Researchers at the Lawrence Livermore National Laboratory (LLNL) in Livermore, California, have developed a new artificial intelligence (AI) model that improves predictions of nuclear fusion experiments, potentially accelerating the search for clean energy.

The model accurately forecast the outcome of a 2022 experiment at LLNL’s National Ignition Facility (NIF), assigning a 74% probability of ignition. This outperformed traditional supercomputer simulations, which achieved only 70% accuracy and required days of processing.

Fusion, the process of combining lightweight atoms such as hydrogen into heavier ones, releases vast amounts of energy with minimal radioactive waste. In contrast, fission, which powers today’s nuclear reactors, produces a greater volume of hazardous byproducts.

NIF’s fusion experiments use 192 lasers to compress hydrogen isotopes inside a tiny gold cylinder, generating the extreme pressures and heat needed to trigger fusion reactions. Achieving a state where energy output exceeds input remains one of science’s greatest challenges.

Existing computer models often provide rough approximations of experimental outcomes. LLNL’s new AI overcomes this by combining past experimental data, expert insights, and high-fidelity physics simulations. Trained on more than 30 million Central Processing Unit hours, the model helps identify potential flaws before running costly experiments.

Kelli Humbird, the study’s co-author and leader of the Cognitive Simulation Group at NIF, said: “This model has the ability to explicitly make choices for future experiments that maximize our probability of success each time.” She added that, as the facility can conduct only a handful of such tests each year, smarter predictions may save significant time and resources.

Although commercial fusion power remains a distant reality, LLNL’s AI breakthrough could help chart a faster path towards abundant, carbon-free energy.

Image Credit: Source