Cheaper than printing it out: buy the paperback book.

While researching prediction and simulation machinery, I had a chance to visit the Jet Propulsion Lab in Pasadena, California, where a state-of-the-art battle simulation was under development. I came to JPL at the invitation of a computer science professor from UCLA who had been pushing the edge of computer power. Like many researchers pinched for support, this professor had to rely on military funding for his avant-garde theoretical experiments. He paid for his end of the bargain by picking a practical military problem to test his theories on.

His test-bed was to see how decentralized, massively parallel computing -- what I'm calling "swarm computing" -- could speed up a computer simulation of a tank battle, an application which only remotely interested him. On the other hand, I was earnestly interested to see a state-of-the art war game.

At the busy front desk of JPL, security clearance was straightforward. Considering that I visited the national research center while American troops were on red-alert along the Iraq border, the bouncers were fairly cordial. I signed some forms swearing my allegiance and citizenship, got a substantial badge to clip on, and was escorted with the professor to his cubbyhole office on an upper floor. In a small gray conference room, I met a long-haired graduate student who used the battle simulation mathematics as an excuse to pursue some far out notions on computational theories of the universe. Then I met the JPL honcho. He was nervously uncomfortable with my presence as a journalist.

Why? my professor friend asked him. The simulation system was not classified; the results were published in the open literature. The JPL honcho replied in so many words: "Well, umm, you see, there is this war going on, and quite inadvertently the generic scenario we have been dry-running for the last year or so -- a game we chose quite by accident, with no thought of prediction -- is being played out now for real. When we first tested this computer algorithm we had to pick some scenario, any scenario, to try out the simulation with. So we picked a simulated desert war with...Iraq and Kuwait. Now we are fighting this simulation. We are a bit on the spot here. It's a little sensitive. I'm sorry."

I did not get to see that war simulation. But about a year after the Gulf War's end, I discovered that JPL was not the only place that serendipitously preenacted that war. The U.S. Military Central Command in Florida ran a second and more useful simulation of a desert battle prior to the war. Cynics interpret the fact that the U.S. government had simulated the Kuwait war twice beforehand as a mark of its imperialist and conspiratorial desire to have that war. I find the predictive scenarios spooky, strange, and instructional rather than diabolical. I use this example to portray the potential power of prediction machinery.

There are about two dozen centers around the world that are playing war games where the U.S. is Blue -- the protagonist. Most of these places are small departments at military schools and training centers, such as the Wargaming Center at Maxwell Air Force Base in Alabama, the legendary Global Game room at the Naval War College in Newport, Rhode Island, or the classic "sand box" table set-ups at the Army's Combat Concepts Agency in Leavenworth, Kansas. Providing them technical support and know-how are academics and savants holed up in the numerous para-military think tanks peppering the beltway of Washington, D.C., or research alleys nested in the corridors of national laboratories like JPL and Lawrence Livermore Labs in California. The toy war simulators, of course, carry acronyms; TACWAR, JESS, RSAC, SAGA. A recent catalog of military software listed four hundred varieties of war games or other military models for sale right off the shelf.

The nerve center for any U.S. military operations is headquartered at Central Command, based in Florida. For its entire existence, Central Command, as an organ of the Pentagon, had been hawking one major scenario to Congress and the American people: Blue vs. Red -- the superpower game where the only worthy opponent was the Soviet Union. When General Norman Schwarzkopf came on the scene in the 1980s, he didn't buy this story. Schwarzkopf -- a thinking man's general -- put out a new perspective, worded in a way that's been quoted up and down the ranks: "The Soviet dog is not going to hunt." Schwarzkopf refocused his planners' attention on alternative scenarios. High on the list was a Mid-East desert war along the border of Iraq.

In early 1989, Gary Ware, an officer at Central Command, began modeling a war based on Schwarzkopf's hunches. Ware worked with a small cell of military futurists in compiling data to create a simulated desert war. The simulation was code-named Operation Internal Look.

Any simulation is only as good as the data it is based on, and Ware wanted Operation Internal Look based on reality as much as possible. That meant collecting a hundred thousand details about current forces in the Mid-East. Most of the work was horribly dull. The war simulation needed to know the number of vehicles in the Mid-East, stockpile strengths of food and fuel, killing power of weapons, climate conditions, and so on. Most of this minutiae was not readily available, even to the military. All bits were constantly in flux.

Once Ware's team worked out a formulation of an army's organization, the war gamers compiled optical laser disc maps of the entire Gulf area. The foundation of the simulated desert war -- the territory itself -- was transferred from the latest satellite digitized photos. When they finished, the war gamers had the countries of Kuwait and Saudi Arabia compressed onto a CD. They were now ready to feed all this data into TACWAR, the main computerized war-gaming simulator.

In early 1990 Ware began running a desert war on the virtual battlefield of Kuwait and Saudi Arabia. In July, in a conference room in north Florida, Gary Ware summarized the results of Operation Internal Look for his superiors. They reviewed a scenario based on Iraq invading Saudi Arabia, and the U.S./Saudi Arabia striking back. Ware's simulation forecast a fairly brief thirty-day war if anything this unlikely should occur.

Two weeks later, Saddam Hussein suddenly invaded Kuwait. At first, the upper echelons of the Pentagon had no idea they already owned a fully operational, data-saturated simulation of the war. Turn the key and it would run endless what-ifs of possible battles in that zone. When word of the prescient simulation surfaced, Ware came out smelling like roses. He admitted that "If we had to start from scratch at the time of the invasion we would have never caught up." In the future, standard army-issue preparedness may demand having a parallel universe of possible wars spinning in a box at the command center, ready to go.

Immediately after Saddam's initial invasion, the war gamers shifted Internal Look to running endless variations of the "real" scenario. They focused on a group of possibilities revolving around the variant: "What if Saddam keeps on coming right away?" It took Ware's computers about 15 minutes to run each iteration of the forecasted thirty-day war. By running those simulations in many directions the team quickly learned that airpower would be the decisive key in this war. Further refined iterations clearly showed the war gamers that if airpower was successful, the U.S. war would be successful.

Further, according to Ware's prediction machinery, if airpower could actually inflict the results assigned to it, U.S. ground forces would not sustain heavy losses. The top brass took this to mean that precise upfront airpower was the linchpin to low U.S. casualities. Gary Ware says, "Schwarzkopf was so adamant on maintaining the absolute minimum casualties of our forces that low casualities became the benchmark upon which all our analysis was done."

Predictive simulations, then, gave the command team the confidence that the U.S. could achieve success with minimum losses. This confidence led to the heavy air campaign. Says Ware, "The simulations definitely had an impact on our thinking [at Central Command]. Not that Schwarzkopf didn't have prior strong feelings, but the model gave us confidence that we could carry through the concepts."

As a prediction, Operation Internal Look got good marks. Despite some shifts in the initial balance of forces, the 30-day simulated air and ground campaign was pretty close to the real sequence, although the percentage of air and ground action was slightly different. The ground battle pretty much unfolded as forecasted. Like everyone outside the field, the simulators were surprised by how fast Schwarzkopf's end run around the front lines went. Says Ware, "I have to tell you, though, that we did not expect to get so far [on the battlefield] as we did in a hundred hours. As I recall, we forecasted a six-day land battle instead of a hundred-hour [four day] battle. The ground commanders had told us that they envisioned moving faster than the simulation indicated they would. So they moved exactly as fast as they predicted."

The war game prediction machinery figured greater resistance from the Iraqis than the Iraqis actually gave. That's because every combat simulation assumes that the enemy will employ all of its available systems. But Iraq never pushed hard at all. The war gamers cheekily joked that no model reflects the white flag as a weapons system.

The war moved so fast the simulationists never got around to the obvious next step in simulations: daily modeled forecasts of the battle in progress. Although the planners recorded every day's events as best as they could, and they could project out into the future from any moment, they felt "it didn't take a genius to figure what was going on after about the first 12 hours."