Ocean Protocol: New Powers for Data Scientists

Ocean Protocol: New Powers for Data Scientists

• 1.
  Trent McConaghy @trentmc0 Ocean Protocol Newsuperpowers for data scientists
• 2.
  1000x more data AILoves Data 1000% less error!
• 3.
  Deep Learning ReallyLoves Data DL models with >> capacity Error 5% .. 0.01% Another 1000x more data Models with limited capacity Error 25% .. 5%
• 4.
  Have lotsa data (1000enterprises) Have lotsa AI expertise & related problems to solve (1000 AI startups)
• 5.
  Have both: Goog, FB,a handful $$ Have lotsa data (1000 enterprises) Have lotsa AI expertise & related problems to solve (1000 AI startups)
• 6.
  Connect 1000+ enterprises With1000+ AI startups All get lotsa data & AI Connecting substrate Connect them!
• 7.
  Ocean incentivizes datasharing (& selling) Ocean public utility network Data permissioning, curation, block rewards for data Data market DM AI commons Data science tools (sklearn, Tensorflow, ..) DM DM DM Data scientists With problems to solve Enterprises, govts, NGOs With data & compute
• 8.
  Why Ocean isdecentralized 1. So data owner retains control of their data “If you don’t have the key, you don’t control the data” 2. Must be a public utility, vs controlled by a single entity. Anti-pattern: FB 3. Allows for block rewards to incentivize data supply Add to data commons, make $.
• 9.
  Roadmap Alpha in Aug,v1 next Mar
• 10.
  Curation of Datasets& Algs. Put your $ where your mouth is!
• 11.
  Manage Your Datasets
• 12.
  Use Ocean viaJupyter * Metamask (Interactive python * crypto wallet)
• 13.
  New superpower #1:More data •Way more data → more accurate models •Way more free data. Via data commons. •Access to data behind others’ firewalls. Via privacy- preserving on-prem compute
• 14.
  New Superpower #2:Provenance & security •Goodbye data honeypots. Goodbye liability. •Data provenance. Where did your data come from? •AI training provenance. AV black boxes, GDPR explainability.
• 15.
  New Superpower #3:More $$ •Generate useful data, make $ •Improve data, make $. Clean it, label it, feature-engineer it •$ for your enterprise’s data, without risk of data escapes. (Compute on it locally.) •Invent an alg, make $ •Curate others’ algs & data, make $
• 16.
  Applications
• 17.
  Q: For autonomousvehicles, need 500G driven miles. How? A: Together. MOBI: 70% of auto production. Toyota, BMW, ..
• 18.
  ConnectedLife: Way more datafor Parkinson’s research
• 19.
  Grow Asia (WEFSpinoff): Help small farmers manage their plots & protect biodiversity
• 20.
  Govt of SingaporeData Authorities (IMDA): Give policymakers optionality to better address privacy, yet benefit from AI data
• 21.
  Conclusion
• 22.
  Ocean gives datascientists new superpowers •Way more data. Data commons. Enterprise data without data escapes. •More $. For generating data. For cleaning, labeling, feature engineering data. For algs. For curation. •Provenance in data & AI training. Goodbye data honeypots. •For applications in AVs, medical, agriculture, govt, … •Owned & operated by the people!