Since computers were invented, we’ve given them information and told them what to do. What if they could decide what to do with the data themselves and become smarter in the bargain? Welcome to the world of artificial intelligence (AI).
Computers are great at giving us the right answer, but terrible at thinking up a unique new answer – we have to teach them how to re-evaluate the question first. The holy grail of autonomous computing is a system that can teach itself about the constant variables humans take for granted, for example when we’re speaking.
But with the almost endless processing capabilities of the cloud, there’s a movement to make computers that are truly intelligent, and you’d be surprised how close we are. What else is iPhone’s Siri but the first steps?
Rather than being simply much smarter versions of brains, computers are actually quite different from people, and we each have very different talents.
Computers are about accurate calculations of huge, absolute numbers. Our powers fall down when it comes to huge maths problems, but we’re really good at looking at things in a more abstract way and extracting meaning.
If we saw long grass moving on the savannah hundreds of thousands of years ago, we didn’t have to know what kind of grass it was, how tall it was and what kind of animal liked to lie in wait for prey in it. We sprang into action in case we were about to become lunch even if it turned out to be a gust of wind.
Artificial intelligence, as the name suggests, is the effort to make computers more like us. Instead of just feeding in huge amounts of information and telling them what we want to do with it, the point of an AI system is to give it the information, tell it what we want to achieve and let it figure out the best way to get there. Even – as some software engineers think – to suggest a course of action we haven’t even considered.
Machine school
It’s achieved through a process called machine learning, which is the ability of a computer to remember something from experience and change its own programming on the fly rather than have us establish new rules for it.
Here’s how it works. Your antivirus software has to know exactly what a particular virus looks like to stop it. If the cybercriminals change just one line of code out of millions, your antivirus program won’t recognise it as a virus and you’ll get infected. Machine learning is where the software has the ability to decide if an incoming file has similarities to a virus it’s seen before, even if it’s completely new.
If a computer virus can be represented as ‘12345’, machine learning tells the program that an incoming file ‘12354’ needs attention. It might be safe, but it’s a bit too much like ‘12345’ for comfort.
Just like in the above example, finding cyberthreats before they become problems is one of the biggest and most urgently needed areas for artificial intelligence research. According to Symantec there was a 42 percent increase in targeted cyberattacks in 2013. Two thirds of companies didn’t even know they’d been breached for more than 30 days, and it took an average of four months to undo the damage to business continuity and the bottom line.
The future of AI is limitless. When computers can process information much more creatively (and much faster) than they do now, they’ll appear to be ‘thinking’ for themselves. That might mean they’ll soon be doing everything from flying planes to having conversations with us, just like in the Spike Jonze movie ‘her’.
Can smart robot shop assistants or secretaries be far away?