Research Seminar by Xiang ‘Anthony’ Chen | No More UI Wrappers! Rethinking HCI's Role in the Era of Generative AI

No More UI Wrappers! Rethinking HCI's Role in the Era of Generative AI

There is a recent surge of HCI work that feels like derivatives of popular generative AI, such as adding UI wrappers to off-the-shelf LLMs and text-to-image models. In this talk, I draw on my own research experience to rethink HCI's role in the era of generative AI. Specifically, I argue that HCI should challenge itself with three important sets of problems. First, while generative AI (e.g., ChatGPT) performs fairly well in most general use cases, it remains unclear whether and how they can support experts' domain-specific workflow, such as drug discovery. Second, one fundamental problem of the current generative AI interfaces is the limitation of expressing and interpreting humans' intent in language alone. Third, as generative AI becomes increasingly capable, HCI should contribute to the long-standing problem of value alignment---specifically, how to enhance human autonomy and AI fairness.