geraldgiam.sg

Politically significant persons

My speech during the Committee of Supply Debate (Ministry of Home Affairs), 29 Feb 2024.

Recently, Mr Philip Chan, a Singaporean businessman and former People’s Association grassroots leader was designated by MHA as a politically significant person (PSP) under the Foreign Interference (Countermeasures) Act (FICA).

MHA said that Mr Chan has “shown susceptibility to be influenced by foreign actors, and willingness to advance their interests”. I understand there are other Singapore residents who are publicly known to be associated with foreign political organisations, but have not been designated as PSPs.

I would like to request for more details of the Ministry’s criteria for designating PSPs. How does the Ministry assess the risk of foreign interference and the public interest in applying FICA’s measures to PSPs? What are the factors that distinguish Mr Chan from other persons who engage in similar activities but who are not designated as PSPs?

I understand the Ministry might be reluctant to be too transparent about the criteria to prevent people from circumventing the FICA provisions. However, MHA should provide better guidance to the public on where the out-of-bound (OB) markers are, so that they can avoid becoming unwitting agents of foreign influence and interference.

National Service pay

My speech during the Committee of Supply debate on the Ministry of Defence’s budget on 29 Feb 2024.

National Service (NS) is an obligation that every male Singaporean or second generation PR must fulfil. However, it is undeniably an opportunity cost that is disproportionately levied upon one segment of society. Due to NS, Singaporean men begin their careers two years behind their female peers, first-generation PRs, new citizens and foreigners. NS poses a significant challenge for aspiring professional athletes, disrupting their training and competition during their peak performance years, and hindering their world-class aspirations.

The current philosophy of giving NSFs an “allowance for basic upkeep” needs to be reviewed in view of the significant increase in cost of living and NSF’s opportunity costs. As a first step, the terminology should be changed from “NS allowance” to “NS salary”, and it should attract CPF contributions.

The median NS salary should be at least the Local Qualifying Salary of $1,600 per month. This still won’t be a market salary, but it will go some way to compensate NSFs for the opportunity cost of two years of service and help them contribute a little to their household income.

I must emphasise that this does not in any way diminish the ethos of national service — just because we pay our NSFs fairly doesn’t mean their service is any less of a national duty.

Insurance coverage for persons with disabilities or special needs

My speech in Parliament during the Committee of Supply debate (Prime Minister’s Office).

Persons with disabilities or special needs often receive inadequate insurance coverage to protect themselves. Even if they manage to get insured, the scope of coverage and insured amount is often limited. MAS is proposing to issue guidelines to insurers that they should not indiscriminately reject an application solely based on declared personal information such as a disability.

Mr Chairman,

Instead, insurers are expected to carry out an objective assessment of every application based on reliable information or data relevant to the risks being insured. However, even under the new proposed guidelines, insurers are not prohibited from declining applications, setting higher premiums or applying conditions in view of the risks presented by an applicant with a disability.

Can the Minister explain when these guidelines will be issued, and how they will be materially different or more beneficial to persons with disabilities or special needs?

Population projections beyond 2030

My speech on 28 Feb 2024 on the Committee Supply Debate (Prime Minister’s Office).

The government has clarified that it does not aim for Singapore’s population to increase to 10 million and is anticipating a population of “significantly below 6.9 million” by 2030. However, with 2030 only six years away and fast approaching, a key question arises: what are the government’s longer-term forecasts?

Population projections are a critical tool for public policy planning.

Given current birth rates and the intake of new citizens and PRs, what are the government’s projections for Singapore’s population in 2040 and 2050? These figures are pivotal, not just for immediate policy adjustments, but for securing a sustainable future for all.

The government claims that it does not seek to achieve any particular population size. I find this rather strange: if there is no population target, how does it decide how much to regulate the immigration tap or how much to invest in pronatalist policies to achieve its desired economic, social and security outcomes over the long term?

Boosting productivity through AI and robotics

Budget Debate 2024, 26 Feb 2024

Mr Speaker,

Fair competition

Singaporean workers aspire towards making a good living and engaging in meaningful work that uplifts not only their own families, but also their communities, their nation and the world. One inescapable reality of work is competition.

We have always had a very competitive culture in Singapore. This has served us well in many ways, from the excellent performance of our students in schools to our efforts to top the global rankings in everything from corruption perceptions to business friendliness.

However, competition also has a darker side. Singaporeans are not looking to the government to shield them from global competition at the workplace. However, we detest unfair competition, where people who do not play by the rules or follow local norms still get ahead.

For example, when Singaporean workers see colleagues getting hired and promoted not on the basis of their ability and hard work, but because their manager prefers working with people who share his cultural background, this creates a profound dissonance in them. Why? Because Singaporeans have been brought up to believe in meritocracy as a guiding principle in our society.

We want a Singapore which rewards workers and professionals based on their competence and hard work, not connections or tribal loyalties.

Nevertheless, while we strive to shape the Singapore that we desire, we are but a small drop in a vast ocean. We have to teach our children and students to deal with the world as it is, not how they want it to be. They must be taught at home and in schools to speak up when they have something to contribute, and not keep silent in the background. They must be encouraged to ask for what is due to them, and not simply accept what others decide for them without question. And they must be willing to network with a wide spectrum of people from different cultures and nationalities, and understand what motivates them.

Singapore is often viewed as a nation of excellence. In local parlance, we do things “swee swee”. Give a Singaporean a task, and when they say they’ve done it, we can trust that it’s been done well. We must never let this culture of excellence slip. It is our advantage in an increasingly competitive world.

Boosting productivity

Artificial intelligence (AI) and robotics have both burst into mainstream consciousness in recent years, with the launch of generative AI and self-driving cars. These technologies could provide a path to boosting Singapore’s productivity by enhancing the speed, accuracy and efficiency of various tasks and processes.

There are many studies of AI’s measurable impact on productivity. For example, a joint-study by BCG and Harvard found that consultants using GPT-4 completed 12% more business tasks, 25% more quickly, with 40% higher quality, than a control group without AI access.

These are amazing opportunities for Singapore to take advantage of. The government can spur a broader uptake of AI and robotics not only for our scientists and businesses, but also for general purpose use by ordinary citizens.

In November 2023, I asked about whether the government planned to develop indigenous capabilities in creating and deploying AI foundation models, including establishing a National Foundation Model Research Institute. In February 2024, I proposed the creation of a national AI healthcare foundation model, which can be used to predict and intervene in a broad spectrum of diseases. I would like to repeat these calls here.

These are not just national initiatives, but possible precursors to greater regional scientific co-operation. Singapore needs the right institutions and opportunities, in order to attract and retain the best minds, including talented Singaporean scientists and entrepreneurs.

I am under no illusions about the potential of these new technologies to cause job losses. This is why in January 2024, I asked the government for its plans to proactively retrain workers who are at most risk of displacement from AI. We need interventions to steel our citizens against AI-driven job redundancy. It is better for us to be the architects of our own disruption, than to allow technology to change us for the worse.

It is tempting to pull out the old playbook of pouring money into training programmes and encouraging workers to attend courses. However, this approach may not succeed in upskilling an entire workforce in disruptive new technologies.

We must also embrace tacit learning through the hands-on use of AI and robotics. Tacit learning is learning by doing.

Using AI tools or robots needs to be made as easy and as commonplace as googling for answers on a web browser or operating a TV remote control. Robots should be deployed more widely in our environment so that the public gets used to seeing and using them every day. Giving every Singaporean hands-on practice with AI and robotics will better ensure that the gains from these technologies go to everyone — manual workers and knowledge workers, civil servants and entrepreneurs, MNCs and SMEs.

SkillsFuture must also support tacit learning. In February 2024 in my adjournment motion on global AI leadership, I called for the subscriptions to cutting-edge AI tools to be subsidised by SkillsFuture Credits. AI tools make workers more productive, and we should give our people more opportunities to use them.

We must strive for the whole breadth of Singapore society to have more contact with the best AI models and robots. Only by trying things out as a first mover, and being willing to accept and learn from failure, can we gain and retain the thousand points of knowledge that no instructor can teach.

New world order

In his Budget Statement, Deputy Prime Minister Lawrence Wong rightly pointed out that we are now in an era of armed conflict, confrontation and terrorism, with major powers prioritising national security over international cooperation, and that there is a diminished willingness to tackle global issues.

These are stark realities. We must work within the realities of this new world order, as the old order is probably not coming back anytime soon.

DPM Wong said that we will pursue better jobs and better growth. He made a commitment to improve wages across professions. In particular, he said that the wages and career prospects of ITE graduates should not be too far behind polytechnic and university graduates.

I fully support this. I hope the wages of skilled tradespersons will come much closer to par with knowledge workers, because of the value they bring to our economy and our society. I will elaborate further on this during the Committee of Supply debate on MOM’s budget.

The introduction of the SkillsFuture Level Up Programme, which injects another $4,000 into Singaporeans’ SkillsFuture Credit, is welcome. DPM Wong said it is to be used for selected training programmes with “better employability outcomes”, including part-time and full time diplomas and undergraduate programmes. In addition, workers 40 years old and above will have the opportunity to pursue another full time diploma at subsidised rates.

How did the government arrive at the conclusion that these diploma and degree programmes have better employability outcomes? Is there empirical evidence to support it? And if there is, are the better outcomes due to the greater skills that these graduates have acquired or because local employers continue to emphasise paper qualifications over skills and experience?

In fact, it has been recognised that one of the most effective ways to pick up employable skills is through on-the-job training (OJT) and apprenticeships. As such, I would suggest that the government subsidise OJT and apprenticeship programmes to the same tune as diploma and degree programmes.

In his Budget Statement, DPM Wong mentioned briefly about how the government will do more to support those whose jobs are made redundant through a temporary financial support scheme for the unemployed. He said the government will be working out the details later this year. Sir, this scheme was announced almost one year ago, at the National Day Rally in August last year. How much longer will it take to flesh out? More importantly, how fiscally sustainable will this scheme be?

Will it include an insurance component, like what the Workers’ Party has proposed through its Redundancy Insurance Scheme, to ensure that premium contributions from employers and employees during times of plenty can be drawn down during economic downturns when retrenchment levels are higher?

Cost burdens for disability community

Mr Speaker, social inclusion must be at the heart of all our economic policies. I am glad to note that the maximum monthly fees at Special Education (SPED) schools will be lowered to $90, with lower fee caps at all centres.

However, Singapore should move towards equalising the fees for SPED schools and mainstream schools. While I am aware that the cost for providing education at both types of schools is different, the school fees should be the same. So if mainstream primary school fees are only $13 a month, then SPED school fees should also be $13 a month, not $120, which is the current average. Even lowering it to $90, while commendable, is still not equitable. This is especially so considering the higher costs that parents of special needs children incur in many other areas besides education. The additional costs of SPED schools should be socialised in the interest of creating a more inclusive and equitable society.

School bus fares remain another significant concern for the disability community. As of 1 January 2024, the Ministry of Education (MOE) has increased the price cap of school bus fares for school bus operators at mainstream schools by up to 13%. But SPED school students face a higher increase in their school bus fares due to the smaller pool of bus operators who are able to meet their more complex needs.

I am aware and appreciate that there are various school bus subsidy assistance schemes like the MOE Financial Assistance Scheme (FAS) and the Enabling Transport Subsidy (ETS).

While I understand the need to keep the school bus operators sustainable, the adverse impact of the cost of living crisis has made the cost of school bus transport an added burden for many parents of students with disabilities. Many rely heavily on school bus transport to commute to and from home, school and social service agencies to attend programmes such as the Early Intervention Programme for Infants and Children (EIPIC), Day Activity Centres, Sheltered Workshops* and Special Student Care Centres. Very often there are additional costs involved. EIPIC, for example, is a half-day programme, so the students need to also be ferried from their school to the programme, creating a double whammy in transport costs.

I would therefore like to call on the government to increase the monthly household income limit for both the FAS and ETS, especially for households with special needs members. In addition, more subsidies can be provided to match the inflation of school bus fares. This would ensure that more families can access and benefit from these subsidies to cover the ever-rising cost of living.

Conclusion

In conclusion, Mr Speaker, as we chart our course through the rapidly changing global and technological terrain, our policies must embody a steadfast dedication to fairness, meritocracy and innovation. By nurturing an ecosystem that champions fair competition, leverages the transformative power of AI and robotics, and places the welfare and progress of every Singaporean at its heart, we can secure a robust and flourishing future. Let’s remain committed to building a society where access to opportunities, a culture of excellence and the value of each individual’s contributions shape the Singaporean journey for generations to come.

Sir, I support the Motion.

* DACs and SW are only 18+.

Transparency of Government Reserves

Parliament, 7 Feb 2024

Mr Speaker,

In our quest to safeguard the Singapore dollar, especially from the volatilities of currency crises, it is important for a country to hold sufficient Official Foreign Reserves (OFR). We’ve seen the dire consequences of not having enough in the tank during Thailand’s ordeal in 1997. The country’s struggle, stemming in part from currency speculation, led to a severe depletion of its OFR. This forced it to float its currency, triggering an economic crisis which affected the whole region, including Singapore.

The Monetary Authority of Singapore (MAS) has assessed that the “optimal” amount of OFR it needs to provide a “strong buffer” against stresses in the global economy and market is between 65% to 75% of GDP. The central bank assessed this level based on internationally-accepted measures of reserve adequacy and its own practical experience in foreign exchange intervention to ensure the stability of the Singapore dollar.

Since the MAS (Amendment) Act came into force on 21 February 2022, MAS has been allowed to subscribe to Reserves Management Government Securities (RMGS). This allows the transfer of OFR to the government for the long-term investment by GIC. This is OFR above what MAS needs to conduct monetary policy and support financial stability. Should there be a need to supplement the OFR on MAS’ balance sheet in the instance of a tail-risk event, it can redeem the RMGS before maturity at par.

Based on a parliamentary reply by DPM Lawrence Wong to me, as at September 2023, MAS’ current OFR stood at S$455.5 billion or 70% of GDP, with the potential to ramp it up to 106% of GDP by redeeming the RMGS holdings of S$237.6 billion.

This means MAS effectively has access to 31% more foreign reserves than the high end of the range it assesses to be optimal to defend the Sing Dollar.

The value of both MAS’ OFR and RMGS have been revealed publicly. So has the value of Temasek Holdings’ net portfolio. What remains is the value of the reserves held by GIC, which it has only revealed is “well over US$100 billion”. It is estimated to be worth many multiples of that.

In a parliamentary reply on 3 Oct 2023, DPM Wong said we need to brace for extreme tail-risk scenarios, including crises of unprecedented scale that could lead to significant capital outflows beyond MAS’ reserves, or emergencies caused by state or non-state actors threatening our economy, livelihoods or national existence.

Sir, the spectre of extreme scenarios cannot be a carte blanche for maintaining a veil of secrecy over the reserves. Is the government keeping the reserves held by GIC a secret only to prepare for extreme tail-risk scenarios, or is it also to avoid questions from MPs and the public on why our substantial reserves aren’t being invested more in Singaporeans?

Greater transparency serves to both improve governance and foster a more inclusive and informed discussion about our nation’s future. Transparency about the size of our reserves will facilitate more robust public discourse and democratic debate about our country’s long-term budget and expenditure plans. It will ensure that decisions regarding the use of our national wealth are made with the participation and understanding of citizens. Most importantly, it will increase accountability, as my Hon. Friend from Sengkang Jamus Lim has explained in his speech, and build more trust in the government.

Conversely, expecting Members of Parliament to scrutinise, debate and approve the government budget and each ministry’s allocation without full knowledge of our reserves is comparable to making a decision on buying a house without knowing how much money they have in their largest bank account.

Mr Speaker, I urge a review of the government’s stance on the secrecy of the reserves.

If the government wishes to maintain its strategic ambiguity regarding the full size of the reserves, the figures could be shared confidentially with MPs or with a cross-party Standing Select Finance Committee that looks into the health of our reserves and government finances, which the Leader of the Opposition has proposed.

This parliamentary committee will then be able to properly consider the specific funding needs vis-a-vis the threats that Singapore must prepare for and make recommendations to the government on how these needs could be funded.

I support the original motion.

Global leadership in AI

5 Feb 2024, Parliament

The government unveiled its National AI Strategy 2.0 report (NAIS 2.0) last December. The report outlines Singapore’s plans to harness artificial intelligence (AI) for the public good, focusing on enhancing AI capabilities, addressing potential risks and fostering a thriving AI ecosystem.

I appreciate the hard work that many policymakers have put into writing this report, including their efforts to consult with industry.

However, I believe that Singapore needs a more comprehensive AI industrial policy and clearer outcomes for each industry. We need a strategy that aims to make our nation the global leader in selected AI fields.

Under NAIS 2.0, the government’s main role revolves around enabling an environment for AI to grow and enhancing the efficiency of public agencies. While these are important, the government certainly has the resources and capability to do much more. In this AI-driven era, we need the government to intercede more proactively to create a world-leading AI industry. Simply leaving things to the free market may not produce the desired results because of the constraints of our local private sector and our small domestic market. It will risk forfeiting some promising economic growth opportunities that AI can bring for our nation.

Singapore’s small population should not deter us from global AI leadership. Historical industrial successes have been built on strategic government interventions, such as Taiwan’s support of TSMC, which played a pivotal role in its journey to becoming a global semiconductor manufacturing juggernaut.

Today, I will address NAIS 2.0’s AI labour policy and advocate for a strong AI industrial policy so that Singapore can aim for global leadership in selected AI domains.

AI labour policy

Let’s begin with the AI labour policy. The goal in NAIS 2.0 to generate 15,000 AI jobs sparked a blend of enthusiasm and apprehension among Singaporeans. Concerns linger that, as has happened in the past, foreigners may dominate lucrative positions — including leadership positions — leaving Singaporeans with mainly the routine and lower-paying jobs. This could impede our citizens’ career advancement alongside the advancing AI landscape.

NAIS 2.0 has three labour planks. Scaling up AI-specific training programmes; scaling up technology and AI talent pipelines; and remaining open to global tech talent.

Will the government commit to ensuring that a sizable majority of at least two-thirds of the 15,000 new “AI practitioner” jobs will go to Singaporeans?

I acknowledge the importance of global AI talent. However, there must be a clear differentiation between exceptional global talent and the average foreign tech worker. We should welcome the former, but should avoid importing too many of the latter, as they may end up competing with Singaporeans who can do the job just as well. Any global talent that we bring in must be expected to transfer their skills to locals, not just use Singapore as a springboard for greater pursuits in other countries. This can be done through tying company grants to the achievement of knowledge transfer, or through limited-term foreign work passes tied to the training of Singaporean workers.

The AI playground is level, with a highly collaborative open-source community. The core techniques and frameworks are mature enough and reasonably accessible through papers and code. With the right training, mentorship and opportunities, Singaporean talent can deliver as well as anyone in the world.

To raise a body of local AI talent, AI-training programme places and talent pipelines must be focused on Singaporeans. We need to plan ahead and start training all our students early in AI — not just students who are academically strong in the sciences and mathematics.

For mid-career workers, hands-on interaction with AI tools is one of the best forms of training. The government should expand the scope of SkillsFuture Credits to cover expenses for subscriptions to AI assistants like ChatGPT Plus or Github Co-pilot to accelerate their productivity. This will level the playing field for Singaporeans with less means to pay for such subscriptions. Paid models like GPT-4 have been assessed to be significantly better than their free counterparts, and we should give our workers more opportunities to use the best models.

Moonshots in industry

Next, on industry. It was once unclear if it were possible for humans to reach the moon. But US President John F. Kennedy made a speech to Congress in 1961, rallying his nation to achieve the goal of landing a man on the moon and returning him safely to Earth before the end of that decade. And NASA’s Apollo 11 mission achieved it ahead of schedule on 20 July 1969.

A “moonshot” is an ambitious, exploratory and groundbreaking target that has the possibility of spurring breakout growth. We need moonshots in AI, which NAIS 2.0 appears to lack.

The government has the resources and capacity to take on more risks on a longer timescale to pursue high-reward moonshots. These could catalyse future engines of growth. But first, the government must catch the vision and have the determination to make our country number one in our chosen AI domains.

Achieving some of these AI moonshots has implications on our economic security and even our sovereignty.

Advances in AI rely on large, high-quality data sets. We must ensure that foreign tech firms and governments do not end up extracting our data overseas to build AI products, which then get sold back to Singaporeans. This will allow such firms to profit immensely while local expertise flounders.

Currently no global framework governs cross-border data flows and ownership. This allows predatory dynamics to continue between countries and companies. Once market dominance is achieved, network effects and the dynamics of chasing a moving target make it almost impossible for new entrants to catch up.

If we are not careful, Singapore may become only a consumer of such platforms, while the economic benefits and the best jobs go overseas. Singapore should avoid this by proactively building comprehensive local data sets for homegrown AI development.

Management of moonshots

We should pursue a few ambitious, publicly-funded moonshot projects. These projects must prioritise transparency and align their outcomes with the national interest, ensuring that economic gains directly benefit our citizens.

A new government-owned AI startup will be needed to catalyse this moonshot, and I will refer to it as the AI Catalyst Corporation. It should be independently run with commercial dynamism, yet be ultimately answerable to Singaporeans.

What constitutes a well-chosen moonshot? I would like to propose five key principles:

First, its products or services must directly benefit Singapore and Singaporeans. Second, it should have export potential and become part of Singapore’s economic growth engine.

Third, it needs to have “moats” — which are durable advantages to prevent it from being quickly outcompeted or swallowed up by global tech giants.

Fourth, there must be a genuine unmet global market gap that Singapore has advantages in tackling. And fifth, the industry should be ripe for fundamental disruption, not just incremental improvements.

Healthcare AI as a moonshot

Healthcare AI could be a moonshot that Singapore can aim for. I will present the case for this, and answer the five questions in reverse order.

Is healthcare ripe for fundamental disruption? Yes. Healthcare systems worldwide are under strain due to ageing populations and chronic disease burdens. Healthcare institutions tend to treat diseases late in their course, when symptoms are severe and care is expensive. Yet major conditions like obesity, heart disease and cancer, are driven by shared underlying factors. This outdated care model no longer aligns well with scientific reality. Healthcare AI, supported by population-scale data, has the potential to predict risks and intervene early to significantly improve health outcomes.

This will require a reorganisation of healthcare delivery that focuses on early prevention and action. Singapore has started this journey through the Healthier SG programme. Let’s turbocharge it with healthcare AI.

Second, is there a genuine unmet global market gap? Yes. Electronic health record (EHR) systems remain fragmented worldwide. Even in the US, no dominant player exists in healthcare AI. The players are fragmented among various entities like EHR providers, tech firms, life sciences companies, insurers and hospitals. Those with vast healthcare data may not have efficient AI models, and vice-versa.

Singapore, on the other hand, has a unique opportunity to build a population-scale healthcare data ecosystem tailored for AI. We can more easily overcome the coordination challenges that may oblige larger ecosystems to build healthcare AI components in a piecemeal fashion. A platform called MOH TRUST already aggregates multiple healthcare research datasets, and the National Electronic Health Records (NEHR) system aggregates clinical data across Singapore. So the government is already collecting and coordinating healthcare data. What is missing is the impetus to use this data to drive the future of AI through an industrial policy.

Third, are there “moats” against global tech giants? Yes. Healthcare AI depends a lot on having local healthcare teams and physical sensors to collect and manage clinical data. AI can serve as an assistant to local healthcare workers and give Singapore an edge over others. Singapore has already signalled its commitment to generating comprehensive data, like in the SG100K genome study. Healthcare also has more durable data moats than other AI spheres like linguistics, where SEA-LION’s defences against global tech giants in low-resource languages is uncertain.

Fourth, does healthcare have export potential? Yes, as a public good, healthcare AI can benefit other countries while facing fewer sovereignty concerns. Debates are taking place globally about where large AI systems are trained and deployed. However AI, when used for healthcare, which can potentially benefit everyone, is less likely to attract controversies or nationalist and protectionist tendencies.

By assuming a leadership role in this field, Singapore can export our healthcare AI innovations, generate much international goodwill and even use this to advance our foreign policy.

And finally, does it benefit Singaporeans? Yes, through improved public health, economic growth and global technological leadership.

Singapore has other comparative advantages in the race for global leadership in healthcare AI. We have a robust healthcare system, the SingPass digital ID, cross-domain talent and a history of government investments and interventions in specific industries.

Singapore’s demographic diversity provides rich healthcare data across ethnicities and ages. Healthcare AI can catalyse the development of adjacent fields like computational genomics and precision medicine.

Singapore has the ingredients for a breakthrough in Healthcare AI. To succeed, the government needs to take the lead in putting these ingredients together.

What should we do in Healthcare AI?

To realise a healthcare AI moonshot, Singapore must combine existing ingredients into a coherent strategy. We should aim to export specialised services and medical diagnoses, not raw data.

We can create an advanced AI model trained on genetics, protein biomarkers, histology and electronic health records. We should aim to radically improve our ability to prevent disease, intervene and make causal inferences.

A multi-modal healthcare AI foundation model moves beyond narrowly specified “point-solutions”. By having a single foundation model for, say, both chest X-rays and retinal image interpretation, we can overcome data fragmentation across medical specialties and make it increasingly possible to uncover foundational principles of diseases.

As for electronic health records, the NEHR system is valuable, but it needs to be AI-ready in order to consolidate complex datasets such as histology imaging, genetic data or protein data. We should create comprehensive longitudinal patient histories spanning years. Existing data needs to be sufficiently standardised to serve AI analysis without a massive amount of preprocessing.

An AI Healthcare Company under the AI Catalyst Corporation could drive this moonshot. This AI Healthcare Company needs its own versatile multi-modal foundation model integrated with the NEHR. This will enable large-scale analysis to identify at-risk groups, conduct preventative screening and perform early treatment. It will also enable high-quality acute care, as a simultaneous expert in genomics, biology, general medicine and the specialties.

The AI Healthcare Company could build the world’s best multi-modal healthcare AI and healthcare dataset. This could establish an unmatched resource — built in Singapore, for the world.

Local AI startups can also benefit from this foundation model to build their own applications to sell to the world.

Just like how OpenAI’s access to ChatGPT queries provides unmatched data for improving their future large language models, the first company to create a versatile multi-modal healthcare foundation model would likely find itself at the frontier of healthcare AI.

Conclusion

Mr Speaker, in conclusion, a healthcare AI moonshot strategy will position Singapore as a global leader in AI by leveraging our unique capabilities in consolidating biomedical and healthcare data. It is a national approach designed to secure data sovereignty, navigate data-privacy concerns, ensure Singapore captures the benefits of AI, and maximise public by-in. This isn’t about picking winners; it is a proactive strategy to ensure Singapore thrives in the AI-driven future to benefit all Singaporeans.

I have presented just one example of a moonshot that Singapore could pursue. There may be other moonshots of equal or greater merit. I welcome open debate on selecting moonshots, but we cannot pull our punches if AI is truly the new industrial revolution.

Singapore possesses the talent, resources and infrastructure needed to compete for the top spot in selected AI fields. Achieving it requires political will, a readiness to embrace risks and proactive intervention by the government. We can do it, and we must do it, for the benefit of Singapore and Singaporeans.

Preventing disclosure of confidential data through ChatGPT

In April 2023, reports emerged that engineers from Samsung Electronics accidentally leaked internal source code. They had uploaded it to ChatGPT, presumably as part of their input prompt to the large language model (LLM). In response to this incident, Samsung took swift action, banning employees from using popular generative artificial intelligence (GenAI) tools like ChatGPT. Additionally, the company urged employees who utilised ChatGPT and similar tools on personal devices to refrain from submitting any company-related information or personal data that could potentially unveil its intellectual property.

Furthermore, a research report released in June 2023, titled “Revealing the True GenAI Data Exposure Risk” by LayerX Security, highlighted a troubling trend. It revealed that 6% of employees had pasted sensitive data, including source code, internal business information and personal identifiable information, into GenAI tools. This concerning behaviour could inadvertently result in organisations unknowingly sharing their plans, product details, and customer data with competitors and potential attackers.

While this research primarily focused on private sector employees, I was concerned about the possibility of inadvertent sharing of sensitive official secrets with GenAI tools by civil servants and government contractors.

In light of these concerns, I raised several questions in Parliament regarding the government’s use of large language models (LLMs) owned by private or foreign companies:

a) How does the government ensure that confidential data is not disclosed in the input prompts for LLMs?

b) Whether the government has signed any non-disclosure agreements (NDAs) with these companies?

c) What are the companies that the government has signed NDAs with?

d) How does the government monitor compliance with such NDAs by these companies?

In response, Mrs. Josephine Teo, the Minister for Information and Communications, provided an explanation of the government’s approach. She assured Parliament that highly sensitive applications and data remain shielded from exposure on the Internet. For instances involving LLMs and sensitive data, open-source models may be customised for use but are strictly deployed on government servers and computers.

For less sensitive data use cases, AI models may be owned and managed by commercial and private companies. The government’s contracts with these companies include clauses pertaining to data handling and security. These clauses encompass non-retention of data and restrictions on data usage for training other products or models. She did not reveal which companies the government has signed NDAs with. She said that the government has implemented a range of technical, visual, and governance measures to ensure data security and enforce compliance. The Minister emphasised the government’s commitment to continuously reassessing the adequacy of these measures as technology evolves.

Here are the original questions raised and answers on 9 January 2024 in Parliament:

REGULATIONS ON INPUT PROMPTS FOR LARGE LANGUAGE MODELS TO PREVENT DISCLOSURE OF CONFIDENTIAL DATA

Dr Tan Wu Meng asked the Minister for Communications and Information whether the Government has plans to develop in-house artificial intelligence capabilities to ensure that input prompts for large language models need not be processed by private firms not under the purview of the Government, or by cloud computing units located in foreign territories or under foreign jurisdiction or control.

Mr Gerald Giam Yean Song asked the Minister for Communications and Information (a) when using large language models owned by private or foreign companies, how does the Government ensure that confidential data is not disclosed in the input prompts; (b) whether the Government has signed any non-disclosure agreements (NDAs) with these companies; (c) what are the companies that the Government has signed NDAs with; and (d) how does the Government monitor compliance with such NDAs by these companies.

Mrs Josephine Teo: Large language models (LLMs), such as those powering ChatGPT, have the potential to enhance the delivery of public services and the productivity of public officers. We adopt a risk-managed approach for LLMs, consistent with the existing public sector framework for the handling of classified information when using technologies such as Internet-based applications and the commercial cloud.

Highly sensitive applications and data are not exposed to the Internet. Where use cases involve sensitive data, open-source models may be finetuned for use but must be deployed on Government servers and computers.

For use cases involving less sensitive data, the artificial intelligence (AI) models may be owned and managed by commercial and private companies. Our contracts with these companies are governed by service agreements which include clauses on data handling and security, such as the non-retention of data, and limitations on the use of data to train other products or models. Beyond contractual safeguards, the Government has also implemented technical measures to screen sensitive data, visual cues to remind users on data security practices, and governance measures to enforce compliance.

We continuously re-assess the adequacy of our measures as the technology evolves.

Source: Singapore Parliament Reports (Hansard)

#ChatGPT #AI #Parliament #WorkersParty #MakingYourVoteCount

Tackling the scourge of scams

The rise of sophisticated online scams and the resulting financial devastation on victims is a critical concern impacting Singaporeans from all walks of life. Across our nation, residents, both young and old, tech-savvy and not, have fallen victim to these fraudulent schemes.

I have met constituents who have lost their life savings to scammers, with even fixed deposit accounts being cleared out by these criminals. Some have joint accounts with their children or parents, doubling the impact on families. These incidents demonstrate a concerning vulnerability that affects us all. While I consider myself relatively tech-savvy, I have to admit that even I feel the looming threat of becoming a victim one day.

Many victims I have spoken with describe a disheartening response from their banks. Upon reporting the fraud, they frequently receive responses that are frustratingly vague and non-committal. They offer little information, citing banking secrecy, and at times, a goodwill payment that doesn’t fully cover the loss. Victims are sometimes told by the police that the funds have been transferred overseas and nothing further can be done to retrieve the funds.

The technical nature of these scams is deeply concerning. “Drive-by download” attacks and the more advanced “zero-day” exploits make it possible for malware to be installed on phones with little or no user action. These methods exploit vulnerabilities in operating systems and applications.

In view of these sophisticated attacks, how far do the authorities investigate each reported scam, especially those involving screen reading and keylogging malware? Without thorough investigations, it will not be possible to ascertain fault and ensure that innocent victims are not held responsible for losses they did not cause. Is the default blame then placed on the victims who have to bear most of the financial loss?

Scams have emerged as a formidable obstacle in advancing digital access for our citizens, particularly in our senior community. Numerous elderly residents I have encountered express a fear of using internet banking or online payments because they are apprehensive about falling prey to scams. Consequently, I find myself hesitant to advocate the use of digital banking to them, despite its convenience, due to the real risk of them losing their entire life savings if they are targeted by scammers.

This situation has precipitated what the Member for Aljunied, Ms Sylvia Lim, aptly describes as a “crisis of confidence” with the digital banking system. Unless the authorities address the issue of scams more effectively and establish stronger consumer protections, our extensive efforts to transition all our citizens into a digitally empowered society will come to nought.

One tool the government has on hand to deal decisively with scams is the Online Criminal Harms Act. This will allow the government to, inter alia, direct online platforms to disable access to accounts suspected to be involved in scams. Parliament passed this Act last July, however it is only set to be progressively rolled out from this year. When will the Online Criminal Harms Act be fully operationalised? Given that an average of 87 scams are taking place every day in Singapore, each day of delay will be one day too late for many scam victims.

Banks must shoulder a greater responsibility in protecting their customers. I echo Ms Sylvia Lim’s earlier call for banks to reintroduce physical tokens as the default measure for multi-factor authentication for all their customers. Multi-factor authentication relies on a combination of “something you know” and “something you have”. However, when phones are compromised by malware, allowing scammers to view screens and keystrokes, this system collapses into a single factor. This allows scammers, who have access to the password entered by the user, to bypass the additional security layer. Therefore, bringing back physical tokens will reinstate the crucial second layer of security.

The Monetary Authority of Singapore (MAS) must more assertively and decisively tackle the problem of scams in the banking system to protect consumers. In my dealings with MAS when advocating for constituents victimised by scams, I have observed that MAS tends to forward these critical cases to the banks for follow up, instead of directly addressing and resolving the issues on behalf of victims. This delegation process then places the onus on the banks to determine who is at fault — the institution or the victim — for the occurrence of the scam. Such a practice raises serious concerns about the impartiality and effectiveness of the investigation.

I have also observed a discrepancy in MAS’ approach to enforcing actions on financial institutions for different violations. On one hand, MAS imposes very punitive measures like restrictions on acquisitions and additional capital requirements on banks when there are brief downtimes in online banking and ATM services. On the other hand, this level of decisiveness and rigour is markedly absent when addressing scam cases. MAS should require banks to tackle scams with the same level of intensity and rigour as they do in safeguarding consumers’ interests for system outages.

Scam victims need a comprehensive explanation from a knowledgeable and impartial entity like MAS about how the scam occurred. This explanation should detail the roles of banks, telcos, customers and other entities in both the occurrence and prevention of such scams. This will determine who is responsible and who should bear the cost of these fraudulent acts.

Furthermore, responsibility should not be limited to financial institutions, telcos and consumers. Social media companies and mobile phone handset manufacturers should be held accountable for securing their platforms against scams. All handsets sold in Singapore should be required to disable side-loading of apps by default and make it difficult for end users to override critical security features. Social media platforms should be required to have processes in place to remove fraudulent posts soon after being notified.

The Ministry of Communications and Information has revealed that a notable proportion of residents, approximately 37%, do not regularly update their devices. Many of these may be less tech-savvy users. It is not reasonable to expect that everyone will have the technical proficiency to keep their devices updated. Therefore, consumer protection strategies must be designed on the premise that a significant number of users will not know how to keep their devices updated, and should incorporate additional layers of security to safeguard these users.

A central agency should oversee all scam investigations and responses. I am aware of the Anti-Scam Command (ASCom) and the important work their officers are doing. However, given that ASCom is a department under the Commercial Affairs Department of the Singapore Police Force, I don’t think they can be held accountable for whole-of-government efforts to combat scams. Who, therefore, is ultimately accountable for the government’s anti-scam efforts?

To summarise, my recommendations are as follows:

First, banks must significantly increase their responsibility towards customer protection, including by providing physical tokens to customers.

Second, MAS should take a more active role in ascertaining responsibility for scams carried out on banks’ digital platforms and supporting victims.

Third, the Online Criminal Harms Act needs to be fully operationalised without further delay.

Fourth, the government needs to hold technology companies more accountable for the security of their platforms and devices.

And finally, a central anti-scam agency should oversee and be ultimately accountable for the government’s anti-scam efforts.

Mdm Deputy Speaker, we stand at a critical juncture in the battle against scams. Our actions in the face of this scourge will define our commitment to protecting our citizens in the digital age. Let’s act swiftly and decisively to protect our people and, indeed, ourselves. I support the Motion.

Medication adherence challenges

I recently asked a parliamentary question to obtain data about patients at public hospitals and polyclinics who declined to collect their prescribed medication. In my question, I sought information on the frequency of such occurrences, the average price of medication declined by patients, the common reasons cited for non-collection, and whether the Ministry of Health (MOH) would consider collecting this data if it is not currently doing so. I was seeking to address a concern that had been brought up to me about whether the cost of medication could be a reason for their non-collection.

In response, Minister for Health Ong Ye Kung explained that MOH does not currently track the number of times patients decline to collect their prescribed medication, nor do they monitor the average price of such medication. Instead, he referred to a research study published in March 2023 that highlighted various non-cost-related factors influencing medication adherence. These factors include concerns about potential side effects, lack of knowledge about the medication and the underlying disease.

I looked up the research study that the Minister was likely referring to. The study mentioned that:

Although financial issues were not mentioned as a factor substantially hindering medication adherence, a minority of participants expressed that paying for regular prescriptions could be a burden for the family in the long term and hence would likely impede medication adherence. As one participant described: “The new oral medication for my diabetes was so costly as it was not covered by subsidies. My family is not well-off. So I stopped the medication.” (#4, M, 71)

The issue of patients declining to collect prescribed medication is a matter of concern. It can lead to worsened medical conditions, increased complications and higher healthcare costs due to delayed and more intensive treatment. Understanding non-collection reasons is vital for healthcare providers and policymakers to craft effective strategies for boosting medication adherence. These include addressing cost-related and other barriers.

The Minister did not answer my question about whether the Ministry will consider collecting such data. I would argue that public healthcare providers should track this to better grasp the medication adherence challenges and enhance patient care and outcomes.

Here is the full question and answer on 22 November 2023

DATA OF PATIENTS FROM SPECIALIST OUTPATIENT CLINICS AT PUBLIC HOSPITALS AND POLYCLINICS DECLINING COLLECTION OF MEDICATION IN LAST FIVE YEARS

Mr Gerald Giam Yean Song asked the Minister for Health (a) in each of the last five years, how many times have patients from specialist outpatient clinics at public hospitals and polyclinics declined to collect their prescribed medication; (b) what is the average price of medication declined to be collected by patients; (c) what are the common reasons cited for non-collection; and (d) whether the Ministry will consider collecting such data if it is not currently doing so.

Mr Ong Ye Kung: The Ministry of Health does not track the number of times patients from specialist outpatient clinics at public hospitals and polyclinics declined to collect their prescribed medication. We also do not monitor the average price of such medication.

A research study published in March 2023 concluded that a wide range of non-cost-related factors influenced medication adherence. This includes concern about side effects, lack of knowledge of the medication and the disease.

Source: Parliament Hansard