top of page

‭Fostering A Trusted Ecosystem For‬‭ Generative Artificial Intelligence‬‭


Image: Credits to Unsplash (Unsplash: Growtika) https://unsplash.com/photos/a-computer-generated-image-of-a-network-and-a-laptop-f0JGorLOkw0

In this Explainer, find out…‬

  • ‭Why is Generative Artificial Intelligence‬‭ (GenAI) important for Singapore?‬

  • What are some issues surrounding GenAI?‬

  • ‭How is Singapore planning to manage‬ GenAI?‬

Introduction


You have probably heard about ChatGPT. For most of our readers, you would likely have thought about using ChatGPT for writing essays too. After all, it is one of, if not the biggest manifestation of Generative Artificial Intelligence (GenAI) that has emerged, transforming some of the ways we think, work and behave.


ChatGPT aside, GenAI has taken the world by storm. Its advancements are salient and global. Singapore, with its growing service and knowledge-based economy, will also be affected by these advancements.


So, what is in store for Singapore? What are some of the key issues surrounding GenAI and how does Singapore intend to manage it? Read on to find out!


What is Generative AI?


Before delving into GenAI, we must first understand what artificial intelligence (AI) is. AI can be loosely defined as the ability of a machine to simulate human intelligence. AI has automated mundane operations such as clerical work and logistical operations. It has also been used to augment human capabilities, for instance to prevent traffic accidents when driving autonomous vehicles.


GenAI is a specific type of AI characterised by its ability to produce new content after analysing patterns and relationships in datasets.¹ Its emergence can be attributed to, among others, advances in natural language processing and the development of large language models. Together, both have enabled machines to interpret and generate human speech and text. Given its design, GenAI can respond more relevantly and effectively to user queries compared to traditional AI. This has facilitated a wide array of applications, including personalising customer experiences and providing help with writing essays.


In fact, the potential of GenAI transcends what models like ChatGPT offer. Its value lies in its potential to uncover new solutions, allowing us to push the limits of science and technology. For instance, researchers of antibiotic resistance have been able to leverage GenAI to suggest new compounds with antibacterial properties and generate instructions to synthesise them. This opens up new areas for research in chemical interaction and antibiotic development, a promising key to overcoming antibiotic resistance.²

How Singapore Can Harness Generative AI


There are many opportunities arising from GenAI. As a growing service and knowledge-based economy, Singapore can benefit greatly from researching and deploying GenAI. Furthermore, GenAI can also enhance public service delivery, thereby playing a constructive role in society.


The Services Economy


GenAI can create many opportunities in the services sector. For one, the financial services industry can harness GenAI to analyse current trends and recommend improvements to financial oversight mechanisms. This engenders increased transparency and accountability, which will help build trust between key stakeholders and attract more financial investments locally and in the region.


Businesses can also tap on GenAI to improve customer experiences. For example, customers purchasing products online may have a more enjoyable shopping experience as GenAI can engage with them and recommend products according to their tastes. Alternatively, customers interacting with customer service agents may benefit from better service if the latter can leverage GenAI to provide more relevant responses given the unique case they are managing.³


The Knowledge-Based Economy


Besides creating opportunities in the services sector, GenAI can also fuel the growth of Singapore’s knowledge-based economy — defined as an economy that produces, distributes and uses knowledge and information. In fact, Singapore’s desire to harness AI (and by extension, GenAI) to nurture a knowledge-based economy is already evident from blueprints such as the Singapore National AI Strategy 2.0 (NAIS 2.0).


Specifically, NAIS 2.0 builds on NAIS 1.0, outlining further plans to deepen the use of both traditional AI and GenAI to transform Singapore’s economy. This is critical, as GenAI is expected to unlock at least US$2.6 trillion in economic benefits globally, and Singapore should cement its expertise and global position in AI research, development and usage to reap these benefits. GenAI will also catalyse advancements in existing areas. This will add between 15 and 40 per cent of growth on top of what traditional AI and analytics can achieve.


The Public Service


Finally, GenAI can also be harnessed to enhance public service delivery. Similar to businesses, governments can leverage the unique strength of GenAI to create content by analysing data and patterns. For instance, government agencies can use improved chatbots to interact with Singaporeans more autonomously and effectively. This will enable government officials to dedicate more time and attention to complex requests.


Government services can also tap on GenAI to improve educational outcomes. For instance, GenAI can be used to create adaptive lessons for students who are struggling with specific subjects. Through personalised lessons, students can learn at a pace that is more suitable for them. This will be beneficial for students who face learning difficulties, yet are unable to access often expensive educational resources.


GenAI can also be harnessed to increase accessibility to political discourse in Singapore. As more Singaporean youths become interested in the local political scene,¹⁰ it is arguably necessary for political discourse to become more accessible. Furthermore, greater awareness of political discourse and policy concerns can empower voters to make informed decisions while championing civic action at the grassroots.


On this note, the Parliament of Singapore has recently introduced a newer Hansard search engine powered by GenAI. Unlike its predecessor, which often produced irrelevant search results, the GenAI-powered search engine has thus far yielded more relevant search results based on user queries.¹¹ This could make political discourse more engaging for Singaporeans, thus reducing political apathy in our society.


Uncharted Waters


While the benefits of GenAI abound, there are also concerns surrounding the design and use of GenAI. As a relatively nascent field, some of these concerns are new and specific to GenAI. They are also poorly understood, warranting further research and collaboration.


Problematic Biases


First, just as predictions made by traditional AI may perpetuate human biases, new content generated by GenAI models may suffer from biases too. After all, GenAI models must be trained using large sets of human-generated data that are ridden with undesirable stereotypes and biases. In turn, such biases will be reflected in newly generated content.¹² To this end, both accountability through safety nets and efforts to consistently review AI models are necessary to minimise GenAI’s perpetuation of biases.


GenAI v. Jobs


Second, the proliferation of GenAI will have profound impacts on a myriad of jobs.


Besides accelerating the decline of low-skilled jobs, a trend which has been previously driven by automation, GenAI could also erode jobs in the creative industries.¹³ Worse, those in the latter group, whom some have believed to be invulnerable, may not be prepared to handle the disruptions caused by AI.


Meanwhile, developers and owners of technology could earn and take an even greater share of profits as GenAI products become more ubiquitous.¹⁴ These trends will only widen the income divide, harming individuals, society and democratic institutions.¹⁵


That being said, there is also evidence suggesting that GenAI may reduce income inequality. This may occur, for example, if low-skilled workers learn how to use GenAI at work, thereby raising their productivity relative to high-skilled workers.¹⁶


Nonetheless, for GenAI to benefit society more fairly, the development and usage of GenAI need to be in the public interest. Workers and industries need to be better prepared to navigate the disruptions posed by GenAI. This could be in the form of upskilling low-skilled workers and encouraging GenAI to augment all types of jobs.


Power and Abuse


Third, the market power amassed by companies who own or utilise GenAI technologies could be abused. With GenAI, companies wield significant influence over consumption behaviour and patterns,personalising ads and adopting creative and novel marketing techniques to increase customer loyalty and boost sales.¹⁷ However, this could encourage overconsumption, indirectly contributing to fossil fuel usage and climate change.¹⁸


Beyond that, abuse perpetuated through GenAI could take a much more sinister turn. As firms concentrate greater market power, they could also become pawns for exerting political influence. Because of the ability of GenAI to generate content like deepfakes, misinformation campaigns have become more convenient and effective. A case in point is the use of deepfake technology in support of Russia’s invasion of Ukraine.¹⁹ Misuses like these can sow confusion and doubt, even forcing us to question our beliefs and condone violence. They fundamentally undermine communities, institutions and the rule of law.


Another avenue for misuse is cyberbullying through deepfake pornography. Almost all deepfake pornography targets women and underage girls.²⁰ Not only is this non-consensual but also derogatory and abusive. While they do not cause physical harm, the mental trauma is permanent. Victims could spiral into shame and self-harm, let alone irreversibly damaging communities and relationships.²¹


Greater oversight over GenAI developers and the wider ecosystem is therefore urgently needed. It is irrational and rash to ban GenAI, forsaking the opportunities for economic growth and beneficial social outcomes. Rather than imposing a blanket ban, platforms should deter misuses and abuses, have guardrails against them, and ensure that GenAI is used in the public interest.


Model AI Governance Framework for Generative AI


Recognising both the benefits and dangers posed by GenAI, the Singapore Government has opted to establish an authoritative and robust framework to harness the benefits of GenAI while protecting Singaporeans from potential harm. Such efforts began in 2019 with the release of the Model AI Governance Framework, later updated in 2020.²²


However, as GenAI becomes mainstream, its unique challenges will require a more relevant set of solutions. To this end, the Infocomm Media Development Authority released the Proposed Model AI Governance Framework for Generative AI (the Framework) in May 2024.²³ It seeks to address these challenges through a multi-pronged approach, targeting nine dimensions individually and in totality (see Figure 1).


Figure 1: Summary of the Proposed Model AI Governance Framework for Generative AI²⁴


This Policy Explainer will focus on three of the nine dimensions, namely:


  1. Accountability;

  2. Safety and Alignment Research and Development (R&D); and

  3. AI for Public Good.


Accountability


First and foremost, the Framework advocates for relevant stakeholders, such as developers and businesses, to be accountable to end-users throughout the development and application of GenAI. This may be achieved through ex-ante or ex-post allocation of responsibility to stakeholders.


Allocation of responsibility occurs ex-ante if it happens during the development process. By identifying each stakeholder and their roles in the developmental process of GenAI, responsibility can be more clearly and fairly allocated. How responsible each stakeholder is depends on their role in the developmental process and how much control they have over potential issues.


Conversely, responsibility is allocated ex-post if it occurs after unanticipated scenarios arise. As the challenges posed by GenAI evolve, some of them may be unexpected. In this case, allocating responsibility is less straightforward; is it justified for any one party to be responsible for unexpected problems? Reviewing the legal framework and mandating insurance clauses are some examples of safety nets to compensate victims fairly.²⁵


Safety and Alignment R&D


Besides accountability to end-users, The rapid speed of AI development necessitates human oversight and intervention to minimise its propensity to perpetuate harmful content. To achieve this, both forward and backward alignment research is required.Forward alignment largely involves developing GenAI models to follow predetermined safety and ethical principles. It also taps into methods such as reinforcement learning, where human feedback is given during GenAI development to enhance the future safety and alignment of GenAI models.²⁶


Backward alignment, in contrast, involves testing models for potentially harmful abilities after their development. For instance, the autonomous replication of AI, where AI can regenerate itself without human intervention, poses numerous safety and ethical risks. Unaligned superintelligent systems without moral constraints can potentially create havoc through criminal activity.²⁷ Hence, there is a need to ensure a cohesive R&D process throughout the development of GenAI.²⁸


AI for Public Good


Finally, to ensure that the benefits of GenAI are distributed fairly across society, it should be designed and used in the public interest. The Framework suggests four areas to harness GenAI for the public good:


Democratising access to technology: Ensuring GenAI is accessible to all across society. This includes increasing digital literacy through partnerships between government and communities.


Public service delivery: Government agencies working to enable and empower the public through AI. This ranges from increasing efficiencies in the public sector to providing solutions to address problems citizens face.


Workforce: Redesigning jobs and upskilling the workforce to enable workers to use AI productively. This could reduce wage inequality within occupations. This also alleviates concerns surrounding the loss of jobs from the rise of AI.


Sustainability: Tracking and measuring the carbon footprint of GenAI, while working with stakeholders to support climate responsibilities. Water and energy resources can be used more efficiently, enabling sustainable economic growth from GenAI without creating significant environmental externalities.


Collectively, they address social concerns such as displacing workers, and environmental ones like the intensive use of energy and water resources. They also facilitate innovation needed for global collaboration on policy approvals related to generative AI.²⁹


Looking Abroad


Beyond Singapore, there is ongoing research in different parts of the world, similarly exploring policies to ensure the proper development of GenAI.


One comprehensive piece of legislation over AI is the European Union (EU) AI Act. It has gained traction in the international community as seen from the likes of Brazil, where a legal framework for AI has been passed in Congress.³⁰ The EU AI Act is notable for its classification of AI according to its associated risks, as well as the checks and balances on high-risk and general-purpose AI.³¹


The Act classifies GenAI as a general-purpose AI, therefore requiring it to adhere to several requirements, including honouring copyright directives.³² This includes works by the creative industry, which could be used as data for training AI models. While using creative work as data can elevate the quality and creativity of search results, creators do not receive sufficient credit for their work. Convenient usage of GenAI creations can also undermine artistic identity, integrity and fair competition in the industry.³³


While the creative scenes are different between Europe and Singapore, several lessons can be gleaned. Contextualising them to fit Singapore’s culture and needs is important and timely as Singapore develops its creative economy and plans how to leverage technology like GenAI.³⁴


The EU AI Act is only one of the many policy tools being developed globally to manage GenAI. While the issues faced by each country differ, fundamental realities and challenges often bear similarities. This underscores the importance for countries, including Singapore, to collaborate with partners abroad. Furthermore, as the technology evolves, international cooperation will help in adapting regulations to address new risks introduced by GenAI.


Conclusion


AI is here to stay, ideally as a force for good. For Singapore, the stakes are even greater for us to remain relevant on a constantly evolving global stage.


Overcoming valid and pertinent concerns with AI is necessary. They should not deter research and development efforts to deprive communities of the benefits of AI. This thus underscores the importance of developing a constructive, accessible and relevant framework like the Model AI Governance Framework for Generative AI. But it does not stop here. AI is a global phenomenon; transnational cooperation is a must.


AI should be used in the public interest. AI should empower and build trust in societies. AI should be one reason humanity will look forward to the future.


MAJU_PE_2024_24_Fostering A Trusted Ecosystem For Generative Artificial Intelligence - Goo
.
Download • 2.80MB
 

This Policy Explainer was written by members of MAJU. MAJU is an independent, youth-led organisation that focuses on engaging Singaporean youths in a long-term research process to guide them in jointly formulating policy ideas of their own. 


By sharing our unique youth perspectives, MAJU hopes to contribute to the policymaking discourse and future of  Singapore.

99 views0 comments

Comments


bottom of page