Usually machine learning is discussed in terms of its impact on marketing or enterprise data, but its next frontier could be considerably more mobile, according to a research director with consulting firm Deloitte Touche Tohmatsu Ltd.
In fact, in its annual technology, media, and telecommunications predictions for 2017, released this month, Deloitte’s global arm has predicted that some 300 million smartphones – more than one fifth of all units sold – will incorporate machine learning, Duncan Stewart, director of technology, media, and telecommunications research for Deloitte Canada, tells ITBusiness.ca.
The present numbers are being driven by smartphone chip manufacturer Qualcomm Inc.’s latest Snapdragon processors, Stewart notes, but Apple is rumoured to be developing a similar technology for their next iPhone too.
“It’s a boon for smartphone users,” he says. “It means you get results faster, use less data, and can even complete certain tasks without being connected to a network.”
More importantly, he says, the increased use of machine learning technology in smartphones will result in enhanced security.
“If your voice data, pictures, and location never leave your phone, it can’t be intercepted,” Stewart says, noting that for the time being users shouldn’t expect their phones to complete overly complex tasks such as translating a series of sentences written in a foreign language into English.
“Obviously if you’re doing something complex, like translating a passage from English to Russian, that you still need access to the Internet,” he says. “But a word or two at a time? Maybe. Recognizing your face versus somebody else’s? For sure. Possible even some limited transcription functions using Google Maps offline.”
More importantly, he says, even complex tasks will soon be utilizing less data than they have in the past – and are only likely to become more efficient from here.
For an example, he cited the GPS-free drones that Qualcomm showed off at this year’s CES.
— Recode (@Recode) January 6, 2017
“Think about the robots currently sorting items in our warehouses or cleaning our floors – they’re still pretty dumb,” Stewart says. “But as machine-learning chips become cheaper and more widely deployed, they’ll become more intelligent, to the point where they won’t just be replacing online tasks, but the number of humans needed to supervise them.”