Unit 1:Introduction to Artificial Intelligence

05 March 2024

What is AI?
AI is machine ability to perform cognitive functions as human do such as learning, reasoning and solving problems.
The benchmark for AI is the human level concerning in term of reasoning speech.
Father of AI: John McCarthy
First AI Machine: Allen Tuning
Intelligence is composed of:
1. Reasoning
2. Learning
3. Problem Solving
4. Linguistic Intelligence

Uses of AI
1. Healthcare: AI is used for medical diagnosis that discover and predictive analysis of disease.
2. Finance: AI help in credit scoring, fraud detection and financial forecasting.
3. Retail: AI is used for product recommendation price optimization and supply chain management.
4. Manufacturing: AI helps in quality control predictive maintenance and product optimization.
5. Transportation: AI is used for autonomous vehicle, traffic prediction and route optimization.
6. Customer Service: AI powered chat-box are used for customer support answer FAQ and handling simple request.
7. Security: For facial recognition, intrusion detection, cyber thread analysis.
8. Marketing: Use for targeted advertising customer segmentation, sentimental analysis.
9. Education: For personalized learning, adaptive testing and intelligent tutoring system.

Goals for AI:

1. To reduce the amount of time needed to perform the specific task.
2. Making it easier for human to interact with machine.
3. Improving the accuracy and speed for medical diagnosis.
4. Helping people learn new information more quickly.
5. Enhancing communication b/w human and machine.

Subfield of AI:
1. Machine Learning: ML is an art of studying algorithm that learn from example and experiences. It is based on the idea that some pattern
in the data work identified and used for future prediction.
2. Deep Learning: It is a subfield of ML. It doesn't mean the machine learn more depth instead it uses different layers to learn from data. The
depth of the model is represented by the number of layer in the model.

3. Natural Language Process(NLP): This model builds upon the human nervous system. This model is used to conduct image
understanding, human learning, computer speech etc.
4. Expert System: It is an interactive system and reliable computer based decision making system that uses facts and heuristics tom solve
complex decision making problem.. It is also considered as the highest level of human intelligence.
5. Fuzzy Logic: It is defined as a many valued logic form that may have true value of variable in any real number 0 and 1. It handles concepts
of partially true. (0 to 1 including 0.1, 0.2 ….0.9)

AI Page 1
 Reactive Machine: Purely reactive machine are the most basic type of AI such as AI system do not store memory or past experience for
future action. These machine focus on current scenarios and react on as per possible best action.
Ex: IMBW

Limited Memory: It can store past experience or some data for short period of time. These machine can use to store data for a limited time.
Theory of Mind: Theory of mind AI should understand human interaction and able to interact socially like human. (These types of
machines still not have developed).
Self-Awareness: This AI is the future of AI. These machines will be super intelligent and will have their own consciousness, sentiments
and self-awareness. These machines will be smarter than human minds.

Forms of AI:
1. Weak AI:- It is an AI that create a particular problem that perform a specific task. (Basically, for one purpose, AI to beat chess
grandmaster)
2. Strong AI: Strong AI is difficult to create than weak AI. It is a general purpose intelligence that can demonstrate human abilities. Like
Reasoning

Approach of AI
1. Acting Humanly(The Turing Test Approach): This approach was designed by Allan Turing, the ideology behind this approach is that a
computer passes the test if a human interrogator, after asking some written question can't identify whether the written responses come from
a human or from a computer.
2. Thinking Humanly(The Cognitive model approach): The idea behind this approach is to determine whether the computer thinks like a
human.
3. Thinking rationally(loss of thought approach): This idea is behind this approach is to determine whether the computer thinks rationally
i.e. with logical reasoning.
4. Acting Rationally(The rational agent approach): The idea behind this approach is to determine whether the computer acts rationally i.e.
with logical reasoning.

Drawback of AI:
1. Biased and Unfairness: It gives results based on previous data that is provided to it.
2. Lack of transparency and Accountability: Complex AI system can be difficult to understand making it challenging to determine how
decision are being made.
3. Job Displayment: AI has the potential to automate many jobs leading to job losses and need a real scaling.
4. Security and Privacy Risk: AI System can be vulnerable to hacking and other security threats and may also posse primary risk by
collecting and using personal data.
5. Ethical Concern: AI raises important ethical questions about the use of technology for decision making, accountability and human dignity.

AI Page 2
Unit 2:Introduction of Searching
15 March 2024

Problem consists of:

1. A State Space: Set of all possible state where you can be.
2. A Start State: The State from where the search begins.
3. A Goal State: A function that look at the current state written whether or not it is a goal state.

• The solution to search problem is the sequence of actions called the plain text that transform the start state to goal state.

The plans are archived through algorithm.

Search Algorithm in AI:

AI is the study of the building agents that acts rationally, most of the time these agents perform some kind of search algori thms in background in
order to achieve their tasks.
Search in AI refers to the systematic exploration of possible configuration or states within a problem space to find a path t o a good state, this process
is goal state.
This process is similar to navigating a maze where the solution is exist but the path is not immediately present. The efficie ncy and effectiveness of
the search algorithm can significantly impact the performance of AI system, especially in complex or large problem spaces.

Search Terminologies:
1. Problem Space: It is the environment in which the search take places.
2. Problem Instance: It is initial state plus goal state.
3. Problem Space Graph: It represent problem state, states are shown by nodes and operators are shown by edges.
4. Depth of a problem: Length of the shortest path or shortest sequence of operators from initial state to goal state.
5. Space Complexity: The maximum number of nodes data store in memory.
6. Time Complexity: Time to solve the problem or the maximum no. of nodes that are created.
7. Amicability: A property of an algorithm to always finds an optimal solution.
8. Branching factor: The average number of child nodes in the problem space graph.
9. Depth: Length of the shortest path from initial state to goal state.

Uninformed Search Algorithm:

The search algorithm in this section have no additional information on the goal node other than the one provided in the probl em definition.
The plain to reach the goal state from the start state differ only by the order or length of the action. It is also called bl ind search.
These algorithm can only generate the successor and differentiate b/w the goal state and non -goal state. The only known factor is the cost of moving
from one node to another.
Each of these algorithms will have
a. A problem graph or tree containing the start node 'S' and the goal node 'G'.
b. A strategy describing the manner in which the graph will traverse to 'G'.
c. A fringe: It is a data structure used to store all possible state(node) that you can go from current state.
d. A tree: The result while traversing to the goal node.
e. A solution path: The sequence of nodes from 'S' to 'G'.

1. BFS(Breadth First Search):

It start from the root node, explore the neighboring node first and move toward the next level neighbor. It generates one tree at time until the
solution is found.
It can be implemented using FIFO (Queue) data structure. This method provide shortest path to the solution. BFS explores the search space
across the Breadth. It expands the shallowest node first ensuring that the shortest path if all the node have equal cost.
BFS is complete and optimal in such condition but can be memory intensive as it store all the generative node in memory.
• Branching factor(Average no of child node for a given node):
b depth=d, then the number of node at level d is equal to bd.

• Disadvantages:

AI Page 3
• Disadvantages:
a. Store large memory
b. Space complexity is high
Complete = yes(if b is finite)
Time = O(bd+1) per step
Space= (bd+1) per step
Optimal=yes(if cost=1)
Where, d is height and b is no. of nodes

• Algorithm of BFS:
○ Start at the entrance: Begin new search from the stationary point.
○ Explore neighbor: Look at all the path directly reachable from the entrance. I graph, these are the nodes directly connected to the starting
node.
○ Mark as visited: Keep track of which path you have explore to avoid going in circle. In graph term, this means marking node as visited.
○ Move to the next layer: Once you checked all the path from the current position move one step deeper into the graph or tree and explore
all the new paths accessible from here.
○ Repeat until found: Continue exploring the graph or tree layer by layer moving deeper only after all the paths in the current layer have
been examined. If you reach the goal, you will get the path.
2. DFS(Depth First Search):
It is implemented in recursion with LIFO(Stack) data structure.
It create the same set of nodes as BFS only in the different order.
DFS dives deep into the search space expanding the deepest node, while less memory intensive than BFS.
It is not guarantee to find the shortest path and get stuck looping in cycle, if the search space is infinite.
DFS is not optimal but can be made complete with modification to avoid infinite loop.
As the node on the single path store in each iteration from root to leaf node. The space requirement to store node is linear with branching factor
'B' and depth as 'M'. The storage space is B*M.

• Disadvantages:
The algorithm may not be terminated and go on infinity, the solution to this issue is to choose a cut-off depth.
If the ideal cut-off depth is 'D' and if chosen cut-off is less than D than this algorithm may fail.
It chosen cut-off is more than 'D', then execution time increases.
It's complexity depends on the number of path and it can't check duplicate path.
Complete = no(infinite loop)
Time = O(BM) per step
Space= (B*M) per step
Optimal=no

AI Page 4