Machine Learning for Web3
summary
Claims an ML pipeline that scores Solidity smart contracts 0–1 for maliciousness, built a public API that parses source into ASTs and returns confidence (demo shows 0.98 for a legitimate “merge” NFT contract). Describes engineering reality: Solidity data is write-optimized not read-optimized, no ABI/schema standardization, poor dev tooling; large LLMs need thousands of GPUs and can cost millions; proposes a censorship-resistant ML stack with decentralized compute, auditable/ZK proofs and privacy-preserving open ledgers.
AI-written summary · about
chapters (16)
- 0:00Introduction and background
- 2:08Evolution of machine learning
- 3:28Killer apps and recommendations
- 5:28The rise of generative AI
- 9:18Compute and training challenges
- 13:00The cost of AI development
- 16:04GitHub Co-pilot and innovation
- 20:01Democratizing AI with chatbots
- 21:49The transformer revolution
- 24:35Machine learning for Web3
- 27:46Demo: Smart contract scoring
- 31:38Defining the Web3 stack
- 35:56The importance of data and privacy
- 37:43Current challenges in Web3
- 41:29Censorship resistance
- 42:43Future of secure e-voting
related talks
01h 01mIntroduction to AI Security
28mML-Driven Test Intelligence at Scale - What Works What Fails and Why It Matters
21mEnd-to-end secure ML Development
21mMachine Learning and Cyber Security - Detecting malicious URLs in the haystack
53mSecurity of Machine Learning Systems
26mRunning Secure AI/ML Experiment in a Unified MLOPS-DevSecOps Pipeline
10mML-based Log Analysis for Faster Debugging
16mLeveraging ML to Create a Reliable Anti-Fraud System