PSC and Collaborators Receive 2 HPCwire Readers’ Choice Awards at SC26 - HPCwire

PSC and Collaborators Receive 2 HPCwire Readers’ Choice Awards at SC26 - HPCwire

Two significant initiatives in high-performance computing have received prestigious recognition, highlighting advancements in data privacy and research accessibility. The POPri method, developed by researchers at Carnegie Mellon University and the Pittsburgh Supercomputing Center, introduces a novel approach to handling user preferences. This technology generates synthetic data that allows for the training of large language models without exposing private user information, thereby significantly mitigating security risks associated with sharing sensitive personal or financial data online. Simultaneously, the NSF ACCESS program has been honored for its extensive collaboration in providing academic users with no-cost access to powerful supercomputing resources. This nationwide effort unites government, academia, and industry partners to deliver essential computational capacity and support services. By integrating resources from leading institutions and technology providers, ACCESS ensures that researchers across various disciplines can advance the state of the art in artificial intelligence and other fields without financial barriers. These achievements are particularly relevant to the open data community as they demonstrate how privacy-preserving techniques can coexist with data-driven innovation. By enabling the use of AI models on synthetic representations of user behavior, POPri offers a framework for utilizing valuable preference data while strictly protecting individual privacy. This balance is crucial for developing trustworthy AI systems and encourages the broader adoption of data-centric research methods that respect ethical standards and security constraints.

Source: hpcwire.com
Published on 2026-11-18