Wrap-Up Session Panel Discussion: Quick Takes
summary
A rapid roundup of concrete tech and bets: Dstack open source plus Fal cloud runs CPU/GPU private inference and aims to pool idle GPUs for cryptographic agents, while a decade-old key product called Data Security Manager acts as enterprise KMS/HSM and integrates with NVIDIA for confidential AI on-prem. Decentralized GPU markets face secrecy, measurement and performance bottlenecks; hardware as a root of trust wins on near‑native performance but requires orchestration, quorum controls and supply‑chain trust. Numbers: company founded 2020, 10‑year vendor, $7B VC fund, 27 years of firm history.
AI-written summary · about
chapters (6)
- 0:00Panelist introductions
- 1:50VC on confidential computing
- 8:22Decentralized networks vs hardware root of trust
- 16:31Q&A on model usage
- 21:11Policy and privacy in secure computing
- 25:30Closing remarks
related talks
44mGPU-less, Trust-less, Limit-less: Reimagining the Confidential AI Cloud
11mIntroductions
42mSponsored Session: Confidential Computing - New Capabilities for New Workloads
36mCTO Panel: The Impact of Confidential Computing with Intel, Azure, AMD & NVIDIA
41mMiguel de Vega, Ravital Solomon, Remi Gai & Pascal Paillier I Pragma Denver 2024
20mTrusted Hardware for AI Workloads: Extending Confidential...- 42mSponsored Session: Confidential Computing - New Capabilities for New Workloads
36mSecurity for AI and Multi-Party Collaboration with Confidential Computing (Oh, and...