Bittensor Guru S2E12 Subnet 2 Omron w/ Dan and Hudson
46 snips
Nov 18, 2025 Explore the cutting-edge of zkML applications with Inference Labs as Dan and Hudson dive into their innovative Subnet 2 Omron. Discover how verifiable AI computation can transform fields like finance, medical auditing, and age verification, all while ensuring data privacy. Learn about their D-Sperse system, which optimizes model slicing for efficient zero-knowledge proofs. Plus, hear insights on competition-driven hardware advances and the future of scaling AI models in a decentralized environment. Get ready for a fascinating leap into the potential of cryptographic AI!
AI Snips
Chapters
Transcript
Episode notes
ZK Proofs As A Trust Layer
- Inference Labs uses ZK proofs to verify that an AI model executed exactly as expected without revealing model internals.
- This provides a cryptographic trust layer for high-value inferences like finance or safety-critical decisions.
Slice Models To Cut ZK Costs
- D-Sperse slices large models into smaller segments and proves each segment separately to massively reduce ZK proving costs.
- Smaller segments need far smaller lookup tables, cutting memory and compute overhead and enabling proofs that were previously infeasible.
Prove Only What Matters
- Selectively prove only the layers you need, such as fine-tuned final layers, to reduce proving work.
- Use random layer selection or partial coverage to strike a balance between cost and security guarantees.