Artificial Intelligence Software Curriculum

Understanding how intelligent systems are built

The Software Curriculum at ARCSA focuses on artificial intelligence as a discipline of understanding, design, and reasoning, not merely tool usage.

It is part of the Artificial Intelligence and Robotics (AIR) program and runs in alternating terms with advanced robotics. While robotics teaches how intelligent systems act in the physical world, the software curriculum teaches how intelligence itself is created, trained, evaluated, and improved.

What Software Means in Artificial Intelligence and Robotics

In our Artificial Intelligence and Robotics Program (AIR), software does not mean web design or app development.

It means:

Machine learning
Computer vision
Natural language processing
Generative AI
Decision-making systems

Students learn how data becomes models, how models become systems, and how those systems behave in the real world.

A Structured Progression, Not Isolated Projects

The software curriculum is organised around increasing levels of abstraction and responsibility.

Students move from using existing AI models to training and fine-tuning their own, and eventually to understanding how modern AI architectures work internally.

The emphasis is always on why a system works, not just that it works.

Computer Vision and Applied AI

A significant part of the software curriculum focuses on computer vision, one of the most important and widely used branches of artificial intelligence.

Students explore how machines can interpret visual information and make decisions based on images and video. Typical application areas include:

Safety and monitoring systems
Intelligent cameras for traffic, workplaces, and schools
Object detection, classification, and tracking
Behaviour and pattern recognition
Measurement and analysis of visual data

Some systems rely on pre-trained models to focus on system design and interpretation. Others require students to train and fine-tune models themselves, introducing them to the realities of data quality, bias, and evaluation.

Beyond Vision: Generative and Language-Based AI

As students progress, the curriculum expands beyond vision into language and generative AI.

Students are introduced to:

How machines process and represent language
How text can be analysed, classified, and generated
How modern generative systems work conceptually

This prepares students for subsequent work with large language models (LLMs), shifting from using AI tools to understanding the principles underlying them.

Learning AI Through Games and Systems

Not all learning happens through abstract examples.

Some AI concepts are best understood through games and interactive systems, where decision-making, learning, and optimisation can be observed directly.

In this curriculum, games are used not for entertainment, but as controlled environments where students can explore:

Learning from feedback
Strategy and optimisation
Reinforcement learning concepts
Trade-offs between short-term and long-term decisions

This approach makes complex AI ideas tangible and intellectually coherent.

The Go Project and Deep Learning

One of the cornerstone software projects in AIR is building an AI that learns to play the game of Go.

This project introduces students to:

Machine learning fundamentals
Neural networks and deep learning
Tree search and decision-making
Reinforcement learning and self-play

Go is used because it played a central role in one of the most significant breakthroughs in AI history and because it provides a rich, structured environment for learning.

The goal is not to create a perfect player, but to understand how intelligence can emerge from data, learning, and experience.

Large Language Models (LLM): From Foundations to Modern AI

In later stages, students explore the foundations of large language models (LLMs).

Topics include:

Natural language processing fundamentals
Word representations and embeddings
Recurrent and attention-based models
Transformers and modern architectures
Training, fine-tuning, and evaluation
Reinforcement Learning with Human Feedback (RLHF)

Rather than treating LLMs as black boxes, students learn how they are built, what their limitations are, and why they behave the way they do.

A Curriculum That Evolves

Artificial intelligence evolves rapidly.

For that reason, the software curriculum is designed to be flexible and forward-looking. While the foundational principles remain constant, specific systems and applications may change over time as new technologies emerge.

What students gain is not familiarity with a single tool, but a way of thinking that allows them to adapt.

How This Prepares Students for the Future

The software curriculum prepares students for:

University studies in AI, computer science, data science, and engineering
Understanding and evaluating AI systems critically
Building and reasoning about intelligent systems
Engaging responsibly with AI as creators, not just users

It complements the robotics curriculum by focusing on intelligence in its abstract form, before it is embodied in machines.

In Context with the Artificial Intelligence and Robotics (AIR) Program

Within the Artificial Intelligence and Robotics (AIR) program, the software curriculum alternates with the robotics curriculum.

Students repeatedly move between:

Thinking about intelligence
Implementing intelligence
Applying intelligence

Each return deepens understanding and confidence.

Concepts explored in the software curriculum are later applied to real robotic systems through the Hardware Curriculum and the BotBox Robotics Lab.