Human Terrain: Pentagon world-sim tool making good progress, say profs

The Register/Lewis Page

Computer suggests a savage bombing campaign, Mr President

Computer boffins in America report that their plans to create a lifelike virtual Matrix-style simulation of the entire world and human race are proceeding well.

Rather than a sinister yet totally implausible plan to farm people in tanks for electric power, the idea here is that America might try out various national-security initiatives – for instance invasions, bombing campaigns, supply of arms and training to regional governments/freedom-fighters/terrorists etc – in silico before doing so in the real world.

The latest news comes in an article written by Professor V S Subrahmanian and his colleague John Dickerson, two computing experts working on such initiatives for the Pentagon.

“Defense analysts can understand the repercussions of their proposed recommendations for policy options or military actions by interacting with a virtual world environment,” write the researchers.

“They can propose a policy option and walk skeptical commanders through a virtual world where the commander can literally ’see’ how things might play out. This process gives the commander a view of the most likely strengths and weaknesses of any particular course of action.”

The boffins say that several new technologies have been developed in recent years to make virtual worlds more accurate. These include stochastic opponent modeling agents (SOMA), artificial intelligence software that models the behaviour of people in different situations; “cultural islands”, which provide a representation of a real-world environment or terrain, populated with characters from that part of the world; and forecasting “engines” which use historical data to predict unknown outcomes.

In future, then, a president and his advisors pondering whether or not to invade a given country might run the plan through the virtual-world-ware to see the likeliest results. Presumably the computer would issue some prognosis along the lines of:

(1) Stable US-friendly democracy with booming economy: 4 per cent ± 3

(2) Long drawn out counterinsurgency quagmire followed by embarrassing withdrawal with tail between legs: 40 per cent ± 6

(3) As (2) but bloodshed eventually dies down somewhat, substantial numbers of US troops remain in country in perpetuity, end result might be declared a success by a later administration: 50 per cent ± 6

(4) Horribly embarrassing military defeat in open battle straight away: 6 per cent ± 5

Or similar. And then the president would do what he felt like anyway.

Naturally it seems certain that the stochastic artificial-intelligence people in the virtual world would be unaware that they were merely software subroutines in a modelling tool. They would think themselves and their possible suffering, misery, liberation etc to be real.

Indeed, groups among them might presumably decide – as part of their internal decision-making process – to develop a simulated world of their own.

Subrahmanian and Dickerson’s article can be read by subscribers to Science magazine here.