It all began to crumble for the brave defenders of the United States when a nuclear warhead launched from an undetected submarine obliterated several West Coast cities. But things were going downhill even before that, with our Middle East assets taking a beating from both conventional and nuclear forces. At least we gave almost as good as we got. In the end, the so-called victor inherited a dying world filled with ash and little else other than stone age technology.
Thankfully, all of this was just a simulation of a war that nobody wants to fight.
The aforementioned battlefield was a nuclear war game called ICBM, which stands for intercontinental ballistic missile. It was created by well-known wargame developer Slitherine and is due to be released in November on the Steam gaming platform for the PC. But a few lucky members of the Pentagon’s wargaming team got an early look in a live battle challenge against the game’s developers.
Defending the United States was four members of the Pentagon’s wargaming team including Mitch L. Reed, Philip Bolger, Sebastian J. Bae and Philip “Doc” Wohlrab. All have experience with military wargames. For example, Reed is the lead designer for the office that develops, conducts and executes all of the Air Force’s Title 10 wargames. Bolger wrote a novel called “The Devil’s Gunman” and works as a defense wargamer in the Pentagon. Bae teaches students at Georgetown University how to research and design educational wargames. And Wohlrab works with a team called The Foxes that fights in wargames for both the Air Force and Space Force.
So the home team was well-represented. The entire event was live streamed to the world, and is now available to watch in its entirety.
They played three games. Although the Pentagon lost twice, they did win one bout. Given that they were playing the game’s developers, it was clear that Team Slitherine had home field advantage. As the Pentagon’s wargamers learned the nuances and strategy of the simulation, their experience in wargaming slowly began to give them the edge. It’s interesting to note that even when winning, the victor would inherit a very broken world with billions killed across millions of miles of ruined cities and irradiated landscapes. So I guess the lesson from the classic 1983 “War Games” movie still holds true today, that the only way to truly win a game of nuclear war is not to play.
DARPA’s Game Breaking Program
The Pentagon team had to compete directly in the ICBM challenge using nothing but their human ingenuity and skill. But if they would have waited a couple of months, they might have been able to deploy a secret new military weapon. The Defense Advanced Research Project Agency just launched its Gamebreaker program in an effort to create an artificial intelligence that can overwhelm any strategy video game’s internal logic, sow chaos on the battlefield and ultimately win every skirmish, battle and war.
Having a magic tool that can destabilize any game and then win sounds pretty cool, though the military is, of course, looking for real world applications. The Department of Defense wants to first upset their own internal wargame simulations, and then apply those lessons to the real battlefield. This would be achieved either through gaining an advantage over an adversary or by countering someone trying to do that to us.
“If we can figure out a generic method to assess and then manipulate balance in commercial video games, my hope is that we might then apply those AI algorithms to create imbalance in DOD simulated war games used to train warfighters for real-world battle,” said Lt. Col. Dan “Animal” Javorsek, the Gamebreaker program manager in DARPA’s Strategic Technology Office.
DARPA selected nine teams to try and meet the Gamebreaker Challenge. Instead of competing directly against one another as in normal DARPA challenges, teams instead picked two different commercial strategy war games to try and destabilize. The idea is that they program their AI to break the first game, and then see if it can also break the second without modifications. Teams were even able to choose which two games they were going to use as part of their experiment.
“We received many incredibly rich proposals from a diverse base of universities and small and large companies,” Javorsek said. “Since the teams aren’t competing against each other, we’re looking forward to a collaborative effort as performers develop their AIs across an array of games.”
There are several video game developers participating as part of the Gamebreaker teams, so their skill in making games should help with destabilizing and breaking them. For example, Northrop Grumman is partnering with well-known wargame maker Matrix Games for their team, while Radiance Technologies is allied with strategy game developer BreakAway Games. Other teams come from both tech companies and universities.
An Impossible Challenge?
I applaud the DARPA efforts on this, though I question whether developing a universal AI that can destabilize and win wargames across the board is possible. Having played computer wargames for most of my life, I can testify that games like that don’t normally have exploitable flaws, and certainly not anything comprehensive that runs across multiple titles. A game’s commercial success is largely dependent on its ability to provide a balanced and fair playing field, and most of them already have comprehensive AIs that are good enough to challenge human players.
There might be a vulnerability or a cheat that could be discovered within a single game, though here too I think that such a thing would be discovered by the gaming community very quickly after a game’s release and patched by the developer. Finding some sort of a golden skeleton key AI that unlocks victory in every single game won’t happen. And even if it could, such a destabilizing strategy would likely be spotted and countered by humans.
For example, let’s say the teams tell their AI to employ a so-called Zerg Rush in StarCraft II, which is one of the games being used in the competition. The idea behind that infamous rush strategy is that the player using the Zerg race spams their opponent with low-level troops in a suicidal rush designed to significantly set their opponent back before they can get established. It exploits one aspect of the game, namely the ability for the Zerg to create lots of cheap, low-level troops and blitz them into an opponent’s base very early in the battle.
When it was originally deployed as a strategy for StarCraft II, gamers who used it in eSports competitions did very well. And players at home could often crush the AI when using it too. But everybody knows about it today. If you suspect that your opponent is going to deploy a Zerg Rush, it’s easily countered, and they are effortlessly crushed afterward since they spent their precious early game time and resources on a fruitless charge.
Very few other strategy games allow anything like Zerg rushes to happen anymore. In fact, each game is so different that I don’t think developing a universal strategy or exploit to win in all cases is possible, much less an AI that could do it for you. Classic games like Age of Empires III play and feel different from modern sci-fi titles like Iron Harvest, and different still from small unit tactics games like BattleTech, to say nothing of titles played on a global scale like ICBM.
I can’t wait to see what the Gamebreaker program comes up with when it releases its results early next year, and I hope that DARPA can find some correlations and ideas to help their own wargames or even to improve their real life strategies. But as an armchair general, I have my doubts. Still, I wish the nine teams in the challenge the best of luck, and look forward to seeing their creations on the virtual battlefield one day.
John Breeden II is an award-winning journalist and reviewer with over 20 years of experience covering technology. He is the CEO of the Tech Writers Bureau, a group that creates technological thought leadership content for organizations of all sizes. Twitter: @LabGuys
source: NextGov
