How Singapore Is Using AI and Data for Public Good
- By DSAITrends editors
- July 13, 2022
Slowly but surely, AI is making its way into our lives. For instance, one AI project at the Housing and Development Board (HDB) in Singapore helped citizens save some two million dollars annually, while the SkillsFuture initiative benefited from the use of AI to flag potentially fraudulent claims.
These are just two examples of how the Singapore Government is driving data transformation and delivering public good through data science and AI, as shared at last week’s Stack-X Data Science Connect virtual conference.
Data and AI for the public good
Dr. Alvina Goh, the director of the Data Science & Artificial Intelligence Division (DSAID) at GovTech outlined a host of projects her team had worked on in a presentation titled “AI for the public good – moving from hype to real applications”.
According to Dr. Goh, AI is used in a variety of use cases, ranging from policy to operational work. As reported by local blog Vulcan Post, the three most interesting ones are probably HDB’s example, its role in fraud analytics, as well as video analytics for urban planning.
Anyone who has purchased a resale HDB flat in Singapore or sold their HDB flat would be familiar with the valuation process. While the mechanics have evolved over the years, it usually entails a visit by an independent surveyor to assess the worth of the HDB unit changing hands.
With help from GovTech, AI was incorporated into the HDB resale portal so that prices deemed reasonable could see the need for the surveyor waived. This helped 90,000 resale flat buyers save about SGD2 million in valuation fees every year, says Dr. Goh.
After a series of egregious and high-profile scams with SkillsFuture Singapore – the national initiative to help Singaporeans and Singapore residents embrace lifelong learning, the DSAID team was approached to help flag potentially fraudulent claims using AI.
According to Dr. Goh, the team eschewed supervised learning and instead adopted an unsupervised learning model to better detect new types of fraud. While not mentioned, the steady stream of prosecution in the news presumably meant the technology was also deployed for other government-sponsored schemes, too.
Finally, an object detector was also built to help government agencies automatically classify the modes of transport used by people: on foot, riding a bicycle, or on a personal mobility device. This contributed to better urban planning, and probably beats manual counts by government employees.
Image credit: iStockphoto/Peera_Sathawirawong