Tournesol An Open Source Project for Collaborative Algorithmic Governance
summary
Tournesol is an AGPL open-source platform that crowdsources YouTube recommendation judgments: contributors paste URLs and use a slider to pick which of two videos should be recommended more often, producing a public CC BY dataset and community-driven recommendations shown via browser extensions (Chrome, Firefox). Backend is Django with SciPy/Numpy for ML, frontend in React. Robust aggregation uses a quadratic-regularized median-like method to limit how much small groups can bias scores; defenses include email certification and a vouching system.
AI-written summary · about
chapters (6)
- 0:00Introduction to Tournesol
- 7:47Problem with recommendation algorithms
- 13:27How Tournesol works
- 28:18Security challenges in ML systems
- 36:22Call for collective action
- 40:29Q&A session
related talks
17mThe Future of AI, from an Open Source Perspective
01h 48mRecommendation engines explained
11mDesigning an Open-Source Platform for Differentially Private Analytics That Is Usable...
32mPlatform-supported Auditing Of Social Media Algorithms For Public Interest
24mCrowd Research: Open and Scalable University Laboratories
23mFrom Data to Recommendations: Building an Intelligent System with ROR
23mML-Powered Search & Recommendation 101
32m"You may also like"
