Why you actually want Skynet (and maybe John Connor)
Remember that movie Stealth? It's the one where Jamie Foxx, Jessica Biel, and the other sexy pilots are forced to fly with a plane that has a computer pilot and, turns out, computer pilots are bad because lightning can strike them and drive them crazy and then they murder all the people?
No? Well certainly you've seen or heard of the Terminator movies. You know, the ones where plucky humans and their hacked robot bodybuilder are forced to fight other robots in order to prevent a future apocalypse ordered by military AI?
They're great films, but they imply that any future where computers are controlling the weapons of war is dystopian AF. In reality, we live in a world that has walls, and those walls are guarded by men with guns. It would be much better if the U.S. could guard those walls with robots with guns controlled by men.
This would provide two advantages. First, if the guards on the walls are robots — not fleshy humans — then people shooting at the walls can only destroy hardware, not kill men and women. But perhaps the bigger factor is that artificial intelligence is enabling robots to become better at some jobs than their human controllers.
In 2016, we wrote about a University of Cincinnati project where a retired colonel and fighter pilot were advising on an artificial intelligence project. The AI was just supposed to control a Red Team well enough to allow fighter pilots to get better experience in simulations, but tweaks to the system made it start winning. And then it beat the general. And then it beat the general even when he was directly piloting one of the aircraft.
ALPHA AI was developed with a team from University of Cincinnati. It can run on a cheap computer but has defeated skilled fighter pilots in simulations.
(Journal of Defense Management)
The adviser, Retired Air Force Col. Gene "Geno" Lee, told the researchers that AI was "the most aggressive, responsive, dynamic and credible AI [he's] seen-to-date."
And that AI, known as ALPHA, ran on a Raspberry Pi computer that can be picked up for under $50. So, for $50, plus probably $30,000 or so in retrofitting charged by manufacturers, we could make 4th generation planes do their jobs perfectly while flying in support of a fifth-generation, human pilot who's calling the shots.
Stealth's artificial intelligence can pilot fighter jets, but, for some reason, needs a special sensor that looks like a robotic eye instead of just using, you know, its radar or even just normal cameras.
This may sound familiar to people for one or both of two reasons. First, the Air Force is actively pursuing this as the wingman concept. But second, Skynet in the Terminator movies got its start piloting stealth bombers where it achieved a "perfect operational record," according to Schwarzenegger's character.
Is this so bad? I mean, sure, we should stop short of handing strategic control of the nuclear weapons to Skynet, but that was never a realistic plot premise. Remember, even during the height of the Cold War, it was rare for launch approval for nuclear weapons to be handed down past the president. If we don't trust generals to make nuclear decisions without the president approving it, why would we ever let a computer have full control?
So, if we develop Skynet and don't give it access to the nukes — if we create safe AI — we're left with a completely new version of warfare where we don't have to risk our own troops at nearly the same level as we currently do. Doesn't sound so horrible now, does it?
And, if the other side gets AI, that's still better for humanity as a whole. Remember when the RAND Corporation anticipated that, by 2025, war with China would be bloody and unwinnable? No? We're the only people who actually read RAND reports? Alright, then.
Here's the thing: World War I was so horrible because it was a nearly unwinnable war for both sides. Once nations committed to the conflict, they poured blood and treasure into a never-ending pit of carnage. Millions died and little was gained for anybody.
AI wouldn't make unwinnable wars winnable — at least not if both sides have it — but it could make them much less bloody, which is still a step in the right direction.
You know what would be even better than sending F-35s up with human pilots to detect enemy air defenses and suppress them? Sending them up with a bunch of fighters that are basically robots with AI. So, if they do get in a fight, they don't need to take the hits.
(U.S. Air National Guard Master Sgt. Joshua C. Allmaras)
So, what about poor John Connor, an excellent small-team leader? What's he going to do when he isn't allowed to kill Skynet but, instead, Skynet is controlling most of the planes and tanks and ships? Well, he'll lead small teams or infantry units on the ground while A Few Good Men's Col. Jessup gives the marching orders. AI can't replace all decision-making at the front, and calm heads under fire will be needed to authorize strikes and targets.
So, yes, we all secretly want Skynet on the wall, even more so than we want Col. Jessup up there. But we also need John Connor, as long as we can keep Jessup, Connor, and Skynet from murdering one another.
- 7 common movie tropes that rely on military stupidity ›
- The U.S. military kindly asks you to trust its death robots - We Are ... ›
- Artificial Intelligence in the U.S. - Backing the Use of Killer Robots ›
- The US military wants to teach AI some basic common sense - MIT ... ›
- Artificial Intelligence and the Military | RAND ›