> Science needs models based on mechanistic understanding of the underlying phenomena. A model that merely predicts is useful for engineers, not scientists
I'm not sure I agree. I'm aware of quite a bit of supercomputing time that is spent doing lattice QCD calculations (which apparently some scientists find useful), and though I'm no quantum physicist I'm pretty sure there is not much of a "mechanistic understanding" in QCD. I think your claim also doesn't apply to a lot of social science - psychology has a lot of functional models, but I don't think there are many mechanisms described.
I'll also state that modern science that doesn't require any engineering is pretty rare nowadays, so if a predictive model helps engineers that can then help scientists, the model has been helpful to scientists.
Ohm's law existed long before there was a mechanistic description behind it, and though it is mostly used for "engineering," I feel confident that a lot of scientists in the 19th century found it useful.
I work with the DOE and was at ORNL before Summit was released (I got to play on Summit-dev). When making these models there is A LOT of exploration happening. There's a whole class of visualization techniques called "in situ" that visualize data as it comes off the press (memory is then dumped because there's neither enough storage space nor can we write to disk fast enough). I'll tell you that there will be a lot of restarting those simulations because the scientists need to explore the data as it is going on. Going in the wrong direction? Made a small mistake that causes cells explode? Realize you're not looking in the right region of interest? You restart the sim (thank god for restart files, right?). Exploration is one of the most important things in research and it is getting more and more difficult. I believe this is what the gp is after. Having these understandings helps you explore the data better. Creating these tools is hard work and takes a lot of collaboration too.
I guess "Mechanistic understanding" was meant to contrast against machine learning, not quantum mechanics. To elaborate, machine learning means fitting of a bunch of data by a given model. In science (eg lattice QCD) one often tries to theoretically (or computationally) explore regimes where data is not yet available. As a (former?) theoretical physicist, I am more than happy to admit that this is not immediately useful, though it will hopefully become useful in the long run.
I'm not sure I agree. I'm aware of quite a bit of supercomputing time that is spent doing lattice QCD calculations (which apparently some scientists find useful), and though I'm no quantum physicist I'm pretty sure there is not much of a "mechanistic understanding" in QCD. I think your claim also doesn't apply to a lot of social science - psychology has a lot of functional models, but I don't think there are many mechanisms described.
I'll also state that modern science that doesn't require any engineering is pretty rare nowadays, so if a predictive model helps engineers that can then help scientists, the model has been helpful to scientists.
Ohm's law existed long before there was a mechanistic description behind it, and though it is mostly used for "engineering," I feel confident that a lot of scientists in the 19th century found it useful.
From https://www.olcf.ornl.gov/leadership-science/physics/:
"New Frontiers for Material Modeling via Machine Learning Techniques" - 40,000 hours allocated on Summit
"Large scale deep neural network optimization for neutrino physics" - 58,000,000 hours allocated on Summit.
Supercomputers typically do not allocate 58 million hours to things which are not useful.