Learn Responsible AI Practices By Debugging Open-Source Models
summary
Shows how the Responsible AI toolbox plugs scikit-learn and Kaggle tabular models or Hugging Face transformers and DBPedia text models into a single dashboard that computes error analysis, interpretability, counterfactuals and causal insights. Demonstrates saving error analysis cohorts, finds one cohort capturing ~40% of errors (married, higher education, low capital gain) and notes the DBPedia run with ~100 samples takes hours to compute locally; notebooks run on a laptop and can also integrate with Azure for production.
AI-written summary · about
chapters (7)
- 0:00Introduction to responsible AI
- 2:13Responsible AI principles
- 5:52Responsible AI dashboard
- 9:54Dashboard components and tour
- 18:45Training custom models
- 26:11Using pre-trained models
- 36:31Resources and next steps
related talks
35mBreaking black-box AI
52mBreaking black-box AI
01h 16mLifting the Hood on Big AI - The Future of Transparency and Accountability in AI
26mResponsible AI-What, Why, How and Future!
46mCan you trust your AI?
33mKeys to Data-Centric AI: Asking AI to Improve its own Dataset.
45mWelcome to the AI jungle! Now what?
42mBuilding and using AI models responsibly Q&A