format?

2. Concepts:

- First of all, let's define the word "understand" a little way: "understanding" an object means that we recognize it, meaning that all or part of the characteristic data of that object we have "learned" and stored in memory, when by sensory organs we receive a specific data of this object (shape, material, utility, nature, language .) and compare with our archived data will identify the object and thereby identify other features of the object not through the sensory system.

- The concept of "decision making" : Our decision comes from the order of central processing (brain), this order comes from two sources: from the outside and from the inside:

+ External words: Receive through senses: images, sounds, or tasks (corresponding to a series of decisions), through the process of analysis (understanding) and we make real decisions show some action.

+ The inner word: Receiving through the inner senses or internal needs, for example: When comparing existing data we may not understand a certain characteristic, a property or concept . and I decided to learn it.

- The concept of "learning" : It is to receive 1 block of data, use existing data to identify (understand) part of the data and analyze the "not understand" part according to the following characteristics Systematically store data. With this cycle we can "define" new features if in our storage category there are no features to describe.

3. Technical solutions:

- An artificial intelligence here is to perform a process: "Learn" - "understand" - "make decisions".

- By programming solutions, we can simulate all the above steps according to the following basic diagram:

[Receiving] - [DL separation system] - [DL identification] - [XL center] - [Executing]

- Reception: Data is received through many channels including 3 main channels:

§ Image

§ Sound

§ Document

- Data segregation: The task of separating receiving data into separate objects and its characteristics

- DL identifier: The task of identifying objects, identifying orders based on archives, storing new data (linked) into the archive according to instructions from the central processor. Summarize the report results on IPs or transfer orders on IP.

- Central processing system: Receive orders, check and compare the order processing process. Provide requirements for 'learning' to specifically dictate. Integrate priority and evaluate the importance of orders.

- HT Executing: Receiving orders from IPs to compare the archives to find out the process of executing orders, outputting the results: via images, sounds, languages ​​or orders.

All top systems use the same storage, synchronized based on data characteristics. The first stage of learning will be very time-consuming, but when the data storage has been relatively learning of this machine will be much faster and simpler.

4. Hardware requirements:

- The amount of data received after separation is very large so the identification and processing system must ensure speed. The special requirement here is a multi-task processing system built on neural networks. In which each object or each feature will be transferred to each individual processing unit and using the same storage. Another advantage of neural networks is the ability to upgrade, we can add more processing units easily.

5. Conclusion:

- My purpose is not to create a thinking machine, to find new things (people have not found out yet), know how to make decisions . It is a form of storing human knowledge, communicating with people more easily and understanding the requirements, commands and enforcement.