As the adoption of artificial intelligence (AI) accelerates rapidly, Singapore has taken steps to ensure the energy efficiency of data centers operated in the country and to ensure that government data used to train models is adequately protected. He said that he had taken the necessary measures.
The Ministry of Communications and Information (MCI) said in a statement that the government will build the necessary computing capacity and grow the data center market in a “sustainable manner” in line with international climate change efforts.
Also: Every AI project starts as a data project, but it’s a long and winding road
The Department will review how the government is balancing the growing demand for computing power with the country’s AI development and sustainability goals, and ensuring that the necessary infrastructure is environmentally friendly. He was responding to a parliamentary question about whether
The Ministry acknowledged that data center AI computing power is a key enabler of Singapore’s national AI strategy, and one of the key strategies is to improve data center energy efficiency and develop efficient cooling solutions. He said that the goal is to promote He cited liquid cooling as an example of what AI computing infrastructure often relies on, adding that it is more energy efficient than air cooling for intensive AI workloads.
There are also measures such as subsidies to support businesses working to reduce greenhouse gas emissions from data centers.
Additionally, Singapore is developing sustainability standards that will pave the way for data centers to run at higher temperatures and use less energy for cooling, MCI said.
Last June, Singapore announced operational standards to optimize energy efficiency for data centers in tropical climates. Recommended standards developed by the Infocomm Media Development Authority provide a roadmap for increasing data center operating temperatures to 26 degrees Celsius or higher. Citing research from the University of Toronto, the agency said the standard could result in energy savings of 2% to 5% for every degree Celsius increase.
A data center set up to operate in such climates began operations last month, providing facilities for researchers and industry players to develop energy-efficient cooling technologies. Touted as the world’s first data center testbed for tropical environments, the new site is hosted by the School of Design Engineering at the National University of Singapore’s Kentridge Campus. The university is leading the effort in collaboration with fellow local university Nanyang Technological University.
Related article: Tropical data center now operational in Singapore
MCI said Singapore is also reviewing its Green Mark certification system for data centers to update its energy efficiency standards.
“Beyond sustainable AI computing infrastructure, it is important to invest in the development of green computing methodologies,” the ministry noted. This includes coding and algorithm optimization, both software and hardware optimization, and standard development of low-data, low-energy AI models.
“The government will continue to deepen its international and domestic partnerships with the research community and industry partners in this area,” it added.
Data used for AI models should be properly protected
Measures are also being taken to manage sensitive information used to train AI models, the MCI said, adding that measures have also been taken to manage sensitive information used to train AI models. In response to a question about whether
“We take a risk-managed approach to our LLM [that is] “It is consistent with existing public sector frameworks for handling sensitive information when using technologies such as internet-based applications and commercial clouds.”
He noted that sensitive applications and data cannot be accessed online.
“For use cases involving sensitive data, the open source model can be fine-tuned for use, but must be deployed on government servers and computers,” MCI said.
Related article: New research initiative aims to build large-scale language AI models for Southeast Asia
When less sensitive data is involved, AI models can be owned and controlled by commercial or private companies, MCI said. Government contracts with these companies are governed by service agreements that include provisions regarding data handling and security, the report said. This includes non-retention of data and limiting the use of data to train other products or models.
The government has also introduced technological measures to monitor sensitive data and provide visual cues to remind users of data security policies. MCI further said it has governance measures in place to enforce compliance, adding that it will continue to re-evaluate the effectiveness of such measures as technology evolves.
Last month, Singapore launched a research initiative to build LLMs that better respond to the demographic dynamics of the Southeast Asian nation. The initiative, named the National Multimodal LLM Programme, will build on the current work of AI Singapore’s Southeast Asian Languages in One Network (SEA-LION). SEA-LION is an open source LLM that the agency says is designed to be smaller, more flexible, and faster compared to LLMs currently on the market. SEA-LION is currently running on his two basic models: a 3 billion parameter model and a 7 billion parameter model.