AI Academy · Advanced learning pilot
AI Foundations & Engineering
Learn how language models work, how people build useful AI systems, and why safety and evaluation matter. Move from basic ideas through vectors and attention to tool use, adaptation and real-world design.
This is an advanced concept-reading pilot, not a complete Grade 10–12 curriculum or evidence of mastery. No account or external AI service is needed. Review the explanations and perform the offline exercises at your own pace.
What you will learn
These lessons are original summaries inspired by eleven supplied topic references. They are rewritten to correct oversimplifications, and they do not reproduce the source charts. Start with basic distinctions before attempting the engineering topics.
Module 1. What Is AI?
AI, Machine Learning and Deep Learning
Explain what makes AI, machine learning, deep learning and generative AI different, and identify the evidence needed to classify a system.
How Generative Models Produce New Data
Contrast autoregressive, diffusion, VAE, GAN and flow-based generation at an intuitive level.
What Happens When You Ask an LLM?
Trace a text prompt through tokenization, embeddings, transformer computation, prediction and decoding.
Module 2. Inside Language Models
Recognize subword tokens, token IDs, model-specific tokenization and the practical meaning of a context window.
Embeddings and Vector Similarity
Interpret a vector representation and calculate the meaning of cosine similarity without confusing similarity with truth.
Transformers and Self-Attention
Describe query, key and value attention, positional information, masking and the distinction between encoder-decoder and decoder-only models.
Module 3. Working with LLMs
Prompt Engineering as Clear Communication
Write a testable prompt with task, context, boundaries, desired output format and a verification step.
Explain the model-tool-host loop, safe schema design and why a suggested function call is not an executed action.
Module 4. Adapting AI
Fine-Tuning, LoRA and Retrieval
Choose between prompting, retrieval-augmented generation and parameter-efficient fine-tuning based on the actual problem.
Module 5. Safety and Application
GenAI Security, Privacy and Reliability
Threat-model an AI application and propose proportionate controls for prompt injection, data leakage, unreliable answers and unsafe tool actions.
Designing Useful Real-World AI Applications
Identify a problem suitable for AI, choose appropriate tools, measure outcomes and keep people accountable.
Further exploration: choosing an AI model
New advanced reference: Eight AI Labels: Model Families and Methods distinguishes language and vision capabilities, model size, architectures, training methods and action-capable systems. Read this after the fundamentals, not as an additional required lesson.
Further exploration: reusable Agent Skills
Read Agent Skills: Reusable AI Workflows to learn the difference between instructions, skills, tools, agents and permissions. This is supplementary advanced reading, not an extra required lesson.
How to use each lesson
Learn: identify one objective. Understand: read the explanation and worked example. Question: check a misconception. Practise: reveal short self-check answers. Apply: try the paper-based design task. Reflect: explain what could go wrong.
For younger learners, begin with the existing AI Discoverers family beta. For current technology developments, visit AI Updates. These are different learning experiences.
Practice · Advanced learning pilot
Check your AI understanding
Try 11 short application questions, one for each lesson. Answer one at a time, read the explanation, and decide which topic you want to review. This is not a grade, certification or proof of mastery. Nothing is uploaded, tracked or saved; answers and the result reset when you leave or reload this page.
Reflect on what you learned
Try a design challenge: Imagine a fictional school-library assistant. Explain which tasks can use fixed rules, whether retrieval is needed, where a human must approve an action, and how you would measure accuracy and safety. Discuss with a teacher or write your thoughts offline.
Use the lessons above to revisit questions you were unsure about. Completing this activity does not record learning progress or unlock access.