100 solutions in 100 days: Singapore government jumpstarts Generative AI capabilities with sandboxes, workshops
By Yogesh Hirdaramani
In partnership with Google Cloud, a coalition of digital government agencies in Singapore is rolling out generative AI sandboxes and hands-on workshops to develop 100 generative AI use cases in 100 days.
At the launch event for the AI Trailblazers initiative, Minister Josephine Teo viewed demos of generative AI solutions developed by startups and public sector teams. Image: Google Cloud Singapore
On 24 July 2023, a coalition of digital government agencies in Singapore announced a joint initiative in partnership with Google Cloud to drive generative AI capabilities in the public and private sector.
The “AI Trailblazers” initiative, led by the Ministry of Communications and Information, Digital Industry Singapore, and the Smart Nation and Digital Government Office (SNDGO), will enable public and private organisations to identify and address 100 real world challenges that would benefit from generative AI solutions.
Over the course of 100 days, representatives from these organisations can use innovation sandboxes and attend mandatory workshops in an effort to rapidly pick up AI skills, build generative AI prototypes and bring these prototypes to production. At the end of the year, participants will demonstrate their products to a panel of technology experts at an awards ceremony.
Driving generative AI solutions in public, private sectors
Through this sprint, the government is driving a coordinated whole-of-society approach to deepening Singapore’s generative AI capabilities across the private and public sectors, with the endgame of advancing Singapore’s position as an open and trusted global AI hub.
Organisations will be able to tap on innovation sandboxes to access Google Cloud’s ML development platform, Vertex AI, pre-trained generative AI models, low-code developer tools, and high-performance graphical processing units (GPU) to build customised AI solutions to address their specific needs.
Public sector organisations can do so with a dedicated sandbox administered by SNDGO through the Artificial Intelligence Government Cloud Cluster, while Singapore-based companies can use a sandbox administered by Digital Industry Singapore.
“These [Generative AI] models can not only be deployed as point solutions to raise productivity, but they can also be embedded deep within our business models and tech stacks, delivering new kinds of value for consumers and citizens,” said Josephine Teo, Minister for Communications and Information, at the launch of the initiative.
But why develop your own solutions, instead of using publicly available AI models like ChatGPT? Karan Bajwa, Vice President, Asia Pacific, Google Cloud, explained in his opening speech that developing custom enterprise AI helps ensure such tools only draw information from accurate, company approved sources, preventing “hallucinations”.
Developing in-house solutions can also help ensure that information fed into these models by end users will not be leaked. In the public sector, this can mean ensuring that unfinalised policy plans remain internal, said Wei Boon Goh, Chief Executive of GovTech, in a panel discussion.
“For generative AI to be truly transformational, we need to look beyond the consumer-facing chatbots that have gained significant popularity. Governments and companies will need to build their own generative AI solutions for their employees, citizens, and customers,” said Bajwa to GovInsider.
Singapore government an early mover
The Singapore government has already emerged as a critical early mover when it comes to the adoption and acceleration of generative AI solutions. In February, Open Government Products announced a virtual writing assistant prototype modelled on ChatGPT customised for government use. This tool is now being used by over 4000 civil servants for writing tasks.
“The speed at which the government has worked with us is absolutely unprecedented and that's why this is a path breaking partnership,” said Bajwa.
During a product showcase, representatives from GovTech’s Virtual Intelligent Citizen Assistant, shared with reporters that the government will be migrating the public service’s 88 chatbots from more simplistic decision-tree structures to large language models (LLMs) that can offer better conversational flow and quality by the end of this year. This will even include the ability to understand and respond to Singlish.
Currently, seven of these chatbots have already been migrated to LLMs. These include public-facing chatbots like those run by the Housing Development Board (HDB) and GoBusiness. They shared that as a result of the move, the development time for these chatbots has dropped from 30 to 40 hours to a mere three hours.
Internal chatbots meant for civil servants have been developed on Google’s Vertex AI platform, which public agencies can access through the AGCC, they explained. Internal chatbots contain more sensitive data and the AGCC can help ensure the development of secure and trusted AI chatbots. Public facing chatbots are run on the Government Commercial Cloud 2.0, and can tap on foundation models and tools developed by other cloud providers.
Unlike the more cautious response to last year’s hype around Web3 concepts like the metaverse and NFTs, the government has moved decisively to tap on the benefits offered by generative AI applications.
In her opening remarks, Minister Teo highlighted that the transformative potential of generative AI “appears to be much more within reach for many more people. And with the expanding possibilities, the scope to harness AI for public good outcomes at speed and scale have multiplied accordingly.”
This could include tackling healthcare and sustainability challenges, from improving preventive and precision medicine to helping organisations meet net-zero goals, she explained.
In the coming months, we can expect to see further generative AI initiatives jointly run by the Singapore government and Google Cloud, including upskilling programmes, start-up accelerator programs, and the operationalisation of responsible AI principles.