Machine Learning
Thanks to impressive developments in consumer technology, we are all familiar with the human-like behaviour of Artificial Intelligence(AI), the ability of machines to learn and modify their own functionality (Machine Learning, ML). Many of us are not concerned about how these technologies work. We are frequently surprised when a human champion is beaten by a supercomputer. Or a person’s appearance is ‘deep faked’ by someone else. Or new works of art are created by software scouring the web.
While these achievements may be headline-worthy and stir the public imagination, enabling these spectacles has historically required an unwieldy mix of computing power and electrical energy. The hard-won machine learning results are gleaned over millions of data points and processed by powerful computers somewhere else. Think of yourself as Dorothy facing the Wizard in the classic film; he is revealed to be a clever illusionist, aided by an immense array of powerful machines behind the curtain.
However, thanks to recent advancements in silicon power and computational efficiency, the challenge of embedding ML technology into everyday devices — like your refrigerator, air conditioner or front door lock — is being met. It’s like putting a ‘brain’ inside a lifeless machine; Embedded systems can develop environmental consciousness and make helpful, personalised decisions autonomously. They do not need to exchange mountains of data with large computers over the internet ( An intensive technique called deep learning). Embedded ML systems can now learn at their place of user operation. This is also known as edge computing (at the edge of the network, as opposed to the central cloud).
What’s even more exciting is that these embedded devices improve over time and with changing environments, generating new data with experience and without human intervention.
All of this may seem overwhelming, but rest assured that ML-enabled embedded systems are already in use and evolving, improving your life.
Let’s take a closer look at the front door lock. The ML technology in Xiotec’s Fortal smart locks facilitates data acquisition from the local environment, such as entry and exit patterns, time of day, motion, tamper detection and other variables. This data becomes the device’s experience to identify usage patterns and make predictions. In conjunction with the user-defined rules programmed in your smartphone, e.g. an individual’s access credentials and time, Fortal’s on-device ML learns the optimal security requirements. This can trigger alerts when unusual events occur e.g. an unexpected visitor or forced entry.
Over time, as the Fortal lock gains more experience and user input, it becomes increasingly helpful and robust in its security and convenience. In short, it becomes more valuable to the user; few products in our homes and businesses can make that claim. A second Fortal lock securing another door will adapt to its unique context, becoming perfectly suited to the specific requirements of that particular entry point.
Indeed, this is the promise of embedded ML — familiar and dependable devices that get better with time without compromising the basic tasks for which they were designed.