For Heads-of · Practitioner
Machine Learning
A branch of artificial intelligence in which systems learn patterns from data rather than being explicitly programmed for every scenario.
- machine learning
- AI
- algorithms
- data
Definition
Machine learning lets a system improve at a task from data and experience rather than hand-written rules. The system identifies patterns in examples and adjusts its behaviour to fit them.
Types
Supervised learning (learning from labelled examples), unsupervised learning (finding structure in unlabelled data) and reinforcement learning (learning from feedback and rewards).
In AI governance
A model learns whatever its data contains, including bias, and its performance degrades as the world moves away from the data it learned from. Machine learning systems therefore need monitoring for drift and degradation and testing for training data bias throughout their life, not only at approval.
Related Terms
Large Language Model (LLM)
A deep learning model trained on very large amounts of text, able to generate and interpret human-like language across a wide range of topics and tasks.
Model Drift
A change in the relationship between a model's inputs and the environment it was fitted to, such that the model's assumptions no longer hold. The model itself is unchanged; the world it was trained on has moved.
Neural Network
A computational model loosely inspired by biological neurons, made of interconnected layers of artificial nodes that learn to recognise patterns in data.
Training Data
The dataset used to train a machine learning model, from which the model learns patterns and relationships.
Related risks and controls
Training data bias and discriminatory outcomes
The most common way a learned model inherits a problem nobody wrote into it.
AI model drift and degradation
A model learned from one distribution and deployed into another one.
Bias testing and fairness monitoring
The test a learned model needs that a coded one does not.