Keynote: Trusted AI: Bringing Trust Back into AI through Open Source
summary
IBM argues that trust in AI requires open source toolchains and shows working projects: the Adversarial Robustness Toolbox with a demo (art-demo.mybluemix.net) that turns a Siamese cat into an ambulance under attack, AI Fairness 360 with roughly a dozen mitigation algorithms and over 70 fairness metrics that can nearly remove COMPAS-style racial bias, and AI Explainability 360 for loan-score explanations. These tools are embedded into an OpenScale dashboard for monitoring and regulatory compliance and are contributed to LF AI.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:30Trusting AI
- 2:37adversarial machine learning
- 4:03bias
- 5:02open source
- 5:12Trusted AI lifecycle
- 5:33Bias mitigation
- 8:09Open Scale
- 9:19Linux Foundation AI
- 10:06Code Aid
related talks
16mKeynote: Open By Design - Open Ecosystem, Governed AI, Trusted Outcomes
26mOpen Source Tools To Empower Ethical and Robust AI Systems
39mThe Search for Transparency and Accountability in the Age of...
16mKeynote: Open Source for AI and AI for Open Source
13mKeynote : The Open Source AI Alliance
46mCan you trust your AI?
35mOpen Source Tools To Empower Ethical and Robust AI Systems
30mStand up for ethical AI! How to detect and mitigate AI bias using Kubeflow