Filtered by: School of Computing
Anthropic made headlines last month when it called for a worldwide pause in frontier AI development, warning that AI systems may be approaching a point where they can autonomously improve themselves without human input.
Writing in Channel NewsAsia, NUS School of Computing's Professor Jungpil Hahn takes the proposal seriously — but asks the question the report leaves unanswered: a pause toward what, exactly? Without defined exit criteria, he argues, it amounts to little more than a deferral. The enforcement challenge is equally steep: unlike nuclear arms control, AI training runs are easy to conceal, and in the context of US-China strategic competition, coordinated compliance would be strategically irrational for either side.
Prof Hahn also raises a dimension the debate has largely missed. Every governance framework depends on human judgment to function — and that judgment is being quietly shaped right now by the AI tools already embedded in how policymakers research, how analysts reason and how the next generation of experts is trained. For Singapore and the ASEAN region, he argues, how we structure AI adoption today will determine whether we retain the institutional capacity to participate meaningfully in AI governance at all.
Professor Anthony Tung from the Department of Computer Science appeared on 狮城有约 (Hello Singapore) to analyse the wave of tech giant IPOs – from SpaceX's record-breaking public offering to the anticipated listings of OpenAI and Anthropic – and what they signal for investor and everyday AI users.
On SpaceX's extraordinary market reception, Prof Tung pointed to something more durable than hype: a long track record of delivering on technically ambitious promises. Rocket recovery technology, once widely dismissed as impractical, has become routine – and that execution credibility, he argued, is what gives investors real confidence. He also noted that SpaceX's integrated hardware-software model gives it a structural edge over pure software players, enabling rapid deployment of AI into physical devices and robotic systems.
As AI companies prepare to meet capital market expectations, Prof Tung pushed back on the assumption that monetisation pressure would trigger a price war. With open-source models and localised solutions already a significant part of the ecosystem, he argued that competition is more likely to play out on accuracy, safety, and functionality than on cost alone. On OpenAI's reported move towards advertising revenue, he observed this follows a well-worn path for internet businesses: free users may see more ads, while paying users gain a fuller, less interrupted experience.
A research project co-led by Professor Abhik Roychoudhury from the Department of Computer Science was featured in The Straits Times in a report on Singapore's national AI-for-Science (AI4S) programme.
The article highlighted the project's aim to ensure that the growing volume of AI-generated code is safe and reliable – developing AI tools to automatically detect bugs, verify that software behaves as intended, and help developers and security professionals audit critical systems. The project, AI for Program Reasoning, is co-led with Professor Christian Cadar from Imperial College London, with partners from SMU, MIT, and ETH Zürich.
Professor Anthony Tung from the Department of Computer Science was featured on CNA's Singapore Tonight live segment, speaking on the AI bubble emerging in manufacturing and the broader cost of deployment without accountability.
Prof Tung identified three patterns of waste in current AI deployment: near-identical foundation models competing within narrow benchmark margins; AI-for-science programmes built on survivorship bias; and AI assistants handed to individual employees for tasks that never aggregate into organisational value.
The root cause, he said, is the absence of AI Deployment Science – a discipline for evaluating return on investment before resources are committed. Without it, capital follows fashion.
"A company of 10,000 people asking the same question 10,000 times is not a learning organisation. It is a forgetting one."
He proposed AI Prudence as the remedy: before every deployment, ask where the value is, how it will be measured, and whether an existing capability could already do it better.
CNA, Singapore Tonight (26 May 2026)
Associate Professor Harold Soh from the Department of Computer Science was quoted in South China Morning Post on Singapore's push to become a global leader in physical AI – robots and autonomous systems designed to operate in real-world urban environments.
Prof Soh noted that Singapore's institutional strength, technical talent, and track record in deploying technology in complex urban settings give it a credible edge in this space. "Singapore is already a trusted international hub, with strong institutions, concentrated technical talent, and experience deploying technology in complex urban environments.
The article. published ahead of ATxSummit 2026, examines Singapore's ambitions to position itself as a living lab for physical AI – from cleaning and delivery robots at Punggol Digital District to longer-term deployment in healthcare, logistics, and manufacturing.
SCMP (22 May 2026) - "Robots at Singapore's AI Zone to Clean, Patrol and Deliver Goods"
In a Straits Times feature on Singapore's national AI strategy, Professor Jungpil Hahn, Provost's Chair Professor at NUS School of Computing highlighted a key concern amid the excitement over AI adoption: the potential for deskilling.
Referring to a study published in The Lancet Gastroenterology and Hepatology in August 2025, Prof Hahn observed that clinicians who frequently depended on AI to detect pre-cancerous lesions gradually lost their ability to identify these growths on their own. This issue goes beyond healthcare, highlighting how professionals in any field might, over time, diminish the very skills AI was designed to enhance.
He suggested setting aside intentional AI-free intervals. "Having explicit days or periods where you know you have to do the task without AI actually forces the institutions, companies or organisations to maintain that capability level," he explained. He also urged organisations to monitor employees' abilities before and after adopting AI – not to restrict the technology, but to make sure human judgment stays sharp.
According to Prof Hahn, the real issue isn’t whether to adopt AI, but how to do so without sacrificing what no algorithm can ever replace.
