At A Glance – The Tokyo Test
Has Artificial Intelligence outgrown the Turing Test?
Fuelled by mass data and sophisticated software solutions, Artificial Intelligence is edging closer and closer to its human equivalent. But how do you test whether AI can truly match a human’s cognitive ability? As we know, in the 1950s, Alan Turing came up with the method to do just that – The Turing Test remains the most renowned process to distinguish humans from machines. In the test, two participants are asked a number of questions to work out which is human and which is a machine. If the results are inconclusive, the machine passes. As AI improves at an exponential rate, Japanese researchers have decided that the standard of testing needs to change.
In 2011, a team at Japan’s National Institute of Informatics began to develop a machine called the Todai Robot to prove that computers can think, and act, in a human way. The machine was expected to pass ‘the Tokyo Test’, a multidisciplinary college entrance exam for the prestigious Tokyo University, by 2021. In order to complete the Tokyo Test, the machine would not only have to pass as a human, but show it could think like a human across a variety of subjects. At the April 2017 TED Conference, AI expert Noriko Arai explained that although the Todai Robot failed, its score was better than 80 per cent of the human participants. While AI has yet to pass the notoriously difficult exam, its level of success signals that this could certainly be possible.
In terms of AI research, the implications of moving beyond the Turing Test are vast. On the one hand, this will enhance the technology to the benefit of those that use it, but at what point do researchers admit that AI has reached human level intelligence? Continually moving the goalposts could be viewed as poking the hornet’s nest. Eventually, it could well be humanity that is under examination.