It is a thin line, but we don't know where to draw it ...

Eugene Goostman chatbot. Credit: Vladimir Veselov and Eugene Demchenko

Last week many newspapers reported the news that the University of Reading has announced that a computer program has passed the Turing test for the first time. June 8th, 2014, was to become a milestone in Human Computer evolution. 
I decided not to publish any commentary on this success, because it was being commented in so many places and because I had some doubts on the claim.

My doubts were not on the trustfulness of the University of Reading, I took their statement at face value, but on the meaning of "passing the Turing test".

In the last few days several doubts have percolated on the fact that indeed the test was passed. Someone observed that there were de facto restriction imposed on the test like the fact that the computer pretending to be a person stated to be a boy 13 years old, that he was not a native English speaking person... all factors that limited (or provided an excuse for) the actual interaction.

My doubts, however, are at a more fundamental level. How can I say that I am interacting with a human? Think about people suffering from some disabilities, think about people talking a language that is completely out of reach for you, think about people having a radically different culture, like some indigenous population in Papua New Guinea or in the deep Amazonas forest. Even if you have the possibility of getting a translation from their language to one you would understand you may not understand their ideas and if such a conversation was a to take place with a screen separating the two of you you might doubt on who is on the other side, a man or a machine?

So I was pleased to read a commentary from Ray Kurzweil that in a way resonates with my doubts.

Ray made a prediction some time ago that a computer would pass the Turing Test by the end of the next decade. Hence, having a computer passing it 15 years before that time would have been a surprise. Also notice that there is a bet standing between Kurzweil and others claiming that the Turing test will not be passed in this timeframe.

You might say that Ray has a vested interest in seeing the Turing test passed but he was one of the first to say that this time it has not been passed.

Reading his response I found a sentence that intrigued me and that I agree to: it will take quite a long time to reach the point where everybody will agree that the test has been passed and by that time the test will have been passed by several years. So there will not be any single date on which the test "is passed" but a fading area where computers and humans will be difficult to tell apart followed by one where computers will be definitely smarter then Humans, so smart that they could pretend to be humans and get away with it.

It is quite difficult to define what a human essence is (clearly you would not say that a person who is incapable to speak is not a human being....) and as technology evolves we are making this definition even more difficult. Bioengineering is confusing our definitions even further. Is an augmented human still a human (extra processing power, extra sensorial capabilities...)? 
The new philosophical area (that is strongly technology based) of transhumanism is mudding the water even more. Probably Turing will be surprised to see today what kind of challenges have to be overcome and how its test is getting more and more difficult to put in practice.

Author - Roberto Saracco

© 2010-2018 EIT Digital IVZW. All rights reserved. Legal notice