#by Deque

Hello guys. 

Johnnyjohnny asked for some sources about AI. I took the time to translate (and adjust to make it understandable) part 1 of my notes from my course of studies into english.

This is just an overview about AI, so don't be disappointed if you are looking for practical stuff. This will be part of the chapter 2 and 3 which I will probably translate too.

This is the course: 
1. AI Overview
2. Fundamentals of logic
3. Knowledge representation
4. Expert systems

Happy learning.

1. AI overview
1.1 The term AI

AI: Automation of intelligent behaviour based on knowledge.

Knowledge: Saved information of a system about

[*]things
[*]connections


Characteristics of knowledge:

[*]difficult to define
[*]extensive
[*]difficult to present
[*]mutable

Thinking: The handling of knowledge, derivation  of implicit knowledge from a knowledge base

[*]prerequisite of thinking: knowledge
[*]theoretical basis: formal (mathematical) logic


Learning: increasing knowledge

[*]adding new knowledge
[*]generalization of experiences (inductive thinking, a finite amount of examples is transmitted to more examples --> incomplete induction)
[*]adding new examples


Intelligence:
Ability of a system to

[*]think
[*]solve problems
[*]abstract
[*]recognize situations
[*]explain actions
[*]operate cooperative
[*]react fault tolerant
[*]learn


So AI are computer-aided techniques to simulate human capabilities, especially thinking, recognizing and operating. 
The purpose:

a) from the user's perspective

[*]computer shall be more useful
[*]better solving of problems


:) from the scientist's perspective

[*]understanding principles that enable intelligence


Problem: Is AI something special within computer science?

a)
		-------------------------
input data	|algorithm		| Output data
--------------->|-----------------------|------------>
		|knowledge base		|
		-------------------------

This separation is characteristic for AI systems. Here AI is something special.

b)
		-------------------------
input data	|			| Output data
--------------->|	algorithm	|------------>
(with knowledge |			|		
	base)	-------------------------

It is nothing special here. This is just the IPO Model.

1.2. Requirements, effects and limits of AI

AI is an unfavorable term which causes misconceptions. AI is not the subject of research but the aim. Let's compare this with aspirations of other artificial systems: 

[*]artificial flower (expectation: no smell, looks good)
[*]plain (expectation: flies, does not land on a telegraph line

This means special requirements are to be accomplished. The artificial systems are often better than the real object. Not important characteristics are ignored.

The aftermath was that the requirements where overdrawn, so that an AI crisis followed during the 90s.

Limits of AI compared to the human intelligence:
Human background knowledge, biochemical and cognitive processes are not formalizable.

Effects of AI:

[*]new content for computer science
[*]knowledge processing is possible in confined fields
[*]more rational problem solving --> cost saving, relief for human work, possibly time saving
[*]new reflection about thinking --> encouragement for human intelligence


1.3. AI fields
1.3.1. Application-oriented enumeration

Deduction systems: 

[*]knowledge is available
[*]producing deductions out of it (logical expressions)


Understanding images:

[*]automatic seeing
[*]"understanding" <--> correct assumptions <--> take action


Speech processing:

[*]"understanding" human speech (written and spoken)
[*]usage: automatic translation

Computer algebra:

[*]calculating with symbolic expressions


Artificial life:

[*]agents
[*]evolutionary algorithms
[*]artificial neural networks


Robotics:

[*]all of the above combined


1.3.2 Method-oriented enumeration

Knowledge representation:

[*]appropriate computerised form to represent knowledge, that is finding suitable data structures
[*]use of logic, OOP, ...


Automated concluding:

[*]deriving knowledge from already known
[*]use and development of logic calculus


Heuristic search/problem solving:

[*]going through several possibilies for an aim (e.g. chess)
[*]blind search --> complexity problem
[*]intelligent search --> much better


Learning:

[*]improvement of the knowledge base
--> enhancements
--> optimization
--> improving the time-complexity

AI tools:
      AI languages
      /		\
     /		 \
   Prolog       Lisp
(logic  	(functional 
programming)    programming)

* special tools to create:
	* expert systems (EE, Babylon, ...)
	* image processing systems (Mathlab)
	* speech processing systems
	...


Artificial neural networks:

[*]using the cooperation of artificial neurons
[*]parallel processing
[*]learning ability