‘Alexa, did you hear a murder?’ Privacy, patience, precision main hurdles for voice technology

Although a U.S. murder case has raised questions about the lack of privacy involving voice activated devices, one Toronto expert says a lack of consumer patience has far more potential to kill the technology’s widespread adoption.

This week Amazon ceded to requests from Arkansas police to hand over sound data recorded by one of its Echo devices in the home James Bates, who faces first degree murder charges in the death of Victor Collins. Police want to see if the data contains audio evidence recorded just before or during the time Collins died in Bates’s house.

That followed the latest WikiLeaks documents purportedly detailing how the Central Intelligence Agency (CIA) can spy on people by listening to them through their smart TVs.

An Arkansas murder case involves sound data recorded by an Amazon Echo device similar to these. (Photo: Amazon)

An American woman fuelled further fears by posting a video combining elements of the other two stories. When she asks Amazon’s Alexa, “Are you connected to the CIA?” her device never answers the question and repeatedly shuts down.

None of these issues surrounding privacy and speech recognition are new to John Wang, who leads the global digital and core banking practice at Deloitte Canada. Whenever such devices are used in close proximity to other people, “there’s no expectation of full privacy because someone can hear you,” he told a recent panel discussion at Toronto’s DX3 conference.

“Voice (technology) has a particular promise around accessibility and convenience. But if I’m doing something private, it may not be voice that I go to,” Wang said.

Survey says …

Privacy concerns are already affecting the way consumers use the technology. When Recode surveyed Siri and OK Google users last summer, only six per cent said they “commonly” use the systems in public areas. The most popular locations – in the car (cited by 51 per cent) and at home (named by 39 per cent) – offer users far more privacy.

Eventually, Wang said, manufacturers will have to bake more privacy features into the devices to ease consumer worries and drive faster uptake.

“There has to be some promise of privacy for full adoption because the idea that something listens to you all the time is kind of creepy.”

Precision trumps privacy

Precision – the ability of devices to accurately understand spoken words – will be an even bigger obstacle than privacy on the path to quick, widespread adoption, said another panel member.

“Especially with voice-to-text (translation), we’re not there yet and people just don’t have the patience for it,” said Greg Apple, head of marketing at voice enabled travel platform HelloGbye.

He said younger people – who care less about privacy than older generations – are more likely to abandon voice systems that don’t work exactly the way they want them to immediately.

Kinu Masaki (SmartEar), Greg Apple (HelloGbye) and John Wang (Deloitte Canada) at DX3 in Toronto.

Fellow panelist Kinu Masaki, CEO of SmartEar, agreed the technology has to work better to win over more users.

“One of the biggest hurdles is that speech recognition is very inaccurate, especially in a very noisy environment,” said Masaki, whose San Francisco firm develops smart earpiece wearables for voice-based applications.

In the same Recode survey, a significant portion of users expressed some frustration with how well voice-activated systems interpret spoken words. Forty-three per cent of those polled said they would use their device more often if they could “speak to it more naturally.”

To address that, Wang said developers must focus on the artificial intelligence (AI) side of the technology by improving its cognitive computing and natural language processing.

Misunderstandings

That could help ward off mishaps like the one last December, when Amazon’s Echo Dot device misunderstood a little boy’s request for a children’s song. The unit answered by reciting a list of pornographic phrases before the kid’s horrified parents screamed at Alexa to stop.

This boy mistakenly got an earful of inappropriate language when he asked Alexa to play him a song last December. (Image: YouTube)

Amazon is reportedly working on one type of AI improvement right now. According to Time magazine, the company is developing a way for Alexa to distinguish between specific voices. At the moment, it responds to queries and requests regardless of who utters them.

Voice recognition I.D. would add a layer of biometric security to systems like Alexa, Siri, Cortana and OK Google by authenticating each user’s identity based on the unique way they sound.

Even security hawks like banks are getting into that type of voice application. On March 7, Canadian bank RBC added Siri-activated e-transfers to the features in its mobile app.

Would you recommend this article?

Share

Thanks for taking the time to let us know what you think of this article!
We'd love to hear your opinion about this or any other story you read in our publication.


Jim Love, Chief Content Officer, IT World Canada

Featured Download

Christine Wong
Christine Wonghttp://www.itbusiness.ca
Christine Wong has been an on-air reporter for a national daily show on Rogers TV and at High Tech TV, a weekly news magazine on CTV's Ottawa affiliate. She was also an associate producer at Report On Business Television (now called BNN) and CBC's The Hour With George Stroumboulopoulos. As an associate producer at Slice TV, she helped launch two national daily talk shows, The Mom Show and Three Takes. Recently, she was a Staff Writer at ITBusiness.ca and is now a freelance contributor.

Featured Story

How the CTO can Maintain Cloud Momentum Across the Enterprise

Embracing cloud is easy for some individuals. But embedding widespread cloud adoption at the enterprise level is...

Related Tech News

Get ITBusiness Delivered

Our experienced team of journalists brings you engaging content targeted to IT professionals and line-of-business executives delivered directly to your inbox.

Featured Tech Jobs