What is machine learning?
Machine learning is the part of AI where a system learns from examples in data instead of following rules someone has written by hand.
In ordinary programming a person writes the rules. With machine learning you show the system many examples and let it find the patterns itself. The result is called a model.
A model that has seen many invoices can learn where the amount and due date usually are. It can then read new invoices it has never seen.
The model is never better than its examples. If a type of case is missing from the data, or the examples are skewed, it shows in the result.
How it works
A model can learn in different ways. With labelled examples, where each one has a correct answer, it learns to predict the answer for new cases. Without labels, it looks for groups and patterns in the data. During training the model compares its guesses with the outcome and adjusts itself a little at a time.
Why it matters
Machine learning suits tasks where the rules are hard to write down but examples exist. Routing cases, spotting anomalies and reading documents that vary in layout are typical. Steps that used to need a person can then run on their own.
What to keep in mind
A model that works well today can get worse as reality changes, for example when new forms or new kinds of cases appear. Keep track of results over time.
The model can also carry forward skews from past decisions, known as bias. It is rarely noticed until someone looks for it.
What it means for you
Ask what data a model has learned from. It says a lot about where it works well and where it can go wrong.
Common questions
What is the difference between machine learning and AI?
Machine learning is a part of AI. Most of what is called AI today is built on machine learning.
Does the system learn from our questions?
That depends on the solution. Some services use customer data to keep training. Ask the supplier and make sure you can control it.
Related terms
Read more
All terms Updated