Applying Responsible AI with Open-Source Tools
summary
Claims that open-source testing bridges the gap between papers and production with concrete tools and numbers: the ACL 2020 robustness checklist finds typos or harmless edits drop accuracy by 10–20%, cloud APIs fail negation and temporal checks ~33–42%, and the BBQ benchmark covers 16 stereotype categories. A deployable NLP Test Library generates hundreds of tests, runs >50 test types across Hugging Face/Spacy/OpenAI models, and augments training data without leaking test cases so regression tests reliably gate production releases.
AI-written summary · about
chapters (4)
- 0:00Intro
- 0:33Current Gaps in Responsible AI
- 8:04Robustness, Exploratory, and Bias Training
- 26:33The NLP Test Library
related talks
43mMachine Learning Testing Ecosystem of Python
16mTransforming Test Automation with AI and Machine Learning
07mKeynote: Trusted AI: Bringing Trust Back into AI through Open Source
31mOperationalizing Responsible AI
38mAI Bias Testing and Mitigation: How to Choose the Right Open Source Tools
46mCan you trust your AI?
30mHow to Properly Test ML Models & Data
17mAI-Powered Integration Testing