Artificial Intelligence SEM7 2025

Section II (Short Answer)

Q.2- Write short answers of the following.

i. Write differences between Strong AI and Weak AI.

• Strong AI (General AI): Refers to machines that possess human-level intelligence, meaning they can understand, learn, and apply knowledge across a wide range of tasks, just like a human being. It possesses consciousness and a genuine understanding of the world. (Example: The highly advanced AI seen in sci-fi movies, which does not fully exist yet).
• Weak AI (Narrow AI): Refers to AI systems designed and trained for a specific, singular task. They do not possess genuine intelligence or consciousness but are highly capable within their narrow domain. (Example: Apple's Siri, chess-playing bots, or Netflix recommendation algorithms).

ii. Explain the structure of intelligent agents and their interaction with the environment?

An intelligent agent is a system that perceives its environment and takes actions to maximize its chances of success.

• Structure: Mathematically, $Agent = Architecture + Program$. The architecture is the computing device or hardware, and the program is the AI algorithm running on it.
• Interaction: Agents interact with the environment using Sensors (to perceive input, like cameras or microphones) and Actuators/Effectors (to perform actions, like robotic arms or a screen display).

iii. Write names of the characteristics of expert systems.

Expert systems are designed to solve complex problems by reasoning through bodies of knowledge. Their key characteristics include:

• High Performance: They offer highly accurate and fast problem-solving capabilities within their specific domain.
• Understandability: They can explain their reasoning and how they arrived at a specific conclusion to the user.
• Reliability: They are consistent and do not make careless errors like humans might when fatigued.
• Domain-Specific: They possess specialized knowledge limited to one particular field (e.g., medical diagnosis).

iv. What is backpropagation in neural networks?

Backpropagation (short for "backward propagation of errors") is the core learning algorithm for artificial neural networks. After a network makes a prediction, backpropagation calculates the error (the difference between the prediction and the actual target). It then works backward from the output layer to the input layer to adjust the "weights" of the network, step-by-step, minimizing the error for future predictions.

v. Explain the difference between informed and uninformed search strategies.

• Uninformed Search (Blind Search): This strategy operates without any domain-specific knowledge or clues about where the goal might be. It only knows the problem definition and explores the search space systematically until it stumbles upon the goal. (e.g., searching every room in a house to find your keys).
• Informed Search (Heuristic Search): This strategy uses problem-specific knowledge (called a heuristic) to guide the search process efficiently. It "guesses" which path is closest to the goal, significantly reducing search time. (e.g., using a map to find the fastest route to a destination).

vi. Write a list of search techniques?

• Uninformed Search Techniques:

• Breadth-First Search (BFS)

• Depth-First Search (DFS)

• Uniform Cost Search (UCS)

• Informed Search Techniques:

• A* Search (A-Star)

• Greedy Best-First Search

vii. How can AI be used in detecting fraud?

AI detects fraud by analyzing massive amounts of transaction data in real-time to identify anomalies or unusual patterns that humans would miss.

• Example: If your credit card is typically used in your home city for groceries, and suddenly a $5,000 transaction is attempted in another country, the AI machine learning model instantly flags this pattern as highly suspicious and blocks the transaction to prevent fraud.

viii. What is the primary goal of problem-solving agents?

The primary goal of a problem-solving agent is to find a logical sequence of actions that leads from an initial state to a desired goal state. To achieve this, the agent defines the problem, formulates a goal, searches for a valid solution path among various alternatives, and finally executes the actions.

---

Section III (Essay Type)

Q 3: Write a detailed explanation of the components of Natural Language Processing (NLP).

Natural Language Processing (NLP) is a branch of AI that enables computers to understand, interpret, and generate human language. NLP is primarily divided into two main components: Natural Language Understanding (NLU) and Natural Language Generation (NLG).

1. Natural Language Understanding (NLU) NLU is responsible for comprehending the meaning of the given text or speech. It breaks down language using the following steps:

• Lexical Analysis: Involves breaking the whole chunk of text into smaller paragraphs, sentences, and words.
• Syntactic Analysis (Parsing): Checks the grammar and sentence structure. It ensures the words are arranged in a way that makes logical sense according to the rules of the language.
• Semantic Analysis: Extracts the exact dictionary meaning from the text. It checks for meaningfulness (e.g., rejecting a structurally correct but meaningless sentence like "Hot ice-cream").
• Pragmatic Analysis: Looks at the context of the sentence to determine its real-world meaning. It considers the speaker's intent and situational factors.

2. Natural Language Generation (NLG) NLG is the process of producing meaningful phrases and sentences in the form of natural language from internal computer data.

• Text Planning: Determining what information to say or communicate.
• Sentence Planning: Determining how to say it, including choosing appropriate words and forming structured sentences.
• Text Realization: Outputting the final, grammatically correct sentences to the user.

---

Q 4: Discuss the role of fuzzy logic in handling uncertainty in AI systems. Gives one real-world example?

Role of Fuzzy Logic in AI Traditional logic systems rely on strict binary values—something is either completely true (1) or completely false (0). However, the real world is full of uncertainty, vagueness, and partial truths.

Fuzzy Logic handles this uncertainty by allowing for degrees of truth. Instead of just "True" or "False," fuzzy logic uses continuous values between 0 and 1. It represents concepts like "slightly," "very," or "almost." This allows AI systems to make decisions based on approximate reasoning, making them much more adaptable and capable of mimicking human-like decision-making in ambiguous situations.

Real-World Example: Automatic Washing Machine A traditional washing machine might only have strict settings (e.g., exactly 30 minutes of washing). An AI washing machine equipped with fuzzy logic uses sensors to detect the "degree" of dirtiness and the "volume" of the clothes.

• If the clothes are somewhat dirty and the load is heavy, the fuzzy logic controller will calculate an approximate, optimized washing time (e.g., 42 minutes) and water level. It doesn't rely on rigid rules; it adapts dynamically based on partial inputs.

---

Q 5: Discuss the principles and components of Genetic Algorithms, and its application areas?

Principles of Genetic Algorithms (GAs) Genetic Algorithms are search and optimization algorithms inspired by Charles Darwin’s theory of natural selection. The core principle is "survival of the fittest." The algorithm generates a population of possible solutions to a problem, evaluates how "good" each solution is, and then combines the best solutions to breed an even better generation of solutions, iterating until an optimal answer is found.

Components (Phases) of Genetic Algorithms

1. Initial Population: The process begins with a randomly generated set of individuals (possible solutions).
2. Fitness Function: This function evaluates each individual and assigns a fitness score, determining how well that solution solves the problem.
3. Selection: The fittest individuals are selected to pass their genes (characteristics) to the next generation.
4. Crossover (Recombination): The selected parents are paired up to swap genetic information, creating new offspring (new solutions) that share traits of both parents.
5. Mutation: To maintain diversity and prevent the algorithm from getting stuck, random, minor changes (mutations) are occasionally applied to the offspring.

Application Areas

• Optimization Problems: Finding the best possible schedule for a university or airline.
• Robotics: Evolving the control systems and walking gaits of robots.
• Route Planning: Solving complex logistical issues like the Traveling Salesperson Problem (finding the shortest route visiting multiple cities).
• Machine Learning: Tuning the hyperparameters of complex neural networks.

---

Q 6: Compare and contrast Breadth-first search and Depth-first search, including their advantages and limitations? How do heuristic functions enhance the efficiency of informed search strategies? Provide examples?

Breadth-First Search (BFS) vs. Depth-First Search (DFS)

Feature	Breadth-First Search (BFS)	Depth-First Search (DFS)
Strategy	Expands all nodes at the present depth level before moving deeper.	Expands the deepest unexpanded node first, going down a single path until the end.
Data Structure	Uses a Queue (First-In-First-Out / FIFO).	Uses a Stack (Last-In-First-Out / LIFO).
Advantages	- Complete: It will always find a solution if one exists.

- Optimal: Always finds the shallowest (shortest) path to the goal. | - Memory Efficient: Requires very little memory space as it only stores the current path. | | Limitations | - High Memory Requirement: Must store all nodes at the current level, which consumes massive memory for deep trees. | - Not Optimal: Might find a longer path when a shorter one exists.

- Not Complete: Can get stuck in infinite loops if trees are deeply nested. |

How Heuristic Functions Enhance Efficiency In informed search strategies (like A*), a heuristic function, often denoted as $h(n)$, is used to estimate the cost from the current node to the goal. Instead of exploring randomly like BFS or DFS, the algorithm calculates which path has the lowest estimated cost and prioritizes exploring that specific direction. This drastically reduces the search space, saving both time and computing power, because the algorithm "ignores" paths that look unpromising.

Example of a Heuristic Function In a map-based routing application (like Google Maps) trying to find the shortest driving path between two cities, a common heuristic is the Straight-Line Distance (Euclidean distance). The algorithm knows the "as the crow flies" distance between the current city and the destination. While roads might twist and turn, the algorithm prioritizes exploring highways that head in the general straight-line direction of the goal, vastly speeding up the search for the optimal route.