18 Sep 2023
Minister for Foreign Affairs Dr Vivian Balakrishnan delivered remarks at an event held on the sidelines of the 78th United Nations General Assembly titled “Artificial Intelligence (AI) for Accelerating Progress on the Sustainable Development Goals (SDGs): Addressing Society’s Greatest Challenges” on 18 September 2023.
The side event, co-convened by Japan, Kenya, Morocco, Singapore, Spain, the UK, and the US, also featured remarks from US Secretary of State Antony Blinken, UK Secretary of State for Foreign, Commonwealth and Development Affairs James Cleverly, and Spanish Secretary of State for Digitization and AI Carme Artigas.
Minister Balakrishnan spoke about the transformative benefits of AI in sectors including health, climate, and food security, and encouraged greater ambition in the use of AI to accelerate implementation of the SDGs. Minister Balakrishnan highlighted that alongside the opportunities, AI also brought with it potential risks. As such, it was essential that the right safeguards were put in place to ensure privacy and transparency, and to build trust and confidence in the responsible development and deployment of AI. On the issue of developing a multilateral, global framework for AI governance, Minister Balakrishnan underscored the need to include diverse viewpoints and advocated for multi-stakeholder engagement involving the commitment of governments, industry, and civil society. Minister Balakrishnan also shared Singapore’s experiences in developing an AI governance testing framework, AI Verify, as well as the launch of the AI Verify Foundation to bring the private sector and industry on board to develop better standards and practices in the use of AI.
The transcript of Minister Balakrishnan’s remarks is appended.
. . . . .
TRANSCRIPT OF MINISTER FOR FOREIGN AFFAIRS DR VIVIAN BALAKRISHNAN’S REMARKS AT THE US’ SIDE EVENT ON “ARTIFICIAL INTELLIGENCE FOR ACCELERATING PROGRESS ON THE SUSTAINABLE DEVELOPMENT GOALS”, 18 SEPTEMBER 2023, NEW YORK
Good afternoon everyone, it is a pleasure to be here.
Singapore is a tiny city state. Our trade volume is three times our GDP, and we only derive global significance by being a critical node in a global network.
Of course, trade, I have mentioned, the flow of goods and products, but equally, if I ask you now to Google submarine cables in Asia, just look at how many radiate into or out from Singapore. So, I say this so that you understand that my views are quite naturally biased. But I am in favour of openness, in favour of connectivity, and later on I want to emphasise the role of regulation and in particular, trust.
Let me first start off with the Singapore Government's attitude to AI (Artificial Intellience). The pace of transformation in the last decade is astounding. Since this is UNGA (United Nations General Assembly) week, have you all noticed the quality of translations has improved significantly within the last decade? To some extent, it is because the UN (United Nations) insists on publishing multilingual, authoritative versions, and the combination of deep-learning, machine learning, against this corpus of curated, high-quality data, enabled pattern recognition and therefore, machine learning, translation, and computer systems which could read, see, classify, translate, and we have all witnessed that within the last decade.
Currently, the latest craze is large language models and generative AI. But even as we speak, I think this is about to be replaced by what I would consider to be the next level in which AI agents have transactional ability – which means to take instructions, to interact with other humans, organisations, and even other AI agents to get into financial transactions, to make contracts. This may sound like pie in the sky, but I think we are on the verge of going beyond generative AI into transactional AI.
Now, before I run ahead of myself, come back to Singapore. Our attitude is, this is a big deal. This is transforming the means of production in the world. For a little place like Singapore to be relevant, we better get on this bandwagon. The first point is that we believe in learning by doing. In other words, we have to practice because we are not going to invent the large language models. We cannot compete with the Silicon Valley. But we can be a vital node which the Silicon Valley needs, even as it spreads its web across the world.
Let me cite you a few examples where we are actually using AI in Singapore. First, in the Singapore Civil Service today, we are using large language models.
Another example. In the eye center, for years, we have been using retinal photographs to predict systemic diseases because it just so happens, the eye is an outgrowth of the brain which is visible and the patterns of the blood vessels, their caliber, the way they cross and exudate, gives us quite a lot of clues about the systemic issues that you have.
Another model, in my previous incarnation in the Ministry of Environment, we are using models, AI models, because we are worried about floods, droughts and extreme weather events which we all know are getting worst because we are behind time in reaching the SDG (Sustainable Development Goals).
Another example, even for fish farming, because you know, we do not have enough land. Again, optimising the fish farms, for security, for biosafety and for improving yield. The point I am trying to make is that we are just doing this, and if need be, doing it within sandboxes because the only way to get on this bandwagon is to do, and to learn from doing, and to learn from those mistakes. Fail quickly and improve it.
The second point is that it is not enough to just get on the bandwagon and do it, but you do need to make sure that safety, privacy, and public acceptability is part of it.
In fact, for governments which want to be early adopters, it is critical to have the right safeguards and guidelines in place for the responsible use of AI and demonstrably so, so that people can judge that you are not just experimenting with their data and taking liberties with their data, but you are doing it in order to improve the quality of services provided to the people.
One approach in which we wanted to be transparent and take a risk-based approach is that in 2019, Singapore released our own Model AI Governance Framework which provides detailed and readily verifiable implementation guidelines to private sector organisations, and to assure people that key ethical and governance issues are taken into consideration when rolling out these solutions. Last year, we went further to launch an international pilot of what we call AI Verify. This is an AI governance testing framework, a software toolkit to help organisations and governments objectively demonstrate that their systems are behaving responsibly, are based on sensible, accurate data, and to do these tests in a standardised form.
The second is that in order to build trust and confidence, we need to ensure that you have wide buy-in and that means, in particular, diverse inputs. And to know that the diversity will pose challenges, but you do need to take into account the diversity in each of our societies in order to arrive at a common approach, based on trust and transparency to the AI frame development. Again, I want to emphasise that it is critical that this not be a purely government venture, but you do need to get the private sector involved. You do need to get civil society and other stakeholders who have real skin in the game as you generate the appropriate regulatory framework.
This year, we launched the AI Verify Foundation. This is our attempt to put this principle into practice to develop better frameworks, standards and share best practices. Our current membership stands at more than 60 organisations, including our friends from IBM, and of course Google, or Alphabet, and Microsoft. The point is this is a public-private-civic partnership. I think that kind of multi-stakeholder engagement and approach is essential.
Third, I just want to end by saying that AI is transformative. It has got great opportunities, but it also has great risks. But these opportunities and risks are issues without a passport, meaning it quickly disseminates and escapes globally.
But the impact, both in terms of opportunities and risks, actually have asymmetrical effects on different countries and different societies. The question is, how ready are we when we face this tsunami of impacts, both positive and negative? We believe that we need regional, and we need multilateral, global approaches. At some point, you cannot avoid the question of governance.
I think we have passed the old paradigm when big tech will say, “do not worry, trust us”. I think that is not enough. I think you now need regulations. The political dialogue and conversations, both domestically and internationally, will have to occur. Within ASEAN, which is the Association for Southeast Asian Nations, Singapore is working with our neighbours to develop an ASEAN Guide on AI Governance and Ethics in order to establish common principles for trustworthy AI and to suggest best practices which we can all learn by doing and sharing.
Let me just conclude that this is an exciting field. Watch this space. But we do need to expand our minds and understand that just as AI has transformed the scene, the old ways of regulating, cooperating and legislating will not suffice and we do need to break new ground.
Thank you all very much for your attention.
. . . . .