With each passing handset release and software upgrade, smartphones get closer to earning their names in truth.
Chances are if you are somewhat up-to-date that your mobile device learns from you, making adjustments along the way to enhance your user experience. However, software can lag sometimes in those situations, requiring connectivity to data centers to provide services that are enhanced by machine learning.
That’s because it still relies on the old computer mainframe model, according to Duncan Stewart, director of technology, media and telecommunications research at Deloitte Canada.
“I was on a hike in the desert last week and there was no cell signal,” he said. “I just wanted to jot something down… my voice recognition didn’t work. It couldn’t transcribe. Why? Because it needs machine learning on board.”
So Stewart had to do it the old-fashioned way: by typing his message with his fingers.
But as the Deloitte Canada tech expert notes in a recent report, that kind of situation will become less common this year.
In his company’s just-released 2017 Canadian Technology, Media & Telecommunications Predictions, in-device machine learning plays a big role in the immediately significant advances of our technology world.
“The hottest device trend is mobile, and the hottest software trend is machine learning so if I’m combining the two of them, it’s going to be really important,” said Stewart, who is co-author of the report.
Machine learning will increasingly take place on devices, away from the data centers it relies on now. That’s similar to the shift that started decades ago—where personal computers, laptops and then smartphones themselves moved computing memory and processing closer to the user.
Deloitte predicts that over the next year in Canada “over 3 million smartphones, or over a third of phones sold in the year will have machine learning capabilities within the device.”
And that is a paradigm shift, according to Stewart.
“We’re actually pushing the intelligence to the edge of the network and that means it’s faster, it’s more accurate, it’s more secure,” he said.
And that’s just the smartphone implications of on-board machine learning. In less ubiquitous devices there is also great potential.
When the technology is rolled out to Internet of Things devices and video cameras, on-board machine learning will help detect takeovers and unauthorized actions from hackers, making everything more secure.
According to another of the report’s predictions, that’s an important feature as we enter the Terabit Era of cyber attacks. After two of these massive attacks occurred late last year, there are more to come in 2017, Stewart said.
The Deloitte report predicts 2017 will see an average of one attack a month on the terabit per second (Tbit/s) scale (and over 10 million lesser attacks in total, with an average size of 1.25 to 1.5 Gbit/s).
Stewart lists three factors that form what he calls the “perfect storm” when they come together.
First, there’s the accessibility of botnet software that can be used to launch distributed denial-of-service attacks.
“A kid in a garage could download the software from the web,” Stewart said.
Then there’s the number of internet-connected devices with higher bandwidth capabilities that can be hijacked and used by the software in the DDoS assault.
“An attack run by an army of toasters is not as serious as an attack run by an army of video cameras,” said the nine-year Deloitte tech vet.
Add to that, an ever-widening pipeline of internet delivery, and those perfect storms start to loom more frequently.
“If [a device] is sitting at home and it’s been compromised and you’re connected over dial-up, it can’t do much damage,” Stewart said. “If you’re connected over 100 Mbps pipe upload, all of a sudden that compromised device is capable of flooding a third party site with malicious traffic.”
So on-board machine learning can help nip attacks in the bud at arguably the only controllable points—our unwitting, vulnerable devices.
Looking further into the future and at different hardware, that same technology will help move digital navigation to its final frontier: indoors.
When machine learning is put on drones, they can use machine learning to recognize objects indoors.
“Why does that matter so much?” Stewart asks. “Because when you’re inside a warehouse flying around, you don’t want to be able to have to do everything through the cloud—it takes too long, it’s too slow. You really couldn’t do it without having it on board.”
Five years from now, Deloitte says no less than 25 per cent of precision digital navigation “will include an indoor portion or be for an entirely indoor journey, compared to less than 10 per cent of all uses in 2017.”
Locating objects and people indoors will be “transformative” and will benefit companies, governments and consumers, according to the predictions.
On a similar, but more important note, on-board machine learning will cut motor vehicle fatalities by 16 per cent by 2022, says the report.
At that point automatic emergency braking will be affordable and widely implemented in cars, so technology can stop a car before a human could react. With fast moving objects and so much at stake, there is no room for latency so machine learning must happen on board vehicles, as sensors scan roads for upcoming objects.
“We could see the adoption of autonomous vehicles occur more slowly than expected, as automatic braking technologies provide an alternative option for Canadians who are attracted to the increased safety that they offer but also still desire to control and operate their own vehicles,” Stewart said.
If trusting technology with their steering wheels is a slower transition for wary consumers, two of the Deloitte report’s predictions reveal that people are opening up their lives to tech in other ways.
Most surprisingly, the resistance to widespread adoption of biometric security is loosening.
Last year, 30% of Canadians had mobile devices with fingerprint readers and 76% of those actually use the feature, says the report.
“And they use them over and over and over again,” Stewart said. “It’s how we turn our phones on, and we turn our phones on 30, 40, 50 times a day.”
“Once you get used to that idea that fingerprints are a good, reliable, accurate, safe way of turning on your phone, why wouldn’t you use it to authenticate a payment or other transactions?”
Looking forward, Deloitte says the challenge is to see how these inputs can be used securely and quickly to deliver proof of identity in other ways.
With the password now accepted as a broken method, there is an authentication void to be filled.
“The fingerprint reader sits on a device we carry with us all the time, that we try always to have with us and if it’s gone we know instantly,” Stewart said. “That actually makes a fingerprint reader on a smart phone close to the perfect authentication tool.”
Fingerprints are hard to fake, we always know where our handsets are, and the devices are usually connected to a network.
The technology that society is already gaining trust for allows us to do multi-factor authentication on a single device.
And another confidence-based trend the Deloitte report highlights is the adoption of IT-as-a-service.
“The traditional way of running a company was that you would buy computers for all your employees, then you would buy software for all your employees and load that software on the machine and they would work away quite happily,” Stewart said.
That was all fine and great. But what if you need to downsize, or double the number of employees that work at your firm?
A scalable model where hardware and software are leased or rented means a more flexible business.
“If you want to hire another 5,000 people using this software, somebody flicks a switch,” Stewart said. “It’s a really elegant model… we’re seeing more and more of this. It’s all enabled by the cloud.”
By the end of next year, Deloitte forecasts over a third of IT spending will be on such “flexible consumption-based business models”, with more growth to come.
“It’s really about convincing the holdouts that this model works,” Stewart said.
Other predictions of the 16th annual Deloitte TMT report:
- 5G networks begin deployment in 2017, providing faster speeds and the longer-term possibility of wireless households.
- Peak tablet demand is surpassed, with consumers favouring laptops and smartphones.
- Vinyl approaches $1 billion sales globally in 2017, but remains “very niche.”
As we leave 2016 behind, it appears the near future of technology will see machines learning with more autonomy, and humans learning to trust in the new opportunities they provide.