Edge AI or “AI on the Edge” is a concept and methodology that brings artificial intelligence right close to the user’s device by deploying the AI algorithms and AI models within the device. Edge AI devices include smart wearables, such as smartwatches, smartphones, industrial robots, autonomous vehicles, and AR/VR systems. What exactly happens in these devices is that the AI processing happens locally within the device, which enables quicker, more real-time responses. And since the data is being processed locally instead of making time-expensive calls to the cloud infrastructure, the edge devices benefit from greater autonomy and data privacy.
Consider a scenario for an autonomous vehicle that has to face changing traffic signals, erratic, human driving skills, uncertainties of pedestrians deviating from road safety rules. In such time-critical moments, it might get too late to knock the door of cloud infrastructure of data centers and get a response. Alternatively, consider a patient in immediate need of first-aid, taking into account their health parameters. In situations like these, delegating the data processing and decision making to the edge device certainly helps.
But that does not mean there is no sync up between the edge device and the cloud. Considering the small size of edge devices and hence, relatively limited compute power, limited storage capacity, and other resource constraints, the edge devices do have a connection with cloud infrastructure to feed in the data to retrain the AI pipelines and deploy the updated AI models. Edge AI is a matter of bringing AI closer to the point where data is generated.