The news around AI mainly reports on the achievements of tools like ChatGPT and Midjourney — powerful, cloud-based tools. But there are plenty of other non-cloud applications where even a little intelligence can make a big difference. Embedded systems have traditionally relied on procedural programming to solve their tasks, especially those constructed for battery operation and using microcontrollers clocked at tens of megahertz and with limited memory. This restricts these devices to tasks that can be implemented by a series of decisions supported by an algorithm such as a fast-Fourier transform or PID control loop. But many tasks are simply a pattern-matching activity. And this is something that AI is really good at. Applying AI to Everyday Medical Diagnostics Take, for example, an electrocardiogram (ECG). Used to monitor the activity of the heart, these time-varying signals have an amplitude of around 10 µV to 5 mV and contain frequencies of around 0.05 to 35 ...