Google has teamed up with nuclear fusion startup Tri Alpha Energy to bring artificial intelligence into the world of plasma research.
High-energy plasma physics is an experimental field with lots of uncharted areas, so to explore all that unknown territory, researchers from both companies created a machine-learning tool based on what’s called the Optometrist Algorithm.
The concept works like an eye exam — researchers are presented with two possible outcomes for the experiment, and asked to choose which results are better.
“While an optometrist asks a patient to choose between lens prescriptions based on clarity, our algorithm asks a human expert to choose between plasma settings based on experimental outcomes,” the authors explain in a research paper published by Nature Scientific Reports.
Plasma physics is different from optometry, in that there are more than 1,000 parameters to consider. But the principle is the same. The choices that result from the earlier settings for plasma shots are factored into the settings for the shots that follow.
The tool has proven successful. Researchers found that they were able to reduce the amount of energy loss by more than 50 percent and keep the plasma hot.
“Results like this might take years to solve without the power of advanced computation to rapidly scale our understanding of the complex properties of plasma,” Michl Binderbauer, Tri Alpha’s president and chief technology officer, said in a news release.
The experiments were run at Tri Alpha Energy’s facility in Foothill Ranch, Calif., using the company’s C-2U plasma confinement machine. Those experiments set new records for confinement time and stability, bringing Tri Alpha closer to a sustainable nuclear fusion reaction.
Since then, Tri Alpha has made the transition to a next-generation, $100 million plasma confinement device that’s been named Norman, in honor of Norman Rostoker, the company’s late co-founder. The Norman machine achieved its first plasma just last month.
Tri Alpha’s investors include Microsoft co-founder Paul Allen’s Vulcan Capital. The company is just one of several ventures aiming to harness nuclear fusion, the energy process that powers the sun. Commercial fusion power holds the promise of cheap, virtually limitless energy from a source that would be more environmentally friendly than nuclear fission or fossil fuels.
It’s a promise also being pursued by General Fusion, a Canadian fusion energy company with backing from Amazon billionaire Jeff Bezos.
Over the past couple of years, General Fusion has turned to the public to crowdsource ways to optimize plasma performance. Now it’s also getting help from big-data experts at Microsoft.
The collaboration, announced at the Microsoft Build conference in Seattle in May, will employ Microsoft’s cloud-based tools to sift through more than 100 terabytes worth of data collected by General Fusion during 150,000 plasma experiments.
Authors of the paper in Scientific Reports, “Achievement of Sustained Net Plasma Heating in a Fusion Experiment With the Optometrist Algorithm,” include E.A. Baltz, E. Trask, M. Binderbauer, M. Dikovsky, H. Gota, R. Mendoza, J.C. Platt and P.F. Riley.