Select Page

Machine Learning – what is it?

Understanding the difference between Machine Learning and AI is a good founding stone.

The emergence of artificial intelligence (AI) and machine learning (ML) has revolutionised how businesses interact with their audiences, analyse data, and create personalised marketing strategies. As more and more discussion develop around the topic of Artificial Intelligence and Generative AI I found a great piece online that describes some of the basics around the fore father of AI, that being Machine Learning.

Below is a brief understanding written in layman’s words of the early adoptions of ML tools and what they did, with some typical examples. Lots of people confuse ML with AI – they are fundamentally different, but ML started the road to GenAI so enjoy this read, written by Assad Shahbaz.

1. Supervised Learning

Technical Explanation:
Supervised Learning uses labelled data to train a model. It requires input-output pairs where the algorithm learns to map inputs to the correct outputs. It’s typically used for classification (identifying categories) and regression (predicting continuous values).

  • Key Algorithms: Linear Regression, Logistic Regression, Decision Trees, Random Forest, Support Vector Machines (SVM), Neural Networks.

Layman Example: Imagine teaching a child to recognise animals. You show them a picture of a dog, label it “dog,” and do the same for a cat. After enough examples, the child learns to distinguish between dogs and cats by associating characteristics (like fur or size) with the label.

Technical Example: A bank wants to predict whether a loan applicant will default on a loan. The bank trains a supervised machine learning model using past data (labelled as either “defaulted” or “paid back”). The model learns patterns in the applicant’s data (e.g., income, credit score) to predict the loan outcome.

2. Unsupervised Learning

Technical Explanation:
Unsupervised Learning deals with unlabelled data. The goal is to find hidden patterns or intrinsic structures in the data without specific output labels. Common tasks include clustering (grouping similar items) and dimensionality reduction.

  • Key Algorithms: K-Means, Hierarchical Clustering, Principal Component Analysis (PCA), t-SNE, Autoencoders.

Layman Example: Imagine you walk into a library where none of the books are organised. You decide to group the books based on their appearance, like placing thick novels together and thin magazines in another pile. This is clustering — no one tells you which book goes where, but you group them based on similarities.

Technical Example: An online retail store wants to group its customers based on purchasing behaviour. With no predefined categories, they use a clustering algorithm to identify different customer segments. These clusters may represent “high spenders,” “bargain hunters,” or “frequent buyers,” helping the store tailor its marketing strategy.

3. Reinforcement Learning

Technical Explanation:
Reinforcement Learning (RL) involves training a model through trial and error, where an agent interacts with an environment and receives rewards or penalties. The goal is to maximise the cumulative reward by learning from the consequences of actions.

  • Key Algorithms: Q-Learning, Deep Q Networks (DQN), Proximal Policy Optimisation (PPO), Monte Carlo Methods.

Layman Example: Imagine training a dog. Each time the dog follows a command, it gets a treat (positive reward). If it does something wrong, it doesn’t get a treat or may be scolded (negative reward). Over time, the dog learns which actions lead to rewards and behaves accordingly.

Technical Example: An AI is tasked with learning how to play chess. It plays millions of games, learning from each win or loss. Through trial and error, it gradually improves and eventually masters the game by maximising its wins (rewards) and minimising its losses (penalties).

4. Semi-Supervised Learning

Technical Explanation:
Semi-Supervised Learning uses a mix of labelled and unlabelled data. This approach is beneficial when labelling data is expensive or time-consuming. The model learns from a small labelled dataset and generalises this knowledge to the larger unlabelled dataset.

  • Key Algorithms: Self-training, Co-training, Graph-based methods.

Layman Example: Imagine teaching someone how to identify apples. You give them a few labelled examples (pictures of apples with the label “apple”) and then give them many unlabelled fruits. Over time, they can start recognising apples based on the initial few labelled examples.

Technical Example: A healthcare system has a small dataset of labelled X-ray images indicating whether they show signs of pneumonia. Since labelling X-rays is time-consuming, they also have a large set of unlabelled images. Using semi-supervised learning, the model leverages both labelled and unlabelled images to improve its diagnostic accuracy.

5. Self-Supervised Learning

Technical Explanation:
Self-Supervised Learning is a special case where the model generates its own labels from the input data. It’s particularly popular in natural language processing (NLP) and computer vision. The model learns to predict parts of the data using other parts as labels (e.g., predicting the next word in a sentence).

  • Key Algorithms: BERT, GPT, SimCLR.

Layman Example: Imagine trying to complete a crossword puzzle. Based on the letters you already have, you try to guess the rest. You’re using the clues and context as your “self-generated labels” to complete the puzzle.

Technical Example: In NLP, a model like GPT is trained to predict the next word in a sentence. For example, given “The cat sat on the,” the model learns to predict “mat” based on the previous words. This approach helps the model understand the structure and meaning of language.

6. Deep Learning

Technical Explanation:
Deep Learning is a subset of machine learning that uses neural networks with many layers to model complex patterns in data. It is highly effective for tasks involving high-dimensional data like images, audio, and text.

  • Key Algorithms: Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Transformers, Generative Adversarial Networks (GANs).

Layman Example: Deep learning is like creating an extremely detailed flowchart that guides decisions step-by-step. The more layers (steps) in the chart, the more complex tasks it can handle, such as recognising objects in a photograph.

Technical Example: In image recognition, deep learning models like CNNs are trained on vast datasets of images. They learn to recognise patterns (such as edges, textures, shapes) across many layers, enabling them to accurately classify objects in images, like identifying whether a picture contains a cat or a dog.

Summary:

  • Supervised Learning: Learns from labelled data to make predictions (e.g., spam detection).
  • Unsupervised Learning: Identifies hidden patterns in unlabelled data (e.g., customer segmentation).
  • Reinforcement Learning: Learns by interacting with the environment and receiving rewards (e.g., game-playing AI).
  • Semi-Supervised Learning: Combines a small amount of labelled data with a large amount of unlabelled data (e.g., image recognition with few labelled images).
  • Self-Supervised Learning: Generates labels from the data itself for learning (e.g., predicting missing words in sentences).
  • Deep Learning: Uses neural networks with many layers to handle complex data (e.g., face recognition, language translation).

Each type of machine learning is tailored for specific tasks and can be applied to various real-world scenarios depending on the data and the problem at hand.

Thanks to Assad Shahbaz —  for this well written brief explanation of ML.
(
You can follow him here)

**********************************

So, in summary says Craig Ashmole, Managing Director of London based Straightalking Ltd consulting services, “There are subtle differences I know, but they really are different.”

Artificial Intelligence (AI)

AI refers to the development of computer systems that can perform tasks typically requiring human intelligence, such as speech recognition, decision-making, and visual perception.

Machine Learning (ML)

A subset of AI, machine learning focuses on developing algorithms that allow systems to learn from data and improve their performance without being explicitly programmed.

Together, AI and ML are capable of processing vast amounts of data, recognising patterns, making predictions, and automating decisions, which are all crucial to modern digital strategies.

“What we use AI for in the business world in these early stages should be tailored around repetitive data gathering or analysis of large data for example or where it’s been used extensively over the years already in website chatbots or within Shared Services centres to speed up enquiries.” Craig goes on to say. “I am reading, that ML and AI is being tailed and used more and more in the world of Digital Marketing and content creation. This will dramatically assist marketing departments to engage in predictive analytics and targeted advertising in their quest get content onto our devices.”

Having spent the majority of my career working with and supporting the Corporate CIO Function, I now seek to provide a forum whereby CIOs or IT Directors can learn from the experience of others to address the burning need to change the way we all work post the COVID Pandemic.

Craig Ashmole

Managing Director, Straightalking Consulting