Without Open Data, there is no Ethical Machine Learning
summary
Claims that open data is necessary for ethical ML, documents companies like Reddit Twitter and Snapchat restricting API or dataset access, and quantifies harm with concrete examples of lost training sources and lobbying wins that limit model choices. Describes tools and standards actually used like CSV APIs metadata schemas and the five C’s of data integrity, shows projects that are building real collections such as the European Federation for Cancer Images and New York City open data, reports a planned dataset release and whether maintainers keep it current, and urges funding and maintainer support as the practical fix.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 3:38Current state of open data and machine learning
- 6:26ML and open data history
- 9:12Personal journey
- 12:05Open data defined
- 14:16Government open data challenges
- 16:23Why open data matters for ethical machine learning
- 18:55Example with my dog Q
- 20:18Public data access loss
- 21:26Worst-case scenarios
- 22:11Solutions for ethical ML
- 26:29Open data organizations
- 28:16Call to action
- 29:19Q&A: Government and companies
- 30:44Q&A: Advice for librarians
- 32:48Q&A: Stigma around open data
- 34:54Q&A: Dangerous datasets
related talks
50mEthical Machine Learning
36mMachine Learning Fairness: Lessons Learned
37mKeynote: Ethical Machine Learning: Creating Fair Models in an Unjust World
47mEthical Machine Learning
38mSocial Implications of Bias in Machine Learning
23mMachine Learning Ethics
54mEthical Machine Learning
39mEthical ML: Who's Afraid of the Black Box Models?