Recommender systems (collaborative filtering to deep-learning feeds)
Algorithms that predict what each person will want from the behaviour of many others; they now decide much of what people see in shops, streaming services and social feeds.
- 1992

First version
David Goldberg, David Nichols, Brian Oki and Douglas Terry (Xerox PARC, Tapestry); later GroupLens, Amazon, Netflix and YouTube
Image: The original uploader was Tvwade at English Wikipedia. · Public domain · Wikimedia Commons - 1994

GroupLens automated collaborative filtering
Predicts each reader's rating of Usenet news posts from people who rated alike, with no manual annotations.
Image: GroupLens Research, University of Minnesota · Trademark logo, used to identify the pr… · Wikipedia - 2003

Amazon item-to-item collaborative filtering
Matching similar items rather than similar users scales recommendations to millions of customers in real time.
Image: Amazon.com, Inc. · Public domain · Wikimedia Commons - 2009

Netflix Prize and matrix factorisation
Latent-factor models trained on 100 million ratings win a $1 million contest and become the industry standard.
Image: Jason Kempin · Rights-managed editorial photograph; co… · Getty Images, published by Thrillist; direct fu… - 2016

Deep-learning recommendation feeds
YouTube ranks its vast video catalogue with deep neural networks trained on viewing history; deep learning becomes the core of algorithmic feeds.
Image: Youtube · Public domain · Wikimedia Commons