Minister Josephine Teo's comments at Meta APAC AI Accelerator Finals 2024

TRANSCRIPT OF MINISTER FOR DIGITAL DEVELOPMENT AND INFORMATION JOSEPHINE TEO'S FIRESIDE CHAT AT META APAC AI ACCELERATOR FINALS 2024 ON 3 OCT 2024

Question: Minister, it is a very exciting week for us in Singapore as the government rolled out a revamp of its Smart Nation initiative. Singapore was one of the earliest countries to put together a national AI strategy, which it has just revamped at the end of last year. Could you tell us the role that open-source AI plays in achieving these goals?

Minister: Thank you very much, and I really want to say a big welcome to Nick again for making his way to Singapore. As it turns out, I met Nick at Davos on the sidelines of the World Economic Forum Annual Meeting in January. At that time, Singapore had just put out our refreshed National AI Strategy. If I were to take a step back and think about the time when we were thinking about our National AI strategy, what was really exciting for us was the "democratising" effect, which meant the ability to give people the power to take advantage of technology and bring about benefits to society. We were also quite struck by the fact that because this whole domain is quite new for many cities and nations, there were no ready playbook. In fact, we were all going to have to co-create and figure out what would really work for society.

What we thought was meaningful for us to do as a nation was to consider how AI could be used for the public good. As we have seen, this technology is being used by the broad community of AI developers who are not necessarily plugged into businesses and do not necessarily have all the resources that businesses are able to bring to bear. Therefore, open source becomes important. Because for anyone to really live up to the democratising impact of limited AI, they have to start with a base that is already quite built up and developed. If they have to start from scratch, the journey would be just too difficult to overcome. So, we are very excited by the fact that over the last few months in Singapore, we have seen the AI ecosystem really taking off, really taking shape, a large part of it powered by the fact that the most sophisticated models, including Llama 3, are being made available in Singapore and the region. We understand that, of course, they are not made available everywhere, depending on the jurisdictions.

We are very happy that the business use cases are taking shape very nicely. If you look at the leading companies in Singapore, whether it is DBS, Singapore Airlines or Prudential, they are all developing models and finding use cases. Between just this handful of leading companies, there are already thousands of use cases and hundreds of models that have been developed. Now, they would only be able to do that if, to a significant extent, they were also using open-source models to start off as it is. We see that there are industry organisations also coming together. For example, in the last month or so, there was an AI Centre of Excellence that was launched for the manufacturing sector. In other words, looking at the potential to use AI across many different manufacturing settings and helping the manufacturing sector to become even more competitive than it already is.

We also have the Centre of AI in Medicine (C-AIM) that was set up for healthcare. I think it is also very interesting that the research community wants to be part of this. The research community is very interested in how AI tools can enable them to make scientific discovery a more efficient process and one that also produces better results. So, when we think about AI for the Public Good for Singapore and the world, it is great that open-source models are enabling enterprise innovation, industry innovation, and also research innovation. I would like to mention the fact that because we wanted to support the research community in terms of the interdisciplinary use of the hybrid tools, as well as the purposes of scientific discovery, we set aside a significant amount of resources. I am sure many of these researchers for biomedical and material sciences are very interested in how to leverage on these resources.

Question: How would the Singapore government want to work with the ecosystem, both developers here in Singapore and also across the region, as well as companies like Meta, who produce these large foundation models, to maximise the positive impacts of AI and minimise the risks?

Minister: Thank you so much, that is a great question. I also want to build on what Nick has shared, Singapore does not have an "not invented here" approach. It does not matter to us where the innovation is from, if you can find usefulness in Singapore through implementation, we are very happy to be part of the new demand. So, when I was looking at the list of innovations that are coming through the Meta AI Accelerator programme, I was equally excited that, apart from the team from Singapore, some of the other ideas are really very exciting in our context. For example, there is this particular project about how you can use AI to make energy consumption within buildings more efficient. That is of great interest to Singapore because we are very built-up city and serious about our climate commitments. We know that buildings contribute to 20% of global greenhouse gas emissions. So, if we could make buildings more efficient in how they consume energy, that is not just good for the businesses that own those buildings, that is good for Singapore. That is actually also potentially a contribution to how the wider world can be more efficient in how we consume energy in buildings.

So, the way in which we think about the AI ecosystem is an open mind - meaning that the innovations can come from anywhere. We want to create an environment where big demand can materialise and then at some point, find support to be further commercialised and spread around the world. Part of the way in which we do this is to ensure that there are enough people in this ecosystem that are regularly talking to each other. For example, there is a group of AI developers that mainly does work for the government, meaning they are using AI models within the government. They meet every Wednesday, and they bounce ideas off each other. Now I am going to hazard a guess that it is not only this community that meets, there will be other AI-related communities whose work is relevant to the research community, healthcare community and financial service community. So, you have many of these communities that are brewing in Singapore, and we welcome more to be set up.

Another way in which we want to support AI adoption, not just for Singapore, but actually in the wider region, is to make sure that we implement AI in a responsible way. Because just as there are many people who see the potential of AI, there are also those who have valid concerns about the risks -the risk of discrimination, the risk that you can have data poisoning, the risk that with these models, you longer need humans in the loop, and they act in ways that are inimical to our interests. How do we create an understanding of the AI risks? Find a way to identify those risks and test for those risks before these AI models are further put into deployment. I think these are some meaningful areas of collaboration that we would also like to pursue alongside the promotion of AI activities in Singapore.