Andrew Trask

Andrew details the most important new techniques in secure, privacy-preserving, and multi-owner governed artificial intelligence. Andrew begins with a sober, up-to-date view of the current state of AI safety, user privacy, and AI governance before introducing some of the fundamental tools of technical AI safety: homomorphic encryption, secure multiparty computation, federated learning, and differential privacy. He concludes with an exciting demo from the OpenMined open source project that illustrates how to train a deep neural network while both the training data and the model are in a safe, encrypted state during the entire process.

Andrew touches on ideas such as Differential Privacy, and Secure Multi-Party Computation, and how these ideas come into play. Many industries are limited by regulations of private data, and in response to the need for greater privacy in AI three cutting edge techniques have been developed that have huge potential for the future of machine learning in healthcare: federated learning; differential privacy; encrypted computation. These modern privacy techniques would allow us to train our models on encrypted data from multiple sources without sharing the data.