Explore cutting-edge approaches to harnessing AI systems for complex tasks in natural language processing during this 53-minute talk presented by Ruiqi Zhong from the University of California, Berkeley. Delve into two key NLP tasks: automatically identifying and explaining goal-relevant patterns in large text collections, and utilizing AI to assist non-programmers in labeling complex SQL programs with expert-level accuracy. Learn how to develop tools that enable effective human scrutiny of AI behavior with minimal effort, potentially uncovering insights beyond human expert anticipation. Gain valuable insights into the speaker's research on AI alignment and his work on enabling human supervision of AI systems for tasks where ground truth is challenging to determine.