Is Malaysia prepared for AI?
SOME years back, Universiti Malaya’s department of artificial intelligence (AI) was on the verge of shutting down due to a lack of student enrolment.
Today, it is an entirely different animal, with student applications far exceeding the number of seats available.
“The AI course started here in Universiti Malaya back in 1998 But it was only in the last five years that students’ interest to study AI spiked, to the point that even those with a cumulative grade point average of 4.00 find it competitive to secure a seat in the department,” says Prof Chan Chee Seng.
Chan is the dean of the Faculty of Computer Science and Information Technology where the department of AI is located. Chan himself teaches AI.
With AI as the new buzzword in town, it is not surprising that more Malaysians are looking to be trained in AI. Businesses have also jumped on the bandwagon.
YTL Group is collaborating with California-based Nvidia Corp – the same company that makes AI chips to train ChatGPT – to build an AI-powered data centre in Kulai, Johor.
Earlier this month, Maxis Bhd said it will be the first telecommunications company to offer cloud-based graphics processing units-as-as-service (GPU-aaS) to local customers.
GPU-aaS provides businesses with easy access to powerful GPUs for advanced AI training and inferencing applications.
Among the smaller companies, Nationgate Holdings Bhd is also part of the AI rush.
UOB Kay Hian Research has said that Nationgate has been appointed an original equipment manufacturer for Nvidia-powered AI servers.
Meanwhile, SNS Network Technology Bhd bagged a RM85.52mil order to supply AI super servers to a data centre in Johor owned by an undisclosed eCommerce platform in South-East Asia and Taiwan.
These are just some of the many examples of AI ventures in corporate Malaysia. However, Malaysia is not the only bright spot in the regional AI space, and this puts the country in tough competition with its neighbours in attracting AI-related investments.For instance, Thailand’s project to create a generative artificial intelligence tool based on its local language, or Open ThaiGPT, has started to show progress.
In Indonesia, its Artificial Intelligence Industry Research and Innovation Collaboration (Korika) has announced it will join hands with OpenAI – the developer of ChatGPT – to build an AI system that aligns with Indonesia’s values.
The republic has, in fact, awarded a 10-year Golden Visa to Sam Altman, the chief executive officer of OpenAI. Indonesia has also introduced the National AI Strategy 2020-2045 to lay the groundwork for AI growth.
Malaysia, meanwhile, is guided by the National Artificial Intelligence Roadmap 2021-2025. The roadmap looks to achieve an AI-led growth that aims for at least 30% increase in gross domestic product growth.
By 2026, the country aims to facilitate the establishment of up to 900 AI startups, nurturing the development of over 13,000 new AI talents.
Grand ambitions aside, Malaysia still needs to ramp up its AI pace, especially in the semiconductor industry, which is the backbone of the domestic manufacturing sector.
HSBC economist for Asean Yun Liu says Malaysia does not have direct exposure to AI chips, unlike South Korea, Singapore and Taiwan.
She notes that Malaysia is more exposed to the lower margin segment of chip assembly, testing and packaging segment.
So, is Malaysia truly prepared to embrace AI? And is the country also prepared to effectively deal with the problems that AI can cause?
For example, it is found that the more employees interact with AI at work, the more they are likely to face loneliness and insomnia.
The research was published last year in the Journal of Applied Psychology, covering samples from Malaysia, Taiwan, the United States and Indonesia.
Also, Japanese newspaper Yomiuri Shimbun reported in August that high-speed computerised trading using AI may be one factor that causes volatility of stock prices on the Tokyo Stock Exchange (TSE).
More than 60% of orders on TSE are placed automatically in high-frequency trading.
It is only a matter of time before Malaysia faces a similar situation where AI’s presence in decision-making may cause unexpected issues.
To address this, the country needs to bolster its army of AI-skilled workers.
Randstad, however, revealed in July that one in three Malaysians has never used AI at work, although 81% of Malaysians feel the impact of AI on their jobs.
Science, Technology and Innovation Minister Chang Lih Kang says the integration of AI in Malaysia’s manufacturing sector is progressing steadily.
Nevertheless, the journey faces challenges, especially in the adoption of Artificial General Intelligence or AGI in Malaysia.
A more advanced version of AI, AGI mimics or surpasses humans’ cognitive abilities.
The limitations to AGI span technological, educational, economic, ethical, and regulatory domains, Chang points out.
“Malaysia’s technological infrastructure, while improving digitally, still lags behind AI nations in high-performance computing, data centres, and research facilities.
“AGI development requires significant computational power and access to large datasets, which Malaysia currently lacks.
“Additionally, the country’s Internet speed and penetration are not sufficient for widespread AGI applications, especially in rural areas.”
The minister also acknowledges that the workforce is not adequately prepared for the complexities of AGI development. This is worsened by a brain drain.
The education system, he says, must adapt quickly to the rapidly evolving nature of AI technologies.
“Addressing these limitations requires multifaceted approaches involving substantial investment in infrastructure, education, and research and development.
“It also means creating a supportive regulatory environment, fostering collaboration between various stakeholders, and ensuring that the benefits of AGI are distributed equitably.”
Chang, however, points out that the Science, Technology and Innovation Ministry (Mosti) has prepared an infrastructure for advanced AI adoption.
“Mimos, a technical agency under Mosti, has launched the My-AI portal which contains all the high-performance computing resources needed to develop AI applications.
“The AI Sandbox initiative from the Malaysian Research Accelerator for Technology & Innovation (MRANTI) provides a platform for local startup companies to develop, test, and implement AI solutions before deploying on a larger scale.
“Preparing the infrastructure alone is not enough to push the adoption of advanced AI in Malaysia. It must be supported by continuous awareness to the public, grants, and funding,” he says.
AI user to AI creator
Universiti Malaya’s Prof Chan and venture capitalist Raja Hamzah Abidin echo a common view that most Malaysian businesses are AI users, without much innovation.
“Many of the businesses merely adopted ChatGPT and just use them. There is no new product creation.
“If you and I rely on the same ChatGPT to offer a service without innovation, what differentiating value can we offer? It will then create a price war as we both fight to prove whose service is cheaper.
“If this trend continues, we risk lagging behind more technologically advanced nations,” says Chan.
Raja Hamzah, the chief executive officer of RHL Ventures, says that in order to develop proper AI infrastructure and expertise, a company needs to first hire capable AI engineers. “This costs a significant amount of money.”
In the case of small and medium enterprises (SMEs) current AI adoption is limited to basic AI deployment such as chatbots, generative AI and vendor-driven SaaS with incorporated AI, says Datuk William Ng.
“These are generally sufficient for most SMEs,” says Ng, the national president of the Small and Medium Enterprises Association (Samenta).
Mosti minister Chang, however, is positive that Malaysia has the potential to evolve from merely a user of AI technologies to a creator of AI innovations
This is given its strategic initiatives, burgeoning talent pool, and proactive governmental policies.
He adds that Malaysia can gain a competitive advantage by leveraging on its unique strength in niche areas such as agriculture, healthcare and manufacturing.
To encourage more SMEs to embrace AI, Samenta’s Ng says the government and vendors can continue to create awareness of the impact of AI to business.
Limited tax breaks on AI deployment can also be provided, he says.
Commenting on incentives, Chang explains that Mosti has a research and development (R&D) funding ecosystem for start-up companies from Technology Readiness Levels (TRLs) 3 until 9.
TRL is a method for understanding the technical maturity of a technology.
The R&D fundings are provided by agencies such MRANTI, Cradle Fund Sdn. Bhd, Malaysia Venture Capital Management Bhd and Malaysia Debt Ventures Bhd.
“These agencies are working hand in hand in facilitating technology-oriented companies by providing the capital for innovation in technology sectors from eCommerce platforms to fintech solutions, AI and robotics to medical devices and green technologies,” says Chang.
No business case for AI?
Research by Chan and his team this year found that AI helps to increase productivity of clerical and secretarial roles by 20% to 30%.As for executives, the increase in productivity is about 15% to 20%.“One should not adopt AI for the sake of it. Industries must first know what they want to achieve from AI and from there, integrate AI accordingly into their operations,” says Chan.
However, economist Geoffrey Williams argues that for most Malaysian businesses, especially micro, small and medium enterprises, there is no obvious business case for AI yet.
“So, pushing AI could distract businesses from perfectly successful business models in search of unproductive and disruptive change.
“The government should free the economic environment to allow access to AI and let businesses choose for themselves.
“If there is a business case, then entrepreneurs will follow the benefit trail,” says Williams.
Beyond incentives and raising awareness on AI, the adoption will only accelerate if there is a proper regulatory framework.
At the moment, Malaysia’s regulatory framework for AI is evolving and lacks comprehensive policies to address the complexities of the technology.
“Clear guidelines on data privacy, security, and ethical use are needed. Regulatory bodies need expertise and resources to effectively enforce these regulations,” says Chang.
RDS Partnership corporate lawyer Nur Shohidah Ramlee points out that one cannot claim copyright for AI-based outputs, as AI is not considered a “qualified person”.
Under Section 10 of the Copyright Act 1987, copyright protection applies to works where the author is a “qualified person,” defined as either an individual who is a Malaysian citizen or a permanent resident, or a legal entity established in Malaysia.
Secondly, under Malaysian patent law, while AI may generate new ideas or innovations, the issue of inventorship is not entirely clear.
Regulation 9 of the Patents Regulation 1986 allows for non-natural persons to be listed as inventors, though the regulation does not define this term explicitly.
“A step forward to address the concerns of AI in intellectual property law is for legislation to include a clear definition of non-natural person and whether AI is included in such definition.
“Additionally, guidelines should be implemented to clarify on how AI-generated inventions and works should be treated under the law, ensuring that there is a clear framework for assigning intellectual property rights and addressing issues of inventorship and authorship.”
A growing concern regarding AI systems is the lack of transparency about how these systems function and which datasets are used to generate their outputs.
This uncertainty raises significant issues, particularly when personal data is involved.
Shohidah recommends that the government implement a multi-faceted approach that strengthens data protection laws.
“This involves amending the Personal Data Protection Act 2010 (PDPA) to include specific provisions for AI systems, mandating rigorous security measures to safeguard personal data, and ensuring that AI systems are designed with transparency in mind.”
In order to protect data, especially private data of national security, Universiti Malaya’s Chan says that Malaysia must have its own infrastructure in place, particularly the GPUs.
GPUs are specialised hardware that can efficiently process huge amounts of data simultaneously, making them ideal for graphics rendering, video processing, and accelerating complex computations in AI and machine learning applications.
“If you do not want your data to be in the hands of an external provider such as OpenAI or Amazon, you need your own GPUs with any open source large language models (LLM).
“With this arrangement, you can train the LLM with your own data securely. This is also where you can commercialise and make money,” he says.
Chan explains that Malaysia can provide licence or access to a model that is already trained with exclusive local data for a fee.
“For example, global companies and countries can pay to access an LLM that is trained with our halal industry-related data.
“Government data must be trained in Malaysia. In addition, the data must also be on a sovereign cloud system for security reasons,” he says.
Currently, Malaysia’s public sector data is on a cloud service called the Government Cloud.
Government Cloud is a strategic collaboration between the government’s cloud service provider (CSP) – MyGovCloud@PDSA – and four commercial CSP panels namely, Microsoft Azure, Google Cloud, TM Cloud Alpha and Amazon Web Services.
“Imagine this, if the data is stored in a cloud system of an American company and the servers are in the United States, how sure are we that the US government won’t have any access to it?
“We need locally owned GPUs and a sovereign cloud to store our data and to train our own AGI models.
“It is all part of setting up the Malaysian AI infrastructure,” says Chan.
30%
minimum increase in
GDP growth targeted
under AI-led roadmap 81% of Malaysians
feel the impact of AI
on their jobs