This week, Amazon Alexa users began reporting that their devices are laughing, creepily, on their own–so frequently that the phenomenon became a Twitter Moment. Sometimes, people reported, they’d be just sitting there in silence before Alexa began to laugh. Others said they might have just asked Alexa to turn off the lights–to be met with an eerie cackle. But in each case, the Alexa seemed unsettlingly amused at the nature of humanity itself.
— CaptHandlebar (@CaptHandlebar) February 23, 2018
Amazon acknowledged the issue quickly. The company implied the problem was that Alexa was just mishearing people, so the simpler command “Alexa, laugh,” was changed to “Alexa, can you laugh?” Furthermore, Alexa will no longer just laugh when you ask her to. She’ll say “Sure, I can laugh,” and then laugh. With these changes, Alexa will never be a disembodied phantom chuckler again.
That fix probably sounds absurd: Amazon’s solution is to make conversation with Alexa more awkward and unnatural, rather than less? But maybe Amazon is actually onto something here. Because Alexa should have never been designed to laugh on command in the first place.
Silicon Valley has been obsessed with making its chatbots feel like our human friends rather than brands run by corporate AIs. The problem is that, while chatbots understand our spoken words with incredible accuracy, they are eons away from understanding the real subtleties of human interaction. So companies like Microsoft, Amazon, Apple, and Google have all invested incredible resources in building puns and jokes into Cortana, Alexa, Siri, and Google Home. The goal? To make this invasive, always-listening technology feel relatable. Like a friend.
These speakers are front-loaded with humor to mask their own shortcomings. Case in point: The Amazon Echo was trained to respond to “Alexa, can you laugh,” because it was a means to amuse a human who was bored and out of ideas as to what to do with their talking speaker. Never mind that you’d never, ever, ever ask a human to laugh–you’d try to make them laugh, instead.
But the Alexas of the world could never really know when to laugh, because they lack the social instinct of humans. Instead, Alexa associates with people by blindly keyword-searching a database for the right answer or punchline. Alexa only ever knows to laugh because she hears “laugh.”
Alexa’s recent case of the “Ha Ha Has” underlines the uncanny valley between Amazon’s Alexa and any real human. It’s why Star Trek poked at the inabilities of the logic-based aliens known as Spock and Data to understand humor. Humor is an insanely complicated topic considered intrinsic to human evolution. It’s why knowing if to laugh and when to laugh, even with a close friend, can still be challenging at times for any of us!
In 10 or 20 years, maybe chatbots will master these social graces. But until they do, Amazon could have avoided its weird laughing problem entirely simply by positioning Alexa’s own personality as a computer you can talk to, rather than a cyborg bestie. Simply put, Alexa should be an “it” rather than a “she” because Alexa is nowhere near human, and bound to let us down again (or scare us) when we expect it to act like us. If Alexa weren’t ever trained to laugh, it would have never laughed like some soulless creep in the first place.
Lying in bed about to fall asleep when Alexa on my Amazon Echo Dot lets out a very loud and creepy laugh… there’s a good chance I get murdered tonight.
— Gavin Hightower (@GavinHightower) February 26, 2018
In a moment of honesty, Alexa laughed aloud to no one–offering an accidental glimpse of the business model sitting just behind her personality. Alexa is not here to be your friend. Ha. Ha. Ha.