Every day we judge between human and automated content. To speak through the fog, we must find our own voices.
US National Library of Medicine
It's one of the rules of modern life: You receive an enthusiastic message from one of your colleagues celebrating Russian pharmaceutical discounts at surprising length. Your response isn't, "Thanks, but I'm good." It's, "You may want to change your password."
In so doing you've found yourself unwittingly and against your will placed into one of the 20th century's most famous and controversial thought experiments -- the Turing test.
The test comes from British mathematician and World War II codebreaker Alan Turing, born on this day 100 years ago. In recent decades Turing has increasingly been seen not only as a war hero but one of the architects of the modern world: the theoretical model of the modern computer, underpinning everything from mainframes to iPhones, remains the "Turing machine" which bears his name. It was in 1950, when computer science was still in its infancy, that Turing famously addressed one of the field's large, looming questions: Can machines think?
Might we someday construct a machine that could think? If so, how would we know? Turing proposed to address the seemingly unanswerable questions with a practical experiment, which which today bears his name. A scientist behind closed doors has a pair of text-only conversations via computer terminal: one with an unseen human stranger, and one with a computer program designed to mimic and impersonate human conversation. Each claims it is the human and the other is the computer. Turing predicted that we would reach a point -- around the turn of the millennium, he believed -- where computers would so routinely fool human judges that we would "speak of machines thinking without expecting to be contradicted."
In imagining the course that the debate and practice of AI would take in the 21st century, Turing was amazingly prescient. But I think the casualness with which we field offers of cut-rate Viagra from our friends and family would have thrown him for a loop. Had Turing, who tragically ended his own life in 1954 at the age of 41, lived to see his 100th birthday today, I think he would be startled, and deeply surprised, at the extent to which his thought experiment has become inextricably woven into the fabric of modern-day life.
What does that fraudulent pharmaceutical email do to alter the real messages we send and receive?
The unavoidable ubiquity of spam today is more than just a nuisance. It genuinely alters the landscape of social trust. Every message we get--every text and email and Facebook comment--we must approach with caution: Is this truly my friend, or is it a spammer? Our day-to-day communication online is a constant, unremitting Turing test.
With this fact comes the danger of interacting in increasingly uniform, standardized ways. If a Facebook friend of mine "likes" my status update, or a Twitter follower retweets my message, I have little by way of style or nuance to convince me the behavior was genuine. The brutal economy of language created by the slow speed of texting (especially on older phones) similarly erases shades of verbal difference. My mother and younger cousin, 40 years apart in age, sound nothing alike -- prosody, syntax, vocabulary -- across a dinner table. But when he borrows her phone to text me while she drives, the "r u here yet" and "K thx" he sends me on her behalf look indistinguishable from the real thing.
On the other hand, as I look back at my instant message history from the past few days, I see that, given sufficiently rich bandwidth and the right frame of mind, an incredible diversity blooms, in first lines alone: my greeting provoking everything from "hey!!!" to "dude." to "yar!" to "yes sir!" to "WHY AM I AWAKE" to "Salutation, Señor Christos." Such diversity not only delights. It authenticates. I have, already, my first reason for confidence that it's not a bot -- or an identity thief, or a friend borrowing their computer -- at the other end of the line. My willingness to speak candidly follows suit.
The scientific literature has often criticized the Turing test for lacking a true "control" -- a stable, objective benchmark with which to chart the progress of AI technology year by year. Chatbot programs' success is measured against the efforts of the human interlocutor and human judge, and their own ability to make conversation. Each year, chatbot developers hone their understanding of the dynamics of human interaction, but the human participants, living ever more closely with technology, themselves become cannier. Though it muddies the test's objectivity, this symbiotic process of co-evolution may offer one of the Turing test's most redemptive aspects: the chance to raise the bar of human interaction.
A bare URL in the body of an email no longer feels sufficient, for example: I increasingly feel compelled to make even the briefest of notes read like my words and mine alone. This may always have been a part of social grace, or of charm; it has now become part of online security. In so doing, the Turing test scores an unexpected victory for inter-personal communication, and becomes an unexpected bridge between the Two Cultures. Poets have always been thinking about how to throw off standard patterns and modes of expression, to find a voice unique, distinct, and singular. Now the computer scientists are too. We all are.
We want to hear what you think about this article. Submit a letter to the editor or write to firstname.lastname@example.org.