DARPA LifeLog (en.wikipedia.org)

DARPA’s LifeLog, proposed in 2003 by the Information Processing Techniques Office, was an audacious lifelogging research effort to build an “ontology-based” system that captured, stored and made searchable the complete stream of an individual’s experiences. The solicitation described aggregating phone logs, emails, web sites, credit-card purchases, faxes, TV/radio choices, books read, location via wearable GPS, and biomedical and sensor data—aiming to trace the “threads” of events, states and relationships and surface preferences, plans, goals and other markers of intentionality. A core objective was predictive: mining timelines to infer routines, habits and social links to support assistants and automated services. Technically, LifeLog anticipated a massively multimodal, longitudinal dataset combining personal sensors, communications, transactional records and semantic ontologies to enable pattern-finding and behavior prediction—exactly the kind of rich training data that powers modern ML personalization and sequence models. Its scope made it a touchstone for debates about privacy, surveillance and dual-use research: while offering powerful capabilities for context-aware AI, it also raised major ethical and policy risks. Intense media scrutiny and political concern led to the project’s cancellation in 2004, but LifeLog’s ambitions presaged contemporary lifelogging, pervasive-sensing research and the privacy challenges of big-data ML systems.