This workshop is a follow-up to the talk “Decentralized AI” by Gene Kogan on December 12, which is about the intersection of artificial intelligence with decentralization technologies such as blockchains and smart contracts. Primary sub-topics include blockchain mechanics and cryptography, federated learning, and security considerations. Relevant platforms will be introduced. The tentative syllabus is as follows:
Cryptography prerequisites
- Public key cryptography
- Hashing and proof-of-work
- Smart contracts
- TUTORIAL: Digital signatures
Blockchains
- How blockchains work
- Bitcoin protocol
- TUTORIAL: Proof-of-work on simple blockchain
- Survey of cryptocurrencies
- Second-generation blockchain applications
- Namecoins, colored coins, smart properties, tokens
- Distributed data, BigchainDB
- Deconstructing blockchains
- Blockchain as a state-transition system
- Ethereum
- Decentralized autonomous organizations (DAO)
Machine learning
- Deep learning
- Applications to computer vision, natural language processing
- TUTORIAL: Basic image classifier in Keras
- Problems with centralized machine learning
- Data security and surveillance
- Problems with economic model
- Federated learning
- Security issues with federated learning
- Differential privacy
- Why are blockchains relevant to machine learning?
Blockchains + machine learning
- Federated learning on a blockchain
- Openmined
- AI + DAOs
- Impact of AI on decentralized applications
- Asset management + “sharing economy”
- Data markets and exchanges
- Mutual funds, insurance, co-ops
- Dangers of AI DAOs
- Art DAOs