Artificial

Intelligence (AI)

Lecture 1:- Introduction to AI

Dr. Hakim Almaghbshy

what machines can do?

 Philosophers have been trying for over two thousand years to

understand and resolve two big questions of the universe:
 how does a human mind work?, and can non-humans have minds?
 Some philosophers accepted the idea that machines can do
everything that humans can do from the computational approach
originated by computer scientists.
 Others said the behavior as love, creative discovery and moral
choice always be beyond the scope of any machine.
 The nature of philosophy allows for disagreements to remain
unresolved. In fact, engineers and scientists have already built
machines that we can call ‘intelligent’.

2
what does the word ‘intelligence’ mean?

 Intelligence is the ability to think and understand instead of doing

things by instinct or automatically.

 Thinking is the activity of using your brain to consider a problem or

to create an idea.
• So, in order to think, someone or something has to have a brain, an
organ that enables learning and understanding things, to solve
problems and to make decisions.

• So we can define intelligence as ‘the ability to learn and understand,

to solve problems and to make decisions’.

3
The definition of the artificial intelligence (AI)

 Artificial intelligence has different definitions, the following are quoted

from textbooks
1. It is a science to make machines think with minds. (systems that like
human)
2. It is a science to make machines perceive, reason and act through the
use of computational models (systems that think rationally)
3. It is a science to make machines do things that would require
intelligence if done by humans(systems that act like human)
4. It is a science concern with the automation of the intelligence
behavior( systems that act rationally)
 (1,2) concern with thinking and reasoning
 (3,4) concern with behavior
 (1,3) concern with the success on human performance
 (2,4) concern with rationality which means it does the right things.
4
Artificial intelligence approaches

1- Thinking humanly, cognitive modeling approach

 A program in machine think like human.
 To make program like this we need to know sufficiently the theory of
mind, and then express that theory as program.
 The program is not designed to have correctly solved problems, its
tracing of reasoning steps likes the human traces for solving the
problems.
 The goal is to create systems that can understand human language,
emotions, and culture and can interact with humans in a natural way.
 Ex.: AI systems that can have conversations with humans using natural
language processing techniques

5
Artificial intelligence approaches

2- Thinking rationally, the laws of thought approach

 The program has to be designed to have correctly solved problems.
 It is based on the inference mechanisms that has been proved to get
the correct result.
 The goal is to create systems that can solve problems and make
decisions in a way that is consistent with the principles of rational
thinking.
 AI systems that can make decisions and provide advice based on a set
of rules and knowledge.

6
Artificial intelligence approaches

3- Acting humanly, Turing’s approach

 Turing defined the intelligent behavior of a computer as the ability to
achieve the human-level performance in cognitive tasks.
 He proposed Turing test to provide a satisfactory operational definition
for the intelligent
 Turning test said that “an intelligent in machine (computer) is tested by
making a conversation between a human interrogator and the machine
and it will pass if the machine deceiving the interrogator that it was a
human”.
 In the Turing test the interrogator does not see, touch or hear the
computer and is therefore not influenced by its appearance or voice.
 The goal is to create systems that can perform tasks such as
recognizing speech, recognizing images, and controlling robots in a
human-like manner.
 Ex. AI systems that can identify individuals based on their facial features
7
Artificial intelligence approaches

4- Act rationally, the relational agent approach

 Agent that perceive and act efficiently
 One way to act relationally is to reason logically to the conclusion that a
given action will achieve the goal, and then act on that conclusion.
 There also ways of acting rationally that does not involve inference.
 The goal is to create systems that can make decisions and take actions
that are consistent with the principles of rational thinking and that
achieve their goals efficiently and effectively.
 AI systems that can play games such as chess

8
The history of artificial intelligence

 The birth of artificial intelligence (1943–56)

 The first work recognized in the field of artificial intelligence (AI) was
presented by Warren McCulloch (the second ‘founding father’ for AI)
and Walter Pitts in 1943.
 Their research on the central nervous system resulted in the first
major contribution to AI: a model of neurons of the brain.

 They proposed a model of artificial neural networks in which

each neuron was postulated as being in binary state, that is, in
either on or off condition. They proved that any computable
function could be performed by some network of connected
neurons.

 In fact, a neuron has highly non-linear characteristics and cannot be

considered as a simple two-state device. 9
The history of artificial intelligence

• Marvin Minsky and Dean Edmonds, built the first neural network co in
1951, von Neumann encouraged and supported them.
• Claude Shannon published a paper on chess playing machines,
which pointed out that a typical chess game involved about 10120
possible moves.
 If computer could examine one move per microsecond, it would take
3 ×10106 years to make its first move.
 Thus Shannon demonstrated the need to use heuristics in the
search for the solution.
• John McCarthy organize a summer workshop at Dartmouth College, ,
this workshop gave birth to a new science called artificial intelligence.

1
0
The history of artificial intelligence

 The era of great expectations (1956–late 1960s)

 computers had been introduced to perform routine mathematical
calculations.
 John McCarthy defined the high-level language LISP – one of the
oldest programming languages for AI.
 He also proposed program is called Advice Taker to search for
solutions to general problems of the world, based on some simple
axioms, and can accept new axioms.
 Advice Taker was the first complete knowledge-based system
incorporating the central principles of knowledge representation
and reasoning.
 Frank Rosenblatt propose a learning algorithm based on adjusting the
connection strengths between neurons.

1
1
The history of artificial intelligence

 Allen Newell and Herbert Simon developed a general-purpose system

(GPS)to simulate human problem-solving methods.
 It was the first attempt to separate the problem-solving
technique from the data.
 It was based on the technique now referred to as means-ends
analysis
 The means-ends analysis was used to determine a difference
between the current state and the desirable state or the goal
state of the problem, and to choose and apply operators to
reach the goal state. If the goal state could not be immediately
reached from the current state, a new state closer to the goal
would be established.

• In this era, They used the general-purpose search mechanism to find

a solution to the problem, such approaches, now referred to as weak
1
methods 2
The history of artificial intelligence

 The impact of reality (late 1960s–early 1970s)

 From the mid-1950s, AI researchers were making promises to build
all-purpose intelligent machines on a human-scale knowledge base by
the 1980s, and to exceed human intelligence by the year 2000.
• however, in this era they realized that such claims were too
optimistic(Because AI researchers were developing general methods
for broad classes of problem).
 In 1966, all AI projects funded by the US government were cancelled.
 in 1971, the British government also suspended support for AI
research.

1
3
The history of artificial intelligence

 The technology of expert systems, or the key to success (early

1970s–mid-1980s)
 when weak methods failed, researchers finally realized that the only
way to deliver practical results was to solve typical cases in narrow
areas of expertise by making large reasoning steps.
 This lead to emerging the expert systems
 The idea of the the expert systems based on the
incorporating of experts skills, experience and expertise into
expert systems to make it perform at a human expert level

 The first expert system was DENDRAL, it was developed at Stanford

University by Feigenbaum to analyze chemicals.
 DENDRAL proved to be a useful analytical tool for chemists and was
marketed commercially in the United States.
1
4
The history of artificial intelligence

• The next major project undertaken by Feigenbaum was MYCIN, it is

an expert system for the diagnosis of infectious blood diseases.
 It was a rule-based expert system (450 independent rules of IF-
THEN)
• In 1986, survey reported a remarkable number of successful expert
system applications in different areas (nearly 200 expert systems) :
chemistry, electronics, engineering, geology, management, medicine,
process control and military science.
• seven years later a similar survey reported over 2500 developed
expert system

1
5
The history of artificial intelligence

 Are expert systems really the key to success in any field?

 Although of a great number of successful developments and
implementations of expert systems in different areas of human
knowledge, The expert system has the following difficulties
1. Expert systems are restricted to a very narrow domain of expertise.
2. Because of the narrow domain, expert systems are not as robust and
flexible as a user might want.
3. Expert systems have limited explanation capabilities.
4. Expert systems are also difficult to verify and validate.
5. Expert systems, especially the first generation, have little or no ability to
learn from their experience.

1
6
The history of artificial intelligence

 How to make a machine learn, or the rebirth of neural networks)

(mid-1980s–onwards)
 By the late 1960s, most of the basic ideas and concepts necessary for
neural computing had already been formulated
 The major reason for the delay was technological: there were no
PCs or powerful workstations to model and experiment with
artificial neural networks. The other reason was financial.
 Hopfield introduced neural networks with feedback –Hopfield networks
 Kohonen published a paper on self-organised maps
 Rumelhart and McClelland invented back-propagation learning
algorithm
• ,

1
7
The history of artificial intelligence

 Evolutionary computation, learning by doing (early 1970s–

onwards)
 Natural intelligence is a product of evolution
 Evolutionary computation works by simulating a population of
individuals, evaluating their performance, generating a new population,
and repeating this process a number of times.
 Evolutionary computation combines three main techniques: genetic
algorithms, evolutionary strategies, and genetic programming.
 Genetic algorithm is used for manipulating artificial ‘chromosomes’
(strings of binary digits), using genetic operations like as selection,
crossover and mutation
 Evolutionary strategies were designed specifically for solving
parameter optimization problems in engineering.
 Genetic programming represents an application of the genetic
model of learning to programming. (its goal is to evolve a computer
1
code that solves the problem) 8
The history of artificial intelligence

 The new era of knowledge engineering, or computing with words

(late 1980s–onwards)
 Expert systems can neither learn nor improve themselves through
experience..
 To overcome this limitation, neural computing can be used for
extracting hidden knowledge in large data sets to obtain rules for
expert systems.
 Experts think in imprecise terms, such as very often and almost never,
usually and hardly ever, frequently and occasionally, and use linguistic
variables, such as high and low, fast and slow, heavy and light..
 Most methods of handling imprecision in classic expert systems
are based on the probability concept.
 The current expert systems handling imprecision based on Fuzzy
logic.
1
9
The history of artificial intelligence

 Integration of expert systems and ANNs, and fuzzy logic and ANNs
improve the adaptability, fault tolerance and speed of knowledge-based
systems

2
0
Rest of the course content

 Rule-based expert systems

 Uncertainty management in rule-based expert
systems
 Machine learning
 Neural networks
 Deep Neural networks

2
1