Artificial Intelligence SEM3 2025

Section II (Short Answer)

Q.2- i. Discuss the differences between Strong AI and Weak AI, and provide examples of each? Difference: Weak AI (also known as Narrow AI) is designed and trained to complete one specific task and operates within a limited context. It simulates human cognition but does not possess true intelligence. Strong AI (also known as General AI) is a theoretical form of AI that possesses human-level consciousness and cognitive abilities, allowing it to understand, learn, and apply knowledge across any domain, just like a human being.

• Example of Weak AI: Apple's Siri, self-driving cars, or chess-playing algorithms.
• Example of Strong AI: Does not currently exist, but examples in science fiction include HAL 9000 or the Terminator.

Q.2- ii. Differentiate between informed and uninformed search?

• Uninformed Search (Blind Search): This strategy searches the state space blindly. It possesses no domain-specific knowledge or clues about how far the current state is from the goal. It only knows how to generate successor states and check if a state is the goal. (e.g., Breadth-First Search).
• Informed Search (Heuristic Search): This strategy uses problem-specific knowledge, known as a "heuristic," to find the goal state more efficiently. The heuristic function estimates how close a state is to the goal, allowing the algorithm to prioritize the most promising paths. (e.g., A* Search).

Q.2- iii. What is the purpose of the Min-max algorithm? Purpose: The Min-max algorithm is a recursive decision-making strategy used primarily in artificial intelligence, game theory, and two-player turn-based games (like Tic-Tac-Toe or Chess). Its core purpose is to find the optimal move for a player by minimizing the maximum possible loss. It assumes the opponent is also playing optimally and calculates the best possible outcome under the worst-case scenario.

Q.2- iv. What is the importance of AI in game playing? Importance: Game playing serves as an excellent, controlled testing environment for artificial intelligence. Games have well-defined rules, clear objectives, and quantifiable success metrics. They require complex decision-making, forward planning, and strategic optimization. By developing AI that can master games, researchers create robust search algorithms and heuristics that can later be adapted to solve complex real-world problems, such as logistics planning, economic forecasting, and robotic navigation.

Q.2- v. Define the term 'state space'? Definition: In artificial intelligence, a 'state space' represents the set of all possible states or configurations that a problem can be in, from the initial starting point to the final goal state. Example: In a game of chess, the state space consists of every possible legal arrangement of pieces on the board. State spaces are typically visualized as mathematical graphs or trees, where nodes represent the states and edges represent the valid actions that transition the system from one state to another.

Q.2- vi. Describe forward and backward chaining? These are two different strategies used by inference engines in AI to draw conclusions from a rule base:

• Forward Chaining (Data-Driven): The AI starts with the known facts and applies logical rules to generate new facts until a specific goal is reached. Example: Fact = It is raining. Rule = If it rains, the ground is wet. Conclusion = The ground is wet.
• Backward Chaining (Goal-Driven): The AI starts with a specific goal or hypothesis and works backward through the rules to see if there is enough data to prove that goal is true. Example: Goal = Is the ground wet? Rule = It is wet if it rains. Let's check the data to see if it is raining.

Q.2- vii. How can AI be used in detecting fraud? AI is a powerful tool for fraud detection because of its ability to analyze massive datasets and recognize complex patterns in real time. How it works: Machine learning models analyze a user's historical data to establish a baseline of "normal" behavior (e.g., typical spending amounts, locations, and login times). When a new transaction occurs, the AI instantly evaluates it against this baseline. If an anomaly is detected—such as a credit card being used in a foreign country immediately after a local purchase—the AI flags the transaction as fraudulent and blocks it to prevent financial loss.

Section III (Essay Type)

Q.3: Explain the architecture and components of a knowledge-based system?

Introduction to Knowledge-Based Systems (KBS) A Knowledge-Based System is a computer program that uses a centralized repository of knowledge to solve complex problems, typically in a specific domain, mimicking the decision-making process of a human expert.

Core Components of a KBS Architecture:

1. Knowledge Base: This is the heart of the system. Unlike a standard database that only stores raw, tabular data, a knowledge base stores rules, facts, relationships, and heuristics about a specific domain. It is typically constructed using "IF-THEN" rules.
2. Inference Engine: This is the logical "brain" of the architecture. It is a processing mechanism that takes the data provided by the user, searches the knowledge base, and applies logical reasoning (using forward or backward chaining) to deduce new information, answer questions, or solve the problem.
3. User Interface: This is the communication layer. It allows non-expert users to interact with the system. The user inputs their queries or problem details here, and the system presents its conclusions or advice back to the user in an understandable format.
4. Explanation Module (Optional but common): A crucial feature that builds trust. This module explains to the user how and why the system arrived at a specific conclusion by tracing the rules triggered in the inference engine.
5. Knowledge Acquisition Facility: A module that allows human domain experts or knowledge engineers to add, update, or refine the rules and facts within the knowledge base, ensuring the system stays current.

Q.4: Discuss the use of case studies like General Problem Solver and ELIZA in understanding AI concepts?

Introduction Analyzing early historical AI programs like the General Problem Solver (GPS) and ELIZA is crucial for students because these case studies highlight the foundational methodologies of artificial intelligence, as well as the inherent limitations of early symbolic AI.

1. General Problem Solver (GPS) - 1957

• Concept Taught: Means-Ends Analysis. GPS was designed by Newell and Simon to imitate human problem-solving protocols. It introduced the concept of "Means-Ends Analysis," a strategy where the AI identifies the difference between its current state and the goal state, and then creates "sub-goals" to incrementally reduce that difference.
• Lessons Learned: GPS successfully solved formalized logic puzzles (like the Towers of Hanoi). However, it taught researchers a valuable lesson about the limitations of AI: logical search algorithms are not enough on their own. GPS failed at real-world tasks because it lacked domain-specific knowledge and common sense, paving the way for the development of Knowledge-Based Systems.

2. ELIZA - 1966

• Concept Taught: Natural Language Processing (NLP). Created by Joseph Weizenbaum, ELIZA was an early program designed to mimic a Rogerian psychotherapist. It worked by recognizing keywords in a user's input and using simple pattern matching and substitution to turn the user's statements back into questions.
• Lessons Learned (The ELIZA Effect): ELIZA was incredibly simple and had absolutely no understanding of what it was saying. Yet, users became emotionally attached to it, believing it genuinely understood them. This case study teaches the "ELIZA Effect"—the psychological tendency of humans to anthropomorphize computers and assume machines possess deep intelligence and empathy when they are merely manipulating symbols.

Q.5: Compare and contrast informed and uninformed search strategies in detail and write 2 search strategies from each informed and uninformed?

Detailed Comparison Search strategies are essential for navigating problem state spaces. They are fundamentally divided by whether or not they use clues to find the goal.

Examples of Search Strategies:

Uninformed Search Strategies:

1. Breadth-First Search (BFS): Explores the search tree level by level. It checks all the immediate neighbor nodes at the current depth before moving down to the next level. It guarantees finding the shortest path if all step costs are equal.
2. Depth-First Search (DFS): Explores as far down a single branch of the search tree as possible until it hits a dead end, then backtracks to the last unexplored node. It uses less memory than BFS but can get stuck in infinite loops.

Informed Search Strategies:

1. A Search (A-Star):* One of the most popular search algorithms. It calculates the total cost of a path using the formula f(n) = g(n) + h(n), where g(n) is the exact cost from the start node, and h(n) is the heuristic estimate to the goal. It finds the optimal path highly efficiently.
2. Greedy Best-First Search: This algorithm ignores the cost traveled so far. It strictly expands the node that is estimated to be closest to the goal based entirely on the heuristic function h(n). It is fast but not always guaranteed to find the absolute best or shortest path.