DeepMind scientists have trained an AI to control nuclear fusion

Scientists at DeepMind have found another real application for their AI program.

London-based artificial intelligence lab owned by the alphabetOn Wednesday, it announced that it had trained an artificial intelligence system to control and sculpt ultra-hot plasma inside a nuclear fusion reactor.

Nuclear fusion, a process that powers the stars of the universe, involves breaking down and fusing hydrogen, a common element in seawater.

This process, which releases huge amounts of energy, has been described as an unlimited source of clean energy, but a number of technical challenges remain to be overcome.

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

The magnets in these tokamaks are used to “contain” volatile hydrogen plasma, which is much hotter than the sun’s core. Magnetic coil control currently requires multiple layers of complex control systems.

Martin Redmiller, Head of the Control Team at DeepMind, and co-author of the book A A paper published in the journal Nature On Wednesday, he told CNBC that it’s a “really complex” control issue.

In collaboration with the Swiss Plasma Center at EPFL, a university in Lausanne, Switzerland, DeepMind said it has developed an AI-enhanced learning system that can control magnets and alter their voltage thousands of times per second.

Learning Enhancement, an AI training technology that DeepMind is particularly trumpeting, involves programming an AI to take certain actions in order to maximize its chance of earning a reward in a given situation. In other words, the algorithm “learns” to complete a task by looking for these pre-programmed rewards.

The unnamed AI DeepMind, which was developed on a virtual simulator, has been used about 100 times on a tokamak at the Swiss plasma center known as the Tokamak Variable Formation. The magnet controls the tokamak for two seconds, which is the maximum amount of time the reactor can run before it warms up.

About 10-20 people from DeepMind worked on the AI ​​system with about 5-10 people from EPFL.

“Fusion as I see it is one of the most fundamental sources of energy that we have in the entire universe,” Federico Felici, a research scientist at the Swiss Plasma Center told CNBC. “Once we have actually mastered this technology… it is a huge feat because you will be able to obtain nearly unlimited energy in the future.”

Damien Ernst, a professor at Université de Liège who was not involved in the work, hailed the research as one of the most important applications of reinforcement learning to date, adding that it “could significantly accelerate the development of fusion reactors and, ultimately, have the potential to fight global change.”

DeepMind has set out to hack Artificial General Intelligence, which is often referred to as the Holy Grail of AI.

The company has begun developing artificial intelligence systems that can master games such as chess and go. Now, she wants to apply more of her technology to real-world applications and science.

While Google has found uses for DeepMind’s AI, its technology has not been widely applied elsewhere.

Demis Hassabis, CEO of DeepMind, said in a statement that the company has demonstrated the ability of AI to accelerate scientific progress and open up new areas of research across biology, chemistry, mathematics and now physics.

DeepMind employs around 1,000 people worldwide, including some of the world’s leading AI research scientists, who can command annual salaries in excess of $1 million. These high profile people, who often have PhDs from the likes of Oxford, Cambridge, Stanford and MIT, can get that kind of money because they are also sought after by big tech companies like Facebook, Apple, Amazon and Microsoft.