With the rise of AI, it is important to strike a necessary balance between promoting open sharing and protecting Intellectual Property, such as AI models. Our ModelGO project tackles the issue of AI model licensing. It is a foundational project of our Institute of Data Science, flagship program on Trustable Model-centric Sharing for Collaborative Machine Learning.

Anomaly Detection is a very challenging problem as we are asking the machine to detect what it has never seen before. I’m delighted to introduce our research project at the Institute of Data Science, focusing on designing new AI methods to improve the detection of data anomalies.

Have you wondered how can AI learn and retain knowledge from video as intuitively as humans do? At the Institute of Data Science, we have developed a 'smart TLDR' model by using AI to identify and focus on key highlights in long videos.

At the Institute of Data Science, we are concerned about the safety of AI and we are constantly trying to discover new vulnerabilities. In this work, we discover that even black-box Large Language Models can be attacked if they reveal their confidence too much.

With AI's ability to generate code, we need to ensure that we have transparent ways to evaluate and improve AI-generated systems. At the Institute of Data Science, we have developed an open online platform to fairly and dynamically evaluate AI coding models.