In the tech industry, it's easy to get caught up in the minutiae. But we should all step back and witness an industry in transition. Here are a few things I believe will shape the next few years.
Continue Reading Below
Smartphone Hardware Runs its Course
As my colleague Carolina Milanesi recently wrote, it's becoming increasingly difficult to innovate in the smartphone space. After a decade of new and shiny gadgets arriving every few months, we are undoubtedly nearing the end of major advances in smartphone hardware—something with which PCs and (more recently) tablets have struggled. Iterative changes will come, but I just don't see any significant smartphone shift on the horizon that truly blows us all away.
Instead, look for the smartphone to serve as the underlying foundation for new hardware innovations in things like augmented and virtual reality. There will also likely be an increased emphasis on software and services experiences.
A few years ago, my son Ben wrote an article called "Our Services Destiny," which notes that new markets always starts in hardware before moving to software. As software matures, the value shifts and ends the cycle in services. This observation, however, has been mostly limited to enterprise case studies. The smartphone is the first time we can apply this dynamic in consumer markets.
This a key reason why we are seeing revenue in the software/apps industry and consumer services pick up steam. As I look ahead, my research is focusing on what consumer services mean for the future and which companies are best poised to own this space.
AI Is in its Infancy
Nothing we have in the market today is really "artificial intelligence." We see some clever algorithms attempting to predict or understand us, but they pale in comparison to AI's potential. The real work being done today is more machine learning than AI, but tech firms of all stripes are in a race to train their networks.
This requires a lot of really good data. I'd argue that most criticism we see from companies talking about AI — Amazon, Netflix, Google, and perhaps even Apple to a degree —is due to the lack of really good data. I'd like to take a deeper look at the weaknesses in every major company's AI strategy, but right now, I'm still baffled by how little these systems actually know about me.
Part of this has to do with two fundamental pieces of the puzzle that are still being worked out. The first is in semiconductors. As I've noted before, we are in the 1980's PC era when it comes to AI chipset technology; it still takes hours or weeks to train a network. The only solution comes from many years of silicon architecture advancements; there is no magic revolutionary breakthrough that speeds this up. Companies like Intel, Nvidia, AMD, Qualcomm, and even Apple have their work cut out for them to solve extremely difficult challenges to give software and services companies the computing power they need to deliver instantaneous network training and true AI technologies.
The second piece yet to come is unsupervised learning. Today, most networks are trained with "labeled data": a human has labeled an image of a dog or a street or a person. Text is, by nature, already labeled but it's tough to teach computers to see this. As the industry gets to a point where machines can be trained without human intervention, we will be one step closer to better training and better AI. This is one reason I found Apple's first published paper on AI interesting since it speaks to a process of unsupervised learning by using graphics instead of physical images to teach computers.
5G: Important But Years Away
One other key development that will drive new innovation is 5G, which will provide desperately needed network capacity to support much of what I outlined above.
We are about six years into the shift to LTE. Qualcomm likes to remind us that network technologies generally live for about 18-20 years and, at about the midway point, we tend to see the next evolution trickle out to the market. If this pattern holds, we should start to see 5G in 2020.
5G will be relevant in many markets beyond computers, particularly cars that will be processing tremendous amounts of data and balancing onboard and cloud processing to enable features related to autonomy, safety, and more. Look for it to power a slew of new connected devices.
These trends will shape what is coming next. The point is this transition will not occur in 2017 or 2018, it might not even happen five years from now. It's important to not get caught up in the hype and view the big picture so we are ready when these major shifts occur.