AI: Characteristics, Expert Systems and Machine Learning

Hello students, welcome to Kwickprep. Examiners ask the same three things about artificial intelligence again and again. Do you know the exact words they want? Today we will learn the characteristics of AI and the four parts of an expert system. Then we will see machine learning in one clear example, and practise exam style questions.

First, the definition. Artificial intelligence, or AI, is a branch of computer science. It deals with the simulation of intelligent behaviour by computers, which means making a computer act as if it is thinking. For example, a chess program studies the board and chooses a strong move, like a human player would.

Your syllabus lists the characteristics of AI in specific words, so learn them exactly. First, an AI system has a collection of data and the rules for using that data. Second, it has the ability to reason, which means it uses the rules to reach a conclusion. Third, it can include the ability to learn and adapt, which means it changes its own behaviour as it gets new data.

Let us spot all three in one medical app. Its data and rules are a list of symptoms and the illnesses they point to. It reasons when a rule links the symptoms to a likely illness. This is only an example of rule based reasoning, not medical advice. If it learns, its advice gets better after it sees more real patient cases.

Next, expert systems. An expert system is an AI program that holds the knowledge of a human expert in one narrow field. It asks the user questions, and then gives advice or a diagnosis. Common uses are medical diagnosis, finding faults in cars, and helping geologists search for minerals. Expert systems are also used for playing chess and giving financial advice.

An expert system has four main parts, and exams often ask you to name them. The knowledge base is a database of facts about the field, collected from experts. The rule base holds the rules, usually written as if this, then that, which link the facts. The inference engine is the reasoning part, which applies the rules to the user's answers to reach a conclusion. The interface lets the user enter answers and see the questions, results and explanations.

Here are two rules from a car fault expert system, written like pseudocode. The first rule says, if the engine does not start and the lights are dim, then the fault is a flat battery. The second says, if the engine does not start and the fuel is zero, then the fault is no fuel. The arrow means the fault is given that value.

Now watch the parts work together. The interface asks the user a question, like, does the engine start? The user enters an answer. The inference engine searches the rule base and the knowledge base using those answers. It asks, does a rule match? If yes, the interface shows the diagnosis and often a probability or explanation. If not, it asks another question and continues.

Pause and predict with the car rules. The user says the engine does not start and the lights are dim. Which rule fires? And which part of the expert system made that decision? The first rule fires, so the fault is a flat battery. The inference engine made the decision, and the interface displays it.

Now machine learning. Machine learning is a category of AI. The program can automatically adapt its own processes, its data, or both. That means a person does not rewrite the program each time it needs to improve. It gets better at its task as it receives more data.

Here is one clear example, an email spam filter. A new email arrives. The filter checks it against patterns it has learned, like suspicious words and senders. If it looks like spam, it goes to the spam folder. If not, it stays in the inbox. The user then marks any mistakes, such as a real email that was wrongly flagged. The filter updates its own data from these corrections, so next time it makes fewer mistakes.

Exams may ask you to compare the two. An expert system gets its knowledge from human experts, while machine learning gets it from data. In an expert system, the rules are written by people, while a machine learning program adapts its own rules. So an expert system improves only when someone updates it, while machine learning improves as it receives more data.

Now exam style questions. First, state two characteristics of AI, for two marks. A good answer is the ability to reason, and the ability to learn and adapt. Second, identify the four main components of an expert system, for four marks. Write them in full: knowledge base, rule base, inference engine and interface. Third, describe the role of the inference engine, for two marks. It applies the rules in the rule base to the user's data to reach a conclusion.

Here is a longer question. Explain how a spam filter is an example of machine learning, for three marks. Point one: it checks each new email against patterns it has learned. Point two: users mark emails as spam or not spam. Point three: the filter automatically adapts its own data from this feedback, so it becomes more accurate.

Avoid three common mistakes. Do not write that the computer thinks, write that it can reason using rules and data. Do not mix up the knowledge base, which holds facts, with the rule base, which holds the if then rules. And do not call every program AI, because a program that never reasons or learns is not AI.

Let us revise. AI has a collection of data and rules, the ability to reason, and can include the ability to learn and adapt. An expert system has a knowledge base and a rule base. It also has an inference engine and an interface. Machine learning means the program automatically adapts its own processes or data. In your exam, always answer with these exact words.