Artificial intelligence promises to radically transform the economy, perhaps faster than any technological revolution in history. According to a report from the McKinsey Global Institute, AI could add $13 trillion to global wealth by 2030, which is equivalent to an additional 1.2 percent of growth per year. However, this brings up an important question: how should organizations integrate artificial intelligence technology into their operations?
One possible solution is to decentralize the process by allowing data to be analyzed by each individual machine or on a local server. This brings AI closer to the “edge”–meaning closer to the source of the data, rather than a centralized cloud. Edge computing, as it’s called, has the advantage of improving efficiency and security. It enables intelligent devices or machines to extract and use data directly from the source on the periphery of the network instead of waiting to hear back from the cloud. For example, the precious milliseconds saved by edge computing could make a big difference for time-sensitive applications on the factory floor such as sorting, inspections, quality control, maintenance, and safety.
However, there is an obvious trade-off: once edge computing devices proliferate in the factory or the workplace, it becomes more difficult to scale up. The problems of complexity, connectivity, and interoperability increase as more nodes are added to the system. Most of these problems will be solved with new technology, but some data analysis may still need to occur in a central or secondary hub. That is why edge and cloud computing are not exclusionary. The two technologies will work together to overlap and complement each other.
Organizations will need to decide what works better on the edge and what works better on the cloud. Although this is not an easy problem to solve, it does have the benefit of flexibility. According to Diego Tamburini, the Principal Manufacturing Industry Lead at Microsoft, edge computing gives organizations the option of “balancing workloads more efficiently based on latency, compute power, or data privacy requirements.” In other words, the edge can perform most of the basic work, while some data is sent to the cloud for deeper processing and analysis, thus improving the productivity of the organization as a whole.