Through the project, the team proposes an integrative program of fundamental research towards a vision in which every human will have such an AI assistant for daily life and work. The overall aim is to build conceptual understanding of human-AI collaboration, to develop representations, models, and algorithms for situated assistance, and to integrate them in an experimental device platform for evaluation. The research program consists of five thrusts: (i) situated language communication with reasoning, (ii) visual-linguistic situation understanding, (iii) human collaboration modeling, (iv) robust situated teaming, and (v) the integrative showcase project The “Other Me”, which develops and evaluates the experimental platform Tom – a wearable situated AI agent that assists the human in creating novel artifacts or diagnosing faults. The research aims to empower a new generation of AI assistants for situated, just-in-time, and federated assistance. They will reinvent the relationship between the human and the device in our daily life and work. This research/project is supported by the National Research Foundation, Singapore under its AI Singapore Programme (AISG Award No: AISG2-RP-2020-016). Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not reflect the views of National Research Foundation, Singapore.
As AI becomes ubiquitous, people’s trust in AI is actually dwindling. The key barrier to adopting AI is no longer technical in nature, but more about gaining stakeholders’ trust. Federated Learning (FL), in which training happens where data are stored and only model parameters leave the data silos, can help AI thrive in the privacy-focused regulatory environment. FL involves self-interested data owners collaboratively training machine learning models. In this way, end-users can become co-creators of AI solutions. To enable open collaboration among FL co-creators and enhance adoption of the federated learning paradigm, this project aims to develop the Trustworthy Federated Ubiquitous Learning (TrustFUL) framework, which will enable communities of data owners to self-organize during FL model training based on three notions of trust: 1) trust through transparency, 2) trust through fairness and 3) trust through robustness, without exposing sensitive local data. As a technology showcase, we will translate TrustFUL into an FL-powered AI model crowdsourcing platform to support AI solution co-creation.
In safety-critical applications (e.g., where human lives are at stake), it is crucial for humans to be well-trained to handle expected and unexpected scenarios of varying complexity. Paramedics frequently respond to life-or-death situations, and they must be trained to handle expected and unexpected situations with respect to patient condition and response to treatment effectively. Manned vessels in maritime traffic operate in environments with many other vehicles, and humans must be trained to safely navigate varied situations. A wide range of critical activities in crime response, healthcare, defence, and construction also require training to improve safety. Through this project, the team intends to develop and assess Explainable and tRustworthy (ExpeRt) AI (or Agent) Training Programs (ATPs) with feedback interfaces to adaptively train human(s) for safety-critical applications with showcase projects on emergency response and maritime navigation. The ExpeRt ATPs will generate safe, unexpected scenarios that adapt to observed learner deficiencies and yet provide fair and comprehensive coverage of all cases/situations. This research/project is supported by the National Research Foundation, Singapore under its AI Singapore Programme (AISG Award No: AISG2-RP-2020-017). Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not reflect the views of National Research Foundation, Singapore.
Computer Generated Forces (CGFs) are artificial entities in tactical simulations modelled after human behaviour. In a previous project with DSO, Intelligent Computer Generated Forces (iCGF) with adaptive capabilities have been developed to provide a viable alternative to script-based CGF. Through reinforcement learning, these iCGFs are designed to acquire knowledge while interacting and operating in a dynamic environment. This project with DSO builds on the capabilities developed in the previous project to enhance the study of iCGF by considering the following two key research questions; 1. How to scale up the applicability of iCGF, in terms of the number of iCGF entities as well as their operability in more complex scenarios and; 2. How to interpret and demonstrate the knowledge learned by iCGF in the form of high-level description.
What determines the coverage of Covid-19 and related political contestation in traditional and new media? In collaboration with the University of Bamberg in Germany, this project sets out to identify hidden power structures between media organizations in contemporary hybrid media systems in Germany, UK, USA and South Korea. This will provide insights about the determinants of information quality and the spread of misinformation during a large social crisis in media coverage.
This project addresses the inefficiencies in the lead generation, prospecting, engagement and qualification processes which are important stages in the acquisition of high net worth and ultrahigh net worth clients in a financial services business. The current process of digital leads generation and prequalifying is inefficient as it involves the manual qualification of large volumes of names collected from social media posts and marketing campaigns. Furthermore, client advisors do not have enough actionable insights on the leads provided by marketing and prefer to follow their own warm leads (e.g. networking) instead of the ones assigned to them. The objective of this project is to significantly improve the efficiency and effectiveness of these activities by adopting modern digital technologies including Artificial Intelligence models and Big Data analytics.
This research project is supported by the National Research Foundation, Singapore under its AI Singapore Programme (AISG Award No. AISG-100E-2020-062).
Current code comparison technology mostly only work in a single language, failing to support inter-language migration. This project looks at closing this gap, through exploring new techniques for comparing code similarity across different languages and paradigms. Through this research, the team aims to improve the quality of software and reduce software development cost.
This project addresses the inefficiencies in the KYC (Know Your Customer) and Due Diligence processes which are critical yet resource-intensive activities in a financial services business. These activities are particularly complex in the context of Wealth Management, as compared to retail banking, where the client profiles of high net worth and ultrahigh net worth individuals are typically associated with a wider multi-national network of other family and business relationships and company structures and entities which need to be considered holistically in order to build a full understanding of their profile. The objective of this project is to significantly improve the efficiency and effectiveness of these activities by adopting modern digital technologies including Artificial Intelligence models and Big Data analytics.
This research project is supported by the National Research Foundation, Singapore under its AI Singapore Programme (AISG Award No. AISG-100E-2020-058).
Under the Fujitsu-SMU Urban Computing and Engineering (UNiCEN) Corp Lab, SMU has undertaken the Digital Platform Experimentation (DigiPlex) project with Fujitsu. The project was carried out using the Digital Annealer (DA), a quantum inspired-technology inspired by Fujitsu. Through the DigiPlex project, certain challenges in solving constrained optimization problems using such technology, and promising methods on tuning of the underlying model parameters to improve run time performance, have been identified. This project aims at developing hyper parameter tuning methodology, machine learning techniques, operations research algorithms, and software tools to enhance quantum-inspired techniques for solving large scale real-world combinatorial optimization problems.
This project aims to develop Fast-Adapted Neural Networks (FANNs) and provide specific solutions to equipping the Advanced Manufacturing and Engineering (AME) systems with FANNs. It considers a wide range of AME application examples such as visual inspection of new product parts and automated identification of product defects. It will improve the yield rate and reduce manufacturing costs, when FANNs-based devices are widely installed in the design, layout, fabrication, assembly, and testing processes of production lines. This research is supported by A*STAR under its AME YIRG Grant (Project No. A20E6c0101).
