Save 37% off PRO during our Black Friday Sale! »

Living at the Intersection of Blockchains and Machine Learning

0b0edcef2f3aee7b51cda719e5f82d7c?s=47 Arpit Mathur
September 20, 2021

Living at the Intersection of Blockchains and Machine Learning

Presentation given at the Philadelphia AI (PHLAI) conference - Sept 2021

0b0edcef2f3aee7b51cda719e5f82d7c?s=128

Arpit Mathur

September 20, 2021
Tweet

Transcript

  1. CONFIDENTIAL Living at the Intersection of Blockchains and Machine Learning

    Arpit Mathur Principal Engineer
  2. AI and Blockchains?! AI Blockchain More data is better Less

    data is better Centralized systems Decentralized systems Black Box Radical Transparency Aren’t Blockchains Antithetical to AI?
  3. AI Challenges • Trust • Bias • Walled Gardens •

    Data Sharing
  4. Can Blockchains alleviate AI’s biggest challenges?

  5. Understanding Blockchains

  6. Blockchains as we know it • A linked list of

    Blocks, each block containing many transactions • Wallets and Smart Contracts announce transactions • Miners combine Transactions into blocks • Miners compete with others to write blocks to the chain via Proof of Work • The winning miner then writes the data to the chain
  7. From Web 2.0 to 3.0 Web 2.x Web 3.0 User

    DNS Application Server App Splunk
  8. Web 2.x Web 3.0 (Ethereum version) User Wallet DNS ENS

    Application Server Blockchain App Contract Splunk Etherscan From Web 2.0 to 3.0
  9. Writing code for Ethereum

  10. Smart Contracts • State Machines • State changes via Transactions

    • Written in Solidity and other languages • Deployed as Bytecode • Must be deterministic
  11. A Hello World contract

  12. Remix and Metamask

  13. OpenZeppelin

  14. Coins and NFTs • Tokens = wallet addresses + balances

    stored in a contract • Standard contract that implements a certain interface
  15. ERC 20 vs ERC 721

  16. Blockchains and AI Data Marketplaces AI products as NFTs Inclusive

    organizations and opportunities
  17. Rise of Data Marketplaces

  18. Bring verifiable data into Blockchains with Oracles

  19. None
  20. “A blockchain comes in and collects all the learnings. I

    emphasize. The learnings, and not the patient data … the parameters, for example, of the neural network weights in the machine learning model. So in this case, no patient data ever leaves an individual hospital” - Dr. Eng Lim Goh, Senior vice president and CTO of artificial intelligence at Hewlett Packard Enterprise.
  21. Renting/Selling Algorithms and ML models on the Blockchain

  22. … as NFTs?

  23. Licensing ML products as NFTs

  24. Democratizing Software Development with GitCoin

  25. Building unbiased performance-oriented organizations

  26. Are we there yet? • Performance and Environmental concerns •

    Motivation
  27. Ethereum 2.0 • Goals o Better energy o Reduced hardware

    requirements o Stronger immunity to centralization o Stronger support for shard chains • 15 TPS to 100,000 TPS Note: Permissionless blockchains will always be slower than databases. That is not the goal.
  28. If you want to go fast, go alone. If you

    want to go far, go together – African proverb
  29. Thank you