1.1.1Introduction to the International Olympiad in Artificial Intelligence

What is IOAI?

IOAI stands for the "International Olympiad in Artificial Intelligence". It is one of the 12 Olympiads that have received UNESCO patronage. It was first held in Bulgaria in 2024 with teams from 32 countries and territories. In 2026, 108 countries and territories took part. The exponential growth is pretty easy to see.

The sponsors include a variety of popular companies and institutions like Google, Jane Street, IMC, and the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI).

The syllabus covers a variety of topics, including machine learning, deep learning, computer vision, language, and audio. That is a lot if you are starting from zero. For some topics, you need to understand how they work. For others, you need to know when and how to use them in code.

But despite the vast syllabus, we are not rebuilding the wheel from scratch here. On the contrary, the problems aim to test more of your understanding and your ability to come up with the right method that would most likely work in the given problem conditions and datasets.

What does "AI" really mean in IOAI?

Now, when you usually think about "AI", I pretty much guarantee that you will think of the AI slop we encounter on various platforms. Or you might think of building models like ChatGPT, complete end-to-end image generation models, and whatnot.

The truth is, the AI we focus on is much more about the technical aspect, the algorithms used for it, and such details. We are not building Claude or ChatGPT in 2 hours. We focus on problems which are sufficiently hard to require knowledge of how AI actually works: the algorithms, the code, and the mathematics. It's basically the nerdy stuff (it's cool after you learn it... trust me on this one).

There is another way of framing what IOAI tests. The competition is built to test the individual's ability to identify patterns, adapt to the constraints, and apply your already known knowledge by connecting what you know with the data/model characteristics you find in the problem. You exploit these patterns and build algorithms and pipelines. This is the crux of the competition.

This breaks the competition into two aspects:

1. Theoretical knowledge about the algorithms and prevalent choices.
2. Practical ability to write code, implement the pipelines and algorithms, and debug code.

Interpretation of IOAI Rules

We won't be going into the details of the rules. Those contain lots of specific details that are only required when you are attempting the competition yourself. But there is one policy that is reflective of the ideals of the committee: the LLM usage policy.

Onwards from here...

Each country has its own national selection process. Each country's process is structured around a certain set of skills. Most of the authors writing this guide are from India. Thus, the guide might have some hidden biases towards skills which the Indian selection process tests.

Nevertheless, the guide is aimed to be a general guide for anyone interested in AI.