Ideally, you can simulate everything from first principles. That works well enough only for a rather limited set of systems.
More typical is that you need to do real world measurements/model (not LLM, but a model of what you simulate) validation before you can reasonably simulate.
Ideally, you can simulate everything from first principles. That works well enough only for a rather limited set of systems.
More typical is that you need to do real world measurements/model (not LLM, but a model of what you simulate) validation before you can reasonably simulate.
And then there is biology and psychology ...