Understanding the Legal Context of AI
Talkat Open Source Summit Europe 2023 · on Tuesday, 2023-09-19· at 11:25· Room 0D-2-0D-3 (Floor 0), Abandoibarra Etorb., 4, 48011 Bilbao, Bizkaia, Spain·
summary
Explains how models are trained (architecture + implementation code + weights are just matrices of probabilities) and why that matters legally: training often reads works to produce statistical measurements, so courts may treat it like Google Books fair use rather than wholesale copying. Shows memorization risk but gives numbers: researchers extract 108 matches from 90,000 images and larger models drop reproductions to about 0.00013%. Notes pending cases (Copilot, Stability, Getty) hinge on outputs not training, and indemnity and copyright office positions remain limited and unsettled.
AI-written summary · about
chapters (6)
- 0:00AI and legal context
- 4:39What a ML model is
- 17:01Copyright in AI training
- 29:23AI lawsuits and cases
- 35:23Trademark of AI outputs
- 39:13Liability for AI content
related talks
34mEthical framework for AI : Creating AI powered products that empower people
54mAI and Ethics: Navigating the Future of Responsible Technology
46mAI for Legal : The story of hype, adoption and what comes next
53mML BuzzWords demystified
41mAI: Q&A
44mKeynote: Tracking the Pulse of Generative AI
01h 16mLifting the Hood on Big AI - The Future of Transparency and Accountability in AI
28mEthical AI?!