AL-304 Artificial Intelligence / Introduction to Artificial Intelligence - Dec 2024 Question Paper

Under each question, How to answer lists the points an answer should cover, outlined by CampusPrep from the syllabus. Check the details against your textbook.

1. 7 Marks
   aWhy is AI considered an interdisciplinary field, and how does it incorporate knowledge from multiple domains?
   How to answer

   • Definition of Artificial Intelligence and its core objective
   • Contributions from Computer Science and Mathematics/Statistics (algorithms, logic, probability)
   • Contributions from Cognitive Science, Psychology, and Neuroscience (modeling human brain/thought)
   • Contributions from Linguistics (NLP) and Philosophy (nature of mind, ethics)
   • Summary of how interdisciplinary collaboration advances AI systems
2. 7 Marks
   bCompare and contrast hill climbing and best-first search algorithms. What are their strengths and weaknesses?
   How to answer

   • Overview/principles of Hill Climbing (local search, greedy, no open list maintained)
   • Overview/principles of Best-First Search (evaluates nodes with priority queue/open list)
   • Tabular comparison (memory usage, handling local extrema/plateaus, completeness)
   • Strengths: Hill climbing is space-efficient; Best-first can backtrack and avoid simple local traps
   • Weaknesses: Hill climbing gets stuck on local optima/ridges; Best-first can have high memory overhead
3. 7 Marks
   aExplore the potential risks and limitations of heuristic search algorithms in AI. How can these risks be mitigated?
   How to answer

   • Definition of heuristic search and reliance on heuristic functions $h(n)$
   • Limitations/Risks: sub-optimality, incompleteness, local minima/plateaus, deceptive heuristics, overestimation
   • Mitigation strategies: designing admissible/consistent heuristics, combining heuristics (max-heuristic)
   • Algorithmic mitigations: random restart, simulated annealing, beam search, memory-bounded algorithms (IDA*, SMA*)
4. 7 Marks
   bList and explain some common problems and challenges in representing knowledge in AI systems.
   How to answer

   • Definition and significance of Knowledge Representation (KR)
   • Key challenges: Frame problem, Qualification problem, Ramification problem
   • Representation of uncertainty, vagueness, and common-sense knowledge
   • Trade-off between expressiveness and computational tractability/efficiency
   • Dynamic knowledge updating and non-monotonicity challenges
5. 7 Marks
   aGive examples of scenarios where non-monotonic reasoning is essential in AI applications. What distinguishes it from monotonic reasoning?
   How to answer

   • Distinction between monotonic reasoning (conclusions remain valid upon new information) and non-monotonic reasoning (conclusions can be retracted)
   • Key properties/theoretical foundations (default logic, circumscription)
   • Scenarios/examples: Tweety the bird (default that birds fly, retracted if Tweety is a penguin)
   • Scenarios in medical diagnosis (symptom changes revising diagnosis) and automated driving/robot navigation
6. 7 Marks
   bEvaluate the strengths and weaknesses of different inference methods (e.g., forward chaining, backward chaining) in AI and provide scenarios where each is more appropriate.
   How to answer

   • Definition of forward chaining (data-driven reasoning from facts to goals)
   • Definition of backward chaining (goal-driven reasoning from hypothesis to facts)
   • Strengths and weaknesses of forward chaining (thorough, but may generate irrelevant facts)
   • Strengths and weaknesses of backward chaining (focused, but may fail if multiple unrelated goals exist)
   • Scenarios: Forward chaining for monitoring/synthesis/production scheduling; Backward chaining for diagnosis/debugging/proving theorems
7. 7 Marks
   aWhat are semantic networks in the context of knowledge representation? Provide an example to illustrate their use.
   How to answer

   • Definition of semantic network (graph structure: nodes represent concepts/objects, arcs represent relationships)
   • Types of relationships (is-a, part-of, has-a, property inheritance)
   • Illustrative example diagram (e.g., Bird -> is-a -> Animal, Sparrow -> is-a -> Bird, Sparrow -> can -> fly)
   • Advantages and limitations of semantic networks
8. 7 Marks
   bAnalyze a real-world problem and suggest how frames can be used to represent the knowledge required to solve it effectively?
   How to answer

   • Concept of Frames in KR (collection of attributes/slots and associated values/fillers, procedures/demons)
   • Selection of a real-world problem (e.g., Medical patient diagnosis or University course registration)
   • Structure of frames for the chosen problem (e.g., Patient frame with slots for Age, Symptoms, History, Default treatments)
   • Demonstration of slot inheritance and procedure attachments (if-needed, if-added)
   • Analysis of how this structured representation solves the problem efficiently
9. 7 Marks
   aInvestigate the role of scripts, schemas and frames in AI chatbots and virtual assistants. Assess their potential for improving user interactions.
   How to answer

   • Definitions: Scripts (stereotypical sequence of events), Schemas (cognitive framework for organizing knowledge), Frames (slot-filler data structures)
   • Application of scripts in conversation flow (e.g., restaurant reservation, flight booking scripts)
   • Application of frames in slot-filling during dialogue (extracting destination, date, time)
   • Application of schemas in context tracking and expectation handling
   • Assessment of impact: higher conversational coherence, fewer repetitive queries, personalized and context-aware responses
10. 7 Marks
    bExplain the minimax procedure in the context of game playing. What is its primary objective, and how does it work in games like chess?
    How to answer

    • Definition and objective of Minimax algorithm (finding the optimal move assuming an optimal opponent)
    • MAX and MIN layers explained with mathematical formulation (maximizing own payoff, minimizing opponent's max payoff)
    • Application in Chess: state space representation, utility/static evaluation function for non-terminal cutoffs
    • Backtracking/propagation of utility values up the game tree
    • Illustration of a small 2-ply game tree
11. 7 Marks
    aPropose a variation of the alpha-beta pruning algorithm that further optimize the search process in game trees and assess its potential advantages.
    How to answer

    • Brief overview of standard Alpha-Beta pruning and its dependence on move ordering
    • Proposed optimization/variation (e.g., Principal Variation Search / NegaScout, Transposition Tables, or Iterative Deepening with Killer Heuristic)
    • Working mechanism of the proposed variation
    • Comparative theoretical complexity: reduction from $O(b^{3d/4})$ to near optimal $O(b^{d/2})$
    • Assessment of advantages: earlier pruning, memory reuse, deeper search within time limits
12. 7 Marks
    bInvestigate recent advancements in robotic systems for solving the block world problem. Discuss the innovations and technologies that have improved performance.
    How to answer

    • Traditional blocks world: symbolic AI, STRIPS representation (ON, ONTABLE, CLEAR, HOLDING, ARMEMPTY)
    • Limitations of classical approaches in real robotics (uncertainty, perception, physical manipulation)
    • Recent advancements: Deep Reinforcement Learning (DRL) for continuous control and policy learning
    • Integration of Computer Vision / CNNs for real-time 3D pose estimation and object detection
    • Simulation-to-real transfer (Sim2Real), tactile sensing, and foundation models in robotic manipulation
13. 7 Marks
    aDefine what Expert Systems (ES) are and explain their role in artificial intelligence and problem-solving?
    How to answer

    • Definition of Expert Systems (ES) as domain-specific knowledge-intensive AI software
    • Key architectural components (Knowledge Base, Inference Engine, User Interface, Explanation Facility)
    • Role in AI: emulation of human expert decision-making and non-algorithmic problem solving
    • Applications/examples (MYCIN, DENDRAL, PROSPECTOR)
    • Benefits in commercial and technical domains
14. 7 Marks
    bCompare and contrast the inference engines used in expert systems: forward chaining and backward chaining. Provide examples of scenarios where each is more suitable.
    How to answer

    • Role of an inference engine in rule-based expert systems
    • Forward chaining mechanism: match-resolve-act cycle starting from asserted premises
    • Backward chaining mechanism: hypothesis-driven subgoaling to discover supporting facts
    • Detailed comparison table highlighting speed, focus, and search space characteristics
    • Specific examples/scenarios: Forward chaining for configuration/monitoring; Backward chaining for fault diagnosis/troubleshooting
15. 14 Marks
    Write short notes on any two of the following: a) Benefits of Expert Systems b) Components of NLP c) Compare and contrast conceptual dependency analysis and semantic networks d) Importance of knowledge representation
    How to answer

    • a) Benefits of Expert Systems: high availability, consistency, reduced cost, permanent documentation of expertise, rapid response
    • b) Components of NLP: Lexical analysis, Syntactic analysis (parsing), Semantic analysis, Discourse integration, Pragmatic analysis
    • c) CD vs Semantic Networks: CD uses universal primitive actions (ATRANS, PTRANS, etc.) focusing on meaning independent of language; Semantic networks model general entity relations (is-a, has-a)
    • d) Importance of KR: bridge between human intelligence and machine processing, enables reasoning, modular knowledge maintenance, decision making