AI: An opportunity for Malaysia to catch up, or slip behind?

  • AI Roadmap, and clear job opportunities show way forward
  • Slow adoption of IR4.0 expected to get boost from National IR4.0 policy

AI: An opportunity for Malaysia to catch up, or slip behind?

AI: An opportunity for Malaysia to catch up, or slip behind?On the one hand, Dzuleira Abu Bakar (pic) said, "There is a huge opportunity for AI to flourish in Malaysia." On the other hand, she also said, "A lot needs to be done to achieve mass adoption of AI."

The two statements are not inherently contradictory, but they do indicate the scale of the work that needs to be done, and the potential rewards the Malaysian government believes it can reap from having a strong base of AI expertise in the country.

Dzuleira, CEO for Technology Park Malaysia (TPM) was speaking as part of a panel titled "The power of innovation with AI as a game changer in the new world", as part of the World Artificial Intelligence Conference (WAIC) 2021. This year, WAIC is having a Southeast Asia inaugural event in Malaysia in conjunction with WAIC 2021 China. It is co-organised by AI company, Skymind Holdings Bhd.

The Malaysian government expects the National Industrial Revolution 4.0 (IR4.0) policy to improve the country’s productivity by 30% across all sectors by the end of 2030, and AI is a significant part of that.

Dzuleira highlighted what was at stake: Through ASEAN, Malaysia has access to a market of 630 million people, and a GDP of US$3 trillion (RM12.6 trillion). The ASEAN market is expected to skyrocket to US$5.2 trillion (RM21.8 trillion)by 2025. The country that can push forward the fastest will stand to gain the most.

There have been a few promising beginnings. In March, the Malaysia Artificial Intelligence Roadmap was launched. Then in April, it was announced that Microsoft will establish its first data centre in Kuala Lumpur, a project estimated to cost US$1 billion and - more importantly - is expected to create 19,000 jobs. "This initiative will help skill a million Malayspians for IT, Artificial Intelligence, and cloud services jobs by 2030," said Dzuleira.

It is against this background that she spotlighted the creation of Malaysia's first AI park under TPM's aegis. Planned to occupy 300 acres of land with a total investment of more than a billion Ringgit over the next four years, it will be a key part of Malaysia's ambitions to implement the nation's IR 4.0 policy. "TPM is bound by the strongest obligation to fully support the National AI Roadmap 2021-2025," she stressed.

 

Malaysia faces a slow adoption rate

However, Dzuleira readily admitted that its obligation comes with challenges. "While we have one of the highest manufacturing output levels in the world, we are actually lagging behind other countries in terms of employee productivity, R&D, and a workforce with higher education degrees," she said. "Malaysia faces a slow adoption rate of IR 4.0 with only 15%-20% of companies having actually embraced it."

She cited a 2019 IDC study where it was revealed that Malaysia's workforce was unprepared for IR 4.0. The common challenges faced by Malaysia include the lack of awareness and the impact of IR 4.0 and its technologies.

Indeed, according to the Malaysian Artificial Intelligence Roadmap Survey conducted by the Ministry of Science and Technology earlier this year, the two top challenges faced by Malaysian companies in implementing AI are a lack of expertise, and limitations in financing them.

Certainly more needs to be done to encourage SMEs to get on board. Sivan Umapathy, Telekom Malaysia’s Chief Information Officer, shared a little about the AI platform that TM is developing together with Skymind Global following a Memorandum of Collaboration signed between them last April. If it goes to plan, the AI platform will eventually make it easier for TM's customers to develop their own AI solutions for their individual business challenges.

 

Costly to train AI models

AI: An opportunity for Malaysia to catch up, or slip behind?But it won't be easy. Adam Gibson (pic), Skymind Global Chief Technology Officer, and creator of the DL4J framework (an open-source distributed deep learning library), talked about challenges facing smaller companies that want to adopt AI for their businesses: it's not cheap.

"It's costly to train these models," he said, especially for problems related to computer vision and natural language processing. "For example, a billion parameter model takes US$1.6 million to train from scratch." It could get even more in the future; In January, Google announced that they had developed techniques to train a language model containing 1.6-trillion parameters.

Gibson explained that although people fine-tune existing models to save time and money, they still are very expensive. And then there's the issue of concept drift, where a solution that works today may need to be updated over time, for example when trying to detect when people are committing fraud. Retraining an existing model can cost hundreds of thousands of dollars. "We need to move towards being cheaper, if that's possible."

Gibson suggested a few future avenues of research. One is Deep Learning Compilers so that these problems can be solved efficiently on a wider range of chips. "This will open up AI applications to smaller companies," he said. And given his history with DL4J, he also unsurprisingly champions open source tools.

Making AI more accessible and cheaper sounds like ideal areas for TPM's AI Park to explore. However, details of what precisely will happen and when are still unclear, and perhaps surprisingly so, given that the park has been in planning since 2019. 

But given that Thailand launched its first state AI centre last November, Vietnam rolled out its roadmap in March, the Philippines theirs in May - and all preceded by Singapore's National AI Strategy in 2019 - perhaps Malaysia should get a move on it.

 

Related Stories :

 
 
Keyword(s) :
 
Author Name :
 
Subscribe to SNAP
Download Digerati50 2020-2021 PDF

Digerati50 2020-2021

Get and download a digital copy of Digerati50 2020-2021