Mr Sato's argument is that the Earth Simulator "has changed the concept of simulation. It has made it possible to deal with the whole [global] system at once because of its memory and speed [capacities].”
Well, firstly allow for a little (justified) pride and exaggeration. It only (!) has 62 super-computers, capable of 35.86 Teraflops, with 'main memory' (?RAM) of 10TB and 2.5 petabytes of storage. (Google is supposed to have 4 petabytes of RAM.) One hour of uncompressed HDV video uses 11.5 terabytes of data. There are 2 billion people in the world, and Ray Kurzweill apparently says that we each have 1.25 terabytes of functional memory. In other words, the Earth Simulator has the equivalent storage to two thousand humans, or a few hundred high quality movies. )
This is not to belittle it. It's a great tool, but life is very complex.
Secondly, the sad fact is that people don't integrate simulations. For instance, how many predictive simulations of the spread of avian flu have there been? Would it not make more sense for the nations of the world to commission just one? (No, even leaving aside political considerations, because there are so many variables and different ways of modelling the issues. No because this sort of simulation can only be a pointer, not a prediction: we want informed debate, not imposed orthodoxy. Remember the 'butterfly effect'? Or Wigner's radical scepticism?)
So we continue to have little islands of simulation, even if those islands are unimaginably large by the standards of 10 years ago.
Einstein spent the last years of his life fruitlessly searching for a theory that would explain everything, and I suspect the idea of a 'world simulation' may prove to be equally elusive. At the same time, our islands of simulation aren't yet as big as they could be, and aren't as connected as they should be.