Edge AI consists of running artificial intelligence models locally on the device itself—a mobile phone, a camera, an industrial sensor, or a car—instead of sending data to remote servers for processing. The term "edge" refers to the periphery of the network, where data is generated, as opposed to the centralized cloud model.
Su relevancia se explica por varias ventajas prácticas:
- Low latency: the response is almost immediate as it does not depend on a connection.
- Privacy: sensitive data does not leave the device.
- Offline operation: it operates even when there is no internet access.
- Lower cost and bandwidth consumption by avoiding the transmission of large volumes of information.
Un ejemplo cotidiano es el desbloqueo facial de un teléfono o los asistentes de voz que reconocen comandos básicos sin conexión. El principal matiz es la resource limitation: edge devices have less memory and computing capacity, so optimized models are typically used through techniques such as quantization or pruning.