DeepMind has discovered another application for its artificial intelligence software.

AI Lab London is controlled by AlphabetWednesday’s announcement by, that an AI-based system has been trained to control and shape superheated plasma in a nuclear reactor.

A process called nuclear fusion that powers stars of the Universe involves smashing and fuse hydrogen. This is the common element found in seawater.

Although this process is capable of releasing large quantities of energy, it’s still subject to technical difficulties.

Here on Earth, scientists use tokamaks — doughnut-shaped vessels surrounded by electromagnetic coils — to try to recreate the nuclear fusion reactions that occur in outer space.

These tokamaks contain magnets that “contain” volatile hydrogen plasma. This is a much hotter substance than the core sun. Complex control systems with multiple layers are needed to regulate the magnetic coils.

Martin Riedmiller is the control team leader at DeepMind and co-author with a paper published in the journal NatureCNBC interviewed him on Wednesday to say that this is a “really complex” problem.

DeepMind announced that, together with EPFL’s Swiss Plasma Center in Lausanne (Switzerland), it had developed a reinforcement-learning AI system which can change the voltage of magnets thousands upon seconds.

Reinforcement learningDeepMind believes that ‘deepminding AI is a very positive technique for AI training. It involves programming AI to do certain things in order maximize the chance of it earning a reward. This means that the algorithm is “taught” how to accomplish a task using pre-programmed rewards.

DeepMind has used the unnamed AI developed in a virtual simulation around 100 times for a Variable Configuration tokamak. The reactor’s magnets were controlled by the AI for just two seconds. This is how long the reactor can operate before overheating.

DeepMind employed approximately 10-20 people to create the AI system, along with about 5-10 EPFL employees.

Federico Felici (a Swiss Plasma Center research scientist) said Fusion is “one of the most fundamental sources of energy that we have in this universe.” This technology will allow you to generate near limitless amounts of energy for the future if we can master it.

Damien Ernst (Universite de Liege professor) praised the research and said that the results could “significantly accelerate the development of fusion reactions and ultimately our ability to combat global change.”

DeepMind was founded to solve artificial general intelligence (often referred to by AI experts as “the holy grail”).

In its infancy, the company developed AI systems to master games like Go and Chess. It now wants to use more of its technology for real-world applications and science.

DeepMind AI has been used by Google, but it hasn’t been widely adopted elsewhere.

DeepMind CEO Demis Hassabis stated in a statement, that DeepMind has shown AI’s ability to speed scientific progress, and opens new avenues for research across biology and chemistry.

DeepMind employs approximately 1000 people around the globe, which includes some of world’s top AI research scientists. These researchers can earn more than $1,000,000 annually. Because these people are highly valued by Big Tech companies such as Amazon, Facebook and Apple, they can also command high salaries.