Scientists overcame a significant challenge by utilizing artificial intelligence (AI) to create nearly infinite clean energy through nuclear fusion. A team at Princeton University discovered how to employ AI models to forecast and prevent instabilities within plasma during fusion reactions.

Nuclear fusion, celebrated as the “holy grail” of clean energy due to its capacity to generate massive quantities of energy while not necessitating fossil fuels nor producing hazardous waste, has long proved elusive to harness. This process mirrors the natural reactions occurring within our sun; however, nuclear fusion energy has remained challenging to master.

In 2022, a group at Lawrence Livermore National Laboratory in California attained net energy production for the first time via nuclear fusion—producing more energy than consumed in the reaction, albeit just enough to boil water. Nonetheless, this achievement marked an important step toward scaling up this technology.

Recently, a groundbreaking advancement occurred when AI became capable of identifying plasma instability 300 milliseconds prior to its manifestation—plenty of time to intervene and maintain plasma stability. According to the researchers, this novel comprehension may pave the way for large-scale implementation of nuclear fusion energy grids.

As per the principal investigator, Eygeman Koleman, a physicist working at the Princeton Plasma Physics Laboratory, AI learns from previous experiences instead of integrating data derived from physics-based models, thereby developing a conclusive control strategy supporting a steady, powerful plasma system in actual reactors, in real-time.

The most recent findings were published on Wednesday in the prestigious scientific journal Nature, entitled “Fusion Plasma Tearing Instability Avoidance with Deep Reinforcement Learning.”

Co-author Sankyeun Kim added, “Our capability to foresee instability ahead of time facilitates responsiveness compared to existing techniques, which are more passive. We no longer need to await instability development and subsequently implement prompt corrective measures before plasma disruption ensues.”