23-09-2020 | | By Robin Mitchell
With IoT devices numbering in the billions, the total collective processing power is truly remarkable. Armed with sensors, these devices produce untold amounts of data that has economic value, and IoT owners of the future could profit from their devices.
With over 20 billion devices deployed worldwide, the Internet of Things (IoT) is one of the fastest-growing industries with one of the widest applications. From basic sensor data gathering to actuator control, IoT devices can significantly improve the performance of any application, and provide engineers with invaluable data. While IoT devices have technically been around for more than 20 years (by definition, a computer can be seen as an IoT device), the modern sense of basic devices with a few sensors and a small amount of processing power (often a microcontroller), has only been in wide circulation for a decade. Even so, IoT devices have seen numerous technological changes and challenges, including rapidly improving microcontrollers, the move to IPV6, and the growing security concerns relating to IoT devices. Now that 2020 has come, the importance of IoT devices continues to grow, and as it does, so do the opportunities presented to them.
While many industries are being improved with the help of IoT devices, the AI industry has by far been the biggest beneficiary. For AI systems to improve their performance, they require data to learn from, but this data must include both input and the corresponding output. For example, a speech recognition system would use spoken words as its input, and the converted text as its output. Through the use of neural nets, an AI system can tune itself to produce the correct output for a specified given input. If large amounts of variations are passed, the system will be not only able to distinguish different inputs, but also be able to understand input data that has never been observed. This works because AI systems can understand differences in inputs and ignore variations. For spoken words, AI can ignore accents, but most likely understand formants that indicate the word being spoken. We say “most likely” because engineers don’t know why AI come to the conclusions they do (researchers are currently working on AI systems that can explain their reasoning). So, how does IoT fit into the picture with AI?
If large amounts of input and output data are what is needed to improve AI, then IoT devices are the holy grail for AI. Individually, IoT devices produce small amounts of data that by itself is not enough for an AI system to improve itself, but the plentiful nature of IoT devices, combined with their ability to stream data 24/7 results in a massive data pool for AI systems to work with. The result has been the development of highly accurate speech to text systems such as those used in Alexa and OK Google, intelligent industrial maintenance systems, and Google Assistant.
It can be seen that sensory data produced by IoT devices is invaluable to AI system, but to anyone who understands the basics behind economics can immediately see that the same sensory data could also be a source of revenue. Large companies, such as Amazon and Google, will have most likely obtained data using their platforms, but those developing AI systems which don’t have access to millions of devices can struggle to train their AI. Therefore, those developers will most likely be interested in buying data from those who have access to relevant IoT devices, and thus providing a revenue stream for the IoT owners. While purchasing IoT data is still in its infancy, the concept of buying data is not, and this has brought about many controversies. For example, companies such as Facebook have been found to sell data gathered by users to third parties who then use that data for advertising and targeting purposes. Thus, those that operate IoT devices who want to sell data need to be incredibly careful as to the nature of the data they are selling. Data such as speech may not be easily sold, and could very easily breach privacy regulations, while data such as temperature readings are perfectly acceptable. Even if data being sold is legally sound, it may draw negative attention from operators who do not want gathered data being sold off.
The value in data gathered by IoT devices is not the only potential revenue stream that IoT devices can provide; their idleness can also be profitable. In the past, large machines owned by businesses would often go unused during the night, and considering that such mainframe systems can cost a small fortune, their idleness during the night is essentially a waste. Instead, businesses could sell their processing power when the mainframe was not in use to outside users who require a powerful system, and charge based on the time that the system was used for. This concept is still in use today, with Amazon offering services through their Amazon Web Services, which charges users based on the processing capabilities they require, and the same for Microsoft Azure. While individual IoT devices do not provide much computing capability (if at all), the combination of many thousands of devices makes for a potentially effective parallel computational system. When IoT devices are not in use (i.e. in an idle state in between sensor readings), their processing time could be sold to external users. They need to execute large parallel tasks that may be too expensive to do on a dedicated machine. While this may cause an issue with regards to portable IoT devices that are reliant on long-term battery operation, those that are powered permanently (e.g. mains power), could readily be used 24/7 to maximise their use.
The strongest feature of IoT devices is not their capability, but their ability to be used in a large group. Their collective ability allows for the training of AI systems as well as the possibility of large scale network computing. While the IoT industry is still in its infancy, it won't be long before IoT devices become a source of potential revenue, and begin to pay themselves off through their idleness.