MuZerio is perceived as an important step forward in the searching for general-purpose algorithms. To put it simply, MuZero is a computer program which was developed by DeepMind company in order to master games and new artificial environments without knowing their rules.
MuZero is one of the newest solutions in the pursuit of methods that can not only learn a model which explains their environment, but also will be able to plan the best course of action. The program masters games like chess, Go, shogi and Atari without being told the rules in advance.
”MuZero really is discovering for itself how to build a model and understand it just from first principles.”
— David Silver, DeepMind, Wired
DeepMind over the last few years came out sequentially with AI programs: AlphaGo (2016), AlphaGo Zero (2017) and AlphaZero (2017). The thing which was common for all of them is that they got the rules of the games they had to master going into their training.

MuZero uses different techniques than its predecessors and therefore overcomes its limitations. The program doesn’t try to model the entire environment, instead it models just aspects that are crucial for AI in the decision-making process.
MuZero doesn’t rely on given knowledge of the environment’s dynamics, such as the rules of the game or an accurate simulator. This ability gives a hope that in the near future we will be able to apply this program to messy and complex real world problems.
Dr David Silver said that DeepMind was already using MuZero to try to develop a new kind of video compression, which could make a massive savings e.g. in data volume.
Moreover, its most advanced predecessor, AlphaZero, has been applied to a variety of complex problems in fields like chemistry, quantum physics and more.
Researchers have never been closer to developing a general-purpose algorithm – MuZero marks a new beginning in AI that can significantly accelerate and facilitate tackling real-world problems
which are typically hard to distill into simple rules. There is no doubt this technology will have a notable impact in tackling new challenges in robotics and industrial systems.
It seems like the ability to plan, allowing humans to generalise gathered experience to make predictions on new scenarios, will not be (sooner or later) the only human domain.
Sources:
https://www.bbc.com/news/technology-55403473
https://www.engadget.com/deepmind-muzero-160024950.html
https://deepmind.com/blog/article/muzero-mastering-go-chess-shogi-and-atari-without-rules
https://www.analyticsdrift.com/deepminds-muzero-marks-new-breakthrough-in-reinforcement-learning/