What is the process by which artificial intelligence automatically acquires knowledge and skills and achieves self-improvement?

The process of achieving self-improvement is "machine learning". Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent; it enables computers to simulate human learning behavior, automatically acquire knowledge and skills through learning, continuously improve performance, and achieve self-improvement. Machine learning mainly studies three aspects: 1. Learning mechanism, the innate ability of human beings to acquire knowledge, skills and abstract concepts; 2. Learning method, based on simplifying the biological learning mechanism, using computational methods to reproduce it; 3. Learning A system that can implement machine learning to a certain extent.

The operating environment of this tutorial: Windows 7 system, Dell G3 computer.

The process by which artificial intelligence automatically acquires knowledge and skills and achieves self-improvement is "machine learning".

Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent.

Machine learning:

Machine learning enables computers to simulate human learning behavior, automatically acquire knowledge and skills through learning, and continuously improve performance. Achieve self-improvement.

Machine learning mainly studies the following three aspects:

(1) Learning mechanism: human’s innate ability to acquire knowledge, skills and abstract concepts .
(2) Learning method: The structure of the machine learning method is based on the simplification of the biological learning mechanism and is reproduced using computational methods.
(3) Learning system: A system that can realize machine learning to a certain extent.

A learning system should generally consist of four basic parts: environment, learning, knowledge base, execution and evaluation.

Classification of machine learning

1. Classification by learning method ( Winston, 1977)

Mechanical learning, guided learning, example learning, analogy learning, explanation learning, etc.

2. Classification by learning ability:

Supervised learning (learning with teachers)

Reinforcement learning (reinforcement learning or reinforcement learning)

non Supervised learning (learning without teacher)

##3. Classification by reasoning method:

Deduction-based learning (explanatory learning).

Induction-based learning (example learning, discovery learning, etc.).

4. Classification according to comprehensive attributes:

Inductive learning, analytical learning, connection learning, genetic learning, etc.

Mechanical learning

Mechanical learning (rote learning) is also called memory learning, or rote learning: it is achieved by directly memorizing or storing information provided by the external environment. The purpose of learning is to obtain the corresponding knowledge through retrieval of the knowledge base and directly use it to solve problems.

The essence of machine learning is to exchange storage space for processing time.
Typical example: In 1959, Samuel (A.L.Samuel)’s checkers program CHECKERS.

Samuel's checkers program CHECKERS

uses the valuation function to score the pattern under a given search depth and calculates it through backward calculation Get the backward value of the upper node to determine the current best move.

When encountering the same situation next time, directly use the backcast value to determine the best move without recalculation.

Main issues with machine learning:

Storing organizational information: To Use appropriate storage methods to make retrieval as fast as possible.
Environmental stability and applicability of stored information: The machine learning system must ensure that the stored information adapts to changes in the external environment.
Trade-off between storage and computation: An important point for machine learning is that it cannot reduce the efficiency of the system.

Guided learning

Guided learning (learning by being told) is also called instructional learning or professorial learning: providing general instructions to the system from the external environment Or suggestions, the system specifically converts them into detailed knowledge and sends them to the knowledge base. During the learning process, the knowledge formed should be evaluated repeatedly to make it continuously improved.

The learning process of guided learning:

Solicit instructions or suggestions from the instructor, convert the consultation opinions into executable internal forms, and add them to the knowledge base ,evaluate.

Ask for instructions or suggestions from the instructor

Simple consultation: The instructor gives general opinions, the system Make it concrete.
Complex consultation: The system not only requires the instructor to give general suggestions, but also specifically identifies possible problems in the knowledge base and gives modification suggestions.
Passive consultation: The system just passively waits for the instructor to provide opinions.
Active consultation: The system not only passively accepts instructions, but also actively asks questions to focus the instructor's attention on specific issues.

Convert consultation opinions into executable internal form

The learning system should have the ability to convert consultation opinions expressed in the agreed form into computer internal The ability to execute the form, and perform syntax checking and appropriate semantic analysis during the conversion process.

Join the knowledge base

During the joining process, the knowledge must be checked for consistency to prevent conflicts, redundancies, loops, etc. question.

Evaluation

Evaluation method: Empirically test new knowledge, that is, execute some standard examples, and then check whether the execution is consistent with the known situation consistent.

Example learning

Example learning (learning from examples, example learning or learning from examples): By obtaining a number of examples related to a certain concept from the environment, often A learning method that induces general concepts.

In example learning, the external environment (teacher) provides a set of examples (positive examples and counterexamples), and then general knowledge applicable to a larger range is summarized from this special knowledge, which will cover all positive examples. example and exclude all counterexamples.

Learning model for example learning

For more related knowledge, please visit the FAQ column!

The above is the detailed content of What is the process by which artificial intelligence automatically acquires knowledge and skills and achieves self-improvement?. For more information, please follow other related articles on the PHP Chinese website!