Stanford University — Machine Learning Security Research Led by Dan Boneh and Florian Tramer

[Navigating Transformative AI] Grant investigator: Daniel Dewey This page was reviewed but not written by the grant investigator. Stanford University staff also reviewed this page prior to publication. The Open Philanthropy Project recommended a gift of $100,000 to Stanford University to support machine learning security research led by Professor Dan Boneh and his PhD student, Florian Tramer. Machine learning security probes worst-case performance of learned models, and we consider work in this area a promising way of ensuring that models are "doing the right thing" in a generalizable way. Our main rationale for making this gift include: We consider Florian Tramer a very strong PhD student who is currently conducting excellent machine learning security work. We expect excellent machine learning security work to be very important for AI safety. Generally speaking, we expect increased funding in areas relevant to AI safety—like machine learning security—to move the field in a direction we consider positive and aligned with our interests in mitigating potential risks from advanced AI; we therefore consider this gift a small nudge in that direction. This gift falls within our focus area of potential risks from advanced artificial intelligence.