Five minutes with Professor Longbing Cao
We chat to ARC Future Fellow Professor Longbing Cao, who recently joined the School of Computing as the Distinguished Chair of Artificial Intelligence.
1. Tell us a bit about your background and what brought you to Macquarie University.
After leaving a chief technology officer role, I joined UTS. There I was motivated to develop actionable research and innovations in artificial intelligence (AI) and data science, which could address real-world data challenges, and some gaps and limitations in existing theories and commercial tools. I established the pioneering UTS Advanced Analytics Institute in 2011 to significantly promote cross-disciplinary, cross-domain research, course development, and industry/government engagement in AI and data science, demonstrating significant leadership and socioeconomic impact.
Since 2005, I have also been promoting ‘data science’ in theory and application (see more in the Data Science Lab). I initiated a series of activities, including the annual Big Data Summit, IEEE Task Force and International Conference on Data Science and Advanced Analytics (CORE A), Springer-Nature’s Journal of Data Science and Analytics, ACM KDD2015 in Sydney, and Master/PhD in Analytics.
With the strategic appointment of being the Distinguished Chair in AI, I am thrilled to join Macquarie University and collaborate with researchers across faculties, disciplines, and domains to further enhance Macquarie’s leadership and impact of AI and data science in research, education, and engagement nationally and internationally.
2. Tell us a bit about your research
The existing statistical, computing and informatics theories often assume objects, data and variables may be independently drawn from identical distributions (IID), which violates and conflicts with real-world non-IID nature, i.e., full of complicated interactions, couplings, and heterogeneities (non-IIDnesses) in complex systems, behaviours and data. This requires non-IID learning, an area I proposed over 15 years ago, to originate new ‘beyond IID’ theories and methods advancing statistics, informatics, and computing.
Deep learning has revolutionised AI, data science, and machine learning, etc., in particular, by the recent ChatGPT and large language models. However, they still suffer from critical issues, including uncertainty, expandability, network vulnerability, and distributional vulnerability. My team has been comprehensively working on new shallow-to-deep statistical and machine learning, including deep non-IID learning, to address such issues. This aligns with the key objectives of my ARC Future Fellow grant, where I aim for original significant theoretical and technical results to quantify and compute complex interactions and heterogeneities that could not be handled (well) by existing state-of-the-art tools.
3. What drives you?
I have been impassioned by fascinating complexities in human, natural, socioeconomic, cyber and virtual systems, behaviours, and data. Understanding and quantifying their complexities requires original methodical computing and informatics theories and tools to address their intrinsic and intricate characteristics and challenges. This has motivated a variety of my proposed concepts and areas, including data science thinking, behaviour informatics, actionable knowledge discovery, and non-IID learning. For example, the age of data and new-generation AI go beyond computational thinking, requiring data science thinking fostering new thinking paradigms to address system complexities, data challenges, digital futures, and enterprise innovations. My sole-authored monograph ‘Data Science Thinking - The Next Scientific, Technological and Economic Revolution’ (Springer, 2018) addresses the United Nation’s Sustainable Development Goals, was translated to Turkish and Russian, will be translated to Vietnamese, and I am writing a new edition in Chinese to be published by Tsinghua University Press. Since joining academia in 2005, such original and actionable data science and AI thinking have inspired my research thinking and results. It has a series of impactful applications such as in AMP, ATO, Centrelink, CBA, IAG and Westpac, reflected by many sole/first-authored publications and the Eureka Prize.
4. Any recent accomplishments you would like to share?
One critical question continuously motivates me: what has happened over the 70 years of AI and 50 years of data science (DS)? With substantial explorations, I have published a series of sole-authored articles to summarise the past, present and future and the good, bad, and ugly of AI and data science. My recent work further concerns decentralised AI enabling edge intelligence and smart blockchain, metaverse, Web3, and DeSci; Trans-AI/DS for transformative, transdisciplinary and translational AI and data science; and AI and data science for smart emergency, crisis and disaster resilience. The COVID-19 pandemic has attracted millions of scientists from about 190 countries and regions. We recently updated our comprehensive analytic results to answer questions like ‘how global scientists have responded to address COVID-19’. For this, we applied AI and data science to crawl, extract and analyse research outputs involving 800,000 publications and discover insights on aspects such as how each country has performed in scientifically fighting COVID-19 under the context of their economy, collaborations and COVID-19 frustrations. In addition, our recent work on deep non-IID learning brings about novel information theories for addressing distributional vulnerability of deep neural networks, accepted by IEEE TPAMI and a tutorial to deliver at IJCAI 2023.
5. What are you focusing on for the rest of 2023?
I will continue my ARC Future Fellowship here at Macquarie, aiming for some meaningful results in deep interaction learning and deep variational learning, toward original statistical-neural-integrative theories and methods to understand, quantify and learn complex non-IIDnesses. I am also devoted to my Chinese book on Data Science Thinking and have agreed to write a monograph on Behaviour Informatics, another area I proposed over 15 years ago, to be published by ACM Books. We will further develop behaviour informatics to quantify, compute, interact, predict, recommend, intervene and manage complex behaviours, actions, and communications and their impacts and consequences. I am also excited to learn great things from relevant leaders and researchers across disciplines and areas at Macquarie and explore strategic collaborations and engagements both within the School of Computing, the Faculty, and across other disciplines and areas, such as AI and data science for science, finance, and health.