AI Revolution — Your Fast-Paced Introduction to Machine Learning

From Basics to Generative AI

Col Jung
Geek Culture
Published in
9 min readApr 21, 2023


The power of generative AI — an image created by Midjourney V5. Source:

Artificial intelligence (AI) is the ability of machines to do things that typically require human intelligence.

This has been enabled by advances in machine learning (ML), which is training machines to ‘learn’ how something works by feeding it lots of data.

How the different pieces fit together

A classic example is training machines to classify patterns. For example, discerning pictures of dogs from pictures of cookies.

Rather than explicitly-program rules of what a dog looks like (hard-coding), for instance, “two eyes, long nose, mouth with a big sloppy tongue”, our machine is simply fed thousands of photos of dogs and told ‘These are dogs. Learn, my cold metallic friend!”

By the end of this training process, the machine has learnt the rules of what a dog looks like by itself. It’s now ready to chomp on new photos and eagerly tell you whether it’s a dog or an imposter.

This is a classic example of supervised machine learning — where the training data is labelled. (We told the machine when a photo was a dog.)

Some common supervised learning use cases include:

  • Handwriting recognition (e.g. writing on your tablet)
  • Medical diagnosis (e.g. imaging for diseases like cancer and COVID-19)
  • Object detection (e.g. face detection on a camera)
  • Email spam filtering.
Types of machine learning problems and tasks. Source: TTPSC

Another common type of ML task is unsupervised learning, where we feed the machine a pile of data without labels and watch it sort through it for patterns. Example use cases include:

  • Clustering (e.g. segment customers based on their behaviour)
  • Anomaly detection (e.g. fraud detection by banks)
  • Generative models. AI platforms like ChatGPT and Midjourney fit here.