Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Watch this one-hour seminar presentation where Carnegie Mellon PhD student Xuhui Zhou discusses research on developing socially intelligent and safe AI systems. Learn about SOTOPIA, a framework for evaluating large language models' social intelligence through interactive simulations, and explore why LLMs struggle with real-world social skills, particularly around information asymmetry. Discover insights about HAICosystem, a new framework for assessing safety risks in user-AI-tool interactions as AI agents become more autonomous. Zhou, whose work has been featured in major media outlets and received recognition at top AI conferences, concludes by examining future directions for creating AI agents that are both socially aware and safe. The presentation draws from Zhou's research focus on aligning NLP systems with human social values and his experience red-teaming GPT-4.
