AI models and systems have become ever more powerful and are on track to automate many tasks that otherwise require human cognitive skills. It is most important that we design and implement responsible AI models and systems that do not amplify societal biases and discriminations; that are faithful to truth and can mitigate misinformation; that have guard-rails against misuse and intentional harm; that are resource-efficient and sustainable; that are transparent about the training data source, model architecture and parameters, evaluation measures and benchmarks, as well as the process of building them. The goal is to build AI models and systems that are trustworthy for users and promote mutual understanding between human societies. These objectives call for responsible and ethical design of algorithms and system processes. We are working on various technical aspects of responsible AI, faithful and trustworthy AI to buttress various governance efforts.


AI and machine learning are powerful tools for humans to solve societal problems. Our team works on projects to construct models and systems to help patients and the elderly, to detect and mitigate misinformation and biases from human sources on the internet, and to apply machine learning to life science research.


Generative models for Conversational AI are less than a decade old,  but they hold great promise for human-machine interactions. Machine responses based on generative models can seem quite fluent and human-like, empathetic and funny, knowledgeable and professional. However, behind the confident voice of generative ConvAI systems, they can also be hallucinating misinformation, giving biased and harmful views, and are still not "safe" enough for many real life applications. The expressive power of generative ConvAI models and their undesirable behavior are two sides of the same coin. How can we harness the fluency, diversity, engagingness of generative ConvAI models while mitigating the downside? Our team is working on making generative ConvAI safer via mitigating hallucinations, misinformation, and toxicity, while improving its factual knowledge and reasoning.


The world is home to an incredible diversity of languages, with vibrant linguistic landscapes. Nevertheless, there are the lack of research attention and technological support for the vast number of languages spoken worldwide, particularly those in Africa, Asia, and the Americas. Despite there being thousands of languages with millions of speakers, the majority of research papers in the field of natural language processing (NLP) continue to focus on English. This language bias persists even after a decade since the need for language independence was emphasized. We emphasizes the urgent need to address this linguistic and geographical imbalance in research and technology development, and we conduct research and development on underrepresented languages. As an example, despite being the second most linguistically diverse country, research attention on languages in Indonesian are often neglected by the research community. To this end, We focus on developing various resources and benchmarks for languages in Indonesia including IndoNLU, IndoNLG, and the first-ever parallel resource for 10 low-resource languages in Indonesia, NusaX. Our works pave the avenue to address the pressing issue of linguistic and geographical representation in research and technology development, in order to create a more balanced and inclusive approach.



Intelligent systems have evolved to the stage where virtual assistants are ubiquitous in smartphones and consumer-grade robots are affordable. Better human-machine interactions depend on machines being able to empathize with human emotions and discover their intent. We are building empathetic machines that are able to recognize and detect meanings and intent during conversations with the human, from their speech, language, facial expressions and context. In addition to understanding the content of speech, the machine needs to recognize if and when the human user is being sarcastic or joking, or if the human user is distressed and needs comforting. The "empathy module" is indispensable for machines and robots of the future that will become caretakers and companions to humans.


The rapidly growing elderly population in Hong Kong, coupled with limited healthcare resources, calls for urgent machine-assisted elderly care solutions. While existing gerontechnology focuses on resource allocation and safety monitoring, it indirectly improves the well-being of the elderly through better caregiver access. To bridge this gap, our aim is to provide a fully integrated elderly care solution, addressing essential healthcare needs like nutrition, medication management, well-being monitoring, remote communication with family and medical professionals, empathetic conversations, and home navigation. By prioritizing human-centered design and conducting rigorous studies, we propose AI, machine learning, and robotics solutions using computer vision and conversational AI. Ensuring safety and trustworthiness of the technology and leveraging anonymized data for insights, we strive to enhance the elderly population's healthcare needs and improve hospital care services in Hong Kong.


We are collaborating with artists, designers and art and design students at the Central Academy of Fine Arts in Beijing, on using AI tools for creative arts, ranging from machine generated art and paintings, to exploring the artistic meaning of AI technology. We are interested in using AI for the future of design and in using AI as a new medium of art creativity. Meanwhile, we are interested in exploring machine creativity and machine learning of aesthetic taste.