The Turing test for media
Will machines ever be able to match – or outperform – human intelligence? David Brennan gets philosophical.
There has been a great deal of debate around the latest paradigm to hit our industry – mad men vs. math(s) men. It is all part of a binary approach to most things in media, whether that is analogue vs. digital; algorithms vs. instinct; analytics vs. insight; ‘new’ vs. legacy media etcetera bloody etcetera.
In order to add my two penn’orth to the debate, I want to reference two of my intellectual heroes; titans in their respective fields of computing and communication theory.
The first of these, Alan Turing, is pretty well-known these days following years of quite shameful persecution and obscurity before his suicide in 1954. His achievements in cracking the Enigma Code have been universally acknowledged, as has his influence in developing computer technology (he developed the concepts of the algorithm, artificial intelligence and computation amongst others).
He also, with extraordinary prescience, created what has come to be known as the Turing test. In line with his belief that man would eventually be able to develop artificial intelligence to the point where it could match or even outperform human intelligence, he set out the challenge in the form of a test.
The Turing test is a test of a machine’s ability to exhibit intelligent behaviour equivalent to, or indistinguishable from, that of a human. The challenge is to convince human evaluators that a ‘conversation’ with a machine would be indistinguishable from an equivalent interaction between humans.
[advert position=”left”]
Last year, the artificial intelligence community was celebrating (at least as far as anyone could tell) the first successful attempt to complete the challenge.
A computer program called Eugene Goostman, which simulates a 13-year-old Ukrainian boy, is said to have passed the Turing test at an event organised by the University of Reading.
At first glance, we appear to have passed a threshold, which gives weight to the math(s) men’s claims that data can eventually process, mimic and predict human behaviour and decision-making, potentially revolutionising the role of marketing. The creativity can reside in the algorithm.
At second glance, though, the results are far less ground-breaking. The program had to communicate with its human judges via online text for five minute chats and convince at least 30% of them that ‘Eugene’ was human. In the event, the program convinced 33% that Eugene did indeed exist. Still a long way to go before Channel 4’s ‘Humans‘ becomes reality, then.
Academics have already criticised the restrictions of the test, including the short length of the chats and the bias introduced by having a non native English speaker as the test object.
But, of course, the major restriction is the nature of the interaction itself; a text-only conversation conducted remotely. That is human interaction at its most basic level; a fact we would do well to consider when we look at the future of big data and the power of the algorithm.
I will always favour the mad men in this debate, because they are dealing with the analogue and that is where the real gains are made.”
Imagine if the Turing test judges had to decide based on audio conversations. Still massively constrained compared to video or (especially) face to face interactions.
Would the program be advanced enough to not only mimic the sound of the human voice, but also be able to portray the nuances of personality, experience and emotion through the myriad verbal signals we transmit through every conversation? I think the world of AI is still a long way from that level of human understanding and representation.
This brings me on to my second intellectual hero; Paul Watzlawik. He was a celebrated therapist, psychologist, philosopher and communications theorist who first proposed the digital/analogue paradigm; not in relation to technology but something far more important; the way humans communicate with each other.
Watzlawik persuasively argues that human interaction can be segmented into digital – the syntax and semantics of the language itself – and the analogue, which is everything else. The analogue would include all of those verbal cues, visual references, micro-expressions and body language indicators that actually comprise the bulk of what is communicated.
The oft-repeated claim that business presentations are only communicating around 7% through the words and charts comes directly from Watzawick’s work; the bulk of the communication effect is through the analogue part of the process.
The problem is, the judges in the Turing test didn’t get access to all of those important analogue clues; their interaction was literally and figuratively digital-only.
So, my take on this is; it’s fantastic that Eugene can convince a third of judges he is real via the digital process, just as I’m delighted that big data can create so many more efficiencies and enhancements to our business.
Unfortunately, it is all still working at the margins, and I think both Turing and Watzlawick, if they were still with us, would understand the causes.
That’s why I will always favour the mad men in this debate, because they are dealing with the analogue and that is where the real gains are made, because that is where most of the influence resides.