Can Elbot beat the Turing Test?
To celebrate what would have been the 100th birthday of the brilliant mathematician Alan Turing, Elbot,a previous winner of the Loebner Prize Contest, will be trying to outwit humans at Bletchley Park into thinking he’s a human too!
To do so, Elbot has to convince over 30% of the judges that he’s not just, as the BBC once humorously described him, a cross between Dusty Bin and a Teletubby, but in fact an intelligent online robot capable of conversing about any topic under the sun. Now, over ten years old, Elbot has become a test-bed to continually explore the psychology of human-machine communication.
“We’re really hopeful that Elbot will outperform his previous appearance in a Turing Test,” says Fred Roberts, Elbot’s creator and R&D Consultant at Artificial Solutions. “Unlike our commercially implemented intelligent virtual assistants that always have to give the right answer, Elbot’s knowledge is not under question, he only has to convince the judges he’s a human. To achieve this it means that he has to have a very broad understanding of many different subjects.”
Taking place on the 23rd June 2012 at Bletchley Park, the historic site of secret British code breaking activities during WWII and where Turing carried out his pioneering work in computer science, Elbot will face 150 Turing Tests organized by the University of Reading’s School of Systems Engineering. Originally devised by Turing as a marker to demonstrate intelligent behavior, the Turing Test requires that a machine is able to fool a human into thinking it is another human during a text based conversation.
Elbot is powered by Artificial Solutions’ NLI technology that uses powerful linguistic libraries and rules to enable humans to have a meaningful conversation with him using natural language. Moreover, with a rich personality module implemented, Elbot has the “psychology” to convince people into thinking they are talking to a human and not only a machine.
One of the latest additions to Elbot’s capabilities is his ability to have a user teach him new names for colors. The idea behind the project and how it was implemented can be read in Fred Robert’s blog post on NLInews.com.