The prospect of artificial intelligence excites and repulsive people in equal measure: will it bring us a kind of paradise or a techno hell? To get a clearer handle of what might happen and when, it’s best to divide A.I. into three categories. The first of these is “artificial narrow intelligence” or what people call “weak A.I.”; this kind of A.I. is already in place; it’s the kind of A.I. that uses big data and complex algorithms to arrange your Facebook timeline or beat you at chess; narrow A.I. has an intelligence that’s limited to one very specific arena; it may not be able to pass the Turing test, but our lives, infrastructure, and financial markets are already very dependent on it. The next step up the AI ladder is artificial general intelligence or strong AI; this is an intelligence that can, at last, think as well as we can; we’re probably about 30 years away from this. The hurdles to creating strong AI are all about building machines that are going to be good at doing things which come very easily to humans, but which machines have, traditionally, really stumbled with. Oddly, it’s so much easier to build a machine that can do advanced calculus than it is to build one that can get milk from the fridge, recognized granny, or walk up the stairs. Our brains are brilliant at so-called “everyday tasks” like decoding 3D images, working out people’s motivations, and spotting casual sarcasm. We’re very far ahead of machines here. Some scientists doubt we’ll ever see strong AI, but the majority of AI experts alive today seem to think that we’ll be there in the coming decades; if you’re under 35 the great probability is that you will be there to enter the strong AI age. So, what will happen to the world once we’ve succeeded in creating an intelligence to rival or equal our own? Well, the rivalry will be extremely short lived for one thing because the key point about strong AI is that it will be able to learn and upgrade itself on its own without instructions. This is what makes it so revolutionary and so different to almost any machine we’ve ever built; the maker won’t be in charge of mapping out all the possibilities of the thing he or she has made. The machine will be given a baseline capacity, but it can then build on this as it develops. It will be a trial and error learner with an infinite capacity to acquire skills; it’ll have what AI professionals call “recursive self improvement”. This is crucial because it means there’ll be no reason for AI to stall once it reaches the human level. The more intelligent system becomes, the better it becomes at improving itself, so the more it will learn and do. This virtuous cycle equates to an exponential growth in intelligence that would leave humanity amazed, but also baffled, dwarfed, and perhaps very scared. It might not take very long at all, only months perhaps, before the machine is cleverer than its creator. This is the moment that gets very exciting. It’s a moment often referred to as “The Singularity”, which is where we encounter the third sort of AI, “artificial superintelligence”. Technically, this is any AI that exceeds human levels of intelligence even slightly, but any self improving superintelligence is going to be sure to improve a lot very fast indeed. AI that reach this level would soon be leagues ahead of us, and statements such as, “well, let’s just switch it off” might be like trying to take down the internet with a slingshot. The prospect of such super intelligence appalls and excites people in equal measure. We’re approaching two alternative futures with the speed and uncertainty of a skydiver who can’t quite remember if he’s wearing a parachute or a rucksack. Some including: Bill Gates, Stephen Hawking, and Elon Musk are so scared they believe that we’re unlikely ever to be able to effectively control any super intelligence we create. Artificial minds will just single-mindedly pursue their aims and these aims may not necessarily coincide with ours. A machine wouldn’t specifically want to kill us, but it’s amorality would mean that it would be willing to cause our extinction if necessary. These critics point out that intelligence is not value loaded. It’s tempting to assume that anything intelligent will just naturally develop vaguely human values, like, empathy and respect for life, but this can’t be guaranteed because ethical values are based on purely human axioms, and given that we find it impossible to agree among ourselves what’s right and wrong in areas like euthanasia or abortion, say, how could we possibly program a computer with a knowledge that could soundly and reliably be deemed moral? Now that’s the pessimistic angle, but there is a more cheerful angle, of course. According to the optimists, in a world of artificial super intelligence, machines will still be our servants, we’ll give them some basic rules of never killing or doing us any harm, and then they’ll set about solving all the things that have long bedeviled us. The immediate priority of super intelligence would be to help us to create free energy, in turn, dramatically reducing the prices for almost everything. We would soon be in the era that Google’s chief futurologists Ray Kurzweil describes as ‘abundance’: everything currently costing would drop to almost $0, the way that data costs now. Work for money would, essentially, come to an end. The real challenge would be not getting miserable with all this abundance, after all, Palm Springs and Monte Carlo already now point to some of the dangers of wealthy people with nothing much to do. The solution here is to develop a side of A.I., that’s been intriguingly dubbed A.E.I., or, Artificial Emotional Intelligence. This A.E.I. would help us with all the tricky tasks at the emotional, psychological, and philosophical end of things. We’d be helped with: understanding our psyches, mastering our emotions, drawing out out true talents- we’d hit what we were best suited to do with our lives-, and guiding us to the people with whom you might form good and satisfying relationships. Most of the many psychological mistakes which allow us to waste our lives could be averted; instead of fumbling through a mental fog of insecurities and inconsistencies, we’d be guided to a more compassionate, happier, and wiser future. Science fiction is sometimes dismissed in elite circles, but we can see now that: thinking twenty to fifty years ahead, and imagining how life will be is a central task for all of us; we should all be science-fiction writers, of a kind, in our minds. We are poised just before a tipping point in human history. We need to build up the wisdom to control which way we will tip, and part of that means thinking very realistically about things that, today, still seem rather phantasmagorical. Humans are toolmaking animals; we’re on the brink of creating tools like no others, so the trick is going to be to stay close to the underlying ancient purpose of every tool, which is to help us to do something we actually want to do more effectively. If we keep our wits about us, there’s no real reason our computers should, necessarily, run away from us; they should just be much much better versions of our earliest flint axes.