← Back to list

The day I understand MuZero.

This year one of the most important gifts has not been something that can be touched: I have understood the essence of the MuZero deepmind…

juantomas · 2021-01-08 11:33 · 188 claps · 1.4 min read paywalled
#machine-learning #abadia #the-abbey-of-crime #muzero #deep-learning
Open on Medium ↗
Wiki topics: ML · Machine Learning EDU · Education & Learning

The day I understood MuZero.

Image from David Foster Post about MuZero

Image from David Foster Post about MuZero


Every year at this time two things happen: a new year begins and my birthday.

This year one of the most important gifts has not been something that can be touched: I have understood the essence of the MuZero deepmind model.

And I can only say that it has been very revealing.

The essence of MuZero is: 

"Learn how to play this game on your own - I'll tell you what moves are legal in the current position and when one side has won (or it's a draw), but I won't tell you the overall rules of the game."

MuZero is a step forward towards AGI. No prior human knowledge, no handmade human rules, no rules about how the environment works, just legal moves and rewards.

MuZero learns the rules, to understand the environment and trains a model to solve it.

On the other hand, I am very proud that my humble investigations around the #AbadIA project have ended up converging on something similar to MuZero and thanks to this birthday gift we can advance our AI. So stay tuned, pull requests ahead in a few days.
Two people have been fundamental in the process to understand MuZero:

Jesus Rodriguez
https://www.linkedin.com/in/jesusmrv/
CEO of IntoTheBlock, Chief Scientist at Invector Labs, Guest Lecturer at Columbia, Investor, Author, AI Fanatic.

His wonderful AI newsletter reminded me that I had yet to understand MuZero.

https://thesequence.substack.com/

And his medium account is a must.

https://jrodthoughts.medium.com/

and 
David Foster
https://www.linkedin.com/in/davidtfoster/
Founding Partner, Applied Data Science Partners | Author of 'Generative Deep Learning' (O'Reilly)

His book on GANs is a must:
Generative Deep Learning 

and his medium channel is awesome and has one of the best explanations on MuZero.

https://medium.com/applied-data-science/how-to-build-your-own-muzero-in-python-f77d5718061a

And of course I need to thank Sngular my employee that allow me to I+D with projects like the AbadIA the AI that will learn to solve the game "The Abbey of crime".

메타데이터
post_id
63f3d4c8b355
slug
the-day-i-understand-muzero-63f3d4c8b355
url
https://medium.com/@juantomas/the-day-i-understand-muzero-63f3d4c8b355
canonical_url
https://medium.com/@juantomas/the-day-i-understand-muzero-63f3d4c8b355
author_url
https://medium.com/@juantomas
status
ok
fetched_at
2026-07-28 14:58:00