Deep learning models such as ConvNet and transformers have made huge progress in real-world applications from image generation to self-driving vehicles. Researchers and developers mainly focus on building larger and deeper models to improve the accuracy and numbers on public benchmarks. However, it is difficult to establish a reliable and trustworthy relationship between humans and AI if there are no meaningful interactions. In this talk, I will present our effort to facilitate the Human-AI collaboration for certain tasks. I will first introduce our work on improving the controllability of deep generative models such as GANs for interactive image editing. I will then talk about Human-AI Copilot (HACO), a new human-in-the-loop reinforcement learning method to efficiently teach agents to drive in uncertain environments while preserving their own curiosity and exploration.
Bolei Zhou is an Assistant Professor in the Computer Science Department at the University of California, Los Angeles (UCLA). He earned his Ph.D. from MIT in 2018. His research interest lies at the intersection of computer vision and machine autonomy, focusing on enabling interpretable and trustworthy human-AI interaction. He has developed a number of widely used interpretation methods such as CAM and Network Dissection, as well as computer vision benchmarks Places and ADE20K. He is an associate editor for Pattern Recognition and has been area chair for CVPR, ICCV, and AAAI. He received MIT Tech Review's Innovators under 35 in Asia-Pacific Award. More about his research is at .