The sci-fi world described by Asimov over 50 years ago is starting to become a reality, although the process is not new. In fact, the first wearable device, a computer for casino use was designed in the late 50s. Technology has come a long way since then. Smartphones, smart cars, smart lights and more, these are all names describing tools with connectivity and sometimes reaction abilities. Gartner estimates that in 2017 over 8.4 billion “things” will be connected to the Internet, and until now the consumer sector is almost twice as large as the enterprise one, but we can expect a shift in the near future.
The technology is already used to provide public safety and energy conservation in Denmark, where public lights have been replaced with movement-triggered LEDs. In New York, the Power authority has employed predictive analytics to be warned about system failures weeks in advance. The introduction of smart buildings is expected to decrease insurance premiums by reducing different risks through close monitoring.
In our homes and cars, it can help to create welcoming and safe environments, perfectly adapted to the needs of the owners by incorporating their preferences alongside good practice rules, such as less energy consumption. Personal applications don’t stop at fitness and health trackers or automatic insulin dispensers. Recently, the IoT development company Itransition came up with the idea of an unlimited-range baby monitor that distinguishes the baby’s voice from the noise and can alert parents.
The Internet of Things is a generic name for any device (a fitness bracelet, a car, a fridge) that can record data and send it to a storage unit. It doesn’t sound too impressive because just saved but not used information only represents a cost. The real exciting part comes from analysing it and acting on the insights it provides to make our lives easier, better or safer.
Deloitte has proposed a framework to describe the IoT architecture based on five steps:
The fourth and fifth levels are the ones that bring value by converting vast amounts of recordings and noise into meaningful patterns that can be used by healthcare, law enforcement or just simply citizens. These are communicated to the users by messages, apps and dashboards.
Yet, given the size of such data, which perfectly fits the definition of Big Data by volume, variety, and velocity, the regular statistical methods are not able to offer a satisfactory answer both regarding speed and accuracy. It’s time machines process the data created by other devices and people only supervise and calibrate the process.
The IoT capacity to create massive amounts of data requires an equal tool to process it. Artificial intelligence is an umbrella term which denotes different algorithms. These can go through terabytes of data, sort it, recognise patterns, create forecasts or estimate probabilities. The outcomes are distilled decisions based on thousands of entry-points.
A computer learns by looking at thousands of tagged pieces of relevant information. For example, it can learn about sentiments by going through thousands of words tagged with a sentiment label, or it can learn to recognise a flower by scanning millions of images depicting flowers. However, the tagging needs to be done manually. After going through a few rounds of training, the computer will be able to classify new items according to the rules it has already apprehended. Yet, this is not feasible in the case of data coming from IoT due to size and the amount of noise. Through a process called deep learning, the computer can also perform the initial tagging step, and it teaches itself. This is done through clustering and classification algorithms.
According to the latest PwC report on IoT enabled by AI, it serves three essential business functions. Prediction decreases costs by anticipating a failure and encouraging replacement or a change of the action course. Prescriptive capabilities of sensors integrated with control panels can help prevent disasters by acting in real time and suggesting on-the-fly changes. At the same time, if systems are built for deep learning, each time they are used they are in fact improving themselves, reducing errors without human intervention.
Like any piece of technology, IoT comes with essential problems that need to be overcome. The most significant ones are related to integration and complexity. The way sensors communicate with the network can affect the speed and the quality of the data. Also, the format of the data is important for further processing. As with most communications that take place via the internet, the safety and security of the transmissions are a significant concern, mainly if the sensors belong to medical devices and a security breach could endanger the life or privacy of an individual. Communication channels attached to energy or communication providers should also be appropriately encrypted to avoid cyber-attacks such as the latest one in Ukraine. Ethical and legislative challenges are also worth mentioning. The tradeoff between comfort and privacy is not embraced by all users. Since this is still an emerging technology, the legal implications have not been adequately addressed yet.
Right now, cities, homes and even our bodies are padded with sensors creating continuous streams of data. The real challenge is to transform these pieces into meaningful rules which can make us better, healthier, stronger and eco-friendlier. But, did IoT cross the chasm towards mainstream adoption already, or it is still more of a cool gadget for geeks? For a technology already adopted in healthcare, agriculture and manufacturing, as well as in personal fitness apps and our homes, it is already too ubiquitous in our lives to be just a fad. We can expect it to grow according to a model similar to the Internet. It is as if we were currently in the 90s of the IoT.