This article talks about machine learning/AI helping to make sense of the data from fusion experiments, and maybe eventually designing and even running the experiments. It’s interesting to think about computers speeding up progress by being able to design and run experiments orders of magnitude faster than humans could. If it works well, they could fail a million or a billion times in short order and there would still be value in a single success. You could also imagine a computer going down a rabbit hole and coming up with a result that humans are not able to explain or replicate, and then you would have to think about what to do with that result. There’s also the question of whether a computer can ever truly “understand” a system, but I guess constructed a model, whether mental or mathematical, testing it against observation, tweaking it, and then testing it against more observation is basically how we do it.