David Wolpert about new developments in computer science
What do you do if there are just too many projects you are involved in and too much research you could talk about? You do it like David Wolpert, narrow it down and split the talk in two parts. Accompanied by questions and a lively discussion, the Professor from Santa Fe Institute held a fascinating talk in front of a highly interested audience.
The first part of David’s talk dealt with the reduction of error of the Monte Carlo Algorithms by learning to control variates. On this, he is working together with Brendan Tracey.
Control variates are a powerful technique to reduce the error of Monte Carlo, but conventionally it is necessary to know a good function approximation a priori. Stacked Monte Carlo (StackMC) is a technique that overcomes this limitation by constructing a control variate from data samples. Possible extensions to StackMC for instance to importance sampling and multiple fitting functions, or discrete input spaces, were discussed.
In the second part David gave an overview on the thermodynamics of computation. Here he is working with Artemy Kolchinsky and Jeremy Owen.
David stated that 5%, tendency rising, of the current energy usage in the US goes to computation. Moreover, 50% of the lifetime budget of a modern high-performance computing center is to pay the energy bill. All of that energy ultimately results in heat. Indeed, Google has placed some of their data servers next to rivers – in order to use the water in the river to cool the servers, i.e., to remove the heat they generate. He mentioned that one of the major challenges facing current efforts to build exascale computers is how to keep them from melting. Clearly the thermodynamics of computation has major engineering consequences, which are becoming a huge challenge to humanity.
Fortunately, the last two decades have seen a revolution in nonequilibrium statistical physics. This has resulted in some extremely powerful tools for analyzing the thermodynamic properties and fundamental limits of far-from-equilibrium systems – like computers. As a result we are witnessing, according to David Wolpert, the genesis of a new field of science and engineering: a modern thermodynamics of computation.
David discussed these developments and highlighted some of the major extensions to theoretical computer science that are arising as thermodynamic considerations are combined with traditional concerns of space/time tradeoffs in computer science.