In the previous sections, we discovered the fundamentals of artificial intelligence. One of the core tasks for artificial intelligence is learning.
Intelligent Agents
When solving AI problems, we create an actor in the environment that can gather data from its surroundings and influence its surroundings. This actor is called an intelligent agent.
An intelligent agent:
Is autonomous
Observes its surroundings through sensors
Acts in its environment using actuators
Directs its activities toward achieving goals
Agents may also learn and have access to a knowledge base.
We can think of an agent as a function that maps perceptions to actions. If the agent has an internal knowledge base, perceptions, actions, and reactions may alter the knowledge base as well.
Actions may be rewarded or punished. Setting up a correct goal and implementing a carrot and stick situation helps the agent learn. If goals are set up correctly, agents have a chance of beating the often more complex human brain. This is because the number one goal of the human brain is survival, regardless of the game we are playing. An agent's number one motive is reaching the goal itself. Therefore, intelligent agents do not get embarrassed when making a random move without any knowledge.
Different goals require different processes. Let's explore the two most popular types of AI reasoning: classification and prediction.
Classification is a process for figuring out how an object can be defined in terms of another object. For instance, a father is a male who has one or more children. If Jane is a parent of a child and Jane is female, then Jane is a mother. Also, Jane is a human, a mammal, and a living organism. We know that Jane has a nationality as well as a date of birth.
Prediction is the process of predicting things, based on patterns and probabilities. For instance, if a customer in a standard supermarket buys organic milk, the same customer is more likely to buy organic yoghurt than the average customer.