As the world waits for construction of the largest fusion reactor yet, called ITER, smaller reactors with similar designs are still running. These reactors, called tokamaks, help us test both hardware and software. The hardware testing helps us refine things like the materials used for container walls or the shape and location of control magnets.

But arguably, the software is the most important. To enable fusion, the control software of a tokamak has to monitor the state of the plasma it contains and respond to any changes by making real-time adjustments to the system's magnets. Failure to do so can result in anything from a drop in energy (which leads to the failure of any fusion) to seeing the plasma spill out of containment (and scorch the walls of the container).

Getting that control software right requires a detailed understanding of both the control magnets and the plasma the magnets manipulate, or, it would be more accurate to say, getting that control software right has required. Because today, Google's DeepMind AI team is announcing that its software has been successfully trained to control a tokamak.

This is NOT allowed according to GDPR.

Researchers at the University of Washington, including Kyle Morgan and Chris Hansen, recently published a study detailing a method that uses machine learning to predict the behavior of a plasma. Their model, which uses a statistical technique called regression, essentially throws out scenarios that lead to nonsensical results, enabling it to use less data, less computational power, and less time. Hansen says that although the model in the study doesn't work quickly enough to use during an experiment, he thinks that it eventually could. The researchers published another recent study that used a single GPU to control a fusion experiment that had previously required several computers. This kind of powerful system, Hansen says, could eventually be used to run the model quickly enough that it would be useful during an experiment.

Today, the most powerful artificial intelligence systems employ a type of machine learning called deep learning. Their algorithms learn by processing massive amounts of data through hidden layers of interconnected nodes, referred to as deep neural networks. As their name suggests, deep neural networks were inspired by the real neural networks in the brain, with the nodes modeled after real neurons — or, at least, after what neuroscientists knew about neurons back in the 1950s, when an influential neuron model called the perceptron was born. Since then, our understanding of the computational complexity of single neurons has dramatically expanded, so biological neurons are known to be more complex than artificial ones. But by how much?

To find out, David Beniaguev, Idan Segev and Michael London, all at the Hebrew University of Jerusalem, trained an artificial deep neural network to mimic the computations of a simulated biological neuron. They showed that a deep neural network requires between five and eight layers of interconnected “neurons” to represent the complexity of one single biological neuron.

Love the enthusiasm here, but how do you see MLC help advance fusion tech?

Machine learning, a technique used in the artificial intelligence (AI) software behind self-driving cars and digital assistants, now enables scientists to address key challenges to harvesting on Earth the fusion energy that powers the sun and stars. The technique recently empowered physicist Dan Boyer of the U.S. Department of Energy's (DOE) Princeton Plasma Physics Laboratory (PPPL) to develop fast and accurate predictions for advancing control of experiments in the National Spherical Torus Experiment-Upgrade (NSTX-U)—the flagship fusion facility at PPPL that is currently under repair.

One side-thought: how much RAM is required for a MLC GPU task?

Brand new install of Ubuntu 1804, latest Boinc - 7.16.16 on an FX6300

Yes this is very strange. Last week I had 48 GPU task on this computer, all ended successfully. Nothing has been changed on this machine since then.

@Jim1348: A write cache is an interesting idea but it's not obvious to me that it actually does reduce writes, not just delay them. I'll need to find more information on how it works in detail.
I'd thought of running BOINC from a tmpfs perhaps, but then I'd have to find a way to make the data persist across restarts.

If you ever wondered why the GPU app took so long, I hope now you begin to see the time and effort that went into it beyond a simple "turn on the flag and recompile". .

Do you know how much faster the gpu task will complete when you write to RAM versus SSD? Thanks.