Ryan Swift‘s research interests are on designing safe and trustworthy AI systems. Ryan believes that key components of safe AI systems are explainability, robustness, privacy, and reproducibility. In particular, he believes these systems should be transparent and comprehensible to a regular person; they should not fail or become misaligned in the event of perturbations such as distribution shifts or attacks; they should provide privacy guarantees for users and their data; and they should not require infrastructure available to only a few companies in the world. Privacy is a uniquely interesting challenge to Ryan, as the strongest privacy guarantees must necessarily also be explainable, robust, and reproducible. He believes privacy is a fundamental human right, and Iis excited to work on building a future where privacy and all other human rights coexist with technology.
PhD student advised by Konstantinos Psounis.