Artificial Agents

Artificial Intelligence

What is an Artificial Agent? An Artificial Person. It can be a hardware (robot) or a software solution.

Humans vs Agents

  • Humans need to perceive various things happening around them (like sound, smell, etc)

  • To do that humans have various sensory organs.

  • Similarly, Agents have various sensors like light, sound, etc.

  • After a human perceives something, they need to execute some action.

  • Humans usually do that using their other organs like legs, arms, etc.

  • Similarly, Agents have something called effectors that carry out tasks on the basis of the inputs received from the sensors.

What do we want Agents to do/ become?

  • The Agents should be Rational. They should be able to carry out reasonable and sensible decisions

  • It should have a good sense of judgement.

Once we have achieved that, the Agent can be called an Ideal Rational Agent.

How do we evaluate an Agent? (PEAS)

  1. Performance measure (P)

  2. Environment (E)

  3. Actuators (A)

  4. Sensors (S)

Types of Agents

  • Simple Reflex Agent - it perceives something and does an action.

  • Model-based Reflex Agent - it has an internal state where it evaluates using a model with the help of perception history.

  • Goal-based Reflex Agent - it has an imposed goal.

  • Utility-based Reflex Agent - if there are multiple goals, it should be able to figure out which goal to complete when.

Environment Properties

  1. Fully observable vs Partially observable

    Example: Fully observable - chess game (because the agent knows the moves made by the opponent and the other properties) & Partially Observable - driving (because there can be unseen properties that the agent might not know of)

  2. Deterministic vs Stochastic

    Example: Deterministic - in a chess game, paths are defined & Stochastic - in driving, they aren't

  3. Competitive vs Collaborative

    Chess game is competitive whereas driving is collaborative

  4. Single vs Multi-Agent

    A simple maze game is a single agent task whereas a game of cricket or football is multi-agent

  5. Dynamic vs Static

    in dynamic, the environment can change in realtime whereas in static it doesn't

  6. Discrete vs Continuous

  7. Accessible vs Inaccessible

  8. Episodic vs Sequential

Last updated