The Roadmap to a User-Owned AI Model
summary
Proposes a practical stack for user-owned AI that uses trusted execution environments (Intel TDX, NVIDIA Hopper/Blackwell) to run models confidentially, with on‑chain node registration and encrypted model weights that only decrypt inside enclaves. Enables monetization of private models, verifiable training provenance, and decentralized training across global nodes while preventing model leakage and poisoning. Adds crypto‑incentivized benchmarks to fund continuous research and produce state‑of‑the‑art models under a decentralized confidential machine learning architecture.
AI-written summary · about
chapters (9)
- 0:00Introduction and background
- 1:29AI control and user experience shift
- 3:55Challenges with current AI models
- 5:18User-owned AI paradigm proposal
- 8:53Decentralized confidential machine learning
- 11:44Building state-of-the-art AI models
- 13:32Decentralized training and monetization
- 14:55Competitions and benchmarks for AI
- 16:19User-owned AI foundation and future
related talks
19mPrivate AI: Building Trust Through Verifiable Computation
47mTowards Confidential AI for the Masses!
37mConfidential AI: Enabling Trustworthy Ecosystems in the Age of Agentic Intelligence - Session
13mThe Value of Decentralizing AI
24mAI, Health & Confidentiality
39mTrust Cards For AI
31mStop Trusting a Black Box: The Economic Case for Open, Sovereign AI
20mBuilding Trusted AI Factories with NVIDIA Accelerated Confidential Computing