Physicist Nobel Laureate Philip Anderson on computers and physics:

The prestige attached to computers and their erudite gimmicks impresses almost everyone, but especially the simulators. They often believe they have proved that a system--like the little crystal of solid helium--can't possibly behave the way experiments show, therefore there's something dubious about the experiments, and not the simulations. Of course, to the casual observer computer simulations are far more impressive than old-fashioned logic and common sense. But we must remember that a simulation, even if correct, can't really prove anything. Computers will always have limits of error in trying to model the world. In the end logic and pure science, independent of the computer, still get us closest to nature...

(That last part should be qualified: simulations can prove things, but those things may have no resemblance to what goes on in the physical world.) In this short piece, Anderson argues that in physics, the tremendous power of computers is more helpful in organizing experimental data than in guiding theorists. Simulations of complex systems that try to account for the interactions of every single atom involved are out of reach of our current computer power, and the shortcuts that simulators make to get around this limitation end up severely biasing the results of the simulation. Thus theorists, Anderson argues, should not be using simulations as a substitute for "logic and pure science". Similar issues crop up in biology. We don't have the computer power to model everything using the principles of quantum mechanics, but good models, at the right level of abstraction, and made with a specific goal in mind, can guide our thinking, help us generate new hypotheses that we might not have come up with using just verbal reasoning, and can even make predictions that we might want to test experimentally. This kind of modeling is in contrast to a type of model I don't like: a virtual cell, that ends up being just a black box containing every last bit of information we know about a cell, a black box which we can shake like a Magic 8 Ball to get out some answer, some prediction that we may be able to test experimentally, but which is not really helpful for better understanding why a cell behaves the way it does. In physics and biology, simulation is not a substitute for hard thinking about the particular system you are interested in.