The Need for Explainable AI
summary
Argues that explainability matters because AI moves into high-stakes decisions and gives concrete methods and results: a super-scalable structure-aware PDF extraction that reads 220,000 papers in 30 seconds to read 220,000 papers and builds knowledge graphs; a chemistry model trained on ~1,000,000 patented reactions that is 87% accurate on 1M reaction corpus and beat humans ~10:6 in a challenge; shows prototype-based interpretations via activation maximization with GAN constraints and cites low business trust (35%) as the adoption bottleneck.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 2:00Need
- 4:18Strategic imperatives
- 4:40Narrow AI
- 5:30Complete AI
- 5:46Broad AI
- 7:39Data Science AI
- 8:11Where do we apply AI
- 9:20How can AI help
- 11:12PDF parsing
- 12:55Organic chemistry
- 13:57Chemical reaction translation
- 14:33Humans against machine
- 16:51Trust
- 22:27Neural Network Models
- 26:57US colleagues
related talks
32mOn the Importance of Explainability
26mUsing Explainable AI to garner trust and overcome adoption inhibitors
35mWhy Do You Need an Explainable Ai?
15mIntroduction to Explainable AI
19mMoving Forward with Explainable AI: Adoption Rate & Essential Dimensions for a Reliable Use
24mWhy and How to Build Explainability into your ML Workflow
48mExplainable Artificial Intelligence (XAI): Why, When, and How?
41mIf We Want AI to be Interpretable, We Need to Measure Interpretability