What is AI — and why should everyone understand it?

Artificial intelligence is behind epic changes that will affect all of us. In this first episode of our AI tutorial series, movingimage CEO Ingo starts at the very beginning: what AI actually is. The spoiler is that there is no single answer — AI is an umbrella term for the different ways computers simulate aspects of human thinking. No prior knowledge required, because this series is genuinely made for everyone.

One term, two ways of learning

In simple terms, AI can be divided into two categories: systems that learn symbolically, and systems that learn sub-symbolically. The difference is not academic — it decides how a system arrives at an answer and how much of that answer a human can follow.

Symbolic AI: if-then rules written by humans

With symbolic learning, the information needed for intelligent reasoning is represented in a language that humans and computers both understand. The simplest way to picture it is as if-then rules that people feed into the system: if the sky is blue and there are few clouds, the probability of rain is low. The reasoning stays readable.

Sub-symbolic AI: learning probabilities from data

The second approach works differently. Instead of rules, the AI receives thousands of past sky images together with what actually happened afterwards — how much it rained 30 minutes later. That pairing is called labeling, and the more data the system sees, the better it gets at spotting patterns. It never explains why it rains; it only recognises how likely rain is. The knowledge it builds is a long list of numbers that no human reader could interpret, which is exactly why this form of representation is called sub-symbolic. The upside: it absorbs the variations that would throw a rigid rule off track, and predictions get more accurate.

This kind of learning happens inside artificial neural networks (ANN) — an abstraction of the human brain and the way its neurons process information. Watch episode 1 for the full explanation, and stay tuned for episode 2, where Ingo shows how these networks actually work and what weightings have to do with it.