IoT and AI: Inseparable

Over the past few weeks we have learned about the development of AI and IoT, both of which are promising to dramatically change the way we interact and use technology. AI and machine learning will provide new ways for us to analyze vast amounts of information. IoT will allow us to gather information from a myriad of new sources, such as cars, appliances, wearables, you name it. Indeed, as IoT devices become even more and more prevalent, the need for AI and IoT together will become more and more evident. With 30 billion+ devices by 2020 [1], all sorts of new devices will start to appear on the network, and all of these devices will generate data from a variety of onboard sensors.

All of this new data will bring a new meaning to “big data.” Gathering insights from several terabytes of data is already hard enough, even when it is well-organized and labeled. With IoT, much of this data will be raw data,, making it impossible for humans to parse through the information for insight. This is where AI comes in. AI and machine learning can help us sift through the mountains of data coming in from these new IoT devices and make meaning of them, unlocking the true potential that IoT promises us.

Consider a couple of possible scenarios where AI will be necessary for IoT to succeed. The commercial aviation industry requires an extremely high reliability rate. Dispatch reliability (the plane being ready to fly on-time) above 98% is essential for airlines to make a profit, and one of the biggest reasons for lower dispatch reliability is unplanned maintenance. As aircraft get older, their parts break more often; with aircraft becoming more and more complicated, failures are becoming more and more complicated to diagnose. A hypothetical IoT implementation would be instrumenting the various systems on the aircraft. However, each system is composed of several more subsystems, and thousands upon thousands of components may be instrumented and providing telemetry data every second. Multiply that by the hundreds of aircraft in each fleet in the air at any given time, and you have a massive data problem. Airlines would want to use this data to predict when an aircraft needs preventative maintenance before it actually has a problem, but without AI to analyze trends and provide analysis, the volume of data would overwhelm any human analysts tasked with extracting meaningful results [2].

Or perhaps we can look at health care. Today we have fitbits and Apple Watches. There are already IoT pacemakers [3] (one of which had the dubious quality of being recalled for hacking fears), and in the future even more medical devices will be connected to the internet. Think of all of the metrics that you get at the doctor. Heart rate, blood oxygen concentration, respiratory rate, blood pressure, and more. We could conceivably see IoT devices measuring many of these metrics, with the hope that they can provide us insights on when we may be getting ill, or when it’s time to go see a doctor. This could provide us with truly “personalized medicine” instead of the short 15 minutes we normally get at the doctor’s office. Again, each person will generate millions of data points individually. Asking a human to analyze the data for dozens or hundreds of patients is an exercise in futility; you will need AI to find the trends that you are looking for.

These are just two examples, and I’m sure you can think of more. Luckily for us, many IoT projects are starting to include AI components as well [4]. IoT promises to change our lives for the better, and adding AI will help IoT truly realize its potential by harnessing the data it produces.

 

[1] https://spectrum.ieee.org/tech-talk/telecom/internet/popular-internet-of-things-forecast-of-50-billion-devices-by-2020-is-outdated

[2] https://www.mro-network.com/big-data/internet-things-aviation-gets-real

[3] https://www.businessinsider.com/fda-recalls-500000-internet-pacemakers-hacking-fears-2017-9

[4] https://www.wired.com/brandlab/2018/05/bringing-power-ai-internet-things/

0