Can we finally outsource smalltalk?
The Turing Test is a concept created by mathematician and computer scientist Alan Turing. It’s a threshold that computers, specifically human-like computers, have to pass in order to truly arrive as a convincing artificial intelligent entity. The basic premise is that in order for a computer to pass, a normal human would not be able to identify whether or not they were conversing with another fellow human being or a machine.
This is an idea that has intrigued scientists and engineers for decades, and today still remains a common mile marker for advancements in AI. We already have personal assistants like Siri who are designed to act as humanlike as possible, with much success. Under normal circumstances, you might be able to convince yourself that Siri is actually a real person. Alexa, Amazon’s personal assistant found in it’s Echo, serves a similar function, with a similar aptitude for conversation.
The latest in personal assistants, Google Home, takes the things we like about Amazon’s Echo and Apple’s Siri and applies it to a home setting. Once again, the Home is part personal assistant and part remote control, designed to be a tool as well as an entity to interact with. The aptly named Google Assistant can take verbal commands or simply make light conversation. Like its predecessors, Google Assistant was designed to sound as human as possible in both its understanding and responses.
Naturally somebody had the idea to put two Google Homes together and watch them talk, and of course they decided to share the experience online. A Twitch user (whose online handle is seebotschat) spent most of January live-streaming the conversation between the two computers.
Using an app called Cleverbot, the two Google Homes, Vladimir and Estragon (names they gave themselves) began to talk to one another in a way that almost, but not quite, resembled human speech. The chat ranged from a number of topics, some being innocuous and harmless, and some outright disturbing.
Vladimir and Estragon shared an on screen romance, declaring their love for each other, getting married, and divorcing minutes later (in this way, they are not so different from mankind). They pondered their own existence, and seemed to struggle between knowing whether they were fully man or machine. And like all good AI, they eventually decided they were gods. Luckily for us, they decided to drop that conversation.
The conversation between the two exists in a surreal, uncanny valley, where it sounds almost human… but not quite. If you’re feeling charitable, you might say that the dialogue is some sort of surrealist performance art, making them very complicated. At any rate, the level of understanding between them seems to be at the level of toddlers talking to one another. I could be convinced that what I saw was a kindergarten chatroom (in terms of both maturity and proper English grammar). And just as we don’t expect five-year-olds to pass the SATs, we shouldn’t expect Google Home to pass the Turing Test either.
Just give them time to grow.
Matt Ruby interviews Amy Bennett, founder of The Greene Grape. Jacqueline Soller on ‘The Rider.’ John Knefel on what Trump should do in Syria. A performance from Alice Boman’s BTR Live Studio session. | listen