TSVP Talk: "Reducing AI Agnostophobia" by Shu Kong

Title: Reducing AI Agnostophobia

Speaker: Shu Kong, University of Macau

Abstract:

Artificial Intelligence (AI) transforms everything from autonomous vehicles to scientific research. Yet, its rapid adoption has triggered "AI agnostophobia", a fear of the unknown exacerbated by unpredicted AI failures, such as autonomous vehicles colliding with overturned trucks and AI assistants exhibiting severe demographic biases. To reduce AI agnostophobia, we talk demystifies the bedrock of modern AI: Foundation Models (FMs) trained on billions of internet data. By analyzing such data, we expose the underlying data imbalances that inherently drive model bias; we present simple post-hoc techniques that can mitigate this bias.

Nevertheless, can we safely rely on AI to analyze our data in highly specialized fields like biology and ecology? To address this question, we introduce AutoExpert, a practical research framework designed to automate domain-data annotation with expert-crafted guidelines. By presenting AI-based approaches and key insights, we delineate both the limitations and the promises of expert-AI collaboration, illustrating how this synergy can accelerate interdisciplinary research.

Profile:

Shu Kong is in the faculty of Computer Science at the University of Macau, with prior academic appointments at Carnegie Mellon University and Texas A&M University. He holds a Ph.D. from UC Irvine. His research spans computer vision, applied machine learning, and interdisciplinary science. He actively promotes the field of Open-World Vision, on which his work earned Best Paper / Marr Prize nomination at ICCV 2021. His previous interdisciplinary breakthroughs include an automated high-throughput pollen analysis system, which was highlighted by the U.S. National Academy of Sciences (NAS) and the National Science Foundation (NSF) as "opening a new era of fossil pollen research." Personal Homepage

Language: English

Target audience: General audience/everyone at OIST and beyond.
Freely accessible to all OIST members and guests without registration.

This talk will also be broadcast online via Zoom:
Meeting ID: 993 1216 5065
Passcode: 603487

※ Please note that this event may be recorded and the videos uploaded. In addition, photos may be taken during the event. These are intended for publication online (the OIST website, social media, etc.)※