In the autumn trimester of 2019, I started a seminar series in Oxford on the application of machine learning and physics. Miha Zgubic helped a lot in bringing it together and Philipp Winsdischhofer will do so for next year. The seminar was very well attended, and we really learned fantastic things about how ML has become a new scientific method for physics.
Often machine learning is described as the answer to the big data challenge, but I think it goes well beyond that. In particular, it has shown to be a viable supplement or alternative to simulation, and possible it might lead to major breakthroughs in hard computational problems surrounding turbulence for example, where little progress has been made for decades.
I gave the opening seminar (slides here), giving a survey of some of the interactions, and some of the relationships with developments in computer science. The entire series can be found on our YouTube channel.
