Category: Analysis

This week’s launch of Android Wear 2.0 and several new smartwatches designed to showcase the new operating system is a useful reminder of the state of the overall smartwatch market, which appears to have struggled outside of the Apple Watch. But it also serves as a useful prompt to consider the general state of the wearables market, which was supposed to be the hot thing just a few years ago but seems to be fizzling now. What is the true state of that market and where will it go from here?

Still Grading Smartwatches on a Curve

It’s illuminating to go back roughly two and a half years to mid-2014, when I wrote a piece about smartwatches being graded on a curve. Most of the smartwatches on offer then were poorly designed, had limited functionality and, consequently, sold in very small numbers. Yet they were generally given pretty decent ratings by the major gadget blogs anyway. At that point, smartwatches really weren’t selling and the reasons were fairly obvious.

Then along came the Apple Watch, instantly selling in much larger numbers than other smartwatches, taking half or more of the unit shipments and the vast majority of the revenues, as well as jumping into the top ranks of all watchmakers by revenue in its first year. I and a number of others had thought Apple’s entry would prompt Google and its OEM partners to up their game and compete more effectively, much as smartphone vendors did following the launch of the iPhone and tablet vendors did following the iPad launch but that hasn’t really happened. Android Wear continues to account for only a small minority of total smartwatch shipments and actually shrank last year. Samsung and Fitbit have taken the second and third slots in the market, mostly without help from Android Wear.

This week’s announcements tweak some of the software facets but they also continue the bombardment of new features which is often characteristic of the Android approach – throw as many radios and other features into the device as possible and don’t worry about the effect on battery life or form factor. What results is a pair of watches from LG which offer either massive bulk and all the features except decent battery life or a slimmer profile but none of the new hardware features. Meanwhile, other Android Wear pioneers like Motorola are sitting this round out.

What we have today is not so much a smartwatch market but three distinct markets with three dominant players:

• The Apple Watch, which really exists in a niche of its own as a premium fashion device and focuses on health and fitness tracking but also does notifications and glanceable information well
• dedicated fitness trackers, a market increasingly dominated by Fitbit in most western markets
• Samsung’s various wearables, many of which are bundled and/or heavily discounted with a smartphone purchase and therefore thrive on a different business model from the rest.

Everyone else basically falls into the cracks between those three markets, falling short of the products offered by the three companies that dominate them — that includes Android Wear.

The Broader Wearables Market has Fallen Short

If you were to go back just a few years, you’d have heard about wearables at every industry conference and seen lots of ambitious forecasts for the category – I found one from 2104 which predicted well over 100 million unit shipments in 2016 including five million smart glasses, 11 million wearable headsets, 15 million wearable cameras, 38 million fitness trackers, and 46 million smartwatches. But, of course, the market was nowhere near that in 2016. There were perhaps half that many smartwatches sold, almost that many fitness trackers, maybe five million VR headsets, and almost no smart glasses.

Why has the broader wearables category not taken off as predicted? Ultimately, it comes down to the fact excitement for the category was driven almost entirely by vendors of gear and components and not at all by the market. The jobs to be done by these new wearables were far from clear and the offerings in the market did a poor job convincing anyone they were worth having. Smartwatches have largely failed because they have borrowed the use case of the smartphone without explaining why performing those tasks was better on a smaller, less powerful device.

The exceptions – those wearable categories that have performed well – are those which have articulated a use case and executed on it effectively. Arguably, the single biggest use case has been fitness tracking. Even the Apple Watch, which does much more, has been most compelling as a fitness tracker and that was reflected in last year’s fall event, which focused on these features more than anything else. But Apple isn’t just making glorified Fitbits – it has also recognized that, when it comes to a watch, people want the device to perform one of the other jobs a watch does for them: look good on the wrist with a variety of outfits. That means a premium device with premium materials and a variety of options for the finish, colors, and so on. Apple has fleshed out the value proposition with additional features like some of the best notification management on a smartwatch and good use of the rectangular screen with complications.

Where Future Growth Might Come From

Having said that, the market served by fitness trackers and premium fitness-centric digital watches is still relatively small. This was never going to be a smartphone-sized market but it’s a heck of a long way from a tablet-sized market at this point and it doesn’t seem to be growing very much. This raises the question of whether it can ever be any bigger and whether there are other categories of wearables that could still find success.

I still believe the smartwatch has potential beyond its current incarnation but I don’t think the technology is ready yet. One of LG’s new watches, and a number of existing ones, feature 3G or LTE connectivity that allows them to operate independently of a phone but the tradeoffs are much like those in early LTE smartphones – they drain the battery and add bulk for relatively little additional value. That will change as the technology matures and becomes both smaller and more battery efficient. The ongoing miniaturization enabled by smartphones will continue to drive other technologies to perform better at smaller sizes and smartwatch CPUs will become more powerful and battery efficient as well. As that happens, we could see more compelling app models on smartwatches, which could finally lead to the kind of innovation that helped smartphones take off and which has been largely missing in the smartwatch space so far.

Until then, though, the most promising wearables categories are those which extend the functionality of the smartphone in some way. Some time ago, I wrote about the fact the smartphone would be increasingly extended in various ways as input and output methods and processing tasks were outsourced to other devices and services. Smartwatches are an example of this, putting input and output permanently out in the open on the wrist rather than buried in a pocket and providing new ways to see notifications, trigger voice assistants, follow navigation directions, and even carry on phone calls. But this extension needn’t stop with smartwatches – the next generation of audio technologies will further extend the smartphone into the ears, providing both audio feedback and input via built-in microphones. If you own an Apple Watch and AirPods today, you can keep your iPhone in your pocket much more, relying on visual feedback on the Watch and audio feedback and input via the AirPods exclusively for some tasks. We’ll see a lot more of this.

When it comes to visual interfaces, augmented and virtual reality have quite a bit of promise but, for today, are mostly occasional-use technologies rather than true wearables, used throughout the day in the way smartwatches are. Virtual reality allows for a much more immersive display than a smartphone ever could and, if headsets weren’t so bulky, they’d make for an interesting way to carry room-scale displays with you everywhere you go as another extension of the smartphone. Augmented reality, especially in a form factor resembling normal glasses, has great promise too, but we’re likely several years from that future as well.

Again, those wearable categories that succeed will do so by solving real pain points or providing real additional value, not by merely sticking a set of smartphone-based components into a new form factor and slapping a logo on it. That approach has characterized too much of what we’ve seen in the wearables category so far and it won’t work any better in the future than it has in the past. But it’s going to be a slow build to meaningful numbers in all these categories, not a massive explosion along the lines of what we’ve seen in previous categories.

Regardless of where you sit on the political spectrum, the maelstrom created by the last US presidential election uncovered a painful reality for the tech industry: a striking gap between it and much of mainstream America.

It’s not that Americans of all socioeconomic levels aren’t using the output of the tech industry. From smartphones to social media and PCs to online shopping, US citizens are, of course, voracious consumers of all things tech.

The problem is a serious lack of empathy and understanding from people who work within the tech industry to those outside their rarified milieu. To its credit, the tech industry has created enormous amounts of wealth and many high-paying jobs. Very few of those jobs, however, are relevant or available to a large swath of the US population. While I haven’t seen any official breakdowns, I’m not aware of many middle income jobs (according to US Census statistics, the average US family income in 2015 was $55,755) in the tech industry. Heck, interns at big tech companies often get paid more than that.

Not surprisingly, that kind of income disparity is bound to create some resentment. Yes, on the one hand, the significantly higher salaries often found in tech jobs do make the goal of working in tech an attractive one for many who aspire to break into the field. But not everyone can (nor wants to) work in tech.

A functioning society, of course, requires people to work across a range of jobs and at a range of income levels. But, it does seem rather disconcerting that an industry that is responsible for driving so much growth across the economy, and that houses the most well-known and well-respected brands in the world, does so little to employ people at mainstream income levels. For all of its focus on social justice and other progressive concerns, the tech industry displays a rather shocking lack of interest in economic inclusivity, which is arguably at the very heart of a just society.[pullquote]For all of its focus on social justice and other progressive concerns, the tech industry displays a rather shocking lack of interest in economic inclusivity, which is arguably at the very heart of a just society.”[/pullquote]

Of course, fixing the problem isn’t easy. But it does seem like there are a few basic ideas that could help and a lot more “thinking different” that might be worth a try. For one thing, the fact that the tech industry notoriously outsources (or subcontracts) nearly every lower and middle-income job to another firm (all in the name of cost-cutting) needs to be re-examined. From bus drivers, to janitorial and security staff to, yes, manufacturing jobs, it’s high time to start making people who do work for a company, employees of that company, with all the rights and benefits that entails. Yes, it could negatively impact the bottom line (though, in the big scheme of things, not by very much), but it would be a tremendously positive step for many. All it takes is some fiscal stamina and a bit of guts.

In addition, the whole mindset of gig-based companies (such as Uber) needs to be reconsidered. Maybe the original intentions for generating a bit of extra income were good, but when millions of people start trying to build their lives around pay-for-hire work, it’s time to start making them the middle-income employees they’ve earned the right to be.

It’s also time to start thinking about packaging and selling technology-driven products in entirely new ways. There might be ways to start building entire new sub-economies around, for example, helping farmers grow their crops more efficiently through the use of sensors and other IoT-based technologies. In addition, building products or services that allow the creation of small businesses, such as a tech franchise, which could help other local small businesses with their tech devices and software. For example, someone who could help local bakers, restaurants, florists or shoe repair shops to run their businesses a bit more efficiently, but provides “door-to-door” service.

Part of the problem is that the tech industry has become so obsessed with only offering the latest, most feature-rich products and services through high-income jobs that they have lost sight of the fact that some people only need very simple “older” tech that could be delivered in a more modest manner through comparatively lower-paying jobs.

Rather than planning for a societal collapse, it’s time to start mapping out a more positive, productive future that links Silicon Valley to Main Street in a useful, meaningful way.

A few weeks ago, we decided to launch a US-based consumer study, focused on understanding how non-techie consumers think about both privacy and security. Our goal was to learn what consumers understand both these terms to mean, what core behavior changes do they make (if any) with products and services based on their privacy or security concerns, and which companies they trust more than others in both cases. Prior to launching the study, I looked at many different studies done by consulting companies, banks, and financial institutions, as well as government studies, to see what kind of questions had been asked before. I also spent some time interviewing consumers to hear how they talk about privacy and security when it came to different products and services. Even with all the prior work put in, this was still one of the hardest studies to get consumer participation. The difficulty of the subject matter itself, along with the high initial abandonment rate we suffered on the study, was a lesson in and of itself. We left comment boxes in certain places of the study and, quite frequently, consumers felt they weren’t adequately informed enough to participate, didn’t have strong opinions, didn’t want to think about it or just didn’t care. The open comments sections were some of the places we received the keenest insights into how consumers view these subjects. With with a few wording changes and adaptions to the study, we finally got enough people to complete it for it to be statistically representative.

Privacy and Security, Same or Different?
We broke the study out into two sections: one on privacy and the other on security. We asked consumers what they felt each term meant and left the same answer options for both questions. Below is the merged chart for both the privacy definition question and the security definition question.

As you can see, consumers felt the strongest definition of privacy was “Not selling personal data, or letting third parties access personal data.” When it came to security, consumers felt the strongest definition was “Secure and encrypt my data so no one can hack or steal it”. But, as our gut sense suggested, there is a fair amount of overlap in how consumers think about privacy and security with the same two answers being quite high in both questions. A core conclusion was, while privacy and security are two separate things, consumers tend to blend their understanding of them into the same definition. In the mind of the consumer, what is private is secure and what is secure is private.

Who Do You Trust With Your Privacy and Security?
Thanks to some other studies, I read quite a bit about how consumers trust things like government and financial institutions. We wanted to look at some of the bigger names in tech and social media as a start.

Apple and Microsoft were nearly neck and neck for the top spot of consumer trust when it came to privacy. Apple squeezed out the top spot overall and, not surprisingly, the top spot among iPhone owners. Microsoft was the most trusted company with privacy by Android owners followed by Apple. Google was in a solid third place regardless of age and smartphone owned, followed closely by Amazon and then Samsung. We asked consumers to rank these companies with a “1” being the most trusted and “8” being the least trusted. Facebook came in at 5.7 followed by Twitter at 6.2 and lastly, Snapchat at 6.7. Interestingly, the ranking did not change much even when we looked at younger consumers 18-25, who are within the Snapchat demographic. Snapchat moved up to the 7th place with this demographic and Twitter was last. Snapchat falling into last place overall is not surprising since a good portion of our respondents did not have a Snapchat account or use the service.

Here is the top line results on company rankings on privacy. The results for rankings with security were not much different.

Reading the comments about why consumers made some of the choices they did proved insightful. It is clear there is a understandable trade off consumers make when they use things which they know are more public, like Facebook. Consumers know what they post is open for anyone to see. Therefore, their feelings around privacy for these services are somewhat less strict. With companies where their actions and behavior are not public like Apple, Microsoft, and Google, they seem to embrace a higher degree of trust since what they do on their phones, PCs, and even what they search for, is not publicly tied back to them as an individual the same way what they do on social media can be tied directly back to who they are. This became clear when we examined behavioral changes they make on social media. The top answer was to be more intentional and careful about what they share/post on social media.

Google was an interesting one for us to explore. We created a few questions just around Google and what consumers believe Google knows about them and what they don’t. While most consumers use Google’s search, they acknowledge the creepiness factor of when you are on a different website seeing ads for things you searched for on Google. Interestingly, while Google was the third most trusted company in both privacy and security rankings, 52% of consumers said they really have no idea how much Google knows about them.

Privacy and Security Fanatics
We know there are some hard core consumers with very strong feelings about their privacy and security and, until now, we didn’t know what percent of the market these consumers made up. We asked some specific questions to help us narrow the field to those who are the most privacy and security conscious. For example, 20.3% of our respondents said they cover their device’s camera with a piece of tape. 13% said they have installed privacy enhancing plug-ins in their devices browser. 15% installed some kind of security software on their smartphone. 11% specifically switched their text/messaging service to one they consider more private and secure. We asked many more to narrow this down but, in each instance, we did not see responses go above the 20% mark. Which leads me to believe the percent of US-based consumers who are the most privacy and security conscious make up around 15-20% of the market, approximately. This demographic tends to skew older — 50+ and heavily female.

While not a large group, it is helpful to get an idea of the size of the market for more privacy and security conscious consumers, especially as more companies are looking to sell products and services with a heavy emphasis on these issues.

As I mentioned at the start, the most interesting takeaway was the difficulty of the subject matter and that it is a difficult topic, one where there is more uncertainty than certainty. I am convinced that any company’s message that over-indexes on the privacy or security angle will only resonate with a portion of the market. Still, I encourage companies to keep pushing both privacy and security forward on behalf of the consumer simply because it is the right thing to do. Consumers will appreciate it, even if they don’t fully understand, or care, about all that is involved.

One of the interesting dynamics to observe in the tech industry is the reputations and relations companies have, not only with the public but also among other companies, including their peers. For the former, Apple, Google, Uber, Microsoft, Amazon, Samsung, and others enjoy very positive relations with their customers. We interact with many of these companies and their products multiple times a day, whether it’s to make a purchase, use one of their devices, or access their services. It’s become so routine, we wonder how we ever got along without them.

But the dynamics vary more widely when we look at how these companies relate with their peers, partners, and competitors, at least from what we are able to see publicly.

Google, Uber, and Amazon, as examples, seem to enjoy strong relations with other companies that have come to depend on them for growing their commerce and selling their products. Amazon has built a huge business with their AWS (Amazon Web Services) that depends upon strong business relationships. Similarly, it’s now working with scores of companies to integrate their Alexa voice technology. Uber and Google have strong relationships with automobile companies collaborating in the development of autonomous cars. Each company benefits from these collaborations, even if they compete in other areas.

Apple, on the other hand, seems to be different. Whether it’s intentional or a result of their history of preferring to do everything themselves, they seem to have more difficult relationships with other companies and do far fewer collaborations. They seem, at times, to be less likely to be trusted, not unlike Microsoft of many years ago. Rarely do you see Apple working in collaboration with Google, Amazon or the car companies on really big initiatives. Apple Mail struggled to work well with Gmail and it took more than a year of haggling back and forth to fix it.

Even with suppliers, the relationships can get very rocky and public, namely their fights with Google, Samsung and now Qualcomm, the latter two suppliers to Apple of critical elements of their products.

Apple’s struggle to work closely with the automobile companies to integrate the iPhone has been far less successful than what they originally set out to do — building the entire user interface for the automobile companies. While the Apple CarPlay integration has found good acceptance along with Google’s Android Auto, the auto companies see them as interim solutions until they can develop their own.

Apple’s inability to build Apple TV into a stand-alone service to compete with the cable companies, a goal once stated, has also failed to materialize, likely a result of the entertainment companies not trusting Apple after what they did to the music business.

Perhaps this is intentional or a reflection of their history of being so innovative and adverse to what other companies were doing. It’s the walled garden approach that’s worked so well for their products. I’m not suggesting what they do is wrong. It’s just different in an industry that’s full of competitors that are also collaborators.

Facebook once again reported very good earnings on Wednesday afternoon, with massive growth in both user numbers and ARPU driving both significant revenue and profit growth. But Facebook also warned of increasing saturation in ad load, which it says will lead to much slower overall ad revenue growth later this year. In that context, it’s worth thinking a little about what other opportunities for revenue growth still lie ahead. Mark Zuckerberg provided something of a hint on the earnings call.

As it currently stands, Facebook is still primarily a social network, albeit one that’s at least as much about content as it is about communication with friends and family. The experience is dominated by a variety of content, from “organic” text posts, photos, and videos shared by its users to a plethora of other content which originates elsewhere and is merely forwarded on by users of the service. That provides plenty of content for users to consume, to the extent Facebook has long since filtered the total universe of content that could be shown to users according to algorithms which prioritize those things likely to drive interest and engagement and therefore keep users on the service to be shown more ads.

However, the fundamental rule on Facebook is still you only really see things your friends have engaged with in some way, whether by actively sharing them or merely liking or commenting on them. Facebook still uses engagement by your friends as an important signal about whether you’re likely to be interested in something too. Your friends are the filters here, along with your own established preferences, both explicit and implicit, about what you’re interested in.

But what if your friends were no longer a filter or limiting factor on what you could be shown? What if other factors could teach Facebook both what you’re interested in and which other pieces of content shared elsewhere on Facebook might be interesting to you? Mark Zuckerberg talked on the earnings call about some of Facebook’s AI efforts aimed at recognizing and understanding the content of not just text posts but photos and videos too. Once Facebook understands the content, it can make a judgment about whether it might be of interest to a given user who has previously engaged with similar content and show it to them, regardless of whether it’s been shared by a friend.

The big advantage for Facebook is it would no longer be limited to the things friends have shared or engaged with – it can show you anything from among a much wider universe of possible content, much of which might actually drive higher engagement because of the subject matter than the things shared by friends. Perhaps you have hobbies distinct from those of your real life friends which are nonetheless shared by many others on Facebook. Perhaps your political views are different from many of those you’re connected to on Facebook. Your personal interests could connect you to a lot more content on Facebook if your friends are no longer a primary filter.

Where this gets particularly interesting is video, which is a key focus for Facebook and one with potential to drive significant additional ad revenue. YouTube has never been limited to showing you content which has only been shared or engaged with by your Gmail contacts. Opening up in this way would allow Facebook to act a lot more like YouTube in showing you recommended videos from outside your personal social graph. Better targeted content, especially content with lucrative ads attached, could drive even higher revenue without necessarily increasing ad load significantly.

But new video-centric experiences could also provide entirely new places for Facebook to place ads and therefore, raise the ceiling on ad load. Facebook is apparently working on an app for Apple TV and similar boxes which would be video-centric – a video service, free of friend-based filtering, could be a great fit for such a service, surfacing the best videos Facebook has to offer based on your interests, combining user-generated and professional content.

Though Facebook has repeatedly warned about saturating ad load, it’s clear it hasn’t given up on finding new places to put ads – mid-roll video ads, ads in Instagram Stories, ad experiments in Messenger and more over recent months demonstrate its commitment to finding new slots to load with ads. This video push seems another obvious way to raise the ceiling. As such, even with ad load slowing growth in 2017, I think there’s still plenty of room for longer term growth, especially around video.

For any major tech battle, we have always had a few names that dominated the field. Some have survived over the years and moved from battle to battle as new names joined in. Artificial Intelligence is the latest battleground — from digital assistants all the way to autonomous driving. While we are still very much in the “recruiting soldiers and training for battle” phase, there are certain companies we are looking at when assessing the market and the progress made: Amazon, Apple, Google, and Microsoft.

Most of the work happening today is on building the foundation for the future. Companies are busy gathering data, training networks, building ecosystems and it seems some brands are also trying to establish themselves as an alternative to names some already see as winners. They are doing that by looking at the bigger picture and trying to build a platform that will become the cornerstone of areas such as voice-first and location.

The thirst for new names of platform providers comes from the desire to not to put all the eggs in one basket and to hedge the bets as to what or who will win in the end. Vendors who embraced Android find themselves struggling to hold hardware margins and differentiate on services, all while competing with their platform provider for the most engaged and lucrative users.

Amazon’s AI Platform

We discussed in our post-CES podcast how this year at the show, AI was everywhere and, when it came to the home in particular, Amazon’s Alexa was everywhere. Amazon was early at the gate with its Amazon Echo products and established Alexa as our primary digital assistant. It did not take long, however, to see Amazon’s interest was much broader than that. Amazon wanted to establish Alexa as the preferred voice-control platform. Amazon did not have a phone to build Alexa on and the home seemed like the most reasonable place to start, given that, in the home, we tend to use several devices and not rely entirely on our smartphones. The location was right and so was the timing, as Amazon rode on the bet many early tech consumers were making on connected homes.

Alexa’s skills have been growing quickly since the release of the API set allowing manufacturers of connected home devices to speak to Echo devices. But Amazon did not stop there. More recently, its Alexa Voice Service allowed vendors to actually put Alexa directly into their devices, taking them from “works with Alexa” to “Alexa inside.” Clearly, Amazon is not doing this out of the goodness of Bezos’ heart. Amazon’s ultimate goal is not selling the most connected speakers but rather becoming the de facto platform for AI in the home.

Let’s be clear. Amazon’s early success is not just because of how open the ecosystem is. Companies are looking at alternative options when it comes to partners and looking for openness but also predictability when it comes to bringing to market products or services in direct competition with theirs. It is early days to see if the trust vendors are putting in Amazon is valid but, for now, the e-commerce giant presents less of a threat to the many brands trying to play a significant role in the connected home and voice-first era.

HERE’s Location Platform

Outside the home, when we talk about AI we often talk about autonomous cars. In this area, we are even further away than in the connected home space from seeing the final impact of AI on transportation overall, not just self-driving cars. Yet, the groundwork done now is what will make everything else possible. When we are talking data gathering and network training in this context, map building is key.

If you tried to play the association name game with any of your friends when you said “map” I bet all of them would say “Google”. The search giant has been in the map business for many years and has won consumers’ preference. HERE is not new to the game either. The mapping and location service company was once owned by Nokia and was built on the acquisition of Navteq. At the end of 2015, HERE was sold to a car-maker consortium of Audi, BMW and Daimler. In late 2016, Tencent, Navinfo and GIC announced their plans to jointly acquired a 10% stake and, in early 2017, Intel acquired a 15% stake. HERE also announced strategic partnerships with Navinfo, Mobileye, Intel and Nvidia all surrounding the topic of maps for automated vehicles.

HERE’s ownership by a car consortium was the first signal that some car makers were starting to consider collaboration over isolationism driven by proprietary technologies. The need to build a broader data set and learn from experts in other areas, all in the attempt to beat tech companies such as Google, Apple, and Tesla to the finish line, has been growing and HERE has certainly benefited from this urgency.

While HERE maps have had somewhat limited traction with consumers, especially in the US, its location platform business has been popular in the enterprise space with big organizations such as Amazon and Microsoft listed as clients.

As carmakers, as well as municipalities, are readying for a self-driving world, I wonder just how much HERE’s core competence, as well as a business model not set to monetize from search and advertising, will play a role in deciding who the right partner is.

When it comes to voice-first and location, Amazon and HERE have transitioned from providing a service to providing a platform. They have done so at a crucial time when players who might be wondering if they can “beat Google” don’t necessarily just want to “join Google”.

The history of the technology industry has seen several swings back and forth between dependence on a network that can deliver the output of centralized computing resources, to client devices that do most of the computing work on their own.

As we start to head towards the Gigabit LTE and then 5G era, when increasingly fast wide-area wireless networks make access to massive cloud-based computing resources significantly easier, there’s a fundamental question that must be asked. Do we still need powerful client devices?

Having just witnessed a demo of Gigabit LTE put on by Australian telco carrier Telstra, along with network equipment provider Ericsson, mobile router maker Netgear, and modem maker Qualcomm, the question is becoming increasingly relevant. Thanks to advancements in network and modem (specifically Category 16 LTE) technologies, the group demonstrated broadband download speeds of over 900 Mb/s (conveniently rounded up to 1 Gb/s) that Telstra will officially unveil in two weeks. Best of all, Gigabit LTE is expected to come to over 15 carriers around the world (including several in the US) before the end of 2017.[pullquote]Gigabit LTE is expected to come to over 15 carriers around the world (including several in the US) before the end of 2017.[/pullquote]

Looking forward, the promise of 5G is not only these faster download speeds, but also nearly instantaneous (1 millisecond) response times. This latter point, referred to as ultra low latency, is critical for understanding the real potential impact of future network technology developments like 5G. Even today, the lack of completely consistent, reliable network speeds is a key reason why we continue to need (and use) an array of devices with a great deal of local computing power.

Sure, today’s 4G and WiFi networks can be very fast and work well for many applications, but there’s isn’t the kind of time-sensitive prioritization of the data on the networks to allow them to be completely relied on for mission critical applications. Plus, overloaded networks and other fairly common challenges to connectivity lead to the kinds of buffering, stuttering and other problems with which we are all quite familiar. If 5G can live up to its promise, however, very fast and very consistent network performance with little to no latency will allow it to be reliably used for applications like autonomous driving, where milliseconds could mean lives.

In fact, the speed and consistency of 5G could essentially turn cloud-based datacenters into the equivalent of directly-attached computing peripherals to our devices. Some of the throughput numbers from Gigabit LTE are now starting to match that of accessing local storage over an internal device connection, believe it or not. In other words, with these kinds of connection speeds, it’s essentially possible to make the cloud local.[pullquote]The speed and consistency of 5G could essentially turn cloud-based datacenters into the equivalent of directly-attached computing peripherals to our devices. [/pullquote]

Given that the amount of computing power in these cloud-based datacenters will always dwarf what’s available in any given device, the question again arises, what happens to client devices? Can they be dramatically simplified into what’s called a “thin client” that does little more than display the results of what the cloud-based datacenters generate?

As logical as that may at first sound, history has shown that it’s never quite that simple. Certainly, in some environments and for some applications, that model has a great deal of promise. Just as we continue to see some companies use thin clients in place of PCs for things like call centers, remote workers and other similar environments, so too, will we see certain applications where the need for local computing horsepower is very low.

In fact, smart speakers like the Amazon Echo and Google Home are modern-day thin clients that do very little computing locally and depend almost completely on a speedy network connection to a cloud-based datacenter to do their work.

When you start to dig a bit deeper into how these devices work, however, you start to realize why the notion of powerful computing clients will not only continue to exist, but likely even expand in the era of Gigabit LTE, 5G and even faster WiFi networks. In the case of something like an Echo, there are several tasks that must be done locally before any requests are sent to the cloud. First, you have to signify that you want it to listen, and then the audio needs to go through a pre-processing “cleanup” that helps ensure a more accurate response to what you’ve said.

Over time, those local steps are likely to increase, placing more demands on the local device. For example, having the ability to recognize who is speaking (speaker dependence) is a critical capability that will likely occur on the device. In addition, the ability to perform certain tasks without needing to access a network (such as locally controlling devices within your home), will drive demand for more local computing capability, particularly for AI-type applications like the natural language processing used by these devices.

AI-based computing requirements across several different applications, in fact, are likely going to drive computing demands on client devices for some time to come. From autonomous or assisted driving features on cars, to digital personal assistants on smartphones and PCs, the future will be filled with AI-based features across all our devices. Right now, most of the attention around AI has been in the datacenter because of the enormous computing requirements that it entails. Eventually, though, the ability to run more AI-based algorithms locally, a process often called inferencing, will be essential. Even more demanding tasks to build those algorithms, often called deep learning or machine learning, will continue to run in the data center. The results of those efforts will lead to the creation of more advanced inferencing algorithms, which can then be sent down to the local device in a virtuous cycle of AI development.

Admittedly, it can get a bit complicated to think through all of this, but the bottom line is that a future driven by a combination of fast networks and powerful computing devices working together offers the potential for some amazing applications. Early tech pioneer John Gage of Sun Microsystems famously argued that the network is the computer, but it increasingly looks like the computer is really the network and the sum of its connected powerful parts.

(Tech.Pinions: Today’s Daily piece, “Adventures in Machine Intelligence” was an Insider post we originally published on December 12th, 2016. We post it today as an example of the daily content for our Insider subscribers. You can subscribe, yearly or monthly, at the page found here)

While I tend to stay away from high-performance computing and data center analysis, I’ve taken up the effort to better understand the soup-to-nuts solutions being developed for machine learning, everything from technical details around chipset architectures, software, network modeling, etc. Luckily, a large number of our clients have assets in these areas so engaging in discussions to help me better understand the dynamics has been straightforward. I’m not going to proclaim to be an expert but my technical background, as well as staying current in semiconductor industry analysis, is proving to be quite helpful. I’d like to share a few basics I find quite interesting.

A great deal of the work up to this point has been around data collection. Large amounts of data on specific subject matter, or around specific categories of data, is the key for machine learning. Simply put, data is the building block of machine learning and intelligence. Interestingly, and somewhat contrary to some opinions, it is not the person or company with the most data who is best positioned but those with the right data. A key part of this analysis about where we go in machine intelligence and how that translates to smart computers (AI) needs to be grounded in collecting the right data. So fundamentally, the starting point is data and the right data.

Lots of companies have been gathering data. Google has been gathering data from searches, world mapping and more. Microsoft has been gathering enterprise data, and Facebook gathers social data. There are a lot of companies gathering data but many are still in the early stages of making their backend data collection efforts into smart machines. In fact, very little of the technology we use is smart. By smart I mean something that is truly predictive and can anticipate human needs. We have a tremendously long way to go in making our machines truly smart. In a recent conversation with some semiconductor architects of machine learning silicon, I asked them if we can estimate a point in time in the history of the personal computer and liken it to where we are today in machine learning. Their answer? No later than the early IBM PCs. This was from folks who have been in the silicon industry for a very long time. The context for this discussion was around how much silicon needs to still advance for machine intelligence and AI to truly start to mature. So it is worth noting their comments on the early IBM days would include their knowledge that the early IBMs ran an Intel 8086’s with 4,500 transistors. Today, we have architectures that have more then 10 billion transistors.

After being convinced we still have a tremendous amount of innovation in semiconductors to get where we need to be in machine learning and AI, I started looking into what is happening today. The next step is to understand how to train a network or how to teach a computer to be smart(er). I stated above it all starts with data, good data, and the right data. Some of the most common examples of network training today are around computer vision. We are teaching computers to identify all kinds of things by throwing terabytes of data at them and teaching them a dog is a dog, a cat is a cat, a car is a car, etc. Training a network is not entirely arbitrary. It is calculated and intentional. The reason is network models have to be built/programmed before they can be trained. Leaning on decades of work on machine learning, many programs exist to train a network in some more common fields that work with large data. Medicine, autonomous vehicles, agriculture, astrophysics, oil and gas, and several others are areas where people have been focused creating this network model. Many hours of hard work and hard science go into the training of these network models so data can be collected and given to the machine so it can learn. Companies playing in this field today are picking their battles in areas where big money is to be made with these training models.

What is fascinating is how long it takes to train a network. With a modern day CPU/GPU and machine learning software, a network can be trained in as little as a few hours depending on the data set. To train a network what a dog is, with roughly two terabytes of data, could take 3-4 hours. However, there are many cases where the data sets are so large it could take several weeks to a month to train a computer just on one single thing. This again underscores the point of how far we still have to go in silicon. I’m reminded of early demonstrations of Microsoft Office running on Pentium chipsets where the demo shined because Excel could process a massive spreadsheet in 30 minutes or less. Today, it is nearly instantaneous. Someday, training a network will be nearly instant as will its ability to query that data and yield insight or a response. Both instant and in real time is the holy grail but we are many years away.

Knowing how early in this stage we are makes it hard to count any company out at this point. But it does emphasize how the right data being collected is key. Companies are right now setting the stage by getting the right data they need to carve out value in the future with AI. What is fascinating is how deep learning algorithms are helping networks learn faster with less data. Expert consensus affirms that having the largest data sets is not necessarily the guarantee of who wins in the future. Because specific network models have to be built, it emphasizes the collection of the “right data” philosophy.

What this means is companies and services can benefit from months or years of the right kind of specific data and still train a network model. Even companies who are starting today and just starting to gather data have a chance in this future leveraging machine intelligence for themselves and their customers — if the data is good.

With some context of where we need to go, silicon architectures — CPU, GPU, FPGA, custom ASICs, as well as memory — are all key to advancing technology in the data center for more efficient and capable backend systems for machine intelligence. But all are still governed by science and we have a relatively good idea what is possible and when. Which is why we know it will still be many, many years and, hopefully, a few new breakthroughs before we get even close to where we need to be for our intelligent computer future.

With news this week about the nomination of Ajit Pai as the next chair of the FCC, much of the attention has focused on his stance on net neutrality and the likelihood the existing rules on net neutrality will be dismantled. However, net neutrality is a complex topic; even the definition of net neutrality is subject to widely differing interpretations. It’s worth breaking down exactly what’s likely to change and what isn’t at a Pai FCC.

Defining Net Neutrality

The first challenge here is defining net neutrality. A very general definition would be it refers to treating all internet traffic equally. It sometimes seems as if some people really do believe net neutrality can only merit the name if it’s really that broad. But that would preclude any sort of network prioritization which puts time sensitive packets above non-time sensitive packets and would also preclude any sort of prioritization by user or content at times of congestion on the network. Most reasonable people seem to at least leave some leeway for sensible network management in order to improve the performance of services subject to delays, such as voice and video calling or live video streaming.

Beyond that, the consensus breaks down very quickly. There are some who insist net neutrality has to bar any and all prioritization or differential charging by content or by user on any basis, whether or not it’s transparent, available to all, or paid for. The best example is the programs introduced over the past couple of years by major wireless carriers, under which some or all content in a particular category is carried without counting against the user’s data plan. T-Mobile and AT&T are the most high profile examples. T-Mobile has two programs – BingeOn and Music Freedom – which “zero rate” video and music content respectively. These programs are essentially open to all comers from a content perspective and there’s no charge to participate. AT&T has recently exempted video from its subsidiary DirecTV from its data caps and says this reflects an internal transfer from the DirecTV division to the AT&T Mobility division in an arrangement also open to any other video provider under the company’s Sponsored Data program.

How you define net neutrality will determine how you see each of these programs. Strict advocates reject both T-Mobile’s programs and AT&T’s, while some others find T-Mobile’s program acceptable but not AT&T’s or Verizon’s. The FCC has never taken a final position on either program but did begin looking into AT&T’s towards the end of last year. The net neutrality rules as presently constituted, however, don’t explicitly bar zero rating programs. Whether you consider either or both of these programs in breach of the principles of net neutrality will determine to what extent you think the new FCC regime will dismantle net neutrality, as I’ll show in a moment.

Rules vs. Motivated Behavior

It’s also worth noting that net neutrality rules were contemplated for many years but only implemented successfully recently. In the time before the rules were finally introduced, there were very few violations of its principles regardless – literally less than a handful of prominent cases existed during that time. The reason is, broadband providers are highly motivated to stay away from controversial breaches of net neutrality principles because they know that such actions would be extremely unpopular with consumers and would invite additional regulatory scrutiny. If we’re talking about actively blocking or degrading any form of content simply because it competes with the carrier’s own content (rather than because it is illegal or against the carrier’s terms of service), that remains very unlikely because there would be an outcry and a backlash and, ultimately, severe financial consequences in terms of lost business if the situation continued.

Net neutrality rules as presently constituted largely lock this behavior in place but the carriers have always made clear they object to the rules largely because they represent additional regulatory encroachments on their freedom to operate rather than because they contemplate any particular action that would contravene them. However, it is clear carriers have other, softer, forms of prioritization and differential treatment in mind, as we’ve seen from the zero rating plans I mentioned earlier. Regulation might or might not bar those zero rating programs but it’s relatively unlikely carriers would ever stoop to systematically blocking or degrading traffic from competing services even in the absence of regulation. Carriers have mostly been willing only to engage in behavior seen as either neutral or beneficial by users. They have much less concern about how they’re perceived by content providers. As such, they’ll zero rate some or all video content because users respond positively to that, regardless of whether providers of other content services like the idea or not. Net neutrality regulation, therefore, mostly helps protect content providers rather than end users, at least in the short term.

Dismantling Net Neutrality

With all that as context, let’s look at what might actually happen in the real world if net neutrality regulations as currently constituted were eliminated. Here are my predictions for what we’d actually see as a result:

• Carriers wouldn’t suddenly (or even eventually) start engaging in discriminatory prioritization or blocking of traffic based on the source – as I’ve said, users would respond badly and even an FCC largely opposed to additional regulation would have to step in and act if this became widespread
• Carriers likely would continue to pursue zero rating programs as a way to both differentiate from competitors and to make their own content services more attractive in some cases (as AT&T and Verizon have already done). With the growth of unlimited data services among the major wireless carriers, this actually wouldn’t have a massive effect on the market
• Some broadband providers are actually bound by terms of merger approvals to abide by fairly strict net neutrality principles regardless of general regulations, with Comcast the prime example through 2018. As such, these companies would have to continue to abide by the rules whether or not they’re overturned, at least for the duration of the commitments they’ve made. AT&T might well be subject to some similar rules as a condition of the approval of its acquisition of Time Warner, putting the two largest broadband providers and the second largest wireless provider under net neutrality restraints

In short, we’re unlikely to see an apocalyptic end to the internet as we know it, even if the FCC begins taking apart the present net neutrality regulations. We will likely see more zero rating and similar programs which don’t prioritize or degrade traffic but merely apply different data pricing to it. If you object to that kind of thing as a breach of net neutrality, you’re likely to be upset but most consumers will be either blissfully unaware or happy about it. If you’re a content provider, you may feel hard done by here too but, again, under the increasingly prevalent unlimited wireless data plans, this will become less of a disadvantage over time. I, for one, am less pessimistic about the outcomes here than many of those decrying the changes on the horizon.

It’s 2020. 5G wireless is being rolled out, with speeds exceeding 10 Gbps, latency below 1 millisecond, and the ability to accommodate vastly more traffic and connections. Will the average household, which doles out some $400 monthly for mobile, pay TV, and broadband today, be able to go wireless only?

The thought is enticing. This could be Cut the Cord 3.0. Version 1.0 cut the landline phone in favor of wireless only, which nearly 50% of households have already done. Version 2.0 has been about cutting the cable (pay TV) cord and going internet only for TV content. It’s happening slowly but steadily. With the 4G LTE and ultimately 5G roadmap promising speeds and latency as good as or exceeding today’s fixed broadband, might customers ultimately look to cut the wire that delivers fixed broadband?

This possibility might get a serious test drive in 2017. Verizon Wireless, which has developed some of its own “pre-5G” specifications, plans to test fixed wireless access in as many as ten cities this year. This would involve getting fiber or other ‘broadband’ infrastructure to within a couple of hundred meters of a household and then use wireless for the proverbial “last mile”. This approach has a lot of appeal to a company like Verizon. First, it allows them to bring broadband service to households without building fiber to every dwelling, the cost of which has slowed the rollout of FTTH services from FiOS to Google Fiber. Second, it allows Verizon to potentially offer a competitive broadband service outside its ‘landline’ footprint. This is compelling because, while the mobile market is very competitive with four national providers, broadband is a near monopoly in many U.S. markets and average speeds are decidedly middle of the pack when compared to other developed economies.

In addition to Verizon, a company called Starry, founded by the folks who did Aereo, has raised $60 million to build out a competitive broadband service using wireless. They’re currently testing in Boston using 28 GHz spectrum, which is one of four bands the FCC has allocated for 5G services. As another example, Redzone Wireless, a Wireless Internet Service Provider (WISP), is offering a “5GX” service to some areas in Maine, using a combination of LTE Advanced capabilities and unlicensed (Wi-Fi) spectrum, promising households average download speeds of 50 Mbps, for $80 per month.

This certainly bears watching. What are the barriers to Cutting the Cord 3.0? The most significant relates to the intertwined challenges of wireless capacity and economics. A typical wireless user consumes about 4 GB per month at an average retail price of roughly $10 per GB, all things considered. Let’s call that 10 GB for a 2.5 person ‘household’. Now, a typical Netflix-watching broadband household consumes some 250 GB per month. That’s a big delta between fixed and mobile consumption.

Mobile operators are acquiring more spectrum and investing in 5G in order to accommodate ~8-10x more traffic, circa 2020. But we should also consider there will be growth in fixed broadband usage too, with 4K, virtual reality, and so on coming down the pike. It’s not inconceivable that a broadband household could approach 1 terabyte (TB) of average monthly use within the next 3-5 years. That would be a tough number for wireless to digest, even with the technology being developed for 5G. Plus, that’s an awful lot of capacity the mobile operators would have to somehow build (or lease) within a few hundred meters of a home or building.

Wireless economics is a related challenge. It costs a wireless operator like Verizon $1-2 to deliver a GB of traffic to a consumer. That’s why everybody’s ‘unlimited’ plan comes with an asterisk, usually kicking in when usage exceeds 25 GB or so. One would have to take a pretty big whack at the costs of delivering those wireless GBs in order to get into the broadband neighborhood. Even if we chop the low-end of today’s $ per GB delivered by half, the math remains challenging.

A final issue many bring up is how do you get the equipment, which might look like a small dish or box, into a building or household? This is something folks have heretofore been hesitant to muck around with. I see this as the least of the barriers. After all, over the past few years, people have become more comfortable installing Nest thermostats, fancier Wi-Fi routers, mesh networks, and femtocells in their home. The equipment piece will sort of look like that by the time these services are ready for prime time.

I think there’s potential here but it might initially be focused on particular market segments. The lowest hanging fruit would be in rural areas, where mobile/internet coverage is currently lacking or sub-par. If we can get enough capacity close to a building, fixed wireless access might be a great solution. Another segment might be the burgeoning multiple dwelling unit (MDU) market, which has been a challenge for broadband anyway. This seems to be one of the target markets for companies like Starry. A third and potentially vital market segment might be households that are budget-challenged. $400 per month for today’s cocktail of mobile, pay TV, and broadband is a tough financial nugget for a lot of households. There’s definitely room for the “Metro PCS” or “Cricket Wireless” of broadband, offering a compelling plan, combining fixed and mobile, with some limits on broadband speeds and consumption, particularly during peak times (to address the ‘Netflix’ problem). Lord knows we could use some competition and pricing options in the broadband world.

One final consideration is how the industry structure might evolve to deliver on this vision. Comcast is aiming to get into wireless, through both an MVNO and apparently having also spent several billion dollars in the recent 600 MHz spectrum auctions. The wireless operators will be reliant on cable and other providers for small cell sites, capacity, and backhaul, in order to build 5G. Then there’s DISH which owns a treasure trove of spectrum that will have to be put to work at some point during our lifetime. And all the big internet players, from Netflix to Amazon to Facebook and Google, are playing in the 5G sandbox in some capacity thinking that, at some point and to some extent, they will need to be at least partial masters of their own network domain. And there’s been a very active M&A market in the fiber biz of late.

So while we are not yet forecasting a significant shift toward Cutting the Cord 3.0, this is the year discussions on the issue will get a serious start.

A great deal has been written recently on the growing importance of voice-driven computing devices, such as Amazon’s Echo, Google Home and others like it. At the same time, there’s been a long-held belief by many in the industry that software innovations will be the key drivers in moving the tech industry forward (“software eats the world”—as VC Marc Andreesen famously touted over 5 years ago).

The combination of these two—software for voice-based computing—would, therefore, seem to be at the very apex of tech industry developments. Indeed, there are many companies now doing cutting-edge work to create new types of software for these very different kinds of computing devices.

The problem is, expectations for this kind of software seems to be quickly surpassing reality. Just this week, in fact, there were several intriguing stories related to a new study which found that usage and retention rates were very low for add-on Alexa “skills”, and similar voice-based apps for the Google Assistant platform running inside Google Home.

Essentially, the takeaway from the study was that outside of the core functionality of what was included in the device, very few new add-on apps showed much potential. The implication, of course, is that maybe voice-based computing isn’t such a great opportunity after all.

While it’s easy to see how people could come to that conclusion, I believe it’s based on an incorrect way of looking at the results and thinking about the potential for these devices. The real problem is that people are trying to apply the business model and perspective of writing apps for mobile phones to these new kinds of devices. In this new world of voice-driven computing, that model will not work.

Of course, it’s common for people to apply old rules to new situations; that’s the easy way to do it. Remember, there was a time in the early days of smartphones when people didn’t really grasp the idea of mobile apps, because they were used to the large, monolithic applications that were found on PCs. Running big applications on tiny screens with what, at the time, were very underpowered mobile CPUs, didn’t make much sense.

In a conceptually similar way, we need to realize that smart speakers and other voice-driven computing devices are not just smartphones without a screen—they are very different animals with very different types of software requirements. Not all of these requirements are entirely clear yet—that’s the fun of trying to figure out what a new type of computing paradigm brings with it—but it shouldn’t be surprising to anyone that people aren’t going to proactively seek out software add-ons that don’t offer incredibly obvious value.

Plus, without the benefit of a screen, people can’t remember too wide a range of keywords to “trigger” these applications. Common sense suggests that the total possible number of “skills” that can be added to a device is going to be extremely limited. Finally, and probably most importantly, the whole idea of adding applications to a voice-based personal assistant is a difficult thing for many people to grasp. After all, the whole concept of an intelligent assistant is that you should be able to converse with it and it should understand what you request. The concept of “filling in holes” in its understanding (or even personality!) is going to be a tough one to overcome. People want a voice-based interaction to be natural and to work. Period. The company that can best succeed on that front will have a clear advantage.

Despite these concerns, that doesn’t mean the opportunity for voice-based computing devices will be small, but it probably does mean there won’t be a very large “skills” economy. Most of the capability is going to have to be delivered by the core device provider and most of the opportunity for revenue-generating services will likely come from the same company. In other words, owning the platform is going to be much more important for these devices than it was for smartphones, and companies need to think (and plan) accordingly.[pullquote]Existing business models and existing means for understanding the role that technologies play don’t always transfer to new environments, and new rules for voice-based computing still need to be developed.”[/pullquote]

That doesn’t mean there isn’t any opportunity for add-ons, however. Key services like music streaming, on-demand economy requests, and voice-based usage or configuration of key smart home hardware add-ons, for example, all seem like clearly valuable and viable capabilities that customers will be willing to add on to their devices. In each of those cases, it’s also important to realize that the software isn’t likely going to represent a revenue opportunity of its own; simply a means of accessing an existing service or piece of hardware.

New types of computing models take years to really enter the mainstream, and we’re arguably still in the early innings when it comes to voice-driven interfaces. But, it’s important to realize that existing business models and existing means for understanding the role that technologies play don’t always transfer to new environments, and new rules for voice-based computing still need to be developed.