Amazon is introducing a privacy-enhancing service, currently in preview, that enables AWS customers to deploy AI models known as “lookalikes” for temporary inter-company collaborations. This service, known as Clean Rooms ML and an extension of AWS’ existing Clean Rooms offering, eliminates the necessity for AWS clients to share confidential data with external partners in the process of constructing, training, and implementing AI models.
Clean Rooms ML empowers users to create a private lookalike model using their collective data without compromising data ownership. According to Swami Sivasubramanian, AWS’ VP for data and machine learning services, this service allows users to maintain control over their models and delete them once they are no longer needed.
With Clean Rooms ML, customers can leverage a subset of customer records to produce a broader set of similar records in collaboration with a partner, resulting in the development of a lookalike AI model. For instance, an airline could utilize information about loyal customers to collaborate with an online booking platform in offering promotions to new users who share similar characteristics.
Moreover, Clean Rooms ML provides tools to customize model outputs according to specific business requirements. AWS intends to introduce settings tailored for healthcare applications in the near future.
In a related development, Amazon has introduced Clean Rooms Differential Privacy, a fully managed service within the Clean Rooms framework. This service ensures the protection of customer data during analysis by generating “aggregate insights.” It enables customers to access consolidated insights regarding advertising campaigns, investment strategies, and medical research without exposing their sensitive data.
