Charting a cohesive national strategy for AI
22 Jul 2024
In 2021, Malaysia introduced its National Artificial Intelligence (AI) Roadmap 2021-2025, which outlined the government’s plan for developing the AI ecosystem. However, the AI landscape has evolved significantly with the advent of generative AI (GenAI), marked by the launch of ChatGPT by OpenAI in 2022. This development has necessitated a refresh of the road map.
Following rapid advancements as countries and players compete for leadership in the AI race, Malaysia has introduced several initiatives such as the Malaysia Digital Economy Blueprint, AI untuk Rakyat and the AI Sandbox 2024. Additionally, the country has established its first AI faculty at Universiti Teknologi Malaysia (UTM).
At present, several ministries and agencies handle different aspects of AI. The Ministry of Science, Technology and Innovation, Ministry of Education and Ministry of Investment, Trade and Industry are responsible for shaping policies such as AI governance and a code of ethics, providing funding and facilitating international partnerships. Other stakeholders include the private sector, institutional investors and regulatory bodies as well as universities and research centres essential for developing AI talent and conducting cutting-edge research.
However, there is a lack of a cohesive strategy, leading to a helter-skelter approach where efforts are fragmented and uncoordinated. This could hamper the overall progress and effectiveness of AI initiatives, say stakeholders.
Malaysia needs a well-aligned approach with a clear general direction that all players can follow, focusing on adopting the technology for economic and social development, says Jun-E Tan, senior research associate at the Khazanah Research Institute.
“There are many initiatives already happening, and the main thing is to ensure they don’t happen in silos. We will need to create an environment that promotes communication and coordination. This is because AI and AI risk touch all sectors,” says Tan.
“The pace of technology is such that we will need a good feedback loop that runs through the system to make sure that vital information on new technology developments and associated risks travels quickly.”
This necessitates bridges to be built across ministries and government agencies, government and non-government actors, AI experts and domain experts, and between local and international experts.
For this, a central AI body, that acts as a focal point for AI development, streamlines communication and provides a clear point of contact for stakeholders needs to be established, says Fabian Bigar, former CEO of MyDigital Corporation. Fabian was appointed secretary-general of the Ministry of Digital last month.
“My opinion is that we need a focal point or institution that is the authority on AI. [This is so] we don’t only [speak] about issuing guidelines and road maps, but you actually have a body to implement and oversee it. It could be an adaptation of what [the country] already has [but] formalised into a structure that everyone recognises as an AI authority,” he elaborates.
“One particular entity can be held responsible for the development of AI.”
The central body would also be responsible for refining and implementing a comprehensive national AI strategy while playing a key role in developing educational programmes for Malaysians to navigate the AI landscape.
This body should be well-connected and built to be the platform that stakeholders can turn to on the latest information on AI across different sectors.
“We will need to form these connections early in the game because the use of AI requires trust, and relationships foster trust. It will also enable us to share resources and cooperate among the different stakeholders, which is more effective than a top-down approach,” says Tan.
“With more information coming from the stakeholders, policymakers will also be able to make more informed policy decisions to make sure that the benefits of AI are distributed and no one is left behind, or left worse off.”
Even so, a more nuanced approach might be necessary as different government agencies have their own leadership and areas of expertise, says Laurence Liew, director for AI innovation at AI Singapore, adding that there is a need for an overarching vision that all the agencies can work towards.
AI Singapore acts as a technical bridge between various government agencies and offers its expertise to overcome any technical gaps.
“It is not for a central agency or body to dictate what we can and should not be doing. I think it’s important that as long as there’s an overarching vision, and as long as everyone is pulling in the same direction, we should leave them in their domain. They know best how to drive their ecosystem,” he says.
Mission statement
Science, Technology and Innovation Minister Chang Lih Kang says there is a need for a whole-of-government approach where various ministries collaborate to develop the ecosystem and achieve this goal. Areas that require development include digital infrastructure, research and innovation, talent development, ethical and legal frameworks and public awareness.
“In developing the AI ecosystem, we have to rope in all related ministries. It has to be a whole-of-government approach. We need a cross-ministry committee to look into that. In fact, the prime minister himself is very enthusiastic about this. He has been talking about AI and digitalisation, but for that to happen, we need to make sure the entire ecosystem is robust enough,” he says.
While countries like the US and China have vast resources, smaller players such as Malaysia cannot afford large-scale experimentation, notes Liew. Therefore, a well-defined national strategy that encompasses clear goals, investment plans and an ethical framework is paramount.
“Ultimately it has to be driven from the top. Leaders will need to rally all relevant agencies for the common vision and goals,” says Liew.
This is because Malaysia’s current AI road map is set to expire in 2025. “Unfortunately, the [road map] was released at a time when ChatGPT had not [been released]. I think it is time to refresh the [road map] because AI then and now [has had] a lot of advancements. We need to come up with an enhanced and renewed version of our strategies with AI,” says Fabian.
“There needs to be more discussion. In fact, a few ministries are discussing among themselves [about] whether this new road map is needed and what the things to pay attention to are. I hope if there is a [new] road map, [it will] provide direction for not only the industry but also government agencies … The road map should also include incentives and initiatives that will help boost the ecosystem,” adds Chang.
The AI ethics and governance framework is also being formulated. Chang says the framework will aim to strike a balance between promoting responsible AI development and avoiding stifling innovation with overly rigid regulations.
Prashant Kumar, head of generative AI and growth markets at Accenture Song, identifies three factors for a country to consider when establishing an AI ecosystem. These are the participatory, productivity and national imperatives. The participatory imperative refers to the level of involvement a country wants to have in the AI value chain, while the productivity imperative highlights how AI can be used to enhance the competitiveness of the country in the AI sector.
The national imperative, meanwhile, concerns aspects such as national security and data sovereignty. This is because developing AI is crucial in sectors such as defence and healthcare. These imperatives can be used to then find Malaysia’s seat at the global AI table.
“Let’s say [in the year] 2035, where are we? You probably have a vision of Malaysia. How do you imagine our citizens will live? How do you imagine our workers will work? If we have something like Malaysia AI 2045, we can steer everyone’s attention towards it and work towards [that goal],” says Fabian.
“AI is not only about technology, it’s about how we are going to prepare for it.”
Too soon for an AI champion?
There is a case to be made for the country to develop its very own AI champion. A national AI champion can drive innovation, create high-value jobs and ensure that the country remains competitive in the global AI landscape.
Just as the space race ignited a global revolution that led to the many technological and scientific advances, the AI race will have its own effect on society.
The term AI champion can refer to a leading company that develops its own foundational model, says Prashant of Accenture Song. Foundational models are AI systems that are trained on large datasets that can be used for various tasks.
As other countries develop their own AI and large language models (LLMs), there is a risk of technology monopolies emerging, says Mahadhir Aziz, CEO of the Malaysia Digital Economy Corporation (MDEC), an agency in the Ministry of Digital. Countries without their own AI champions may become dependent on foreign technologies, potentially leading to limited access and higher costs.
Moreover, countries with AI monopolies could exert significant control and influence over the development of AI technologies globally. This can lead to a concentration of power and decision-making in the hands of a small group, which will hinder the technological progress and economic competitiveness of other countries.
“A monopoly could stifle innovation, as dominant players might prioritise their own interests and inhibit competition. This lack of competition can slow down technological advancements and reduce the diversity of AI solutions available in the market,” says Mahadhir.
To mitigate these risks, it is crucial for Malaysia to invest in its own AI capabilities to foster a competitive yet collaborative ecosystem.
Mahadhir says there are a few ways to do this. This is by providing funding, mentorship and resources to small and medium enterprises that focus on AI to allow them to scale and innovate; offering incentives, tax breaks and grants to companies investing in AI research, development and deployment; and focusing on education and training programmes to build a strong pool of AI talent.
“We should identify Malaysia’s niche markets that generate the best return-on-project outcomes and investment amounts. Through a laser-focused investment strategy, Malaysia will be able to see near-term results, which will have a ripple effect on the local economy and job opportunities. Simply put, invest in something where we can have immediate wins and grow from there,” notes David Lim, the founder and CEO of enterprise AI solution firm Wise AI.
Finding Malaysia’s niche in the AI value chain is crucial since competing across the entire spectrum is unrealistic, says Prashant. This can be done by focusing on narrow AI models, prioritising intellectual property, boosting export competitiveness and capitalising on the country’s strengths.
“We need to find out what we want to own. We can choose narrower spaces where we want to be better than most people in the business so that in due course, we are able to actualise the benefits to the economy from this … I think that the government having a proactive master plan is important. But the master plan should have sufficient agility to allow winning horses to run big races,” he adds.
“Aerodyne is one company that has done a phenomenal job in terms of application related to visual AI-based drone photos. Is that a space where we could be among the best in the world?”
Ultimately, developing a foundation model is not essential for participation in the AI economy but it can foster local innovations, through the development of technologies such as LLMs. For instance, foundation models trained on local data will be able to respond to regional nuances and cultural specifics.
However, these technologies come at a high cost. Building and maintaining AI models require millions of graphics processing units and specialised data centres that are very expensive.
This means the feasibility of developing new models in Malaysia is low because of funding limitations, says Chang. Focusing on industry-specific models where AI adoption is high could be the way forward. While Malaysia may not be at the forefront of AI development, the country can still benefit from effectively utilising AI in areas like medicine, agriculture and education. This will allow it to remain competitive in the global AI race.
“We are not looking at developing cutting-edge technology. But we are hoping we can utilise [existing technologies] and that we are not left behind on that. If we want to develop new technologies, it will take time,” he explains.
Collaboration with other Asean nations including Singapore could be a viable strategy as it would allow the countries to combine resources and expertise while sharing the financial burden. The Sea-Lion LLM is an example of how regional cooperation can encourage AI advancements that reflect the region’s unique cultural and ethical considerations.
“AI Singapore led the efforts to build Sea-Lion, which is a Southeast Asian LLM. It wouldn’t have been successful if we didn’t have the cooperation of our Southeast Asian [counterparts]. Given the way LLMs work, if Southeast Asian does not have one, then we are going to be very dependent on either the Americans, the Europeans or the Chinese to have a language model, which may not reflect the Southeast Asian culture, ethics or biasness,” AI Singapore’s Liew points out.
Ready for lift-off
Chang has a different take on an AI champion. He believes that the champion does not need to be a developer of AI technology. Instead, he wants to see industry leaders utilising existing AI technologies to address their needs.
“I think if we encourage one or two corporations as AI champions, and we invest resources into these two [entities], slowly the tech will be monopolised by them. We don’t want that. Everyone should be an AI champion,” says Chang.
“I would say industries that use the most innovative way to adopt AI, they are the AI champions.”
Meanwhile, Fabian sees AI becoming a tool that empowers humans through human-AI collaboration, augmented creativity and upskilling the workforce.
AI has permeated different sectors ranging from healthcare and education, to manufacturing and agriculture. Now, these sectors are all scrambling to implement the technology to get ahead in the AI race. As such, industry participation is key to leveraging the potential of AI across different fields.
“The biggest strength is that we have done it before. Malaysia is now the world’s sixth largest semiconductor producer, seeking RM500 billion in investments for its semiconductor industry and to train 60,000 Malaysian engineers to meet market demands. Although most of it focuses on the lower end of the value chain, Malaysia was able to gain from this niche,” says Lim of Wise AI.
“With its substantial capacity, Malaysia can diversify and move higher in the value chain. The same applies to the AI landscape in Malaysia.”
Chang says there is a need to build skill sets and that the government is working on incorporating AI training into universities and vocational training programmes. For instance, MDEC has implemented talent development programmes such as the Cikgu Juara Digital to train teachers and students in AI, robotics, coding and computational thinking.
MDEC has also onboarded close to 200 local companies that have been identified as AI technology service providers to nurture and scale the AI ecosystem.
In addition, UTM has established the Malaysian AI Consortium consisting of researchers, educators and practitioners. This complements the AI Talent Roadmap for Malaysia 2024-2030, which aims to formulate and develop AI curriculum and research programmes.
“Malaysia must invest in the development of digital skills among its workforce to remain competitive in the digital economy. Such investments will ensure that Malaysia cultivates a skilled workforce capable of meeting the evolving demands of the digital economy,” says Mahadhir.
A strong foundation in maths, science and language is important in nurturing future AI talent, notes Siti Fauziah Toha, professor of AI at the International Islamic University Malaysia. This foundation needs to be developed early to ensure a steady pipeline of AI developers. To do this, there has to be a shift from a use-based approach to a curriculum that encourages innovation and problem-solving skills.
This is to avoid Malaysia becoming a perpetual user of AI technologies developed elsewhere.
“It’s going to forever be a black box, meaning that [the country] will not be the owner of the technology, but dictated by those who own the technology, if we don’t go out from the rat race of [merely] being a user,” says Siti.
“We should nurture our talent to not just understand the knowledge, but to be able to creatively come up with AI technologies.”
Source: The Edge Malaysia
