Part 4: Actions, Reasoning & World Models – DeepMind’s Era of Experience Explained
Автор: Saurabh Anand
Загружено: 2025-05-19
Просмотров: 68
Part 4 of this series digs into: How agents observe, plan, and reason in the era of experience.
We cover:
Why LLMs only get a teaser of experience through APIs and tool use
The limits of human-style reasoning (yes, even “Let’s think step-by-step” can be flawed)
How flawed historical reasoning reveals the need for AI to evolve its own ways of thinking
The rise of world models — simulated realities where AI learns, adapts, and builds intuition
0:00 – Recap of Part 3: Streaming Experience & Rewards
2:04 – Actions & Observations: LLMs, APIs & Game Environments
5:10 – Planning & Reasoning: Beyond “Let’s Think Step-by-Step”
11:44 – World Models: Simulated Experience as a Birthplace of Intelligence
16:45 – Teaser for Part 5: Philosophy, AI Safety & Goal Alignment
This video is made for educational and commentary purposes.
All trademarks, logos, and copyrighted content belong to their respective owners. This content falls under fair use as it is transformative, non-commercial, and intended to inform and educate.

Доступные форматы для скачивания:
Скачать видео mp4
-
Информация по загрузке: