AI curriculum

Course outline for Artificial Intelligence Foundations

AI labels often hide the real decision: should a task use a fixed algorithm, a learned model, a person, or a combination? Build a practical model of how AI systems learn and generate output, then justify where each kind of AI fits and where it does not.

Technology: AI Foundations
Duration: 4 hours
Modules: 2 · 2 hours each
Level: Level 1

About Artificial Intelligence Foundations

Artificial Intelligence Foundations Course Objectives

Distinguish algorithmic, machine-learning, deep-learning and generative AI.
Explain training and inference with examples.
Allocate calculation, summary, recommendation and approval tasks across people, software and AI.
Identify capability limits and state when a task should not use a model.

Pre-requisites

Basic computer literacy.
Interest in evaluating AI applications; coding is not required.

Before this course

Complete the required courses or bring equivalent experience. Recommended courses add useful depth; they are not enrolment gates.

No prerequisite course is required. The entry skills above still apply.

Computer literacy and interest in AI decisions; coding is not required.

Find your place in the learning map

Lab Setup

Computer with a browser and a text editor.
Use supplied model output examples or an optional locally available model; no paid service is required.

Detailed Course Outline

Module 1: Kinds of AI and how systems learn

Algorithmic and symbolic AI: Algorithms, search, explicit facts, and inference rules; distinguish deterministic calculation and rule-based recommendations from predictions learned from data.
Machine learning foundations: Examples, features, training objectives, and generalization; contrast fitting patterns from data with specifying every decision as a rule.
Neural networks and deep learning: Layers, weights, and learned representations; position deep learning within machine learning rather than treating neural models as all AI.
Generative AI and foundation models: Broadly trained models, generated text, and task adaptation through context; distinguish plausible generation from deterministic calculation or verified evidence.
Tacit knowledge and the limits of rules: Implicit expertise, ambiguous examples, and incomplete procedures; identify decisions that cannot yet be expressed as reliable explicit rules.

Practical outcome: Classify example tasks by the kind of AI they need.

Module 2: Choosing and constraining an AI approach

Training and inference: Learning model parameters versus applying learned parameters to new inputs; distinguish development data from the context supplied during use.
Observe, interpret, evaluate and act: Inputs, candidate interpretations, evidence checks, and permitted actions; use an application loop without claiming it describes neural internals.
Allocating work across people, software and AI: Failure cost, evaluation feasibility, and action authority; assign calculations, summaries, recommendations, and approvals with explicit escalation ownership.
Model capabilities and limitations: Unsupported claims, incomplete evidence, and task suitability; specify when rules or human judgment are preferable and when escalation is required.

Practical outcome: Allocate tasks using failure cost, evaluation feasibility and action authority, and define an escalation path.

Practical exercise

Classify a tax calculation, a document summary, a rule-based recommendation and an approval request.
Choose an algorithm, a learned model, a hybrid approach or a human process for each task and justify the choice.
Record the failure cost, the available evaluation check and the action authority for each task.
State one task where a model should not be used and the condition that would change that decision.
Name one case that escalates to a person when evidence is insufficient, and record the escalation owner.
Produce a one-page allocation table that a reviewer can check against the stated criteria.

What you will produce

Justify an algorithmic, learned, hybrid, or human approach for calculation, summary, recommendation, and approval tasks using failure cost, available checks, and action authority.

How the work is checked

The one-page allocation table covers all four supplied tasks and justifies each selected approach.
Each task records failure cost, an available evaluation check, and the action authority.
An insufficient-evidence case names a human escalation owner.
One no-model decision states the condition that would change the choice.

A four-hour, coding-optional introduction with a one-page allocation exercise; the application loop is not a description of neural internals. Model training, quantitative decision theory, and organization-wide adoption are outside scope.

When another approach is better

Use an explicit algorithm instead of a generative model for fixed arithmetic or rules that can be tested directly.
Do not delegate a high-consequence decision when its outputs cannot be checked and responsibility is unclear.

How we train

Ask about exercises, instructor feedback and the prior knowledge you need. Tool-specific courses marked provisional may become modules in a broader course.

Discuss your learning goals