Google’s AI Learns to Control the ‘Artificial Sun’

The new artificial intelligence developed by Google's DeepMind company learned to shape plasma by controlling nuclear fusion inside a tokamak. Artificial intelligence will act as a major building block for nuclear fusion to enter our lives.
 Google’s AI Learns to Control the ‘Artificial Sun’
READING NOW Google’s AI Learns to Control the ‘Artificial Sun’

Google’s ‘learning artificial intelligence neural network’ company, DeepMind, has achieved an important success for the future of artificial intelligence the other day. The company taught the artificial intelligence it developed to control the nuclear fusion reaction. Artificial intelligence has been trained to control this complex process in partnership with the Swiss Plasma Center.

Humanity’s greatest tool in nuclear fusion research today, tokamaks create an environment in which hydrogen atoms collide with each other at extremely high temperatures. Thanks to this reaction, a continuously rotating plasma emerges that is even hotter than the surface temperature of the Sun. Controlling this plasma will be a key task to harnessing the potential of nuclear fusion. This is where DeepMind comes into play.

Artificial intelligence has learned to control the shape of the plasma:

Controlling the plasma will be one of the biggest engineering challenges of our time, while DeepMind will take on the task of autonomously controlling the plasma. According to the research published in the journal Nature, artificial intelligence will control the 19 magnetic coils inside the tokamak, thanks to the ‘deep reinforcement learning system’ developed.

The system started to be trained in a simulation at first. The first thing he learned was how the shape of the plasma was changed by changing the setting of the 19 magnetic coils. After that, some shapes were given to the system where the plasma should be, and artificial intelligence tried to recreate these shapes. Eventually, artificial intelligence learned how to create shapes most accurately. Moreover, he showed his success in simulation on the tokamak named ‘Tokamak à Configuration Variable (TCV)’ in real life.

Artificial intelligence’s attempt to control nuclear fusion is not actually the first in history. Since 2014, Google has been working with a company called TAE Technologies on applying machine learning to a different type of fusion reactor. In addition, the United Europe Torus in the UK is getting help from artificial intelligence to predict the behavior of plasma.

In addition to the scientific world, this concept had actually appeared before millions before. In the Spider-Man universe, Doc Ock was developing an AI and using his artificial arms to keep the fusion reactor under control – trying to ensure…

Comments
Leave a Comment

Details
159 read
okunma45175
0 comments