What it means
Traditional software does what a programmer told it to do, step by step. Machine learning inverts that: instead of writing the rules, you supply examples and let the system derive the rules itself. Show it enough labeled photos of cats and it learns what distinguishes a cat, without anyone ever writing down a definition of a cat.
The learned rules live in a set of numbers called parameters, adjusted repeatedly during training until the system's outputs match the examples closely enough. Almost everything marketed as "AI" today is machine learning of some kind, and deep learning — machine learning using large layered neural networks — is the branch that produced modern language and image models.
Why it matters
The distinction is practical, not academic. Rule-based software fails on anything its author didn't anticipate; a machine-learning system degrades more gracefully but can also be confidently wrong in ways nobody wrote down and nobody can fully explain. It also explains why data quality matters so much: a model learns whatever patterns are actually in its training data, including the ones you'd rather it didn't.
In practice
When a vendor says a product "uses AI", the useful follow-up is whether the learned component is load-bearing. A tool that learns from your data and improves is doing machine learning; a tool with a hand-written decision tree and a chat interface is not, whatever the marketing says.
Where this shows up
Tools and models in our catalog.
