Towards an AI Commons
3 likes1,739 views
The document discusses the potential of AI to address various societal challenges through a public utility network model, emphasizing the need for connection between problem owners and data owners. It proposes an AI commons facilitated by blockchain technology, incentivizing data sharing while preserving privacy through private data modeling. Use cases include applications in environmental monitoring, healthcare, and agriculture, ultimately aiming to scale AI for good initiatives.
Towards an AI Commons
- 1. Roadmap Alpha Aug ‘18, Mainnet Mar ‘19 Towards an AI Commons @oceanprotocol Trent McConaghy
- 3. Q: can AI help?
- 4. ✓ Predicting deforestation ✓ Tracking livestock to reduce conflict ✓ Micro insurance for crops AI + Satellite projects
- 5. ✓ Tailored smart city ✓ Give a voice to the voiceless AI + Smart Communities ✓ Epidemic modelling
- 6. 2017: AI FOR GOOD SUMMIT
- 7. Q: can AI help? A: Yes! #AIforGood
- 8. Q: can AI help? A: Yes! #AIforGood But…
- 13. Connective tissue: public utility network Problem owners Problem solvers Data owners Storage & compute
- 14. Blockchains are public utility networks
- 15. Blockchain public utility network for the AI Commons Problem owners Problem solvers Data owners Storage & compute
- 16. Service provider frontends For data commons, Storage & compute markets Public utility network incentivizing data: • Block rewards for data • Privacy-preserving AI modeling Problem owners Problem solvers Data owners Notebooks, data, storage, compute AI Commons Frontend Specify Problems Storage & compute Data science tools Data, storage, compute Network rewards Jupyter notebooks
- 17. Catalyze the commons and profit-making, via a substrate for a thousand marketplaces to bloom. Data marketplace DM DM DM DM DM DM DM DM DM DM DM DM DM DM DM DM DM DM DM Have AI (Want data) Have data Have AI (Want data) Have AI (Want data) Have data Have data Have AI (Want data) Have data Have data Have AI (Want data)
- 18. Block rewards to incentivize data supply Add to data commons, make $
- 19. Public utility network Preserve privacy: bring AI compute to the data f(x) private data modeling algorithm privately train model private model model predictions Data stays behind firewall
- 20. Framings
- 22. “Inter-Service Network” Algorithm Tribe Storage Tribe analytics further processing curation Compute Tribe Data Tribe curation curation curation
- 23. “Network of Networks, for AI Stack” Crypto perspective: Ocean is L2 on L1 AI perspective: Ocean is L1
- 24. “SABRE for data” SABRE is metadata/substrate for airline tickets. Ocean, for data itself. Consume data: Data scientists, Biz analysts Supply data: Biz’s, NGOs, Governments Data market DM AC DM AI common s DM DM AC Public utility network Consume tickets: Consumer travelers, biz ppl Supply tickets: Airlines Expedia Kayak 3rd party reseller Easyjet website Travel agent Hip munk BA website Travel agent SABRE
- 25. User Experience
- 26. Keep Data Science Tooling, Add Crypto *
- 27. Every dataset is an asset. Every Jupyter notebook is an asset (!)
- 28. New powers for data scientists •Way more data. Data commons. Enterprise data without data escapes via on-premise compute. •Provenance in data & AI training. Goodbye data honeypots. •More $. For generating data. For cleaning, labeling, feature engineering data. For algs. For curation.
- 29. Applications
- 30. Use case: water quality Satellite data in the commons to verify impact of forest restoration on water
- 31. Better predict Parkinson’s by connecting data across hospitals & borders; promoting & promoting research collaboration (ConnectedLife)
- 32. Grow Asia (WEF Spinoff): Help small farmers allocate their fertilizer better, to increase yields
- 33. Govt of Singapore Data Authorities (IMDA): Give policymakers optionality to get benefits of AI, while addressing privacy
- 34. AI Commons: 40+ UN Agencies, AI XPRIZE Using Ocean in #AIforGood
- 40. Conclusion
- 41. Q: how to scale #AIforGood? A: Blockchain public utility network for the AI commons Trent McConaghy @trentmc0