Foster customer trust One technique is infusing privacy directly into The ability to have and contain this intelligence Still another approach that protects data from the way that edge devices are architected. on the edge is powerful. But there are cases being seen by a central system—even when it’s Using edge products themselves to host AI where deriving insights requires a central analysis in fact sent to one—is homomorphic encryption. lets enterprises keep user data local, rather or combining larger pools of data from a cohort This technology enables computation on than sending it to a central service for analysis. of partners. Rather than openly share data between encrypted data without the need to decrypt it first; Previously, the necessary computing power and all partners in an ecosystem, companies are turning data can be transported, processed, augmented technology to support this approach were not to machine learning approaches that can combine or changed by a third party, without letting that able to fit into smaller edge devices, but advances insights from multiple, disparate datasets without party see the data they are working with. Even the in chips and processing power made it a viable directly sharing and combining the data first. most sensitive data can now be shared for analysis option. Simcam, an in-home security system, A consortium of 10 large pharmaceutical companies without actually exposing it. Travis County, Texas is uses an “intelligent edge” approach to deliver created the Machine Learning Ledger Orchestration developing a voting system that uses this approach 29 an AI solution on its in-home cameras. A key for Drug Discovery, a blockchain-based distributed to monitor the voting process ahead of the 2020 30,31 32 point of value for a home security system is the learning model. The system is allowing partners US presidential election. Without ever touching ability for a camera to detect events or objects to to train a drug discovery algorithm without each the actual votes (which could open up the system trigger a security response. But in order to provide company directly sharing its data with the others. to bad actors manipulating data or voter fraud), privacy, all facial recognition occurs on the Simcam Raw data is never seen by the whole ecosystem, analysis of voting data can be performed while device itself, rather than sending data to the cloud while the traceability of the blockchain ensures that it is still encrypted. for analysis. Simcam’s cameras use Intel’s Movidius every operation is seen and accepted by all partners. Visual Processing Unit (VPU), and analysis is run locally using Intel’s hardware. This allows for more control and privacy—and ultimately fosters more customer trust.