Most technologies go through a stage when everything seems possible. Personal computers in the early 1980s, the internet in the late 1990s and mobile apps around the beginning of this decade were like that.
But so was the first unboxing of a Galaxy Note 7. In time, either suddenly or gradually, reality sets in.
The internet of things still looks promising, with vendors and analysts forecasting billions of connected devices that will solve all sorts of problems in homes and enterprises. But the seams are starting to show on this one, too. As promising as the technology is, it has some shortcomings. Here are a few.
IoT systems are only as good as the data they capture, and some of it is not great.
As much as 40 percent of data from IoT sensors may be wrong, redundant or useless by the time it gets to the cloud, according to General Electric. That makes data collection and processing harder.
Harsh environments raise the odds that a sensor will generate bad information: Weather, vandalism and pests are among the many dangers. For better results, enterprise IoT users may need to calibrate their sensors, install redundant nodes or use one type of sensing device, like a camera, to monitor another.
Artificial intelligence can help solve the problem by weighing inputs from multiple sensors to reach accurate conclusions. For example, doctors can monitor a patient with wearables that measure different vital signs and can be checked against each other. Also, filtering out readings that aren’t needed — like 1,000 consecutive reports that a pipeline hasn’t cracked in the last five minutes — is a big part of what edge computing is designed to do.
INSECURE CONSUMER DEVICES
Many home IoT devices ship with security holes, and those that don’t will probably get them eventually. And it’s not enough to send out patches and ask owners to install them, because most won’t, the Broadband Internet Technical Advisory Group warned in a report issued late last year.
Hangzhou Xiongmai Technology learned this the hard way last year when its DVRs and connected cameras got pulled into a Mirai botnet that took down large parts of the internet. Weak default passwords opened them up to the attack. The company had stopped shipping units with those weak passwords the year before, but for older products it was reduced to asking owners to update the firmware and change the password by hand.
This means two things: IoT devices attract hackers, because they’re plentiful and overlooked, and it’s up to vendors to build in security and allow for automatic over-the-air updates.
IoT is still a moving target, and it will be for a while yet. That makes it hard to choose technologies, because some may not survive for the long haul.
If you’re a consumer, orphaned products like Nest’s Revolv hub can feel like a ripoff and an inconvenience. If you’re an enterprise with millions of sensors built for a network that’s fallen out of favor, maintenance and migration could be very expensive.
Some specifications that are only a few years old, like IoTivity and AllJoyn, are already merging. Having fewer standards is probably a good thing, and vendors will try to make the old work with the new, but consolidation may still hold surprises.
Low-power, wide-area networks raise this issue, too. There are many to choose from now, but analysts say there probably won’t be enough of a market for all in the long run. While the towers might not come down before it’s time to replace the devices that use them, a network in survival mode won’t be expanding, either. So for now, it’s best to proceed with caution.
YOU CAN’T JUST PLUG IT IN
Connecting a bunch of machines and suddenly getting automation and business insights sounds great, but IoT doesn’t work that way.
Because it crosses the line between computing systems and physical infrastructure, like machine tools and air conditioners, IoT forces IT to collaborate with operations people they may not even know. In fact, a Technalysis survey last year found operations departments were in charge of IoT projects more often than IT shops.
And while pilot projects are a good way to start, ad hoc deployments by each department keep enterprises from getting all the benefits of IoT, Strategy Analytics found in a survey last year. Data collection projects need to go hand in hand with analytics, including knowing what questions to ask and which tools can answer them. Haste can cause more confusion than convenience: Fifty-one percent of enterprises Strategy Analytics surveyed weren’t sure whether IoT was paying off.