Back to Study Material Artificial Intelligence

AI agents

An AI agent is any entity that perceives its environment through sensors and acts upon that environment through actuators to achieve goals. This definition...

AI agents

What Makes Something an Agent

An AI agent is any entity that perceives its environment through sensors and acts upon that environment through actuators to achieve goals. This definition, introduced by Russell and Norvig in their landmark textbook, is deliberately broad. It covers simple thermostats, complex autonomous robots, and software bots that browse the web and take actions on behalf of users.

The key properties of a good AI agent are described by the concept of rationality. A rational agent selects actions that are expected to maximise its performance measure given the evidence from its perceptions and whatever built-in knowledge it possesses. Rationality is not the same as omniscience. A rational agent does the best it can with the information available, even if that information is incomplete or uncertain.

Agents are categorised by their internal architecture. A simple reflex agent acts solely on the current percept using condition-action rules, such as a robotic vacuum that turns when its bump sensor is triggered. A model-based reflex agent maintains an internal state of the world, allowing it to handle partially observable environments. A goal-based agent considers future states and plans sequences of actions to reach a desired outcome. A utility-based agent goes further, using a utility function to choose among competing goals based on expected benefit.

Learning Agents

A learning agent adds a fourth component: a learning element that improves the performance element based on feedback, a critic that measures how well the agent is doing against a performance standard, and a problem generator that suggests exploratory actions. Self-driving car systems combine all four agent types in their overall architecture.

Interview Tip

In interviews, drawing the agent loop, percept goes in, agent function maps it to an action, action changes the environment, new percept generated, shows conceptual fluency. Mentioning PEAS, which stands for Performance measure, Environment, Actuators, and Sensors, as the standard framework for specifying an agent task will impress interviewers asking about agent design.