The Death of Data: Risk, Retention, and Rot
summary
Argues that indiscriminate hoarding poisons machine learning and everyday lives, citing a nine‑year old job application profile that still authenticates accounts and real harms like misidentified toddlers barred from flights, wrongful background matches, and hiring denials. Notes massive data projects: 15 TB/night raw from LSST swelling to 200 PB processed and shows how noisy business logs and duplicated records bloat risk and cost. Recommends concrete steps: hire archivists, set retention policies, automate deletion and deduplicate to shrink threat surface and downstream bias.
AI-written summary · about
chapters (18)
- 0:00Intro
- 0:37Our friend ergot
- 2:21Data has a sell-by date
- 5:18Data Hoarding
- 12:03The Death of Data
- 12:20Hold on a minute!
- 15:27Archives are not backups
- 21:11Problems
- 25:20800-588-2300
- 26:48Threat Surface
- 30:30BMI Case study in poisoned data
- 35:10Hire an Archivist
- 35:42Constrain storage space
- 36:20Train Data Retention
- 36:41Automate Deletion
- 37:30De-Duplicate
- 38:43Human Rights
- 40:43Accept Death
related talks
14mData Lifecycle Management as Foundational Pillar to AI Agents
07mThe Data Imperative
47mKeynote - The Evolving Role of Data in Software Development
24mData’s Rotten Scoundrels – How To Avoid Becoming One
48mControlling a Data Footprint – How to Build a Data Disposition Framework
27mNavigating Data's Next Great Shift
24mStop Making Your Data Team the 'Data Police'
16mGarbage Data, And What To Do About Them