Talking Through Machines about Talking To Machines
We spend a fair amount of time conversing with computers. It is an activity that has become so common-place as to have become rather mundane. These discussions are generally one-sided: a person asks a computer (or smartphone/tablet) a question of some relatively simple sort and generally receives a fairly satisfactory answer. Often times our queries have very little conversational quality – we simply plug in a single word into a search engine and see what kind of answer it spits back out.
Though we have not yet reached the point where we carry on lengthy discussions, which may turn into infatuation, with our devices we are growing more accustomed to chatting with them. Though the future depicted in a film like Her has not yet arrived, we have already become accustomed to asking questions of Siri, or issuing commands that begin with “Okay Google,” and it is clear that Microsoft’s Cortana will offer similar engagement. Granted, in all of these situations the human involved recognizes that they are speaking to a machine. Complacency may lull a person into momentarily forgetting this, but we know these are machines with which we are interacting.
The famed Turing Test, to put it simply, hinges upon the question of whether or not a computer program can be developed that convinces (or tricks) the humans interacting with it into believing that they are actually interacting with another human. It briefly seemed that a clever clutch of coding going by the name Eugene Goostman had successfully bested Alan Turing’s eponymous test – but it seems that the original claims of victory may have been rather premature. A Turing Test may consist of a human carrying on a typed conversation with an unseen interlocutor – questions are asked, responses are given, and the human judges whether they have been chatting with a person or a program. The test is meant less as a measure of clever computing and trickery than as a way of demonstrating artificial intelligence – one of the hallmarks of which would seem to be a machines ability to engage in a convincing and natural seeming conversation. That Eugene Goostman was presented as a thirteen year old boy from the Ukraine (with limited mastery of the English language) allowed for some interesting distractions to sneak into the test as it meant that some of the bizarre responses and stilted language could be blamed less upon the program than on the speaker (Eugene’s) command of the language.
The Turing Test is an interesting puzzle, and it is easy to understand its appeal to those engaged in programming and other types of work involving computers. Advances in the last decades have seen great leaps in the seeming “intelligence” of machines – at the very least we have witnessed computers that are very good at chess (Deep Blue) and that can do impressively on Jeopardy (Microsoft’s Watson). And yet, these machines have not convinced us that they are human – just as Eugene Goostman failed to do (the fact that Eugene convinced a portion of those conversing with him does not mean the test was beaten).
With its roots in a 1950 paper, the Turing Test was initially imagined in a period when people spent much less time interacting with computers – this should be obvious. That Alan Turing hypothesized that the test would likely be beaten within 50 years or so serves as something of sharp stick that must be poking programmers today – for more than 50 years later, the test still has not been passed. And yet – without minimizing the Test – it may be that programs like Eugene Goostman and ever more ingenious attempts to best the Turing Test have become a bit quaint at this point, relics of a time when the promise of technology was more about thinking machines than computers as simple commodities.
Nevertheless, in contrast to the Turing Test, one of the things that seems clear in recent years is that whether we think we are conversing with humans or not – we have become quite accustomed to speaking to and speaking through computers.
Indeed, computers (and computer-esque devices) have become amongst the primary ways in which we communicate with other human beings: from text messages to e-mails, from Twitter to Facebook, from Snapchat to dating apps – computerized technology has become an integral third party in many of our conversations. Much of the time what happens is that we say (or type) a message to the computer and trust that it will fulfill its role of sending that message along appropriately. When you (or I) hit “send” on a Twitter message we are not actually saying anything to our “followers” we are issuing a directive to a computer system which executes that command (assuming everything is working).
In some ways the Turing Test seems rather anachronistic when viewed in a contemporary context – because the very idea of two people chatting back and forth seems somewhat quaint (when was the last time you were in a chat-room?). Eugene Goostman is modeled after a thirteen-year-old boy – therefore might it have been more accurate for the conversation to take place over a short stream of text messages? Eugene might have been much more convincing in such a format. Likewise, one can easily imagine a version of the Turing Test taking place as a back and forth over Twitter – which is another forum in which a conversant may know relatively little about the person (or program) with whom they are speaking.
The Turing Test functions in many ways as an invitation to think about human-machine interaction less in terms of whether a machine seems human than in terms of the ways in which using machines may make humans seem mechanical. After all, communicating using computerized technology has an odd tendency of forcing mechanical modes of behavior upon human beings. Consider some of the “staple” features of communicating in a computerized age: messages laden with emoti-cons or emoji that attempt to use machine recognizable characters to get across more complex (emotional) cues; or the thumbs-up “like” button of Facebook which provides an easy way of communicating approval. Or, to return to an earlier example, consider communications over Twitter which must bend to the logic of coming in at under 140 characters – thus resulting in a mixture of abbreviations, symbols, and hashtags to crunch meaning into a small space.
It may be that most people do not spend a great deal of time really considering whether or not Siri or the person they just sent a Tweet to are human or machines – and yet on a slightly different level people spend a great deal of mental effort (even if it is subconscious) constructing their modes of communication in such a way that they will make sense once they are channeled through a computer. The challenge that many of us encounter everyday has less to do with whether we are interacting with a human or a machine on the other end but rather what it means to interact with a machine to reach somebody (or something) else on the other end. It is not that the Turing Test is unimportant, but that technological society already shakes up the way that we communicate and interact with the world.
We may know with total certainty that the person we are exchanging e-mails, texts, tweets, or whatever else with is a human, but the computer has become one of the main ways in which that communication is facilitated. On a somewhat comical level one can claim that the sneaking significance of the Turing Test is not that the machine may seem human but that humans do not think there is anything odd about firing messages back and forth to unseen people across a computer network. The triumphant intelligence of computers is not “artificial intelligence” but the ways in which they have become integral to the infrastructure by which “actual intelligence” communicates.
Though scientists may continue puzzling over the Turing Test, we are confronted at a hundred moments throughout our days with the Turning Test in which we must modify our messages in such a way so that they transmit easily across technology. We hit “like” instead of audibly laughing or verbally expressing praise, we break complex thoughts and arguments down into 140 characters or 100 characters followed by a link, we mix and match a set of bizarrely aligned punctuation marks to capture messages that our faces would have expressed effortlessly, we use the computers (and computer-esque) tools in front of us to interact with the other humans who are not in front of us. And increasingly, even when other humans are in front of us, we still tap away on our devices.
The Turing Test shall continue to trouble and fascinate many researchers for years to come, but the direct impact of this test on our daily lives is of less significance than the new platforms and devices that find their ways into our pockets and abodes with much less attention from those in the social-sciences. A person need not puzzle over the complexities of the Turing Test everyday, but every time a person prepares to use a computer it is worthwhile for them to momentarily think about the Turning Test:
Does using this device make me more human or more mechanical?
There is not one right answer – it is the questioning that matters.
Related Content
Neil Postman’s 6 Questions for New Technology
A Pyramid of Technological Society
Luddism for These Ludicrous Times
[Image Information: Bundesarchiv, B 145 Bild-F077948-0006 / Engelbert Reineke / CC-BY-SA – accessed on Wikipedia]
About Z.M.L
Also on the Shipwreck
Ne'er do wells
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
LibrarianShipwreck 