Op-Ed: Scientists getting techno fear about intelligent machines?

Perhaps the oldest story line in science fiction is intelligent machines dominating the world. From Daleks to drones to C3PO and Robbie the Robot, machine intelligence has always had an audience. Now, it’s got real scientists worrying about it.

A story in The New York Times about the threat from intelligent machines is near enough to plausible media news to rate some attention. The story itself, however, is about what scientists attending a think-fest conference organized by the Association For The Advancement Of Artificial Intelligence see as real threats:

Impressed and alarmed by advances in artificial intelligence, a group of computer scientists is debating whether there should be limits on research that might lead to loss of human control over computer-based systems that carry a growing share of society’s workload, from waging war to chatting with customers on the phone.

Their concern is that further advances could create profound social disruptions and even have dangerous consequences

. It’s debatable exactly how much more disrupted this non existent society can get, but arguable that homicidal, chatty robots wouldn’t help much. As people trying to live with homicidally chatty corporate answering services have noticed, it's already quite difficult enough. The scientists have gone beyond the average story line with the interesting and entirely credible idea that criminal use of artificial intelligence is a real, and all too likely, possibility. That’s a real threat. Another is that machines can take over human manual jobs like driving trucks (already in use in a less dramatic form by the US military MULEs) and other roles. This is a pretty ethical argument by the people trying to develop artificial intelligence, and the Association itself deserves some attention. Their website is well worth a look, particularly the “Trust me, I’m a robot” piece which looks at the brute logic of laws related to robots. Seems the arguments are advanced enough to discover nobody is prepared to insure a robot, anyway. Even the ethics of creating an intelligence get a look in, which is something even the theologians don’t seem to have dealt with much. Getting taught ethics by humans may be like getting taught hairdressing by the Grim Reaper, but it’s nice to know someone’s willing to try. The scientists’ sense of humor hasn’t entirely left them. One scientist referred to technologists replacing religion, and refers to the idea of super intelligent machines running amok as their equivalent of the Rapture. The Association is trying to put together a working concept of “the loss of human control of computer-based intelligences.” Even defining that concept isn’t likely to be easy. Bearing in mind the developments of the last ten years, anyone in any hurry to predict exactly what could go wrong in the next twenty? This is where ideas meet reality, and it’d be nice if some people read a bit more on subjects which were very thoroughly examined decades ago. Isaac Asimov covered all of this in his robot books in the 1950s. Asimov didn’t oversimplify the issues, nor claim omniscience, nor claim to have all the answers. What he came up with, however, was a very interesting set of principles, called the Three Laws of Robotics:

1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.

2. A robot must obey any orders given to it by human beings, except where such orders would conflict with the First Law.

3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

Note the use of a very modern word in computer and robotics terminology: “Conflict”. A conflict in a program is where one instruction conflicts with another, and prevents functions. The point here isn’t to quote Asimov per se, but to point out something to programmers and others: These are comprehensible principles, which both humans and robots could understand. Even legislators could understand it. These three laws condense the entire issue into working principles with clear objectives, which is so far what seems to be lacking in this debate. So far we’ve got an unspecified future threat, and no obvious objectives or methods for dealing with it. There are huge difficulties inherent in explaining to a modern machine what the words “harm”, and “existence” mean. Try coding that. You could try the materialist approach, “If you bump into something it must be real,” for existence, for example. You could define “harm” as anything which would reduce whatever you use to describe the quantification of the thing being harmed, sort of a points system where somebody in danger of being hit by a truck is likely to be less than they were, if hit. Ponderous logic, but that's what we've got at the moment. Let’s start by defining artificial intelligence. Do we mean something with recognizable human intelligence, or something with a totally different form of intelligence? Human type intelligence is definable, to the extent of its functions. What about something which by definition isn’t human? How do you control something you can’t define? Asimov worked on clear concepts of robot intelligence, which was both recognizable and ideal, a classic machine able to interact with humans, communicate effectively, and comprehend higher and extended logic. These were “machine people”. The artificial intelligence which is being developed isn’t “people” yet, but it’s definitely getting more intelligent in a hurry. I could go on forever, but the point here is that the practical solutions have to be built in to the artificial intelligence. There has to be an “Off” switch if anything goes wrong.