Machine learning
In short
Machine learning (ML) is a way of building AI in which a computer learns patterns from examples instead of being given step-by-step rules. It then uses those patterns to make predictions or decisions about new data.
- Also called
- ML
- Topic
- The basics
- Level
- Beginner
- First described
- Arthur Samuel used the phrase in the title of his July 1959 IBM Journal paper, "Some Studies in Machine Learning Using the Game of Checkers"2.
In ordinary programming, a person writes the rules. In machine learning, a person supplies data and a learning method, and the system works out the patterns itself. The ICO, the UK data regulator, describes ML models as building "from the past information contained in collected data to identify patterns"1.
The ICO names three main approaches1. In supervised learning, the model trains on labelled examples and learns to map inputs to the right labels. In unsupervised learning there are no labels; the model finds clusters and outliers by measuring how similar data points are. In reinforcement learning, a model learns by trial and error, rewarded or penalised for the steps it takes1.
Machine learning has replaced hand-written "expert systems" as the dominant AI technology, according to the ICO, because rule-based systems were labour intensive, expensive to scale and inflexible1. A trained model then applies the rules it honed during training to new inputs1, so it can only be as good as the data it learned from.
The idea is decades old. In 1959 Arthur Samuel of IBM reported that a computer could learn to play draughts better than the person who wrote the program, after 8 to 10 hours of play2.
Deep learning, the method behind modern chatbots and image tools, is one branch of machine learning. It uses neural networks with many layers that learn their own representations of data3.
Example
A spam filter is a classic case. It is trained on emails people have already marked as spam or not spam, then sorts new mail using the patterns it found. In the ICO's terms that is supervised learning on labelled data1.
Why it matters
Nearly every AI product you meet is built by learning from data, so its strengths, gaps and biases come from that data.
Questions people ask
What is the difference between AI and machine learning?
AI is the broad goal of machines doing tasks that need human thinking. Machine learning is one way to get there, by learning from data rather than following hand-written rules. The ICO calls data-driven machine learning the dominant AI technology today1.
What are the main types of machine learning?
Supervised learning, which trains on labelled examples; unsupervised learning, which finds patterns in unlabelled data; and reinforcement learning, which learns by trial, reward and penalty. These are the three main approaches named in ICO guidance1.
Is deep learning the same as machine learning?
Deep learning is a type of machine learning. It uses neural networks with many processing layers, each learning a more abstract representation of the data3.