Year of the machines: 2017 will be a breakthrough year for machine learning

19 Jan 2017

Machine learning is to grow beyond smartphones to land in drones, and AR and VR devices. Image: Sarah Holmlund/Shutterstock

2017 will be a breakthrough year for machine learning, with capabilities evolving from smartphones to also arrive in drones, tablets, cars, internet of things, and AR and VR devices.

According to Deloitte’s Technology, Media and Telecommunications (TMT) predictions, more than 300m smartphones – or more than one-fifth of phones sold globally in 2017 – will come with machine learning capabilities.

But in addition to smartphones, other mobile devices will be able to perform machine learning tasks even without connectivity.

Machine learning capabilities are forecast to arrive in tens of millions of drones, tablets, cars, VR or AR devices, medical tools, internet of things (IoT) devices and unforeseen new technologies.

“Machine learning is fascinating as it will revolutionise how we conduct simple tasks like translating content, but it also has major security and health consequences that can improve societies around the world,” said Richard Howard, TMT partner at Deloitte Ireland.

“For example, mobile machine learning is a strong entry point to improve responses to disaster relief, help save lives with autonomous vehicles and even turn the tide against the growing wave of cyberattacks.”

ITaaS to take 35pc of IT spend

As well as the evolution of machine learning, Deloitte predicts that by the end of 2018, spending on IT as a Service (ITaaS) for data centres, software and services will reach in excess of €520bn worldwide, up from €344bn in 2016.

Although flexible consumption-based business models will not be ubiquitous by 2018, at over a third of all IT spending (35pc), they’re expected to reach almost half a trillion euro and grow rapidly. This shift will begin to transform how the IT industry markets, sells and buys technology across businesses worldwide.

“While Ireland is ahead of the curve in the adoption of IT as a Service, we expect to see [this] increase at the large enterprise level, where adoption to date has been behind the industry average,” Howard said.

“While Ireland has benefited from the need for increased global storage capacity in the form of data centres, we are seeing the indigenous tech sector benefiting from ancillary services such as analytics and security software.”

5G road opens up

The fifth generation of cellular networks – 5G – is also set to be transformed in 2017 due to planned upgrades to 4G networks.

The first limited 5G deployments should acquaint users and operators with several of the most important features of 5G networks, including significantly higher speeds, lower latency, and support for low-power low-bitrate IoT devices and sensors.

While 5G will provide multi-gigabyte download speeds, the cost of upgrading networks and the need for a larger number of mobile cells may mean the technology will be deployed in large urban areas initially.

Security in the terabit era

2017 will also be a pivotal year for cybersecurity as distributed denial of service (DDoS) attacks become larger in scale and harder to mitigate.

Last year saw DDoS attacks occur at an unprecedented scale, including attacks harnessing IoT devices on Deutsche Telekom, data centre firm OVH and US infosec journalist Brian Krebs of Krebs on Security.

But these are just the tip of the iceberg.

Deloitte is expecting attacks to reach an average scale of one terabit per second each month, more than 10m attacks in total, and an average attack size of between 1.25 and 1.5 gigabits per second.

This escalation in the DDoS threat is largely due to the growing number of IoT devices; online availability of malware methodologies, which allow relatively unskilled attackers to corral insecure IoT devices and use them to launch attacks; and access to ever higher bandwidth speeds.

Deloitte also predicts that the active base of fingerprint reader-equipped devices will likely top 1bn for the first time in early 2017, with each active sensor used an average of 30 times a day, implying over 10trn aggregate presses globally over the year.

With the rapid pace of access and adoption of this technology, the challenge is to determine which additional applications could use fingerprint readers and other biometric inputs to provide rapid and secure authentication.

John Kennedy is a journalist who served as editor of Silicon Republic for 17 years

editorial@siliconrepublic.com