Events
Seminar @ Cornell Tech: Avi Schwarzschild
Safe and Reasonable AI
Even with the widespread and increasing use of AI, the safety implications and the reasoning capabilities should be better understood. In this talk, Avi Schwarzschild will first discuss memorization in LLMs through the lens of copyright infringement. Then, Schwarzschild will discuss recent research on reasoning in neural networks. While AI continues to impress us, Schwarzschild will focus on the ability to extrapolate from easy to hard by looking at iterative components that give neural networks the ability to think longer to solve harder problems.
Speaker Bio
Avi Schwarzschild is a postdoctoral researcher at Carnegie Mellon University where he works with Dr. Zico Kolter. His work focuses on safe and secure machine learning as well as reasoning in artificial intelligence systems. Prior to starting at Carnegie Mellon, he earned his Ph.D. in the Applied Math and Scientific Computation program at the University of Maryland where he was advised by Dr. Tom Goldstein on his work in deep learning. His ultimate goal is to build capable AI systems that can be deployed in a wide range of applications with clearly understood and managed risks. To this end, his work focuses on two main topics: reasoning and safety.