CSE 332: AI and Intelligent Agents Intro

• The lecture introduces CSE 332: Artificial Intelligence and Intelligent Agents.
• The instructor is Dr. Belal Badawy for Spring 2025.

Assessments Weight

• Attendance, participation and assignments are 10% of the final grade.
• Quizzes also account for 10%.
• Midterms comprise 15%, with one midterm exam.
• Practical exam and project contribute 25%.
• The final exam is worth 40%.
• Total assessment weight equals 100%.

Outlines

• Covers introduction to AI, agents, rational agents, challenges of rational agents, PEAS (Performance measure, Environment, Actuators, Sensors), and environment types.

Introduction to Artificial Intelligence

• Intelligence is defined as the ability to learn about, understand, and interact with one's environment.
• Artificial Intelligence involves making a computer think.
• AI focuses on designing intelligent computer systems exhibiting characteristics associated with human behavior.
• AI requires learning: acquiring knowledge and rules, and reasoning: applying rules.

AI Principles

• AI principles include the data structures, algorithms, and programming techniques for knowledge representation and implementation.
• Goals of AI research include reasoning, knowledge, planning, learning, natural language processing, perception, and object manipulation.
• Problem reduction is breaking down one hard problem into multiple simple problems.

Characteristics and Applications of AI

• AI has a high societal impact, affecting billions of people.
• AI is diverse, covering areas like language, vision, and robotics.
• Considered complex
• Applications include game playing, speech recognition, natural language understanding, computer vision, and expert systems.

Types of AI

• Artificial Narrow (Weak) AI is task-specific and currently exists, such as Siri, Amazon’s Alexa, and Open AI’s Chat GPT.
• Artificial General (Strong) AI is theoretical, able to perform any intellectual task a human can.
• Super AI is also theoretical, surpassing human cognitive abilities in reasoning, learning, and judgment.

Agents

• An agent is anything that perceives its environment through sensors and acts through actuators.
• In a human agent, eyes and ears are sensors; hands, legs and mouth are actuators.
• In a robotic agent, cameras and infrared range finders are sensors; motors are actuators.
• The agent function maps percept histories to actions, represented as f: P* → A.
• The agent program runs on the architecture expressed as agent = architecture + program.

Vacuum-cleaner World Example

• Percepts include location and contents, like [A, Dirty].
• Actions are Left, Right, Suck, and No Op.
• The agent uses a look-up table.

Rational Agents

• Rationality involves measuring success, agent's knowledge, possible actions, and percept sequence.
• A rational agent selects actions maximizing performance measure using evidence from percept sequence and built-in knowledge.
• Rational decisions are made by searching for the best path.

Challenges of Rational Agents

• Rationality differs from omniscience because percepts might not supply all information, and card games hide cards.
• Rationality aims for expected outcome, while perfection aims for actual outcome.
• Autonomy is the extent of behavior determined by experience, not designer knowledge.

PEAS (Performance, Environment, Actuators, Sensors)

• Setting must be specified for intelligent design
• PEAS involves performance measure, environment, actuators, and sensors.
• Example: automated taxi driver with safety and profit. It drives on the road using steering wheel and sensors.
• Part-picking robot maximizes correct parts using arm, hand and camera sensors.
• Interactive English tutor maximizes student’s test score using keyboard and screen displays.

Environment Types

• Fully observable vs. Partially observable: Is everything an agent requires to choose its actions available to it via its sensors?
• Deterministic vs. Stochastic: Does the change in world state depend only on current state and agent's action?
• Episodic vs. Sequential: Is the choice of current action dependent on previous actions?
• Static vs. Dynamic: Do static environments not change while dynamic environments change?
• Discrete vs. Continuous: Is there a limited number of distinct, clearly defined percepts?
• Single agent vs. Multi-agent: an agent operating by itself or many operating together?