Introduction to Artificial Intelligence (AI)

• The document provides an introduction to Artificial Intelligence (AI).

Introduction to AI

• AI involves making computers think.
• It automates activities associated with human thinking, such as decision-making and learning.
• AI is the art of creating machines to perform functions requiring intelligence
• Also considered a branch of computer science for automating intelligent behavior.

"Thinking" in AI, 4 types

• AI can involve thinking like people, acting like people, thinking rationally, and acting rationally.
• "Thinking like people" is the effort to make computers think with minds, "in the full and literal sense."
• Automation of activities associated with human thinking such problem solving, learning etc
• "Thinking rationally" involves the study of mental faculties using computational models.
• Furthermore studying computations that enable perception, reasoning, and action.
• "Acting like people" means creating machines performing functions that require human intelligence.
• "Acting rationally" is designing intelligent agents by being concerned with intelligent behavior in artifacts.
• Human minds are effective at rational decisions but have limits.
• The brains are difficult to reverse engineer because they lack modularity like software

Rationality in AI

• "Rational" in AI is used in a technical sense.
• Rationality in AI is assessed based on decisions using utility
• Rational decisions are those that optimally achieve predefined goals.
• Rationality concerns decisions themselves, not how these decisions are reached
• Being rational means maximizing expected utility
• Computational rationality is key

Maximising Utility in AI

• Maximizing includes consequences, context, and other actors.
• Utility represents an individual's preferences over a set of objects or options.
• Expected utility is crucial, focusing on the right action on average, not just success.

Rational Agents

• These agents can perceive and act, selecting actions which maximize expected utility,
• The agent's percepts, the environment, and the action space impacts rational action selection.

Defining AI,

• Al is the study of ideas enabling computers to be intelligent
• AI design of computer systems exhibit human intelligence

The Two Major Roles of AI

• Study intelligent aspects like humans.
• Represent intelligent actions via computers.

History of AI

• 1940-1950: Early days, Boolean circuit model of the brain.
• 1950: Alan Turing publishes the “Computing Machinery and Intelligence”
• 1950-1970: Early programs, include Samuel's checkers, Newell & Simon's Logic Theorist.
• 1956: Dartmouth meeting, term “Artificial Intelligence" coined
• 1965: Robinson's complete algorithm for logical reasoning developed
• 1969-79: Knowledge-based system developed early
• 1980-88: Expert systems industry booms
• 1988-93: Expert systems industry busts, known as “AI Winter”
• 1990: Resurgence of probability, agent-focused and general increase technical depth.
• Agents and learning systems lead to speculation of an “AI Spring”.
• 2000 : Computation with artificial neurons, modeling the brain with artificial neurons.
• AI systems are now easier to train with examples, reducing manual programming
• Most AI breakthroughs rely on Machine Learning.

Common AI Applications

• Search (includes Game Playing)
• Representing Knowledge and Reasoning with it
• Planning
• Learning
• Expert Systems
• Natural language processing
• Interacting with the Environment (e.g. Vision, Speech recognition, Robotics)

Types of Natural Language Applications

• Speech technologies like Siri using Automatic Speech Recognition (ASR).
• In addition, uses Text-to-Speech Synthesis (TTS) and Dialog systems
• Language processing tech for answering questions, Machine translation
• Web search tools
• Text classification or spam filtering

Applications using Computer Vision:

• Autonomous rovers that can traverse environments through visual perception.
• Identifying Iris patterns and matching photos

Expert systems

• Used for tools, predictions, decisions

Robotics

• This field uses mechanical engineering and AI for real world automation
• Applications include use in vehicles, rescue operations, and home assistance.

Core AI Technologies

• Inference,Reasoning
• Knowledge representation
• Natural language processing
• Speech recognition and synthesis
• Computer vision, image processing
• Multi-agent system
• Machine Learning
• Data mining

Intelligent Agents and AI

• An "agent", an entity that perceives and acts in an given environment
• The entity will use sensors and actuators
• Actuators impact the environment, which affects the input.
• Example are cars
• Sensors used are camera, lidar, speed gauge
• Actuators used are gas petal, steering wheel, brake petal

Agent types

• Reflex agents choose actions based on the current input, and memory
• Planning agents, where decisions are based on (hypothesized) consequences of actions.

Rational Agents

• Choose actions to maximize the expected utility.
• Characteristics are based on percepts, environment, and dictate action selection.

The Task Environment (PEAS)

• Performance measure: How well do actions achieve results
• Environment: Defines the surroundings.
• Actuators: Parts that act
• Sensors: Parts the perceive
• In designing an agent, one must specify PEAS as fully as possible

Properties of task environments

• Fully/partialy observable or agents may request memory.
• A single-agent and multi-agent must behave randomly - otherwise contingencies arise.
• Discrete or continuous - All environment states must be accounted

Agent Structure

• AI designs agent programs to implement agent functions like mapping.
• Programs run on computing devices, combining physical sensors and actuators to set architecture.
• Agent = architecture + program.

Kinds of agents

• Simple reflex agents (tác tử phản xạ đơn giản)
• MOdel based reflex agents (tác tử phản xạ dựa trên mô hình)
• Goal-based agents （tác tử dựa trên mục tiêu）
• Utility-based (tác tử dựa trên lợi íc）
• Learning agents （tác tử có khả năng học）