Machine learning (ML) and artificial intelligence (AI) are at the forefront of technological development. As a technology, ML is inherently flexible and has the potential to revolutionize industries and the world. Many of the potential applications of ML exist on small, low voltage circuits. Tiny Machine Learning or tinyML, as it is referred to, is a sub-field of machine learning. TinyML is where machine learning and embedded internet of things (IoT) collide. This emerging technology has the potential to be revolutionary for multiple industries.
The tinyML Foundation describes its discipline as the following:
“a fast-growing field of machine learning technologies and applications including hardware, algorithms, and software capable of performing on-device sensor data analytics at extremely low power, typically in the milli-watt range and below, and hence enabling a variety of always-on use-cases and targeting battery-operated devices.”
In previous articles, we’ve extensively discussed the Internet of Things (IoT) - both how to keep IoT devices safe and how 5G and the IoT have the potential to create smart cities. IoT devices are some of the most obvious and exciting applications for tinyML. While these devices often aren’t battery-run, their ideal is to dissipate as little power as possible to decrease the cost-over-time of one’s device.
For reference, a typical household lightbulb dissipates 60W of power meaning these machine learning circuits operate extremely cheaply. These devices often involve some degree of ML, and with advanced tinyML technology, both the scope and efficiency of these devices could be improved. Alongside the development of battery technologies, many of these devices may eventually be fully wireless, run-on batteries, and fully portable.
While IoT devices seem like the obvious first step for tinyML, countless other applications are possible. The tinyML foundation presents many potential applications on its website. For example, they show how tinyML technology could make agriculture more water-efficient and sustainable.
The author, Ravi Rao, argues “for the implementation of the precision agriculture model, connecting all the devices to the network and passing data to the cloud is not always feasible … using microcontrollers interfaced with moisture sensors and water control valves, it is possible to implement a simple automatic irrigation system that turns the irrigation system on or off depending on a static value of soil moisture levels.”
TinyML has the possibility not only to revolutionize our consumer devices but could help promote sustainability and has the potential for many more applications. The key benefit of tinyML is its efficiency and decentralized nature. While current ML systems send data to the cloud for processing, tinyML systems can perform data analysis on-device, making for less data flow, more efficient applications, and a world of possibilities.