Artificial Intelligence Is Better than None

Humans are having so much fun hating each other these days that other concerns rarely remain in the forefront of public consciousness for long. One such concern that had a flurry of attention in the press in 2018 but then evaporated was the long-term danger posed by Artificial Intelligence (AI) and robotics. Perhaps the death of Stephen Hawking, who warned that AI was an existential threat to humanity, had something to do with the passing interest. Other respected figures added their voices including Elon Musk whose robotic cars roam the highways. Musk warned, “The risk of something seriously dangerous happening is in the five-year timeframe.” I’m sure another press flurry will come soon, for robots aren’t going anywhere – or, more accurately, they are going everywhere.

The kind of dangerous happening Musk had in mind was, in scifi terms, less Skynet, which consciously chose its own objectives, than HAL, which unexpectedly resolved conflicting mission objectives it was given by trying to kill the crew. There is a difference. As we become ever more reliant on AI to route electric power, drive our cars, fly our airplanes, manage our industrial production, and even just control the temperature in different rooms of our homes, the risk to the systems grows. The worry is not so much about narrow AI, such as the algorithms that tell Amazon or Google what products to recommend to you, though even these sometimes behave unpredictably. The concern is about more generalized higher level AIs that reprogram themselves as they learn about the world by interacting with it; they already have done very strange things such as learn (by themselves) to cheat at games by changing the rules. As they get ever more complex and ever better at simulating consciousness they are bound to spring surprises on us – whether or not by “intent.” There are growing numbers of robotic weapons with the capacity to act autonomously, whether or not they are allowed to do so, with some self-evident dangers, but even autonomous machines built for benign purposes can go badly awry. Reliance on AIs also makes us vulnerable to nefarious fellow humans who aim to disable them with EMP or simpler methods.

A particular kind of AI has intrigued us the most, at least since Karel Čapek’s 1920 play R.U.R. (Rossum's Universal Robots) put the word “robot” in the dictionary: androids. Autonomous robots need not be humaniform (android). Scarcely any are at present, and those few are not very high functioning, but they are clearly the end game. We love android robots, sometimes literally. Brothels with robotic love dolls are in business in Toronto and Paris. The appeal is bound to grow as the machines become not only more animate but better conversationalists. They also scare us. Recently the idea that robots might end mankind through love not war has gained some currency: the notion being that we might so prefer machines to actual fellow humans that we don’t bother to reproduce. (See Saturn’s Children by Charles Stross set in an entirely robotic post-human solar system in which the humans weren’t killed but just faded away; I wrote a short story called Circuits Circus set in the waning days of human-to-human sociality.) It is true that many people already have deeper relationships with their smart phones than with their families, and it is also true that online games such as Love Plus that let the player construct a virtual lover (rather than deal with the pesky real thing) are wildly popular. However, the old-fashioned homicidal robot uprising (pioneered in R.U.R.) is still the classic nightmare.

Roboticist Daniel H. Wilson writes science fiction (including the novel Robopocalypse) when he isn’t tinkering together the mechanical successors to humanity. Some years ago, however, he wrote a tongue-in-cheek guide titled How to Survive a Robot Uprising: Tips on Defending Yourself against the Coming Rebellion. In it he reveals the various ways your smart house and your car might try to kill you, and he escalates the possibilities up to the full-out terminator-style assault. He discusses the ways to defeat various tracking sensors and recognition technologies. He explains how to exploit the quirks of AI decision-making; they think far faster than you in familiar situations, for instance, but struggle with wholly novel ones. He tells how to recognize a robot if one calls you on the phone by the way it constructs speech. Though he advises running and hiding whenever possible (with tips on how and where), he tells how best to fight back. Blinding cameras by throwing mud on them is a good first step, though it won’t help if the robot can access real-time drone or satellite imagery. All-in-all, it is a fun little field guide, which despite the silliness really does cover many of the peculiar strengths and weaknesses of present and future advanced AIs and robots. He says in the intro, “Behind every bit of advice exists an area of real research with genuine answers that have been culled from extensive interviews with robotics experts…You probably found How to Survive a Robot Uprising in the humor section. Let’s hope that is where it belongs.”

So, when the AI assistant in your smart house tells you to change a light bulb, wear shoes with well insulated rubber soles to guard against a sudden power surge. Further, enjoy your robotic inamorata when Amazon delivers the crate to your door, but after you activate him or her, keep track of the kitchen knives.

Jefferson Airplane - Plastic Fantastic Lover