Already champion of Chess and Go (under the name of AlphaGo), the IA DeepMind also participates in much less symbolic work. Thus, the artificial intelligence created by the engineers at google was able to “imagine” – thanks to reinforcement learning technology – a simulation system able to modulate the behavior of the coils (modification of the magnetic field) in order to modify the configurations of the plasma inside the TCV (tokamak of the Swiss Plasma Center). In other words, DeepMind participates in current work on nuclear fusion!
The control of plasma in a tokamak is absolutely essential within the framework of the research which must lead us to current tests on the first “commercial” fusion reactor. The plasma configurations envisaged by DeepMind could be tested on the tokamak of the Swiss Plasma Center, and it turns out that the artificial intelligence had been right in its modeling. Experts in the field now say that in the face of the reported results, DeepMind’s participation will be a real asset in fusion research.
The analytical power of DeepMind could thus be used to optimize the performance of the most recent tokamaks, such as the ITER JET which is already beating records after records.