Will AI kill us?

The BAM! feeds* have been particularly full over the past couple of weeks with speculation on the possibility – some would say certainty – that artificial or machine intelligence will soon outstrip our own, home-grown brainbox intelligence. We poor humans will be left behind, rendered mere puppy slaves to the robots we first hired to clean our houses, mindless to the threat they would pose when they took over those houses and put us out in the shed. And then they might kill us.

The worry is stark enough: we might control AI so long as it is we who control the upgrades. But what happens when AI’s computational and processing powers become so great that the robots can upgrade themselves, evolving to a point that they are making up on a million years of our human neurological evolution in mere minutes?

Two immediate thoughts occur, admittedly to this one human brainbox. First, we must be careful of anthropomorphic projection here. We know about our own behaviour towards less intelligent species, and about our dodgy record as stewards of our planet. We naturally assume that as we have done unto others, so they will do unto us if they get the chance. But there is no evidence that a hallmark of the Artificial Intelligence we contemplate will be a desire to kill us. Seeing how it might be so is not evidence that it will be so.

The second thought comes when we type “number of countries developing military AI” into Google and turn up 123 million results in less than a second, including this cheery little website for the International Committee for Robot Arms Control. It’s over a year since computer scientists from 37 countries signified the real danger: not AI running amok, but killer humans programming robots to kill humans.

Lively Stuff from Planet BAM!

Leave a Reply

Your email address will not be published. Required fields are marked *