AI-404: explanation not found
summary
The talk examines why modern neural-network-based AI systems produce outputs (e.g., generated images) without meaningful, image-grounded explanations when asked to justify them. It discusses how the race for faster, more complex, black-box models leads to systems that give dictionary-like or opaque answers rather than concrete justifications, and asks whether current AI can be explained or has already become uncontrollable.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 2:32Data scientist role and tech watch
- 6:23AI definition and example
- 12:04Explainable AI
- 15:11Ethical and legal stakes
- 19:25AI explainability methods
- 26:24Global explainability tools
- 31:01Visualization and limits
- 34:08AI evolution and regulations
- 39:22Best practices and ACT compliance
- 43:15Conclusion
related talks
45mAI-404: Explanation Not Found
32mOn the Importance of Explainability
41mThe Explainability Problem: Towards Understanding AI
43mBring in the Lawyers: Explainable AI Driven Decision-making
35mWhy Do You Need an Explainable Ai?
56mAI: It AI-n't What You Think!
37mInteractive Explainable AI
59mIt AI n't What You Think!