A Morning with Deloitte: TMT Predictions 2018

On the recommendation of my former colleague Dan Deeth, I spent this morning down at the Tannery listening to Deloitte’s Duncan Stewart (@dunstewart) prognosticate on Technology, Media, and Telecommunications.

Dan was unable to attend, as he was in Barcelona this week at Mobile World Congress, and he tells me that Duncan’s been a long-time reader of the Global Internet Phenomena reports that Dan and I wrote. So hey, why not make the fifteen-minute drive to Kitchener?

Coincidentally, the timing comes on the heels of several posts on this blog about the challenges of and best practices for prognostication, so I was in the right frame of mind – skeptical, but interested – for this event.

Of course, I took notes.

From his corporate bio, “Duncan is the Director of Technology, Media and Telecommunications Research for Deloitte Canada. He is a globally recognized speaker and expert on the forecasting of consumer and enterprise technology, and media and telecommunications trends. Working worldwide and across all industries, Duncan advises clients on the impact of new and existing technology, and demographic and regulatory changes to their business strategies.”

Absent from the bio, but arguably more interesting: he was once a contestant on Jeopardy, where he “finished in the top three”.

(I’m not going to write about everything in the report – just go to the link above and download it – but I’ll write about Duncan’s commentary and will probably add my own)

Duncan started off by showing a report card that graded last year’s predictions, which was a nice surprise! I’ve sat through many session like this one, and read many reports that make predictions, but rarely (if ever?) have I seen the prognosticator actually show his own performance. Keeping score is really the only way to know when and how to refine one’s ability, so things were off to a good start.

I was also pleased to discover that Duncan is an engaging, energetic, concise speaker, who certainly appears to have deeper knowledge/familiarity of the subjects than the slides alone. For the audience, this meant that simple points on the slides were backed up with solid explanations delivered in an entertaining manner, and we got more than we would have from only reading the report.

Things That Didn’t Quite Make It

Duncan spent a couple of minutes quickly running through things that had lots of hype, but didn’t quite make it. Beyond the tactical achievement of knocking these things off in a manner that explained why they aren’t covered in detail in the report, his explanations also provided good indications of the evidence that is used to determine what is real and what is hype:

• eReaders: remember those?
• 3D TVs: no one makes them anymore
• Fitness bands: the manufacturers are struggling, engaging in massive lay-offs; “the high-tech equivalent of the January gym membership”
• Smart devices in the home (e.g., smart kettles, coffee machines, rice cookers, sous vides, etc.): people have realized that it’s easier to just flip the switch than to unlock your phone, open the app, etc. to turn the thing on
• Consumer/home 3D printers: manufacturers are abandoning/have abandoned this space, in favour of enterprise
• Consumer/home drones: similar situation as 3D printing
• VR Headsets: a Duncan Stewart rule of technology is, “people don’t like tying computers to their face”; gamers will do it, sure, but at the mass consumer-level VR is not an important technology

What many tech folks forget is that we aren’t representative of the general population.

I expect some in the audience were a little surprised by some of these claims – after all, we live in a hotbed of technology, and many might have careers that depend on some of the above. However, what many tech folks forget is that we (e.g., this region, these demographics) aren’t representative of the general population. I feel like, in January 2017, half my office had fitness bands…and many are into home automation, and so on, but Duncan’s predictions are based on mainstream demographics, and that shifts things.

General Trends

Next, Duncan flashed up Gartner’s July 2017 Hype Cycle.

The main commentary here is that, if you look carefully, you’ll notice that there are no grey circles on the cycle: i.e., nothing listed by Gartner is expected to reach the “Plateau of Productivity” within the next two years.

Conclusion? “The technology world of 2020 and 2018 are going to look almost exactly alike.”

“The technology world of 2020 and 2018 are going to look almost exactly alike.”

Again, that might be a controversial, or at least a counter-intuitive, statement…but again, it’s one that stands up to a bit of scrutiny.

In part to support this assertion, and in part to continue the narrative, Duncan reviewed a few important technology markets. I’ll just summarize briefly:

• Global Consumer Technology Spending: Citing data from the Consumer Technology Association (the group that runs the Consumer Electronics Show), Duncan showed that spending on consumer electronics peaked in 2013 (at $1.045 trillion) and has been declining fairly steadily since (to $929 billion in 2017). One take-away is that people are shifting their spend from hardware to software and services.
• How is that money spent? The big three of Smartphones, Computers (PCs and Laptops) and TV Sets account for 81%. Paraphrasing Duncan, these are the devices where the eyeballs and the people and the money are…where people spend two-to-five hours per day.

With that introductory and general material out of the way, we were ready to dive into the specific items.

Augmented Reality

Augmented reality’s received a great deal of hype, even if the halcyon days of Pokemon Go are long gone. Ha, remember Pokemon Go?

Despite the promise (OK, potential) of augmented reality to impact our day-to-day lives, the applications with which most people are familiar: adding ears, tongues, noses, and other appendages to videos.

Hardly groundbreaking, revolutionary stuff.

But Duncan reminded us that seemingly trivial use cases are still a way for technology to gain a foothold – he invited his audience to recall that, for many, Pong contained their household’s first computer chip. Although, looking at the room’s demographics, only about half would’ve been in a position to actually remember that.

Next, he showed IKEA’s impressive augmented reality app, which lets users picture and plan furniture purchases; he followed that up by showing IKEA’s pretty good augmented reality app from 2013. OK, so if IKEA’s had a good enough app for five years, then why hasn’t it taken off? Well, it turns out that most people want to experience physical furniture before buying it.

So augmented reality – while still full of promise – is still in its early days. It’s a few killer apps and a few years away from becoming an everyday thing.

So augmented reality – while still full of promise – is still in its early days. It’s a few killer apps and a few years away from becoming an everyday thing.

Smartphones

In Duncan’s view, hardware innovations are really running out: “Going forward, our phones are going to look – from the outside – almost exactly the same from now until forever.”

Instead, we’ll get ‘invisible innovation’, meaning that phones will continue to improve, but internally.

Next, Duncan showed that smartphones – on average – are violating a rule of technology that it always gets cheaper over time. The average selling price of phones actually bottomed-out in 2015, at $305 globally. Things have crept up since.

Why? Mostly, it’s because people are willing to pay gobs of money for phones, and are willing to wait to make the upgrade until that magical new phone is released.

Finally, Duncan talked about “the enterprisation of smartphones”, which is a poetic little successor to “the consumerisation of IT”. Smartphones are becoming increasingly enterprise-capable, and that will have a real impact on how people do their jobs in future.

Machine Learning

And now onto probably my favourite topic from the day. Context: I’m a computer engineer (by training) who’s dabbled in artificial intelligence…so this one was right up my alley as it was the most technical.

Duncan pointed out that much of the hype/conversation about machine learning focuses on the software; instead, he thinks people should be talking more about the hardware chips, specifically FPGAs and ASICs.

In 1943, a human neural network was modelled with electrical circuits for the first time. Yeah, I know…wow.

Since then, there’s been a very slow – albeit significant – advance in the state of the art. However, even as recently as the 1990s and early 2000s, machine learning via neural networks, and deep learning in general, was seen as a bit of a dead end. Even with some pretty impressive achievements, the computational and data needs for continued and general-purpose advancement just weren’t being met by the technology that was available.

However, around 2009, folks started to employ GPUs, rather than CPUs, as the platform of choice for machine learning…and there was something like a 100-1000x increase in performance.

Next, the era of Big Data arrived, so these information-hungry machines finally got the huge amounts of data they needed to learn, and the machine learning really took off on the trajectory we see today.

As a specific example, Duncan cited DeepMind’s WaveNet (yeah, the same folks as AlphaGo).

Machine learning has a bright future, owing to ongoing developments/advancements in:

• Hardware: CPUs to GPUs delivered several orders of magnitude of performance improvement, and things are poised to deliver a similar result as FPGAs and ASICs (like Google’s TPU) become commonplace
• Automation: automate routine tasks, so data scientists can focus on innovation
• Data Reduction: using approaches like synthetic data to reduce the raw data needs of machine learning systems
• Model Interpretability: the ability to peer inside the ‘black box’ of machine decisions is fundamental in many areas, as we try to fight the bias that arises from a reliance on (unintentionally) biased data sources and emulations of human decision-making; for more on the risks associated with black box decisions, I strongly recommend you check out Weapons of Math Destruction; Duncan’s view is that advances in model interpretability – that is, our ability to understand how the model works – will lead to wider applications of machine decisions in fields where the risk is currently too high (e.g., banking, insurance, etc.)
• Local Machine Learning: for instance, Apple’s A11 bionic chip includes neural network hardware, so machine learning capabilities are being built into devices rather than needing to rely on access to the cloud; as chips advance, getting faster and with amazing power performance, the potential applications are mind-blowing (e.g., piezoelectrics, implanted devices, etc.)

Duncan’s analysis shows that large- and medium-enterprises have doubled their use of machine learning in the past year, and will double again before 2020. Which companies? All of them. Which functions within the company? All of them.

Machine learning has become – or, at the least, is on the cusp of becoming – general purpose.

Duncan’s analysis shows that large- and medium-enterprises have doubled their use of machine learning in the past year, and will double again before 2020. Which companies? All of them. Which functions within the company? All of them.

Q & A

There were a few other focal topics, but nothing came up that you can’t get out of the report…so onto the Q and A. Stuff in quotes are me paraphrasing Duncan’s reply; non-quotes are my editorializations:

• What about self-driving cars? Duncan’s reply cited the different levels of automation and the relative challenge of achieving each. “Still a very long way to go before they’re on the road in sufficient numbers as to change society. Like, 2035.”
• What about 5G? Ah, 5G. I wrote a whitepaper on this topic. “There aren’t that many applications that need millisecond latency and high-bandwidth. Early 2020s for major impact, although as an alternative in rural areas and as an alternative to laying wire there’s promise for immediate use.”
• Technophobia / people getting worried about technology? Not a big deal. “People like the low-tech solution of just putting the phone away.”
• Electric vehicles? “There are no current battery technologies that will be available within a decade that will materially change the range, durability, and price of electric vehicles” If people could get over their damned range anxiety, then things would be fine!
• Smart cities? “A lousy buzzword, in that it means so many things to so many different people. Smart homes? Smart sewers? Smart transportation infrastructure? Smart public transport? Smart energy? Many questions, each with a different answer.”
• Hyperloop? More hype than loop. [Update: I’ve seen a few people take and share this comment, so let me add that Duncan’s full/actual response praised Elon Musk’s capabilities and contributions, while also acknowledging his obvious skills with showmanship and PR; ultimately, Duncan downplayed the immediacy of the hyperloop and pointed at a few engineering challenges that remain to be convincingly solved at large scale].

And that about does it for today.

Also, yes – amazingly enough, we went the whole session without any mention of blockchain, cryptocurrency, or quantum computing…which, wow.

My thanks to Duncan for a compelling session and for supporting his predictions with more evidence than one usually gets in a pontification presentation.

If you were at the event, I’d love to hear your thoughts!

About Lee Brooks

Lee Brooks is the founder of Cromulent Marketing, a boutique marketing agency specializing in crafting messaging, creating content, and managing public relations for B2B technology companies.