British artificial intelligence scientist and DeepMind CEO Demis Hassabis.
OLI SCENE | AFP | beautiful pictures
Scientists at DeepMind have found another practical use for their artificial intelligence software.
London-based AI Lab, owned by Alphabetannounced Wednesday that it has trained an AI system to manipulate and generate a superheated plasma inside a nuclear fusion reactor.
Nuclear fusion, a process that powers the stars of the universe, involves the smashing and fusing of hydrogen, a common element of seawater.
This process, which releases large amounts of energy, is considered a potentially limitless source of clean energy, but several technical challenges still need to be overcome.
Here on Earth, scientists use tokamaks – doughnut-shaped vessels surrounded by electromagnetic coils – to try to recreate the nuclear fusion reactions that occur in space.
The magnets in these tokamas are used to “contain” the volatile hydrogen plasma, which is hotter than the core of the sun. Controlling current magnetic coils requires multiple layers of complex control systems.
Martin Riedmiller, control group leader at DeepMind, and co-author of a The article was published in the journal Nature on Wednesday, told CNBC it was “a really complicated control issue”.
Together with the Swiss Plasma Center at EPFL, a university in Lausanne, Switzerland, DeepMind says they have developed a reinforcement learning AI system that can manipulate magnets and change their voltage thousands of times. times per second.
Reinforcement learning, an AI training technique of particular interest to DeepMind, involves programming the AI to perform certain actions in order to maximize the chances of earning a reward in a particular situation. In other words, the algorithm “learns” to complete a task by looking for these pre-programmed rewards.
DeepMind’s unnamed AI, developed on a virtual simulator, was used about 100 times on a tokamak at the Swiss Plasma Center called the variable configuration Tokamak. It drives the magnets in the tokamak for two seconds, which is the maximum amount of time the reactor can run before overheating.
About 10-20 people from DeepMind worked on the AI system along with about 5-10 people from EPFL.
“Fusion as I see it, it’s 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’ve really mastered the technology…it’s a huge achievement because you’ll be able to have virtually limitless energy ahead of you.”
Damien Ernst, a professor at the University of Liege 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 dramatically accelerate the development of fusion reactors and ultimately resistance to global change.”
DeepMind set out to disrupt artificial intelligence in general, often referred to as the holy grail of AI.
The company started out by developing AI systems that could master games like chess and Go. Now, it wants to apply more of its technology to real-world applications and science.
While Google has found applications for DeepMind’s AI, its technology has yet to be widely adopted elsewhere.
DeepMind CEO Demis Hassabis said in a statement that the company has demonstrated AI’s potential to accelerate scientific progress and open up new avenues of research in biology, chemistry, mathematics and science. Now it’s physics.
DeepMind employs around 1,000 employees worldwide, including some of the world’s top AI research scientists, who can command an annual salary of more than $1 million. These top people, often with PhDs from Oxford, Cambridge, Stanford and MIT, can earn this money because they are also sought after by big Tech companies like Facebook, Apple, Amazon and Microsoft.
