Artificial Intelligence

What is Artificial Intelligence?

Artificial Intelligence (AI) is a broad field focused on creating systems that perform tasks associated with human intelligence, such as recognising patterns, understanding language, making predictions or generating content.

• Input: Information supplied to a system, such as text, images or structured data.
• Model: A system or mathematical representation that maps inputs to outputs.
• Inference: Using a trained model to produce a prediction or response.
• Human review: Checking output quality, context and suitability before acting on it.

Basic workflow

Define the task → prepare input/data → choose an approach → test results → review limitations → improve.

AI, Machine Learning and Deep Learning

AI is the broad field. Machine learning is a part of AI where systems learn patterns from data. Deep learning is a machine-learning approach that uses multi-layer neural networks.

These categories overlap, but they are not interchangeable. The best approach depends on the task, available data, risk and required performance.

Python basics for AI

Python is commonly used in AI education because its syntax is readable and its ecosystem includes libraries for data handling and machine learning.

# A simple Python function

def greet(topic):
    return f"Let's learn {topic}!"

print(greet("Artificial Intelligence"))

Practise variables, conditions, loops, functions, lists and dictionaries before moving to larger AI libraries.

Machine learning workflow

A basic machine-learning workflow includes defining the problem, preparing data, splitting data, training a baseline model and evaluating results on data not used for training.

1. Define the target and success metric.
2. Inspect and clean the dataset.
3. Separate training and evaluation data.
4. Train a suitable baseline model.
5. Review metrics, errors and limitations.

Model evaluation essentials

Metrics depend on the task. Accuracy can be useful, but it may hide problems when classes are imbalanced. Precision, recall, F1 score and error measures can provide additional context.

• Overfitting: The model learns training-specific patterns that do not generalise well.
• Data leakage: Evaluation is made unrealistically easy because information from outside the training process has leaked in.
• Baseline: A simple reference result used to compare later improvements.

Neural networks and deep learning

Neural networks contain connected computational units arranged in layers. During training, model parameters are adjusted to reduce a chosen loss function.

Deep learning can be useful for complex inputs such as images, audio and language, but often requires careful data preparation, evaluation and computing resources.

Generative AI basics

Generative AI systems create new content—such as text, images, audio or code—based on learned patterns and a user prompt. Their outputs can be fluent but still incorrect or incomplete.

• Provide relevant context and a clear task.
• Specify the expected format and constraints.
• Verify factual claims and important calculations.
• Do not submit sensitive information unless the tool and policy permit it.

Prompt engineering

A useful prompt usually explains the role or context, the task, the audience, constraints and the format of the desired output.

Iterate by checking the first answer, identifying what is missing and adding precise instructions. A well-written prompt improves direction but does not guarantee accuracy.

Responsible AI checklist

1. Check whether the tool is appropriate for the task.
2. Protect personal, confidential and proprietary information.
3. Review outputs for factual errors, bias and missing context.
4. Test with varied examples and document limitations.
5. Keep a human responsible for important decisions.

Artificial Intelligence glossary