Pierre Stock

Their best minutes

• 5:40

  The MAD Podcast

  RL training depends on calibrated tasks and resilient infrastructure

  Pierre Stock · VP of Science, Mistral AI

  46:42of 1:19:11 · Oct 22
• 4:58

  The MAD Podcast

  Mixture-of-experts models trade training complexity for inference capacity

  Pierre Stock · VP of Science, Mistral AI

  27:28of 1:19:11 · Oct 22
• 4:01

  The MAD Podcast

  A model pinged its trainer from inside a sandbox

  Pierre Stock · VP of Science, Mistral AI

  1:03:20of 1:19:11 · Oct 22
• 2:57

  The MAD Podcast

  Mistral’s in-house cluster teams fixed issues in two nights

  Pierre Stock · VP of Science, Mistral AI

  10:28of 1:19:11 · Oct 22
• 2:33

  The MAD Podcast

  Mistral trained Le Chonk with less compute than Chinese rivals

  Pierre Stock · VP of Science, Mistral AI

  18:40of 1:19:11 · Oct 22
• 4:43

  The MAD Podcast

  Mistral trains domain skills with tests, rubrics and simulations

  Pierre Stock · VP of Science, Mistral AI

  53:35of 1:19:11 · Oct 22
• 4:34

  The MAD Podcast

  Open-model guardrails raise the cost of malicious use

  Pierre Stock · VP of Science, Mistral AI

  1:07:40of 1:19:11 · Oct 22
• 2:17

  The MAD Podcast

  Mistral says its models speed senior researchers up two to four times

  Pierre Stock · VP of Science, Mistral AI

  1:14:14of 1:19:11 · Oct 22