Your datasets, under your control: Introducing the Mozilla Data Collective
summary
Claims that the web has hit a token crisis get quantified: an estimate of 114 trillion tokens used for a single large model versus 60–160 trillion web tokens available, implying human data scarcity. Commons datasets like Common Voice remain CC0 but aren’t always suitable for communities, so a platform is built that lets contributors keep ownership, set terms, track usage, publish data sheets for 316 languages, and provide discoverability and supply‑chain transparency for text now, with audio and images planned.
AI-written summary · about
chapters (7)
- 0:00Introduction and context
- 3:46Overview of talk structure
- 5:11Tokenomics and data value
- 9:48The token crisis and AI slop
- 13:59Exploitative token harvesting
- 18:15Mozilla Data Collective
- 21:58Q&A and closing remarks
related talks
44mCommon Voice - Building Multilingual Voice Datasets
32mSpeech Technologies for a Multilingual World
36mCataloging the world's data for great good
38mWithout Open Data, there is no Ethical Machine Learning
46mPanel Discussion: What is Hot in Data World in 2022?
10mBig Data and the Hypocrisy of Privacy
33mWrangling Internet-scale Image Datasets
33mWhat’s the true value of tokens?