Robot with a Bomb

A few days ago, I had an interesting conversation with a man who chairs a working group that evaluates the effectiveness and reliability of autonomous systems. The gentleman in question is a semi-retired physics professor with an IQ well into the genius range, as well as more than four decades of practical experience as a technical consultant to government and the defense industry.  He also happens to be blessed with the kind of hardnosed common sense that can sometimes be lacking in members of the academic elite.  In other words, this man is just about the smartest person I’ve ever met.

So I was a bit surprised by something that came up when our discussion wandered to the topic of arming autonomous machines. The learned gentleman showed a solid grasp of the technical challenges involved in creating self-governing hardware and software systems, and he had some thought-provoking things to say about the ethical implications of trusting actual battlefield weapons to the decisions of self-controlling robots.

We talked about HAL 9000 from 2001 a Space Odyssey, Skynet from the Terminator films, and the WOPR (whopper) from the movie War Games.  My professor friend offered the opinion that, after many years of fictional speculation, reality has finally caught up with the imaginary threat.  That was the part that surprised me—his assumption that armed autonomous systems are a brand new problem.  But the problem isn’t new.  It isn’t even particularly recent.  Killer robots have been around by the tens of thousands for well over half a century.

Anyone who has ever worked with acoustic homing torpedoes will know exactly what I’m talking about. They’re fire-and-forget weapons.  We point them toward the enemy and turn them loose.  It’s their job to seek out the target ship or submarine, evaluate its characteristics, and then get close enough to destroy the unlucky vessel with a massive blast of military-grade explosives.  Along the way, the torpedo must make a complex series of life-or-death decisions, with no human guidance whatsoever.  We depend completely on the weapon’s ability to make the right choices at the right time, and there is literally no way to stop the torpedo if it decides to go after the wrong target.

Some torpedoes can be guided remotely by a human operator through a trailing wire, but once it’s off the leash, a modern acoustic homer is nothing more (or less) than a robot with a bomb. There’s no way to call it off, no way to alter its direction or priorities, and—despite what you may have seen in certain submarine action movies—there’s no remote destruct mechanism.  It’s a fully autonomous system with a lethal cargo that makes most drone strikes look like small arms fire.

Here’s an excerpt from the prologue of my first novel, Sea of Shadows, that sums up the situation pretty clearly… 

It had no name for itself. It was not even aware of its own existence.  It waited in its shipping canister, cradled as snugly in the cylindrical steel container as a high-powered bullet in the chamber of a rifle.  Cold.  Sightless.  Unfeeling.  Not sleeping, merely unawakened.

R-92 was a state-of-the-art acoustic homing torpedo. It was a cybernetic predator: an electro-mechanical killing machine.  Fast.  Smart.  Unbelievably lethal.  Every component, from the shark-like hydrodynamic form of its fuselage—to its multi-spectrum acoustic sensors—to the axial-flow turbine that formed its engine, was optimized for the undersea environment.  Its brain was a fifth-generation digital computer, hardwired for destruction with a machine-driven relentlessness that no living predator could match.  R-92 and its brethren had been honed for the chase and the kill by two and a half centuries of technological evolution.

But R-92 knew none of these things. It simply waited.

Fiction? Yes.  But only in the sense that the particular torpedo called ‘R-92’ is a product of my imagination.  The description above could apply—with near-perfect accuracy—to any one of a hundred torpedo models that are currently in use by navies around the world.

The earliest models saw combat in the early 1940s, when Nazi U-boats used them to attack convoys of Soviet merchant ships. In the seven or so decades since, the acoustic homing torpedo has become smarter, faster, and a hell of a lot more deadly.

I’m not suggesting that the emerging debate over autonomous killing machines is somehow unimportant, or even irrelevant. Far from it.  This is a serious topic, with implications far beyond anything I can imagine.  I’m saying that the debate should have started decades ago, because the killer robots are already here, and they don’t look anything at all like Arnold Schwarzenegger.