Maybe I'm a little biased, but I think we're headed toward a new revolution in science. My work in molecular dynamics has convinced me that we aren't too far a

Maybe I'm a little biased, but I think we're headed toward a new revolution in science. My work in molecular dynamics has convinced me that we aren't too far away from enough computing power to completely simulate many phenomena that are too complex to otherwise understand. Biology in particular is very complex, and I'm not entirely sure researchers can come up with models that are subtle and complete enough to accurately describe significant processes without becoming too mind-boggling for a human mind to grasp. On the other hand, if you can just simulate all these processes, drug development becomes a series of algorithms that optimize the desired effect while minimizing undesired effects. Some types of technology will require quantum computers. To simulate a system of N qubits using a classical computer requires 2^N states which quickly overpowers even the most advanced supercomputers. But if we can make even rudimentary quantum computers our simulation efforts will be significantly advanced (hey, we can factor the number 21, so it's a start!). To give you an idea of the current state of simulation research, we still require a lot of approximation models. What would you do if you wanted the most accurate simulation of a system you could get? Well, you would create one wavefunction of all the particles in the system and simulate its time evolution (Dirac equation). Of course, we can barely do this for simple atoms right now, so instead combinations of experimental data and basic QM simulations are used to get information about atomic/molecular interactions. For a large molecular system, these interactions are studied to create a new approximation model (like coarse-grain simulations). The problem is that each higher abstraction loses touch with experimental reality. Current biological simulations are really only "suggestive" instead of predictive. On the other hand, argon and nitrogen simulations match up almost exactly with experimental data. Again, as computing power develops, less abstractions will be necessary and simulations will become closer and closer to their experimental counterparts. To give an analogy for some of you computer graphics people, consider the huge number of "tricks" used in rasterization engines for video games to make the images look photorealistic. However, even a simple global illumination algorithm (bidirectional path tracing, photon mapping, Metropolis light transport) give incredibly realistic results but require far more time. It's a tradeoff between models (hard to come up with good ones) and computational capacity (but doesn't require much thinking).