3.3 Types of AI Agents

4. Utility-Based Agent

Utility-based agents not only aim to achieve goals but also consider how well those goals are achieved. They use a utility function to measure the desirability of different outcomes and choose actions that maximize overall satisfaction or performance. This type of agent is especially useful when there are multiple possible solutions and trade-offs.

Real-world examples:

•  A self-driving car that chooses routes based on safety, travel time, and fuel efficiency.
•  A recommendation system (e.g., Netflix or YouTube) that selects content to maximize user satisfaction.
•  An investment trading system that balances profit against risk.

5. Learning Agent

Learning agents are capable of learning from experience and improving their performance over time. They start with basic knowledge and adapt their behavior as they receive feedback from the environment.

A learning agent consists of:

•  Learning element
•  Critic
•  Performance element
•  Problem generator

Real-world examples:

•  A spam email filter that improves its accuracy as users mark emails as spam or not spam.
•  A virtual assistant (e.g., Siri, Alexa) that adapts to user preferences over time.
•  A game-playing AI that improves its strategy by learning from previous games.