How are large language models trained?

From the creator

DeepMind’s Nikita Namjoshi breaks ​​down the two core phases of how large language models are trained: pre-training and post-training. Learn how next token prediction builds a model's foundation, and how supervised fine-tuning, reinforcement learning, and Autoraters refine it into a safe, helpful assistant. Subscribe to Google for Developers → https://goo.gle/developers Speaker: Nikita Namjoshi Products Mentioned: Google AI, Gemini

Published 2026-07-22 · Watch on YouTube ↗
Availability and playback restrictions are controlled by the original channel.