“The Turing test was passed for the first time in 2014 when a chatbot posed as a 13 year-old Ukrainian boy, Eugene”“Funding for AI-based projects has been growing rapidly in recent years”
Louise Flanagan investigates the growing field of artificial intelligence.WHEN most of us think of artificial intelligence (AI) our minds readily conjure up sci-fi fuelled images of super-intelligent, humanoid robots hell-bent on replacing the human race and taking over the world. This is not exactly true (at least not yet).In its driest terms, AI is a device or agent that can display intelligence on par with, or greater than, a human. AI is far from a dry research area though. Funding for AI-based projects has been growing rapidly in recent years. AI research has produced devices that can beat chess grandmasters, recognise faces, diagnose diseases, test terrains for danger in war zones, and even autonomously vacuum floors.But where exactly is AI heading? Are we even in a position to see that far ahead just yet?There are two types of AI: “strong AI” and “weak AI”. “Weak AI” is all around us, in our GPS devices, the voice recognition technology in our smartphones, and even in a standard school calculator. These devices copy a narrow range of human brain functions.“Strong AI” on the other hand, is about capturing all the subtleties of human thought and behaviour, from learning and memory to perceiving other’s emotions and being able to make sense of things it hasn’t seen before. “Strong AI” doesn’t exist yet, but “weak AI” is slowly building its strength. How will we know when machines do reach this landmark?One way is the Turing Test, devised by computer scientist Alan Turing back in 1950. The test is simple: a human judge has a text conversation with a computer and a human, but the judge can’t see either of them and doesn’t know which is the computer and which is the human. If the machine can fool the judge into believing it’s human 30% of the time, it passes and is deemed artificially intelligent.The test was passed for the first time in 2014 when a chatbot posed as a 13 year-old Ukrainian boy, Eugene. While Eugene technically succeeded, its persona did provide a convenient disguise for any grammatical errors and undeveloped answers.Although it’s still useful there are drawbacks to the Turing Test. Firstly, machines are being created purely to pass the test, without being concerned about true intelligence. Secondly it only focuses on the machine’s ability to mimic human behaviour – it does nothing to account for the machine’s thought processes.Other approaches to AI, not so concerned with the Turing Test, also exist, focusing on creating machines that mimic how humans think. Alternatively, there are approaches that don’t use humans as models for intelligence at all, instead focusing on rational solutions to questions and situations.So how far away are we from the “strong AI” like we see in 2001: A Space Odyssey and Ex Machina (pictured above)? The short answer is a long way. It is possible we could create machines which are sentient and autonomous and which have the ability to create things themselves. However in the near future we’ll more likely be dealing with AI that can do a small number of tasks very well, better than humans, but they will not be humans.