3 ms·

This type of result is why the buzz (at least in academic circles) about the value of "computational research" as a "third pillar" of research is wrong-headed.

This type of result is why the buzz (at least in academic circles) about the value of "computational research" as a "third pillar" of research is wrong-headed. Running computer simulations is a great way to compare theory to real-world data or to extrapolate out the implications of a particular theoretical idea, but it's still just another tool of theoretical research, and not a separate type of research in and of itself. The computational results are only as good as the model, which is all theory.

chm 13y ago
From my experience, there seems to be two theoretical/computational research scientist types: the one who doesn't really care about the numbers but focuses on equations and what they really tell us, and the one who only cares about numbers. The latter is dangerous.
- rotskoff 13y ago
  Indeed, in computational chemistry there seems to be two classes of theorists. The first group attempts to simulate a realistic system using high performance computing and achieve, typically, agreement with existing experiments. The second group tries to think through problems and hopes to provide explanations that don't need the brute force approach. The former model often lacks both creativity and quantitative accuracy. The latter might be impossible for some chaotic or complex systems. I have to agree that having "theorists" who simply are trying to compute some quantity to match existing data must do so very carefully.
greeneggs 13y ago
That doesn't seem fair. The article doesn't say, but in this case it sounds like the theories were based on previous experiments, not on computer simulations. And the deeper, underlying theories are correct. The problem is that we can't efficiently simulate large enough systems to discover phenomena like this computationally. But in the future, with better simulation algorithms, or possibly even large quantum computers, it should be possible to get results like this purely from simulations. (Of course, I qualified my argument by saying "in the future." So maybe you are right that computational research isn't yet a third pillar of research. But I do think it will become one, as our simulation tools incrementally increase in power.)

chm 13y ago

From my experience, there seems to be two theoretical/computational research scientist types: the one who doesn't really care about the numbers but focuses on equations and what they really tell us, and the one who only cares about numbers. The latter is dangerous.

rotskoff 13y ago
Indeed, in computational chemistry there seems to be two classes of theorists. The first group attempts to simulate a realistic system using high performance computing and achieve, typically, agreement with existing experiments. The second group tries to think through problems and hopes to provide explanations that don't need the brute force approach. The former model often lacks both creativity and quantitative accuracy. The latter might be impossible for some chaotic or complex systems. I have to agree that having "theorists" who simply are trying to compute some quantity to match existing data must do so very carefully.

greeneggs 13y ago

That doesn't seem fair. The article doesn't say, but in this case it sounds like the theories were based on previous experiments, not on computer simulations. And the deeper, underlying theories are correct. The problem is that we can't efficiently simulate large enough systems to discover phenomena like this computationally. But in the future, with better simulation algorithms, or possibly even large quantum computers, it should be possible to get results like this purely from simulations. (Of course, I qualified my argument by saying "in the future." So maybe you are right that computational research isn't yet a third pillar of research. But I do think it will become one, as our simulation tools incrementally increase in power.)