Commentary: Smart speakers will take over the world? Don't make me laugh

Commentary: Smart speakers will take over the world? Don't make me laugh

Most artificial intelligence devices are unable to respond accurately to even the most basic human commands, one Financial Times observer points out.

NEW YORK: For a few days last week, it really seemed as though the robot uprising might be upon us. 

Owners of Amazon’s smart speakers reported spontaneous bouts of creepy giggling from their Echoes. Uninvoked, it seemed, Alexa was laughing at us - we who are supposed to be her masters.

“Lying in bed about to fall asleep when Alexa on my Amazon Echo Dot lets out a very loud and creepy laugh,” tweeted @GavinHightower. “There’s a good chance I get murdered tonight.”

One Reddit user described Alexa’s laugh as “witchlike”, while another said it was “chillingly creepy”.

Right on cue, Elon Musk turned up on a panel at the South by Southwest film festival in Texas this weekend to ramp up the terror factor.

“I’m close to AI and it scares the hell out of me,” Mr Musk said. “The danger of artificial intelligence is much greater than the danger of nuclear warheads, by a lot.”


Amazon insists Alexa’s laugh is not a menacing prelude to mass destruction. It is just a big misunderstanding. Its virtual assistant was “mistakenly” hearing the phrase, “Alexa, laugh”, Amazon has said. 

To fix the problem, it is changing the verbal command to “Alexa, can you laugh?” and adding something extra to the response.

“Sure, I can laugh,” Alexa now says in flat and joyless tones, before exclaiming: “Tee hee!” I am not convinced this would be any less sinister if activated by accident.

I find myself shouting at and muting Alexa as often as it gives me the right answer.

The coverage of this bug reveals a genuine and recurring problem with virtual assistants, and Alexa in particular. Far from being on the brink of taking over the world, Alexa’s random laughter shows just how primitive AI helpers are.

They are still unable to respond accurately to many human commands that roam beyond their limited scripts. They can barely maintain more than a sentence or two of conversation. And they have an annoying habit of chipping into conversations uninvited.

Once, when my wife and I were discussing the privacy pitfalls of putting an internet-connected listening device in our home, our smart speaker interrupted us to say: “I’m listening.”

From left, the Apple HomePod, Google Home and Amazon's Alexa are displayed for a photograph in
Apple's HomePod (left), Google Home and Amazon's Echo (right) on display. (Jason Henry/The New York Times)


The problem became particularly acute when I tried to take AI into my car.

I was keen to try Alexa in the car to give me hands-free access to iPhone apps that Apple’s Siri does not talk to, such as the Waze navigation app, Spotify for music or Audible for audio books.

Spotify support is still listed as “available soon”, so I have not yet tried it, but Audible — which is also owned by Amazon — works really well. All I have to do is say, “Alexa, play my audio book” and it starts streaming from where I last left off on any other Audible app or Echo device.

However, I soon discovered the problem I had created for myself. With the audiobook playing, there were now a lot of words being spoken out loud in the car. It took Alexa, always so eager to believe that someone is talking to her, about 15 minutes to start babbling in response to an imagined invocation.

The trigger here is particularly annoying because after Alexa has stopped talking, she rewinds a few seconds of the book — then thinks she has heard her name from the same passage all over again. 


I was tempted to leave it muted. But then I would have had to reach down to the cigarette lighter every time I wanted to ask Alexa something, which seemed rather to obviate the point of hands-free control.

Being able to summon directions through any other navigation service than the still-limited Apple Maps is probably worth US$50 by itself. 

But it took me a long time to figure out where in the Alexa app I had to enter my home address in order to have me routed home, and almost as long to find the right form of words that would get me there.

It is hard to tell if the problem here is with the Alexa service itself or my microphone system. But as a user, the end result is the same — a device that I find myself shouting at and muting as often as it gives me the right answer. 

No wonder most people use their smart speakers for playing music and setting timers.


The advance of the virtual assistants will doubtless continue despite their shortcomings. Amazon this week announced that Alexa is now coming into our workplaces too. 

The real arms race is not between the robots and the humans but between Amazon, Google and Apple, who are too deeply invested to stop now.

The funny part is that the only way the assistants will improve is to use them more often — AI systems live or die on data. 

Yet while we continue to train them, the joke rather seems to be on us.

© 2018 The Financial Times Ltd. All rights reserved. Please do not copy and paste FT articles and redistribute by email or post to the web.

Source: Financial Times/sl