Trustworthy AI & Open Source
summary
Panel reports concrete results: small neural nets trained with tanh show boolean and predicate‑like rules and an interpreter network can compose pre-trained modules — neural cells showing propositional calculus. Europe’s AI4EU on‑demand platform already hosts 61 assets and is moving toward Eclipse open source governance. ANITI pursues a hybrid logic+statistical strategy, translating learned models into clauses solvable at scale and using optimal transport to debias. DFKI stresses uncertainty handling and synthetic testing for autonomous driving; CEA ships N2N/2D2 tooling and integrates pharmacy for partial formal validation.
AI-written summary · about
chapters (16)
- 0:00Start
- 5:01German Research Center for Artificial Intelligence
- 9:02Perceptual Ai
- 11:10Three Levels of Communication
- 28:08The Ai on Demand Platform
- 28:54Ai on Demand Platform
- 30:33What Is Trustworthy Ai
- 30:48Guidelines for Trustworthy Ai
- 33:26Ai Platform
- 35:11What Is Aniti
- 41:48Interpretable and Explainable Ai
- 46:25Robust Ai
- 48:14Image Recognition
- 53:31Ethical Legal and Epistemological Consequences of Biases
- 1:12:10Overview
- 1:26:31Embedded Neural Networks
related talks
58mBuilding Trust and Safety in Artificial Intelligence with Abstract Interpretation
42mReasonable AI
46mCan you trust your AI?
29mTrustworthy AI for the Physical World
01h 00mThe Promise and Limitations of AI
32mHow to Build Trustworthy AI with Open Source
34mWhy Transparent AI is the only AI that can be trusted
30mTrustworthy AI - opening up the black box