Machine Learning + Human
summary
Google Design's Human-Centered Machine Learning (HCML) lays out seven concrete steps and checklists: ask whether ML is necessary, catalog assumptions, map user impact versus ML lift, and fake it/Wizard of Oz prototypes before costly models. Methods include tracking false positives/negatives, diverse labelers, canarying and planning for change. Claims and numbers: datasets can amplify bias (studies flag ~117 million faces in surveillance databases, roughly one in two US adults) and early IBM face models show ~35% error on some groups before remediation.
AI-written summary · about
chapters (4)
- 0:00Start
- 6:165. Plan for change, and adapting the system over time
- 10:41How would you guide the "human" to improve their outcomes?
- 26:58Models take time and expense to train, it is important to understand intentions before creating complex ML systems
related talks
38mDesigning Machine Learning for Humans
36mMachine Learning Fairness: Lessons Learned
50mEthical Machine Learning
23mMachine Learning Ethics
38mSocial Implications of Bias in Machine Learning
25mThe Uncanny Valley of ML
41mGhost in the Machine: The Unexpected Results of Not Understanding Your Users
01h 03mUsing Machine Learning and AI to serve people and communities