This AI Model Predicts Whether Fusion Power Experiments Will Work
1 min read
Summary
Researchers at Lawrence Livermore National Laboratory in California used an AI modelling technique to predict the likelihood of success of a nuclear fusion experiment.
They used so-called generative machine learning to accurately predict a 74% probability that US National Ignition Facility’s experiment would be successful.
The experiment, which used a variant of the more common doughnut-shaped tokamak design, involved firing powerful lasers at a millimetre-sized capsule of hydrogen fuel.
As a result, the hydrogen fused and generated more energy than was inputted to start the process.
The team suggested that the AI modelling technique could be used to make designing future experiments cheaper and more efficient.
