Earlier this year more than 3,000 Google employees signed a letter to chief executive Sundar Pichai demanding the company halt work on the Defense Department’s Project Maven, which applies algorithms to warfare. The disgruntled employees also wanted their boss to pledge “that neither Google nor its contractors will ever build warfare technology.” In June the company announced it would not renew its Project Maven contract. This is incredibly shortsighted and will increase the likelihood of war and civilian deaths.
Past warfare was described primarily by tonnage and throw weights, because precision was almost nonexistent. But ever since humans started dropping bombs out of airplanes, they’ve been aiming for more precision.
On June 15, 1944, a squadron of 75 American Superfortress B-29s left China to destroy the Imperial Iron and Steel Works in Yawata, Japan. The site manufactured about a quarter of Japanese steel at the time. The 47 bombers that made it to Yawata dropped more than 365 bombs. One accidentally destroyed a power house more than a kilometer away from the complex. The rest missed.
It wasn’t for lack of trying. In 1943, behavioral scientist B.F. Skinner demonstrated new guidance technology to track simulated Japanese destroyers. He then revealed that inside the nose cone of the bomb were three pigeons trained to peck away at silhouettes of Japanese warships. But real technology advances. Wartime news reports claimed the highly complex Norden bombsight could hit a pickle barrel from 20,000 feet above. But in 1944 bombardiers recorded that “75% of Norden bombsights fell short of specifications,” missing by more than 300 feet.
For the rest of the war, the city of Yawata was firebombed in an unsuccessful campaign to destroy the iron and steel works. Notably, it was a target on Aug. 9, 1945, three days after Little Boy was dropped on Hiroshima. A B-29 carrying the atomic bomb Fat Man made runs over Yawata, but thick smoke from the ground made targeting impossible, and the bombers headed to the next target on the list, the Mitsubishi Steel and Arms Works in Nagasaki. The bombardier, Capt. Kermit Beahan, used a Norden bombsight to target the factory. He wasn’t even close—off by almost two miles. Horseshoes and hand grenades!
During the Cold War, the U.S. and Russia often negotiated over throw weight, the payload that a missile carries—power over precision. In hot wars, conventional bombing continued. The U.S. dropped seven million tons of bombs during the Vietnam War, 10 times as much as during the Korean War and twice as much as the Allies dropped in World War II. There are no good statistics, but most think more than 90% of bombs missed across those wars.
During the first Gulf War, the paradigm changed. Tomahawk cruise missiles would read the terrain preloaded from satellite imagery and accurately find their targets. Even aerial bombing improved. Some 17,000 precision bombs, mostly laser-guided, almost always hit their targets. But 210,000 conventional bombs still missed by an average of 300 feet.
By the late 1990s, the Joint Direct Attack Munition kit was developed. It was bolted onto conventional 500-pound bombs, which updated their position via GPS as they dropped. They almost always landed within 10 feet of their target. It was first used in Kosovo in 1999 and perfected in the Iraq War.
There is so much more to do. The Israelis successfully deploy the Iron Dome missile interceptor. They claim it is 90% accurate, but many dispute this. Last week a new interceptor, known as David’s Sling, co-developed with Raytheon ,was launched against two Syrian SS-21 Tochka missiles headed toward Israel. Neither missile interceptor hit its target.
Precise weapons set expectations that war can be fought without civilian casualties. Which brings us back to Google. The U.S. has used drones successfully against al Qaeda and Islamic State. Drone pilots identify and eliminate terrorists from thousands of miles away. Collateral damage and civilian deaths still happen, but they’re often the result of faulty intelligence, not the bombing technology.
Project Maven, or the Algorithmic Warfare Cross-Function Team, was set up by the Defense Department last year. Working with private industry, the group applies machine learning and artificial intelligence to read drone footage and “autonomously extract objects of interest from moving and still imagery.” This probably doesn’t include facial recognition. It eventually will.
I don’t think we’ll ever get to fully autonomous weapons, the fever dreams of Terminator’s Cyberdyne Systems and the Campaign to Stop Killer Robots. Treaties will be signed to outlaw them.
Still, it’s naive to think war is obsolete—bad actors will always be around. But technology will soon surgically zap bad guys. Project Maven gets us closer. Which is why Google engineers’ refusal to develop precision technology means more civilians will die or the military will be reluctant to use force. That’s wrong. They should Google Si vis pacem, para bellum.