Open Source is Not Enough: The Importance of Algorithm Transparency
summary
Argues that open source code alone fails because users still cannot inspect what an algorithm actually uses or why it produced a result; demands that systems see all inputs, have control surfaces described, and document assumptions and internal state. Gives concrete examples: GPS “drive” definition (drive = leave a place after >10 minutes, arrive and stay >10 minutes), Google published PageRank in 2004 then moved on, proposes a free open index of the web, and cites a neural OCR achieving ~90% accuracy as explainable.
AI-written summary · about
chapters (14)
- 0:00Algorithm transparency intro
- 1:54Good vs bad algorithms
- 4:01Opaque algorithms problem
- 8:00Transparent algorithm needs
- 11:10Old information access
- 18:21Black-box dangers
- 20:07Demanding transparency
- 22:13Open source limits
- 24:25User feedback for transparency
- 27:19Documenting algorithms
- 31:35Explaining algorithms
- 35:50Bus problem sustainability
- 40:03Why transparency matters beyond convenience
- 43:12Transparency and equality
related talks
39mA Hopeful View of Truly Open Source AI
21mKeynote: "Open as in:"
48mOpen Source: What's Next
37mIs Your AI Truly Open ? Cutting Through the Hype
26mOpen Source is Dead
17mKeynote: From Artisans to AI: The Evolution of the Open Source Ecosystem
23mDefining Open Source AI: Can the “Judgement of Solomon” Help the Open Source Community Find Success?
17mThe Future of AI, from an Open Source Perspective