A chat bot finally beats the Turing intelligence test


By Anand Patel

Have you ever wondered when the inevitable story of Terminator 2 will come to fruition? Where your Facebook machines and twitter devices will one day become self-aware and decide they will no longer be the slaves to our internet addiction and instead decide to rise up and enslave the human race as we know it? Me neither, that’s crazy talk.

Scientists and technologists however have been trying to crack a challenge set by mathematician and marathon runner Alan Turing for over half a century. The Turing test basically boils down to “a test for intelligence in a computer, requiring that a human being should be unable to distinguish the machine from another human being by using the replies to questions put to both.” Thank you Oxford Dictionaries.

Supposedly the Turing test was recently successfully passed by a computer program called Eugene Goostman, which had been designed and built by three programmers and was showcased during an event at the Royal Society in London, held on the 60th anniversary of Turing’s death. Eugene Goostman takes the form of a 13 year old boy and has been engineered to “think” and “feel” just like a real boy.

eugene goostman

Eugene Goostman, the artificial intelligence (AI) system - better known as a chat bot or chatterbot - was put to the test in front of 30 judges at the event. It was questioned for 5 minutes, and managed to convince 10 out of the 30 judges that it might not be a computer program but an actually 13 year old Ukrainian boy who could hold a “meaningful” conversation with an actual human being. Goostman is a combination of two operations - a language processing system and a keyword matching engine - using large databases of predefined messages to choose appropriate responses.

Now don’t expect a robot butler anytime soon, although this is a major milestone in terms of AI, it’s still a way off the likes of HAL or C3-PO.

After the initially result was announced, some controversial points were raised over the accomplishment. Firstly, despite some of the program’s responses appearing at times to be nonsense - compared to a humans typical response - this was overlooked because of his supposed age and background. Secondly, it could be argued that Goostman did not actually fully understand the questions the judges asked and simply just provided the closest suitable responses when matching the input against a database of possible responses. However again the program’s “background” helped disguise Goostman’s level of intellect and understanding to the questions being tested upon.

Although Eugene Goostman passed the Turing test with a score of 33 percent, it cannot be considered to be a strong AI machine as it does not show any true signs of independent thought and does not do any sort of machine learning from past experiences. But this does highlight a question - if Goostman is the best thing to be produced so far, will we ever see a sentinel machine in which we can have a casual conversation with in our lifetime? I’d like to think so but I would rather someone hurry up and make a real life hoverboard. But thats just my personal priority.

If this type of news gets your blood pumping you might have a whale of a time at our first ever AnalogFolk Hack Festival. If you think you could have done a better job than a chat bot you should come along and test your might, or just come along for the music and free food.