Foundations of Artificial Intelligence
48 Questions
0 Views

Choose a study mode

Play Quiz
Study Flashcards
Spaced Repetition
Chat to lesson

Podcast

Play an AI-generated podcast conversation about this lesson

Questions and Answers

Which stage of Computer Vision involves preparing data for analysis?

  • High-Level Processing
  • Acquisition
  • Decision Making
  • Pre-Processing (correct)

    • What is one of the primary applications of thresholding in Computer Vision?

  • High-Level Processing
  • Data Acquisition
  • Legal Compliance
  • Anomaly Detection (correct)

    • What key concept in AI highlights the issue of pre-existing biases in algorithm outputs?

  • Accountability
  • Transparency
  • Fairness and bias (correct)
  • Privacy Rights

    • In the context of AI ethics, what does transparency refer to?

    <p>Understanding decision-making processes</p> Signup and view all the answers

    Which of the following is NOT a component in the stages of Computer Vision?

    <p>Regulatory Compliance</p> Signup and view all the answers

    What does the EU AI Act primarily focus on concerning high-risk AI applications?

    <p>Strict regulatory measures</p> Signup and view all the answers

    Which application of Computer Vision is primarily concerned with identifying defects in products?

    <p>Defect Detection</p> Signup and view all the answers

    What main concern is associated with the Privacy and Data Rights concept in AI?

    <p>Data collection and consent</p> Signup and view all the answers

    What function does A* Search use to determine the best path?

    <p>f(n) = g(n) + h(n)</p> Signup and view all the answers

    Which local search algorithm allows worse moves early on to avoid local maxima?

    <p>Simulated Annealing</p> Signup and view all the answers

    What are axioms in the context of a knowledge base?

    <p>Sentences that were not derived from other sentences</p> Signup and view all the answers

    Which local search algorithm is most focused on exploring the solution space?

    <p>Genetic Algorithm</p> Signup and view all the answers

    What type of operation is TELL in regard to a knowledge base?

    <p>Adds new sentences to the knowledge base</p> Signup and view all the answers

    Which property means that an inference system can derive all sentences that are entailed?

    <p>Completeness</p> Signup and view all the answers

    Which informed search algorithm expands nodes based on the smallest heuristic value?

    <p>Greedy Best-First Search</p> Signup and view all the answers

    What does the term 'satisfiability' refer to in knowledge representation?

    <p>A sentence true in at least one model</p> Signup and view all the answers

    Which of the following best defines Artificial Intelligence?

    <p>The development of computer systems able to perform tasks requiring human intelligence.</p> Signup and view all the answers

    What distinguishes narrow AI from general AI?

    <p>Narrow AI solves specific tasks while general AI can apply knowledge across various tasks.</p> Signup and view all the answers

    In the context of AI, what constitutes an intelligent agent?

    <p>An entity that can perceive its environment and act autonomously.</p> Signup and view all the answers

    What is included in the main research areas of AI?

    <p>Reasoning, Learning, Problem Solving, and Perception.</p> Signup and view all the answers

    Which of the following is NOT a stated application of AI?

    <p>Fashion Design</p> Signup and view all the answers

    What characterizes a rational agent in AI?

    <p>An agent that aims to achieve the best possible outcome based on knowledge and goals.</p> Signup and view all the answers

    What is the main goal of satisficing planning in AI?

    <p>To find a solution that meets basic requirements without seeking the best outcome.</p> Signup and view all the answers

    What does the percept sequence refer to in the context of intelligent agents?

    <p>A chronological record of all data received by the agent from its sensors.</p> Signup and view all the answers

    Which technique is primarily used to ensure the optimality of a solution in AI planning?

    <p>A* search with an admissible heuristic.</p> Signup and view all the answers

    Which aspect of AI is primarily focused on creating routines to handle unexpected scenarios?

    <p>Problem Solving</p> Signup and view all the answers

    What does the closed world assumption in propositional STRIPS state?

    <p>No facts outside the known set can be true.</p> Signup and view all the answers

    What is contingent planning primarily concerned with?

    <p>Dealing with the uncertainty of actions' outcomes.</p> Signup and view all the answers

    Which of the following represents a challenge faced by an AI planning agent?

    <p>The agent may encounter stochastic outcomes from actions.</p> Signup and view all the answers

    In the context of plan-space search, what is the primary objective?

    <p>To find a valid plan from a graph of partial plans.</p> Signup and view all the answers

    Which statement best describes optimal planning?

    <p>It seeks to identify the best possible solution as per specific criteria.</p> Signup and view all the answers

    Which aspect of an AI planning system is affected by incorrect knowledge?

    <p>The reliability of action outcomes.</p> Signup and view all the answers

    What is the primary goal of a Markov Decision Process (MDP)?

    <p>To maximize rewards over time.</p> Signup and view all the answers

    Which of the following is true about Passive Reinforcement Learning (RL)?

    <p>The agent learns to evaluate a fixed policy only.</p> Signup and view all the answers

    In the context of exploration vs exploitation, what does exploration allow an agent to do?

    <p>To visit states that have not been adequately studied.</p> Signup and view all the answers

    What characterizes Active Reinforcement Learning compared to Passive RL?

    <p>The agent learns both the policy and the value function.</p> Signup and view all the answers

    Which aspect of Greedy Temporal Difference Learning is particularly emphasized?

    <p>Always selecting the action currently believed to be the best.</p> Signup and view all the answers

    How does the utility of a state factor into Reinforcement Learning?

    <p>It determines the value of applying an action in that state.</p> Signup and view all the answers

    In Reinforcement Learning, what does the term 'rewards' refer to?

    <p>The points earned for reaching a particular state.</p> Signup and view all the answers

    Which scenario best describes the limitation of Passive Reinforcement Learning?

    <p>The agent may never explore certain states due to sticking to one policy.</p> Signup and view all the answers

    What is a characteristic of a valid sentence in First Order Logic (FOL)?

    <p>It is true in all models of the environment.</p> Signup and view all the answers

    Which elements are part of the model of the environment in FOL?

    <p>Set of objects, functions, and relations.</p> Signup and view all the answers

    How do planning and scheduling differ?

    <p>Planning identifies tasks while scheduling identifies the timing of those tasks.</p> Signup and view all the answers

    What distinguishes Domain Specific planning from Domain Independent planning?

    <p>Domain Specific planning is tailored for particular problems within specific domains.</p> Signup and view all the answers

    Which term in FOL refers specifically to an object?

    <p>Constant Symbols</p> Signup and view all the answers

    What does AI Planning primarily involve?

    <p>Making autonomous decisions to achieve goals.</p> Signup and view all the answers

    In FOL, what does a term represent?

    <p>A logical expression that refers to an object.</p> Signup and view all the answers

    Which of the following is NOT a real-world application of AI Planning?

    <p>Decision-making algorithms for stock trading.</p> Signup and view all the answers

    Study Notes

    Foundations of Artificial Intelligence

    • Artificial Intelligence (AI) is a field that studies the synthesis and computational agents that act intelligently. AI focuses on creating computers and programs that act intelligently.
    • AI is defined as the theory and development of computer systems able to perform tasks normally requiring human intelligence.
    • Examples of intelligence include logical reasoning, problem-solving, creativity, and planning.
    • Narrow AI requires reconfiguration or a new algorithm for each task (e.g., chess, speech recognition, facial recognition).
    • General AI applies intelligent systems to any problem, understanding, learning, and applying knowledge across a broad range of tasks.
    • Main research areas of AI include reasoning, learning, problem-solving, and perception.
    • Applications of AI include robotics (industrial, autonomous, domestic), industrial automation, healthcare (drug design, operating theatre robotics), games, virtual and augmented reality, education, agriculture, and personal assistance.

    Intelligent Agents

    • An agent is anything that perceives its environment through sensors and acts upon that environment through actuators.
    • Intelligent agents act autonomously to achieve goals.
    • A percept sequence is the complete history of all data an agent receives from its sensors.
    • Agents are judged by how they act.
    • Rational agents act to achieve the best possible outcome, maximizing performance based on knowledge and goals.
    • Agent rationality involves prior knowledge, performable actions, the percept sequence to date, and a success criterion.
    • Observability is when not all necessary information is available for an agent to decide, leading to partially observable environments.

    Stochasticity and Discrete/Continuous Environments

    • Stochasticity refers to randomness or unpredictability in a system or process.
    • Deterministic actions have consistent outcomes.
    • Stochastic actions have variable outcomes.
    • Discrete environments have a finite number of action choices and states (e.g., chess).
    • Continuous environments have infinite possibilities (e.g., tennis).
    • Adversarial environments involve competing agents (e.g., chess).
    • Benign environments have no competing agents (e.g., predicting the weather).
    • Types of agents include reflex, model-based reflex, goal-based, utility-based, and learning agents.

    Search Techniques

    • Search is a powerful technique for solving AI problems.
    • Problems are often formulated as directed graphs.
    • Solutions involve action sequences achieving the goal and satisfying constraints.
    • Blind search (uninformed) algorithms explore the space without prior knowledge, often expanding each state to find all successors until the desired state is found.
    • Breadth-first search explores the search space level by level, using FIFO.
    • Depth-first search explores deeply, using LIFO.
    • Informed search uses heuristic functions (e.g., A*, Greedy Best-First Search) to prioritize more promising paths.
    • Iterative Deepening Search combines depth-first search's space efficiency with breadth-first search's completeness, limiting depth until the goal is found.
    • Weighted A* gives a weight to the heuristic component of evaluation function.

    Modelling Challenges

    • Algorithmic complexity measures computational resources needed by an algorithm in relation to input size (often expressed with Big O notation).
    • Includes time complexity and space complexity.

    Knowledge and Reasoning

    • Knowledge in AI encompasses data, information, and concepts used by AI systems to understand the world and solve problems.
    • Reasoning is using knowledge to draw conclusions or make decisions.
    • Knowledge-based agents use a knowledge base and an inference engine to make decisions and solve problems, updated with new knowledge.
    • A knowledge base contains statements in a knowledge representation language.
    • Axioms are statements not derived from others.
    • TELL operations add sentences, ASK queries the base. Inference rules derive new sentences.
    • Sound inference derives only entailed sentences. Complete inference derives all entailed sentences.
    • Satisfiability means a statement is true in at least one model. Validity means a statement is true in all models.
    • First-Order Logic (FOL) represents objects, functions of these objects, and their relationships in an environment.

    Planning

    • Planning involves devising strategies to achieve a desired goal state by choosing actions to maximize the probability of achieving it.
    • Planning involves identifying tasks to achieve an objective (goals). Scheduling involves choosing the best time to perform these tasks.
    • AI planning refers to intelligent systems making decisions to achieve goals. This can be applied in various contexts (e.g., industrial automation, autonomous driving).

    Planning vs Scheduling

    • Planning focuses on identifying actions and tasks that need to be performed to achieve a goal, whereas scheduling specifies when each action or task should occur.
    • Planning without scheduling is not actionable.

    Domain-Specific vs Domain-Independent Planning

    • Domain-specific planning focuses on solving problems within a particular domain, tailoring the techniques to that specific area.

    Knowledge-Based Agents

    • A knowledge-based agent maintains a knowledge base updated and queried to infer new information about the environment. Algorithms use to solve planning problems.

    Reinforcement Learning

    • A machine learning technique that explores and observes the effects of actions to discover methods that maximize rewards. Actions are evaluated by the effects on the environment.

    Stochastic Games

    • Stochastic actions lead to multiple possible states with varying probabilities.

    Game Theory

    • Studies strategic interactions between players, helping understand and predict decision-making in competitive or cooperative scenarios.

    Multi-Agent Systems

    • How multiple intelligent agents interact within an environment.
    • Dynamics between agents can be benign (interfering but not competing), cooperative (working towards shared goals), or adversarial (competing).

    Perception

    • Understanding how systems can interpret and use data from sensors, including the reliability and persistence of information.
    • Raw data from sensors needs processing. AL algorithms interpret raw sensor data into meaningful information.

    Online State Estimation

    • Determining a system's most likely current state in real time.
    • Methods include filtering algorithms, such as Bayes filters, that estimate a probability distribution over all possible states, using sensor data and previous state estimates.

    Particle Filters

    • Help a robot figure out its location by using many guesses (particles) of possible positions and refining them over time.
    • Particles representing possible locations.
    • Particles consistent with sensor data are kept and duplicated.
    • Over time, these cluster around the robot's current position.

    State Representation

    • How the current state of a system or environment is described.
    • Kinematic states represent position, orientation, velocity, and acceleration without considering forces or masses.
    • Dynamic states represent position, orientation, velocity, acceleration, and forces or masses.
    • Monte Carlo localization is used in robotics to estimate a robot's location in a map using particles.

    Planning to Perceive

    • Sensing actions: Actions related to gathering information from the environment.
    • Importance: Critical for accurate decision-making and interaction in robotic systems.
    • Anchoring process: A method to keep consistent the correspondence between symbolic objects and sensor data.

    Computer Vision

    • Computer vision automatically extracts, analyzes, and interprets images or videos.
    • Converts image data into numerical arrays allowing using machine learning algorithms which can generate new images.

    Law and Ethics in AI

    • Fairness and algorithmic bias: AI can perpetuate existing social biases.
    • Transparency and explainability: Complex algorithms are often “black boxes”, making it difficult to understand how decisions are made.
    • Privacy and data rights: Collecting and using data raises concerns about individual rights.
    • Accountability and responsibility: Determining who is responsible for AI actions if they cause harm.
    • EU AI Act: Provides rules and regulations for high-risk AI applications in Europe.

    Mechanism Design

    • The set of techniques used to design the rules of games to incentivize agents to behave in ways that achieve specific goals.

    Machine Learning

    • A subfield of AI, focusing on discovering models from data, and making use of statistical methods.
    • Example types of machine learning: supervised learning (data is pre-classified, learned from examples), unsupervised learning (data not pre-classified, learning by observing similarities), reinforcement learning (observe the effects of different actions).
    • Classification: Assigning categories.
    • Regression: Predicting continuous values (e.g., pricing).
    • Prediction: Forecasting future data.
    • Recommendation: Suggesting relevant items.

    Studying That Suits You

    Use AI to generate personalized quizzes and flashcards to suit your learning preferences.

    Quiz Team

    Related Documents

    Foundations of Artificial Intelligence PDF

    Description

    Explore the basics of artificial intelligence in this quiz. Understand the key concepts, types of AI, and their applications across various fields. Test your knowledge on narrow and general AI, as well as the main research areas within AI.

    More Like This

    Use Quizgecko on...
    Browser
    Open
    Browser
    Browser