Artificial intelligence
To understand AI, we must first distinguish it from traditional software. Traditional software follows strict, pre-written rules. If the input matches the rule exactly, it produces a specific output. It cannot handle ambiguity or new situations it hasn't been programmed for.
Artificial Intelligence (AI) is a branch of computer science focused on creating systems that can perform tasks requiring human-like intelligence. Unlike traditional software, AI does not just follow fixed rules; it uses data to make decisions, often in uncertain or complex environments.
| Feature | Traditional Software | Artificial Intelligence (AI) |
|---|---|---|
| Basis | Fixed algorithms and hard-coded rules. | Data-driven; learns patterns from information. |
| Decision Making | Deterministic (same input = same output). | Probabilistic (makes best guesses based on data). |
| Adaptability | Static; requires manual code updates to change behavior. | Dynamic; can adapt and improve over time without reprogramming. |
| Complexity | Struggles with ambiguity or unstructured data (e.g., images, speech). | Handles uncertainty, pattern recognition, and natural language. |
Key Distinction: AI simulates intelligent behaviour such as reasoning, learning, and problem-solving. For example, a traditional calculator always gives the same result for 2+2. An AI system might 'learn' that in a specific context, 'two plus two' refers to a brand name, not a math problem. This ability to interpret context and adapt is what defines AI.
- Expert Systems: Simulate human expert decision-making using rules.
- Machine Learning: Allows systems to learn from data without being explicitly programmed for every scenario.
- Natural Language Processing (NLP): Enables computers to understand and generate human language.
Understanding AI requires recognizing that it is a collection of techniques aimed at replicating aspects of human cognition, rather than a single algorithm.
Learning Objective 2: Describe the main characteristics of AI.
AI systems are defined by three core characteristics. You must describe all three to fully answer this objective:
- Collection of Data and Rules: AI systems require a large dataset to learn from or operate upon. They also utilize rules (either hard-coded in expert systems or learned via algorithms) to process this data.
- Ability to Reason: The system can draw conclusions, make decisions, or solve problems based on the available data and rules. This mimics human logical thinking.
- Ability to Learn and Adapt: Many AI systems (specifically those using Machine Learning) can improve their performance over time. They analyze new data, identify patterns, and adjust their internal processes or rules to become more accurate.
Learning Objective 3: Explain the basic operation and components of AI systems.
A common type of AI system examined is the Expert System. It simulates the decision-making ability of a human expert.
Key Components:
- Knowledge Base: Stores facts and rules about a specific domain (e.g., medical diagnoses, car faults). This is the 'memory' of the system.
- Rule Base (or Inference Engine): Contains the logical rules (IF-THEN statements) that allow the system to reason. The inference engine applies these rules to the knowledge base to derive conclusions.
- Interface: Allows the user to input data and receive explanations or recommendations from the system.
How it Operates (Forward Chaining Example):
- The user enters initial data via the interface (e.g., 'The car will not start').
- The inference engine checks the rule base against the knowledge base.
- It asks follow-up questions to gather more specific data (e.g., 'Are the lights on?').
- It applies logical rules to narrow down possibilities.
- It outputs a diagnosis or recommendation based on the accumulated evidence.
Example: Robot Vacuum Cleaner
A robot vacuum uses AI (specifically machine learning) rather than just pre-programmed paths.
- Data Collection: It gathers data about room layout, obstacle locations, and dirt levels during cleaning.
- Reasoning/Learning: It analyzes this data to identify patterns (e.g., 'dirt accumulates near the dining table').
- Adaptation: It updates its internal map and cleaning rules. Next time, it prioritizes that area or avoids a newly placed obstacle, demonstrating the characteristic of learning and adapting.
Correction: The question asks for characteristics (what AI does), not components (what AI has).
- Correct Answer: 'AI has the ability to learn from data,' 'AI can reason logically,' and 'AI uses a collection of rules.'
- Incorrect Answer: 'It has a knowledge base and an inference engine.' (This describes structure, not characteristic).
Error: Describing AI as simply 'automated software' or 'following instructions'.
Correction: AI must be distinguished from standard automation. Standard automation follows fixed rules. AI involves handling uncertainty, learning from experience, or simulating human-like reasoning. Always use words like 'adapt', 'learn', 'pattern recognition', or 'probabilistic' to show you understand the difference.
Examiner Expectation: Examiners look for the cycle of data and adaptation. Generic statements like 'it learns' are insufficient. You must specify:
- Data Input: What data is gathered? (e.g., 'user clicks', 'voice samples').
- Pattern Recognition: How is it analyzed? (e.g., 'identifying common phonemes', 'finding correlations').
- Adaptation/Output: How does the system change? (e.g., 'updating the algorithm weights', 'refining search results for future queries').
Why this works: This directly addresses the definition of machine learning as a process of improving performance through experience.
Example Phrase: 'The system gathers data on user search history, analyzes patterns in frequently clicked links, and adapts its ranking algorithm to prioritize similar content for future users, thereby improving relevance.'
Examiner Expectation: Use the correct terminology for the components and their interaction. Do not say 'the computer thinks'.
Why this works: It demonstrates technical knowledge of the architecture.
Example Phrase: 'The inference engine applies rules from the rule base to the facts in the knowledge base to deduce a conclusion, which is then presented to the user via the interface.'
- Ability to learn/adapt: The system can improve its performance over time by analyzing new data or feedback.
- Ability to reason: The system can draw logical conclusions or make decisions based on available information and rules.
- Collection of data and rules: The system operates using a large dataset and a set of rules (or algorithms) to process that data and simulate intelligent behaviour.
- Knowledge Base
- Rule Base (or Inference Engine)
- Interface
- The system gathers data from various users speaking different words or accents.
- It analyzes/identifies patterns in the audio data (e.g., phonemes, pitch, speed).
- It stores successful and unsuccessful recognitions to refine its internal model.
- It adapts its processes/algorithms over time, allowing it to better identify a specific user's voice or handle new vocabulary, thus improving accuracy.
- The user enters symptoms (data) into the interface.
- The inference engine retrieves relevant rules from the rule base and facts from the knowledge base.
- It asks follow-up questions via the interface to gather more specific data if needed.
- It applies logical reasoning (IF-THEN rules) to narrow down potential causes.
- It outputs a diagnosis or recommendation based on the accumulated evidence.