Smart devices are considered smart because they can automate their services. They often communicate using wireless media to provide required functions to users. A smart device can be any device that can process data; from microcontrollers to smart vehicles. Smart devices often lack storage and compute power to process most algorithms which use Artificial Intelligence (AI) methods. However, many algorithms which are used for AI have been optimized such that they can be used in these devices.
Where is AI?
AI models are used in processes for computation beyond usual application-based processes. Using AI models, different types of AI methods are used to process data. These models are usually lightweight algorithms which can be used within smart devices. Model training and testing are usually carried out on low memory and computation power such that they could be used in smart devices.
Classification is one of the methods used by AI models in smart devices. Using classification, objects which are of similar type are classified into groups. For example, AI can be used to process images and recognize patterns to detect groups of data in images. Most devices which could process images within the device use these methods to identify people from photos for tagging users, authentication using biometrics and classifying images to detect security-related attacks.
Prediction is another method used in smart devices which use AI to process data. Prediction can be used to predict an output based on a history of pattern-based output. For example, many apps within smartphones such as those for health care and weather forecasting use prediction-based models to predict output and provide recommended steps to users to improve their health or predict weather information.
Most large scale devices such as smart vehicles and devices used for pattern identification using cameras and videos may use advanced AI methods like deep or convolution-based neural networks for their AI models. These vehicles may use advanced models for classification, vision and mitigating accidents. They may also use short-range and long-range wireless media such that AI processes are handled in groups of nearby devices using distribution of processes.
Why use AI within smart devices?
Remote or cloud server based AI model were used prior to using models within devices. With this method, data is transferred from a smart device to a server and the AI model(s) in the server processes and responds data back to the device. This method can benefit from the high compute power and storage available in remote servers, process data faster and use advanced models to process data. The issue with this method is the latency of transferring data between the device and the server. Therefore, it is rarely useful for real-time processing of data. With AI models used within the device real-time output could be obtained as processing does not need to use a remote server or communication method.
Using AI within smart devices also could reduce costs for maintaining servers and costs for communication. With the former method of data processing, remote servers need to be allocated and AI models are deployed within remote storage. Also, network and hardware devices along with those for load balancing and backups in case of disasters need to be integrated. These costs are often not billed with use of AI models within devices.
Image Courtesy: https://www.pcmag.com/