Program facts
- Study fee
- Free
- subjects
- 5
- lessons
- 20
- Estimated time
- 12.5 hours
- Access period
- 60 days
- Study language
- Arabic and English
About the program
When an AI system produces a recommendation, classification or generated paragraph, what happened between the input and the output? This diploma builds a structured answer, beginning with the meaning of an AI system and task type, then moving through data, training and generalization, neural networks and generative models, evaluation and risk management. The aim is not to memorize product names or treat AI as if it understands like a person. Learners distinguish fixed-rule automation, machine-learning models and generative systems; interpret probability, error and thresholds; and examine why a fluent output can still be wrong. The pathway ends with a responsible-use plan defining acceptable data, human verification points, suitable metrics and monitoring for the context.
Practical relevance
Concepts transfer to everyday tasks such as sorting messages, forecasting demand, summarizing documents or drafting text. Applications focus on task selection, metrics, verification and boundaries through fictional cases that require neither confidential data nor an external service account.
Why this program
Products change quickly, but the foundational questions remain useful: What problem are we solving? What data does the model learn from? What does “good” performance mean for this task? Which mistakes matter most? The program builds a shared vocabulary for answering those questions before jumping to tools or marketing claims.
Career relevance
The content is useful to professionals who work with AI-enabled tools, projects or decisions because it supports more precise discussion of data, evaluation, limitations and verification. This is educational relevance, not a promise of employment, proof of programming proficiency or eligibility for a specialist role.
What you will learn
- Distinguish AI systems from fixed-rule automation and distinguish machine learning, deep learning and generative AI.
- Explain how data, features, labels, representation and sample quality affect model development.
- Explain training versus inference, generalization, overfitting and train-validation-test splits.
- Understand neural networks, embeddings, transformers, attention and language models at a beginner-appropriate conceptual level.
- Interpret probability, thresholds, confusion matrices, accuracy, precision and recall through simple examples.
- Evaluate generative outputs without treating them as automatic sources of truth and identify when human or source verification is required.
- Analyze bias, privacy, security, transparency, accountability and human-oversight risks.
- Select a suitable use case and build a simple evaluation, monitoring and responsible-use plan.
Learning outcomes (10)
- Explain what an AI system is and distinguish it from fixed-rule automation without attributing human-like consciousness or understanding to the system.
- Distinguish machine learning, deep learning and generative AI and connect them with suitable task and output types.
- Explain a simplified AI-system lifecycle and the difference between training, inference and post-deployment monitoring.
- Analyze how data, features, labels, representation and sample quality affect model learning and performance.
- Explain the conceptual differences among supervised, unsupervised and reinforcement learning, and interpret loss, parameters, generalization and overfitting.
- Use probability, confusion-matrix, accuracy, precision, recall and threshold concepts to interpret a simple classification model.
- Explain the conceptual role of neural networks, embeddings, transformers and attention in modern models without requiring advanced mathematical derivation.
- Explain how language models generate outputs probabilistically and identify reasons for hallucination, uncertainty and verification needs.
- Select an evaluation approach for the task and compare metrics and human or automated checks according to error costs and use context.
- Analyze bias, privacy, security, transparency, accountability and human-oversight risks and design a simple responsible-use plan.
Who it is for
- Learners and professionals seeking a structured AI foundation.
- Users of AI tools who want to understand more than interface actions or prompt recipes.
- Business or technical team members who need a shared vocabulary for use cases, evaluation and risk.
Prerequisites
- No advanced AI or programming experience is required; general digital literacy and readiness for conceptual examples and simple metric calculations are sufficient.
What it does not offer
- Not an advanced course in programming models from scratch or deriving algorithms mathematically.
- Does not prove proficiency in a named commercial product and does not grant a professional licence or university degree.
- Does not present AI outputs as reliable facts without verification.
Curriculum
- Subject 1 Open for free
Understanding AI and Intelligent Systems
Establish a shared vocabulary: what the AI system is, what output it produces, and when behavior comes from written rules or a learned model. It closes with a simplified lifecycle separating training, inference and monitoring.
- Subject 2
Data and Machine Learning
Focus on what a model learns from and how: data, features and labels, learning paradigms, loss and parameters, then generalization, dataset splits, overfitting and leakage.
- Subject 3
Neural Networks and Generative AI
Move from neural networks and embeddings to transformers and language models. Explain probabilistic generation, context limits and hallucination without portraying the model as a conscious actor.
- Subject 4
Evaluation, Probability and Practical Selection
Connect evaluation to the decision: baselines, errors, probability, thresholds and classification metrics, then generative-output evaluation and deciding whether AI is suitable for the task at all.
- Subject 5
Responsible Use, Risk and Human Oversight
Integrate bias, privacy, security, transparency, accountability and human oversight into a responsible-use plan covering controls, monitoring and when to stop or escalate to a person.
Subjects unlock in sequence as you progress.
Strengths
- Progression from system definition and data through training and evaluation to responsible use.
- Explicit separation of conventional machine learning and generative AI, including where they overlap and differ.
- Simple quantitative examples for model metrics and thresholds without turning the diploma into an advanced mathematics course.
- Decision cases that balance benefit, error, privacy and human verification rather than stopping at principle lists.
Evidence basis
The design draws on the OECD AI-system definition, the public overview of ISO/IEC 22989, open Google machine-learning learning materials, the NIST AI Risk Management Framework and Generative AI Profile, and UNESCO AI ethics principles. Source boundaries are explicit: a public standards page is not treated as having read the full standard.
Reference organizations
- OECD
- ISO/IEC
- Google for Developers
- NIST
- UNESCO
Frameworks used
- OECD AI Principles — updated 2024
- NIST AI Risk Management Framework — 1.0
- NIST Generative AI Profile — NIST AI 600-1
- UNESCO Recommendation on the Ethics of Artificial Intelligence — 2021 recommendation
Methodology
The blueprint starts with core definitions, then data and learning, generation, evaluation and risk. Official sources anchor terminology and boundaries; examples, exercises and questions are original, and any time-sensitive fact is rechecked before use.
How the route works
Short Diplomas are relatively comprehensive learning pathways that cover a major subject through five connected learning subjects and structured assessments, with a diploma transcript available with the credential.
Self-paced sequential study. Content unlocks progressively and each subject combines reading, examples, exercises, knowledge checks, and assessment.
- Build integrated knowledge in a major subject.
- Study through five clear, connected learning subjects.
- Complete a defined assessment for each subject.
- Receive a diploma transcript showing subjects and results when the credential is issued.
Assessment
Each Short Diploma contains five subjects, each with an approved assessment. A sixth final exam is not required when the five assessments adequately cover the learning outcomes.
Assessment attempts: 2 per assessment, 7 days apart
The pass mark is announced before the program is published.
Certificate
Successful completion makes the learner eligible for the Short Diploma certificate and Diploma Transcript under the credential policy.
Every certificate has a public verification code. Verify a certificate
Disclaimer: Training completion credential, not a university degree or licence to practise a regulated profession.
Program questions
Do I need programming before starting?
No. The design is foundational and can be understood without writing code. Simple formulas for metrics such as accuracy and recall are explained step by step. Training models programmatically requires additional technical study.
Is the program focused on ChatGPT or one product?
No. It teaches concepts that help learners understand categories of systems and models, avoiding dependence on an interface, price or feature that may change quickly. Any product-specific example added later would require separate current verification.
What is the difference among AI, machine learning and generative AI?
The diploma treats AI as the broader field, machine learning as a way to build models that learn patterns from data, and generative AI as a class of systems that generates new content probabilistically. Their boundaries are developed in Subjects 1 and 3.
Will I learn how a model is trained?
Yes, conceptually: data, features, labels, loss, parameters, optimization and generalization. The program does not require calculus derivations or building a deep architecture from scratch.
Why does the program include probability and metrics such as precision and recall?
Because one number cannot describe every kind of error. You learn to connect a metric to the decision and to recognize when a high-looking result can mislead because classes are imbalanced or error costs differ.
Does generative AI know whether its answer is true?
The program does not assume that. Language models generate probable token sequences and can produce unsupported or incorrect statements, so verification methods and task-appropriate confidence requirements are taught explicitly.
Does the program cover privacy and bias?
Yes. Subject 5 connects bias, data quality, privacy, security, transparency, accountability and human oversight with practical cases without turning general principles into an assumed national law.
Can I apply the exercises to my work?
Yes, while keeping confidential data out of external services. Each exercise includes a self-contained fictional case, and you can transfer the decision logic to your environment while respecting the policies and laws that actually apply there.