Connected cars are probably the most visible aspect of the Internet of Things (IoT) to many people, with self-driving cars showing up in the media every week. And, we've seen several concerns related to smart cars, some of them involving security, with people hacking cars remotely and taking control. This exposure is not incidental: cars are a big part of our lives. However, a nuance of IoT that is not being discussed enough, in my opinion, is simply a consequence of its nature: the need to interact with people in a pervasive, seamless manner—which is no simple task to get right!
For instance, see what happened outside my window today... out of nowhere, my car notified my watch that it was going to start its engine. I could see the car outside my office window and, sure enough, its lights were flashing to indicate ignition. Why had this happened? Well, I don't know.
Quickly after noticing the ignition message, I forwarded the notification from my watch to my desktop monitor, where I fiddled with passwords and such in hopes of stopping these devices from communicating with each other. (Now, I don't mind devices communicating with each other as long as they don't plot against me—turning on my car to drain my battery and waste my gas!)
I suppose the major flaw in my plan was that my car was parked right outside my window where it could "see" everything I was typing!
OK, so maybe I am over-exaggerating just slightly. There is no evidence that any gadgets plotted against me, no machines achieved self-awareness, and no sign of the Terminator around. But my smart car did communicate with my smart watch via my smart phone via the cloud. And, this resulted in the car starting at an unscheduled time.
None of this can be really be called "smart" then can it? In fact, it's quite dumb... starting my car was neither a requested nor desired action. If it had happened inside a closed garage, it could've been dangerous to someone (possibly me).
So, here's a system that I would consider being smart: an unintentional request to start my car should be stopped if the car identifies that it’s actually inside a closed garage—a harmful place. But, how could the car detect that? By communicating with the garage, of course! Or, perhaps the car might ask for confirmation from me before starting itself because, in this particular case, it would use my phone's GPS to "know" that I parked at my office and realize I'm scheduled to have meetings all day at the office because it's pulled that info from my calendar. It would understand this action to be irregular and trigger a confirmation before acting on its own. If, on the other hand, my car were to start up around noon, when I usually leave for lunch, and adjust the temperature inside to suit my liking, then perhaps no confirmation would be needed, as it would be considered a routine request. This would be an example of what I consider an IoT device being pervasive to my life in a good way: adaptive enough to make me feel comfortable that "things" are making decisions on their own to help me, not plot against me!
So, the real value that IoT brings to the market is its potential to change people's lives by making things easy. My little car story above is an innocent example of how this revolution is still an evolution. Computer Science needs to focus on developing "smartness" that combines all of these different, unstructured elements into a well-coordinated system that acts intelligently from a human's point of view. Machine Learning and Artificial Intelligence have produced some limited results for sure, such as self-driving cars and robot trading, but their analytical and mathematical approaches to solving problems have clear limitations that may not be easy to overcome. Ultimately, what we expect from these systems is to have them interact intelligently in our favor, and what we consider intelligent might not adhere to an easily-defined mathematical formula at all.
Sadly, the majority of the discussions about IoT follow a vector of a different direction: How can we monetize the information that is collected? This is very obvious when the only focus of the data is the business, and it involves the act of consuming: place, life habits, health. So far, the focus has been on the business needs instead of how to benefit the people who use the devices. This may explain why most IoT initiatives have been centered on the cloud, instead of a peer-to-peer approach: With the cloud, it’s much easier to have an app that forces people to connect to a central location, a place where data can be harvested and ads can be inserted (the good-ol’ Point of Purchase marketing concept). This would be much harder to achieve on a peer-to-peer network, although from the engineering point of view, it makes way more sense to have your thermostat talking directly to your phone, and vice-versa, rather than both of them connecting to the cloud.
The most common business model at the moment is one in which apps give you something for free so long as you agree to share your private data with the vendor. Or often worse, you buy something that forces you to share your private data. I truly believe that the power of IoT lies way beyond this.
Internet of Thing 1, Thing 2, and Thing 1 Million: Up Up Up With a Fish...
The goal should be for devices (the "things" in the Internet of Things) to communicate with each other as a system that collectively produces value. By "value" I mean impacting users’ lives for the better, which is the ultimate monetization of the data.
This challenge is not being completely ignored by the major players. Any IoT conference today talks about the difficulty of collecting, connecting, and processing the data being pushed by the "things." And contrary to Dr. Seuss’ famous story, it’s not only Thing 1 and Thing 2—which by the way, made a good mess in a very short time while interacting with humans—but Thing 1 Million.
These Seussian similarities do not stop with this story: the new topology maps that are being discussed combine the chaos of Thing 1 and Thing 1 Million with the Sneetches, with billions of messages coming in and out of the Star-Off and Star-On machines. Concentrators, hubs, and other elements are being introduced as IoT platforms attempt to glean insights from the tremendous influx of data that is or may be produced... but, this is all being handled way more tactically on a per objective basis rather than being used to craft an overall strategy.
Vendors have been leading this transformation. And as good vendors, they are more concerned with the "what" and "how" than with the "why." I think this is simply following technology’s natural evolution—faster sensors, more connectivity, more CPU power with the cloud—but, I also see that we need more than just engineering here. We need strong support coming from the theoretical side. My guess is that Machine Learning will have to combine with probabilistic algorithms, such as stochastic processes, to be able to deal with all uncertainties that a normal person’s life presents. And when it does, a new research field will emerge.
As Alvin Toffler defined very well, consumers have become Prosumers, with daily life producing more digital content than ever. Most of this data, such as videos and photographs, is not well structured. If you throw the googol of data produced by sensors into the mix, you get a glance at the real revolution that is about to start.
This is a non-stop wave. New paradigms, new algorithms, and new disruptive ways to implement complex information technology systems are all in demand. Otherwise, Thing 1 and Thing 2 will never really see Thing 1 Trillion as part of their gang, and we will always feel like Sally and her brother: up up up with a fish...