Simulating everything.

A new system called Sentient World Simulation is being pitched to the US Military by Purdue University and Simulex. Curiously naive advertising, and havent we heard all this before, eg the Earth Simulator and MNE?

The whole thing is the brainchild of a Purdue Professor, and Simulex CEO, called Chaturvedi, but its writeup from Purdue is downright naive:

“Sentient World Simulation (SWS) will be a continuously running, continually updated mirror model of the real world that can be used to predict and evaluate future events and courses of action. SWS will react to actual events that occur anywhere in the world and in corporate newly sensed data from the real world. SWS will provides the ability to examine the likely progression of the status-quo as well as explore any “what if” scenarios.

SWS will consist of a synthetic environment that mirrors the real world in all it key aspects-Political, Military, Economic, Social, Information, and Infrastructure. Within the synthetic world are models of individuals, organizations, institutions, infrastructure, and geographies that represent the traits and mimic the behaviors of their corresponding real world counterparts. As the models influence each other and the shared synthetic environment, behaviors and trends emerge in the synthetic world as they do in the real world. Analysis can be performed on the trends in the synthetic world to validate alternate worldviews. The key computational components of SWS are: Virtual Execution Environment (VEE), SWS Virtual Model Repository (SVMR), and Dynamic Resolution Manager (DRM). Virtual Execution Environment (VEE) will enable users to configure a unique execution environment for their own requirements. Many such environments can run simultaneously. SWS Virtual Model Repository (SVMR) is a warehouse consisting of validated simulations, models, and model-components – at varying levels of resolution, data, and inter-model coupling relationships. These essential elements can be selected, sustained, and appropriately combined to build any desired simulation through conformance to model standards.”

It adds: “Simulations can be presented at varying levels of fidelity.” Well, yes, Anything from low to zero, Id have thought. No amount of Virtual Repositories and Dynamic Resolution Managers can conceal the fact that each model has to interact with the others. You cant see them as neat black boxes. If I suddenly decide to have my social model interact with my infrastructure model, how, honestly, do I do this? What is the connection between them? Is there an equation, a model, of the process? Or do I just check them out of the repository and set them running?

Also, the idea of different levels of fidelity reminds me of the famous puzzle, how long is the coastline of Britain? The answer, of course, depends entirely on the level of detail at which you measure it.

“Another example of scale-dependence is the length-measurements of borders between countries. The problem of measuring them has been known for a long time and in most cases the deviations do no harm: e.g. the circumstance that the length of the border between Spain and Portugal is given differently by the officials of these two countries has no consequences. The difference results from the fact that the official maps of Spain have a bigger scale than those of Portugal. Here we find the same phenomenon as with the coastline above – the maps of Portugal show more edges and corners, which mean that the border is longer than on the maps of Spain. The length given in the Spanish encyclopedia is 987km and that in the Portuguese one is 1.214 km.”

Most coverage of this project is on conspiracy sites and blogs, and over-dramatic.

According to The Register, which seems to have broken the story, “Chaturvedi is now pitching SWS to DARPA and discussing it with officials at the US Department of Homeland Security, where he said the idea has been well received, despite the thorny privacy issues for US citizens…. In fact, Homeland Security and the Defense Department are already using SEAS to simulate crises on the US mainland. The Joint Innovation and Experimentation Directorate of the US Joint Forces Command (JFCOM-J9) in April began working with Homeland Security and multinational forces over “Noble Resolve 07“, a homeland defense experiment. SEAS (as will SWS) provides figures for specific economic sectors, and helps military, intel and marketing people visualize their global connections. Users can vary export and import figures for manufactured goods, for example, to gauge the potential impacts on other sectors. … JFCOM-J9 completed another test of SEAS last year. Called Urban Resolve, the experiment projected warfare scenarios for Baghdad in 2015, eight years from now….. Jim Blank, modelling and simulation division chief at JFCOM-J9, declined to discuss the specific routines military commanders are running in the Iraq and Afghanistan computer models. He did say SEAS might help officers determine where to position snipers in a city square, or to envision scenarios that might emerge from widespread civil unrest.SEAS helps commanders consider the multitude of variables and outcomes possible in urban warfare, said Blank. “Future wars will be asymetric in nature. They will be more non-kinetic, with the center of gravity being a population.”The Iraq and Afghanistan computer models are the most highly developed and complex of the 62 available to JFCOM-J9. Each has about five million individual nodes representing things such as hospitals, mosques, pipelines, and people.The other SEAS models are far less detailed, encompassing only a few thousand nodes altogether, Blank said. Feeding a whole-Earth simulation will be a colossal challenge. “(SWS) is a hungry beast,” Blank said. “A lot of data will be required to make this thing even credible.”

Exactly. Its a big step from where to position snipers to modelling the whole world, or even just the bits youre fighting in.

Simulexs website says: “In just 8 years, Simulex, Inc. has established itself as a visionary in the technology industry. We have doubled in size, now employing a team of over thirty experienced, multi-disciplined programmers and project managers. Our portfolio continues to grow with the support of the U.S. Department of Defense and several Fortune 500 companies.” Their clients include the US Military, Lockheed Martin (which is pretty much the same as the US Military) and Eli Lilly.

Incidentally Urban Resolve is a human in the loop (HITL) experiment. (Thank goodness they didnt call it a Human In The Loop Experimental Reaction experiment….) “which put human decision-makers into a scenario generated by supercomputers.” HITL1 took place in August 2006 and was blogged (live) by a USJFFCom public affairs man. This is worth reading.

However it is almost entirely the same as what came out of MNE4. Are these linked in any way, or are the simulations of everything so full of data that they dont notice each other?

Leave a Reply

Your email address will not be published. Required fields are marked *