AI training needs to move from being product-centred to outcome-driven

Big Tech has been scooping up a large slice of the emerging tech training pie, with academies such as IBM Skills Build, Google AI Academy, Huawei ICT Academy and more driving artificial intelligence (AI) upskilling initiatives.

These academies have made some of their courses complimentary and provide participants with digital credentials branded after the Big Tech firm.

However, such training is usually centred around their products – “a means to get customers onto their platforms,” says AI Singapore (AISG)’s Director of AI Innovation, Laurence Liew.

While these courses have allowed many professionals to ramp up their AI skills on paper, this may not necessarily translate to real-world uses.

To go beyond “learning how to click on the platforms,” AISG has taken a pioneering approach in closing the AI skills gap - based on training, not just theory.

This was in 2018 before a slew of training providers emerged during the ChatGPT boom, which triggered a demand for AI-related courses.

While Southeast Asian governments are increasingly partnering with Big Tech to roll out AI training, Liew questions the genuine depth of these skills and to what extent they translate to job placements and productivity benefits.

“You can learn how to use the platform. But the danger is also when you don’t understand the fundamentals,” he adds.

To subscribe to the GovInsider bulletin click here.

Very little stickiness in acquired skillsets

From his interactions with learners who engage in product-centred training, Liew has observed poor learning outcomes as learners are unable to apply these skills – many of which are picked up over a six-month period and assessed only via multiple-choice questions (MCQs).

To better align skills with industry demand, AISG plays a dual role as both matchmaker and capacity-builder.

By combining two of its programmes - the AI Apprenticeship Programme (AIAP) and 100Experiments (100E) - AISG aims to align AI training with organisational demand for AI products.

100E gathers problem statements from organisations, while AIAP taps into AISG's in-house expertise to train talent to leverage AI to solve these problems.

As of 2020, the number of use cases has doubled to 200 and includes a diverse range public and private organisations.

Some public sector organisations who have submitted problem statements include the Singapore General Hospital (SGH), Inland Revenue Authority of Singapore (IRAS), Ministry of Foreign Affairs (MFA), and more.

AISG is funded by the National Research Foundation (NRF) to build Singapore’s capacities in AI development.

The organisation’s work has since drawn international interest from countries in Europe, the Middle East, and others, according to The Straits Times.

AI literacy for the public

AISG takes a multifaceted approach to close the AI skills gap in Singapore, targeting engineers, researchers, product managers and even the “man on the street.”

“We have always been focusing on the PhDs but suddenly we’re realising that while we can have the best large-language models (LLMs), what if 99 per cent of our citizens have no idea how to use it or what it is all about?” he explains.

Countries are beginning to understand that closing the skills gap is not just about having engineers and researchers, and that there is a wider literacy component, he adds.

The use of natural language in generative AI (GenAI) applications has democratised AI for the wider public. AI is not just for the AI engineers, Liew says, as any Singaporean can now tap into the technology to improve their work.

AISG has partnered with Enterprise Singapore to fund up to 50 per cent of the Microsoft Copilot licence costs for small and medium-sized businesses (SMEs) for a year’s worth of use, which includes additional training support and resources by Microsoft.

The organisation rolled out the “AI for everyone (AI4E)” video series in 2021 - “20 minutes, condensed version, four languages” says Liew – but efforts and funding to publicise the series beyond their own platforms remain constrained.

AISG is considering promoting the series more on popular social media platforms such as TikTok and Instagram.

To subscribe to the GovInsider bulletin click here.  

More flexible key performance indicators needed

Liew shares that their approach has evolved since its founding, especially when it comes to AI products.

Funders may expect to fund a large quantity of projects, but this is not always necessary or practical, he shares. 

For example, instead of building multiple LLMs for different use cases, AISG focused its efforts on SEA-LION, an open-source model that supports the larger Southeast Asian ecosystem to build and train their own models.

Businesses can use it to build AI tools like chatbots, coding assistants, and meeting summarisers, said AISG’s Senior Director for AI Products, Leslie Teo, previously told GovInsider. 

Also, with a new LLM coming out every other week, funders cannot wait for six months after a funding cycle to measure the outcomes, he adds.

As AISG receives public funding for its programmes, the organisation has to balance keeping up with the rapid pace of AI innovations while being accountable to taxpayers’ money.

The evolution of AI goes hand in hand with that of AISG’s existence.

“If AISG still exists in its current form in the next five years, we have failed because our role is to enable the [AI] ecosystem... When [the ecosystem] is capable [enough], companies don’t need to come to us to do 100E anymore. They also do it faster and cheaper.

“We must work ourselves out of the job then we can go onto the next big thing, right?”

Liew recently published a GenAI-assisted book titled “AI-First Nation” that details AISG’s initiatives in nurturing Singapore’s pool of AI talent and the country’s journey in AI. Find out more about his book here >>