DARPA to develop anti-Credit Crunch software
Ain't no more horses gettin' outa that stable
DARPA, the famed Pentagon boffinry bureau occasionally beset by marble-localisation issues, has struck again. This time, the agency seems to be seeking new mathematical tools which could - among other things - prevent any future collapses of global capitalism.
The new project comes to us under the name "Survivability of Interdependent Systems", and the maverick military madscience office will hold a workshop to explain the idea in June. However they offer the gist of it thus:
Currently, the metric for assessing system survivability uses a fixed and known reliability. As reliability is a physical propensity, it cannot be known. Also, propensity needs to be time indexed. Therefore, the existing method is limited.
This potential program would seek to develop and validate a new procedure for the realistic evaluation of the survivability of complex interdependent systems and networks. The development of this procedure would involve developing stochastic processes with novel properties and a stochastic calculus for the processes that will enable new capabilities for system risk assessment ... software for implementation would have to be created and validated.
Of course such stochastic calculus ware could be used for many things. You could use it to determine the likelihood of complicated multi-part gizmos working as desired - for instance missile defence interceptors dependent on boosters, sensors, communications, kill vehicles and so forth. You could check out complicated military logistics networks, or the fiendishly complex supplier pyramids which support at their apex things like stealth fighters.
But, as with a great many advanced technologies originally developed for military purposes, there would seem to be wider applicability. "Complex interdependent systems and networks" might as easily describe global financial markets, and DARPA's desired new calculus might - when assessing the "survivability" of balance sheets based on complex derivative bits of paper - be quite a handy thing to have.
One notes that the new DARPA push is administered by Dr Nozer Singpurwalla, whose research interests "include probability, statistics, and decision theory". Dr Singpurwalla is also a Professor of Statistics and Decision Sciences and the director of the Institute for Reliability and Risk Analysis.
DARPA will, as ever, be able to explain many possible military uses for this programme in discussions with their uniformed masters. But, as with IP networking - a DARPA project which became a runaway success in distributing smut before the military ever got much use out of it - the agency seems to be interested in answering questions first and shooting afterwards. ®
Sponsored: Hyper-scale data management