The importance of representative data
summary
Claims that representative, high-quality data prevents cascading AI failures and that poor data already hurts 92% of practitioners. Describes an experimental crowdsourcing tool Task Mate (beta) deployed in India, Mexico and Kenya that collected over 600 images for an agriculture project. Shows no-code visual task editor, Google Sheets sync and API, location, audio and photo steps, peer validation and automated checks (noise, blur, GPS) plus face/license-plate blurring and paid taskers.
AI-written summary · about
chapters (6)
- 0:00Start
- 0:11Building for everyone The importance of representative data
- 7:05Intro Task Mate
- 9:56Task Mate Demo
- 14:14Conclusion
- 15:12Session Developer Guide
related talks
46mData Discussion - AI for Good
11mData-centric AI: A new paradigm
55mEthics and Bias in data science
47mData for good: Lessons from the frontline
55mKeynote - Big Data's Ethical Drought
26mHow to develop useful and unbiased AI
31mPanel Discussion : Ethics in Data Science
58mImproving disaster response with machine learning