How DeepSeek Rewrote the Rules of AI Training: MOE & Efficiency Deep Dive
Автор: Fridman Way
Загружено: 2026-03-15
Просмотров: 79
Описание: Dylan Patel and Nathan Lambert break down the technical breakthroughs that allowed DeepSeek to achieve frontier-level performance at a fraction of the cost. They discuss Mixture of Experts (MOE) architecture, Multi-head Latent Attention (MLA), and the 'Bitter Lesson' of scalable learning. This deep dive explains why 2025 is the year of architectural efficiency in AI. #LexFridman #DeepSeek #AIArchitecture
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке: