Institute for the Theory of Computing

As Lance and others blogged, Simons Foundation has called for proposals for a new Institute for the Theory of Computing. This is great and we should be grateful to the foundation.

What got my attention:

Modes of Operation: To help attract the best-established researchers and the strongest postdocs, the Institute must provide excellent working conditions for collaboration among its residents, excellent scientific leadership to determine the planned activities, and an excellent director with strong administrative support to manage it. To have full impact on the field, the Institute must host a frequently changing group of computer scientists as well as mathematicians and scientists from other fields. Suitable models may be found in the Kavli Institute for Theoretical Physics in Santa Barbara and in the mathematics institutes such as the Mathematical Sciences Research Institute in Berkeley or the Institute for Mathematics and its Applications in Minnesota. The Institute plan may call for a small core of long-term members, perhaps with a structure similar to that of the Kavli Institute for Theoretical Physics in Santa Barbara; or the plan may call for a membership that changes entirely every year, perhaps with a structure similar to that of the Mathematical Sciences Research Institute in Berkeley.

Question: are these efficient models for 2010's and beyond? DIMACS, IMA, MSRI and many other institutes were set up several decades ago, and operate in a mode quite similar to each other (when funding allows). Now my new premises are:

• it is difficult to get senior researchers to travel far for sabbatical at various academic Institutes, and
  
• given the Internet, we can very effectively video conference, network and collaborate at far distances.
  

Given these two premises, shouldn't new institutes go where researchers are (instead of other way around) and be a totally distributed center across geographic US regions? There could be a lightweight, but intellectually strong center running say very high profile series of talks, brainstorming meetings, a periodic physical meeting of all distributed centers, etc., providing leadership; any group of researchers across the country can propose and run a special activity of 2 or 3 years (say when some of them are on sabbatical). We need to balance the obvious benefits of being in one place for collaboration and the convenience of being away for collaboration. :)

Disclaimer: In the spirit of the fast Internet world, I am rushing out this post before debating pros and cons, eg., didn't discount for the angst of leaving a tried model for the unknown.