Posted by Jeevan Jayaprakash

Image source: ExtremeTech

In this issue, we look at what is at stake for Amazon with Alexa, the update to the TensorFlow machine learning library and its implications as well as Ofcom’s plans for the rollout of 5G in the UK.

We love user feedback, so if our digest is too long, short, amazing or boring — just let us know at

Have a great week,

Hi Mum! Said Dad

Just how much does Alexa matter to Amazon? The answer is very.

In the last digest, we talked about how Alexa is aiming to become the operating system of the home and so far, it is making a real fist of it but how did Amazon — an e-commerce company (remember?) — get here and how integral is this play to the company as a whole?

To answer that question, we need to take a little trip down memory lane. Back in 2014, Amazon was forced to write off $170m as its foray into the highly competitive smartphone market with the Amazon Fire phone failed. This was a crushing blow to Amazon who had harboured hopes of being able to reach customers directly without having to go through an intermediary such as an Apple or a Google. The Fire phone was meant to be the vehicle for achieving this but it blew up in their face. However, only two weeks later, the world was introduced to the Echo speaker complete with what can now be called a relatively primitive Alexa.

With the pain of the Fire phone still so raw, Amazon has thrown everything behind Alexa. In order to accelerate adoption, Amazon deliberately sells Alexa devices such as the Echo and the Dot at a loss and is said to price them at 10–20% less than the cost of production. Investment banking firm, Evercore, estimates that Amazon made a loss of $330m on Alexa in 2016 and this is expected to double in 2017. The costs are only set to increase as Amazon is currently advertising >500 job openings for its Alexa team. Also, the brains behind Alexa, Alexa Voice Services, is given away for free to developers, allowing them to integrate it into any device that has a speaker and a microphone and this openness has been crucial to its advances as a platform over the last two years.

In its race to ‘own the customer’, Amazon has also managed to get one up on the competition with its recent deal to integrate Alexa into the dashboard of Ford cars. The rationale for the deal was simple — Ford was happy to partner with Amazon as they aren’t a threat to the car industry. The same cannot be said of Google and Apple, who are both developing their own self-driving cars that will inevitably run the Google Assistant and Siri out of the box, respectively. It seems that Google and Apple’s long and far reaching tentacles are proving to be a boon for Amazon in their pursuit of penetration.

Penetration is all well and good but ultimately, Alexa needs to pay off for Amazon and its investors. The focus will surely have to shift to monetisation and demonstrating an ROI for Alexa soon.

The Christmas period was particularly lucrative for Amazon. Image source: Financial Times

Google updates its machine learning library; opening the door for mobile

Google has announced a release candidate for version 1.0 of Tensorflow, its open source machine learning library. However, the most interesting change is the inclusion of an experimental compiler for linear algebra called XLA (Accelerated Linear Algebra).

So, what the hell does that mean and why does it matter?

The XLA comes with improvements in memory usage and will also speed up the rate at which mathematical computations are performed by TensorFlow, producing machine code (the 0s and 1s that a computer actually understands) that can either run on a CPU or GPU, i.e. mobile devices. This is significant because machine learning has typically required very powerful computers or a 24/7 cloud backend. When you couple this with the fact that mobile devices are getting more powerful year on year anyway, this opens up a whole array of exciting possibilities.

Machine learning on mobile could herald a new generation of applications that can understand, learn from and react to consumers and the environment around them without requiring an internet connection — i.e. ‘on-device machine learning’. For example, we could see indoor navigation apps that use computer vision to track and guide users in places like museums, airports and high-street stores without requiring the use of beacons. We could also see applications that can diagnose skin-related health conditions through a smartphone’s camera or via the analysis of photos.

Google’s latest Android Wear 2.0 wearables will be the first devices to benefit from the improvements to Tensorflow. Wear 2.0 wearables will be able to use its own internal hardware to undertake natural language processing and suggest one tap ‘smart replies’ for any incoming message.

Real-time image recognition via a smartphone’s camera (values indicate the probabilities of being said object). Image source: J Alammar’s GitHub

Ofcom sets out plans for the launch of 5G in
the UK

The UK telecoms market regulator, Ofcom, has released a report entitled ‘Update on 5G spectrum in the UK’ in which it has outlined its plans for the awarding of licenses for the 5G spectrum. The report states that early trials will begin in 2018 (the actual rollout is expected to happen in 2020) and Ofcom has stated that one of its key objectives is to ensure “that the spectrum is made available in the most appropriate and timely way to enable investments, innovation and competition in the development of 5G services to benefit consumers and businesses”. The Chancellor of the Exchequer, Philip Hammond, has gone on record saying that he wants the UK to be a “world leader” in 5G alongside the likes of South Korea and China, who are both expected to launch 5G in 2019 — a year earlier than the UK.

So, why is this significant?

5G will bring greater speeds (multi-gigabits per second) and significantly improved bandwidth, which will be an enabler for technology such as the Internet of Things (IoT), VR/AR and Drones. Indeed, Ofcom said the following with respect to IoT in its report: “5G should help the evolution of IoT services and applications and improve interaction between different platforms. Possible future applications could include real-time health monitoring of patients; optimisation of street lighting to suit the weather or traffic; environmental monitoring and smart agriculture”.

The notion of this connected, smart world full of things like self-driving cars and drones has been floated for a number of years now. The technology itself is showing promising signs but the constraint has always been the lack of ubiquitous, high-speed internet. The arrival of 5G promises to be as transformative as 3G and 4G, if not more so — potentially ushering in a world that could feel very different to where we were are now.

The promise of 5G. Image source: Ofcom

Originally written as part of Hi Mum! Said Dad’s Weekly Digest.

If you would like to receive our Weekly Digest straight to your inbox, please drop us an email at