Artificial Intelligence Lecture 3 Quiz

Lecture 3: Artificial Intelligence 1

• PEAS Example 2: Spam Filter
  • Performance measure: Minimize false positives and false negatives
  • Environment: User's mail account, server's mail account
  • Actuators: Mark as spam, delete
  • Sensors: Incoming message, user's account profile

Goal-Based Agents vs. Cost-Based Agents

• Goal-based agents: Actions depend on the goal. For example, a robot moving from one room to another will have different actions than a robot moving to a different room.
• Cost-based agents: Goal is to minimize the cost of erroneous decisions over the long term. For example, a spam filter that aims to put all emails in correct categories.
  • Agent 1 makes fewer errors (12 out of 1000 emails) than Agent 2 (38 out of 1000). But the errors made by Agent 1 are more severe than those made by Agent 2, hence Agent 1 is better.
  • Errors in Agent 1 lose the user 11 potentially valuable emails. Errors in Agent 2 do not have as high of a negative impact on the user.

Specifying the Task Environment

• Discusses how to define task environments for AI agents

Environment Types

• Fully Observable vs. Partially Observable:
  • Fully observable: Agent's sensors provide complete environment state. All variables are known.
  • Partially observable: Agent's sensors do not provide complete environment state. Not all variables are known. Example is shown using robots: one robot can see the entire field, the other one doesn't.
• Deterministic vs. Stochastic:
  • Deterministic: Next state completely determined by current state and action.
  • Stochastic: Probability distribution over possible successor states.
  • Strategic: Deterministic except for other agents' actions
  • Example: Chess, where the next move is fully determined by one player's move, whereas in dice-based games (stochastic), the future action is not predetermined.
• Episodic vs. Sequential:
  • Episodic: Agent's experience divided into unconnected single decisions. Example: spam filter sorting emails (not dependent on prior decisions)
  • Sequential: Coherent sequence of observations and actions. Actions depend on past actions and experiences. Example: robot navigation, or games like chess or Pac-Man.
• Static vs. Dynamic (vs. Semi-dynamic):
  • Static: Environment doesn't change while agent is thinking
  • Dynamic: Environment changes while agent is thinking
  • Semi-dynamic: Environment doesn't change with time, but agent's performance score does change.
• Single-Agent vs. Multi-Agent:
  • Single-agent: Agent operates by itself in the environment.
  • Multi-agent: Environment has multiple agents interacting (cooperative or competitive). Example: an autonomous taxi driver
• Known vs. Unknown:
  • Known: Rules, rewards, and transition models associated with environmental states are known to the agent. Example: a board game with fixed rules
  • Unknown: Rules, rewards, and transition models are not known to the agent. Example: a maze with unknown paths, or a real-world environment with uncertain outcomes.

Exercises and Summary

• Categorized examples (Chess, Dominos, Medical Diagnosis, Self-driving Car) showing characteristics of environments.
• Questions to self-assess understanding of different types of task environments.
• Distinction between episodic and sequential environments.
• Communication as a rational behavior in multi-agent environments.

Test your understanding of Artificial Intelligence concepts covered in Lecture 3. This quiz focuses on PEAS examples, goal-based, and cost-based agents. Evaluate your knowledge on spam filters and decision-making in AI agents.