Ergebnisse
11 ArtikelSortieren
1:04:26EN
Sequoia Capital· 6. Okt. 2026PodcastAt 100,000 Accelerators, Something Fails Every HourGoogle AI infrastructure chief Amin Vahdat tells Sequoia's Sonya Huang why FLOPS is a vanity metric, why the TPU line split in two, and why he would rather wait on a utility than build his own gigawatt.
1:34:36ENNate Foster: Treating the Network as a ProgramYaron Minsky, Nate Foster · Jane StreetFrom programmable switches to BGP, a programming-languages researcher explains how higher-level policies and formal models can make networks easier to change.
45:33ENSequoia Capital· 26. Mai 2026PodcastCursor Ships 1TB of Composer Weights Across Oceans in Under a MinuteFederico Cassano and Fireworks' Dima walk through the globally split RL stack that trained Composer 2 on a sparse trillion-parameter MoE.
41:32ENSequoia Capital· 28. Okt. 2025PodcastMoore's Law Is Slowing. Nvidia's CTO Still Wants 10× a YearMichael Kagan on why Nvidia's $7 billion Mellanox deal, not just more transistors, is how you turn a data center into one computer.
1:56:38EN
Cognitive Revolution "How AI Changes Everything"· 8. März 2025Podcasto3 took ARC from 20% to 87%. Nathan Labenz sketches early superintelligenceModeration: Nathan LabenzOn Consistently Candid, the Cognitive Revolution host argues that math and coding will be superhuman by 2026, and that a reasoning core plus specialist “intuitive physics” may already be enough.
2:16:17ENCognitive Revolution "How AI Changes Everything"· 5. Feb. 2025PodcastA 10B Model Trained Across Three Continents — and an H200 for $1.49 an HourModeration: Nathan LabenzPrime Intellect's founders lay out a four-step plan to turn spare global GPUs into collectively owned intelligence, and argue that reinforcement learning just made their job easier.
1:03:21ENHorace He on Building ML Systems for Trillion-Trillion OperationsHorace He · MetaHow GPU bottlenecks, compiler design and distributed parallelism shape the systems behind large-scale model training.
46:35ENSequoia Capital· 13. Nov. 2024PodcastThey trained a real-time video model in 20 hours and skipped the game engineDecart CEO Dean Leitersdorf tells Sequoia how Oasis runs live on H100s, why ChatGPT is a fallen wall rather than a pain point, and why the next trillion-dollar company may be a generated experience.
42:42EN
No Priors: AI, Machine Learning, Tech, & Startups· 22. Aug. 2024Podcast · 2 StimmenEthereum Once Held Two Thousand GPT-3 Clusters of GPUsModeration: Sarah Guo, Elad GilFoundry CEO Jared Quincy Davis on idle H100s, why today's AI cloud is still colocation, and how compound systems move work off giant interconnected clusters.
1:01:26ENCognitive Revolution "How AI Changes Everything"· 25. Juli 2023PodcastAlexNet Used Two GPUs. GPT-4 Took Twenty Thousand. The Shortage Will Be Back.Moderation: Nathan LabenzRun:ai CTO Ronen Dar on filling idle accelerators, why H100s can be cheaper per experiment, and why Llama 2 undercuts compute governance.
53:37ENNo Priors: AI, Machine Learning, Tech, & Startups· 9. März 2023Podcast · 2 Stimmen30 models, 42 scenarios, 7 metrics: Percy Liang's public checkup for language modelsModeration: Sarah Guo, Elad GilA PubMed-only model once led the U.S. medical licensing exam — until a system roughly 200 times larger overtook it.