reading time of blog

Singapore has carved out dedicated cloud resources so its government agencies can deploy artificial intelligence (AI) applications more efficiently and securely. 

The cloud cluster has been established with the aim of driving up AI adoption in the public sector and supporting research into how AI can be applied, said the Smart Nation and Digital Government Office (SNDGO). It will also support local AI startups, according to the government agency.

Called the AI Government Cloud Cluster, the platform runs within a dedicated environment on Google Cloud, where the U.S. vendor's AI technology stack and partner applications are available for deployment. These resources include A2 supercomputers running on Nvidia's A100 GPUs and a repository of AI models, running first- and second-party as well as open source platforms, which government agencies can customize for their specific requirements. The AI models span multi-language text translation, audio-to-text conversion, and software coding. 

Doing things differently with AI

Several global "shifts" have underscored the need for a new approach to technology, with AI and cloud playing key roles, according to Chan Cheow Hoe, SNDGO's government CTO and senior advisor for the Singapore Economic Development Board. 

Speaking at the Google summit, Chan pointed to environment and technology changes and said people now want to know the impact digitalization has on the environment and on their personal security. More people are exposed to cybersecurity risks and investors want to see results, with businesses having to work harder for every dollar. 

There are more significant trust and safety concerns compared to a couple of years ago, where fewer people cared about these issues, he said. There is also zero tolerance for downtime, where online services and apps that go down for an hour will make headlines. 

With technology now omnipresent and touching every facet of daily life, Chan stressed the need to safeguard customers and citizens and ensure systems and services are trusted -- otherwise no one will want to use the technology. 

There have also been big technological shifts and he pointed to cloud, alongside SaaS (software-as-a-service), and generative AI as the key ones that hold significance. 

Heralding the Singapore government's cloud journey, which began seven years ago, Chan said the move was important not just for cost efficiencies, but also to provide access to a global ecosystem of the best technology resources. Organizations that do not open up to the cloud also risk running bespoke applications that might not be able to scale quickly and support faster times to market, he added. 

However, many organizations remain straddled with legacy systems, including Singapore's public sector, according to Chan. There is also a dearth of relevant IT capabilities, without which the journey toward digital transformation will be difficult. 

In addition, policies need to catch up to an environment that is increasingly powered by the cloud, he said. Unless Singapore adopts a progressive attitude toward policies, brings in the best talent, and gets rid of legacy systems, any step forward could mean another two steps back. 

This situation creates a compelling case to do something new, Chan said, pointing to the government's efforts to boost AI and cloud adoption. He said a new iteration of GCC will be equipped with higher security measures to enable more confidential and critical workloads to be moved to the cloud infrastructure. 

Agritech company Jiva is also taking these kinds of precautions in its adoption of AI, including generative AI. Its mobile app recently introduced a new feature, called Crop Doctor, which lets farmers load images of crops to diagnose diseases and recommend relevant treatments that can be tailored to their requirements. 

Asked if the company had concerns about tapping generative AI amid the reported security issues, Jiva's head of business strategy and partnership Aditya Thareja acknowledged there were worries about the potential risks. "We want to ensure things like hallucinations [occurring] are reduced to a small percentage. The last thing you want is to provide the wrong advice to farmers," Thareja said during a media briefing on the sidelines of the summit.

Jiva, which has operations in Singapore, Indonesia, and India, built Crop Doctor on Google's Vertex AI, tapping computer vision and image processing to identify crop diseases and causes, such as potassium deficiency. It recommends treatments based on an in-house data library 

To mitigate potential risks, Thareja said the large language AI model is trained on only data that is sourced and vetted by Jiva. The generative AI model does not scour public data from the internet, he stressed, adding that the company continuously finetunes the prompt-engineering capability to ensure the right answers are provided. 

Crop Doctor was tested on WhatsApp with 25,000 farmers before the feature was launched on Android in Indonesia, where the mobile OS has the highest penetration. Jiva currently has 125,000 farmers on its registry in the Asian market. 

An earlier iteration, called Crop Care, also provides treatment recommendations based on crop images, but does not customize these results based on a farmer's specific access to the required components.  

According to Jiva, Crop Doctor currently clocks an accuracy rate of above 90% for common major crop diseases, if the picture is in focus and the subject matter is correctly identified. The agritech vendor told ZDNET it continues to collect images from real-world cases from farmers to further improve the accuracy rate.