MDDI 演講稿 · 2025-03-13
高階政務部長Janil Puthucheary在Meta Llama孵化器計劃新加坡啟動儀式上的致辭
要點
- • Meta在新加坡啟動Llama孵化器計劃，收到逾100份申請，為期六個月，設初創企業、中小企業及政府機構三條專屬賽道。
- • 新加坡於2023年釋出《國家AI戰略》第二版，在首版及智慧國家議程基礎上，納入大型語言模型領域的最新進展。
- • 孵化器內嵌「負責任AI設計」理念，將藉助資媒局旗下開源治理工具AI Verify與Project Moonshot，提升使用者對AI產品的信任。
- • Llama模型已部署於新加坡政府商業雲平臺，公共部門人員藉此開發並向全體政府機構交付AI產品。
- • Meta孵化器匯聚AI Singapore、SG Innovate、e27及德勤等合作伙伴，進一步深化新加坡與Meta的AI戰略合作關係。
- • 新加坡將「以最優技術促進公共利益」定為AI核心戰略願景，以此統領所有合作、生態建設與人才發展工作。
完整譯文（繁體中文）
MDDI 英文原文譯文 · 翻譯日期: 2026-06-21
數字發展與信息部高階政務部長賈尼爾·普圖切利博士（Dr Janil Puthucheary）在新加坡 Meta Llama 孵化器計劃啟動儀式上的開幕致辭（2025年3月13日）
1. 歡迎出席 Meta Llama 孵化器計劃在新加坡的啟動儀式。
2. 本計劃收到了逾100份申請，其中不乏許多極具創意的提案，令我深感振奮。
3. 在座的各支團隊，接下來六個月，你們將與導師及合作伙伴攜手踏上一段意義非凡的旅程，藉助領先的 AI 工具開發應對現實挑戰的解決方案。這將是一段令人振奮、充滿意義的旅程！
4. 與你們一樣，新加坡政府也一直在踏上與 AI 同行的旅程。這一旅程始於多年前，是彼時智慧國家之旅的延伸。最新的迭代著眼於技術驅動下社會、經濟與政府的全面轉型。我們認識到，AI 的運用有潛力改變新加坡的運作方式，並幫助我們突破過去60年來一直面對的現實物理侷限。
5. 隨著技術驅動發展的步伐不斷加快，我們必須努力確保自身處於前沿。這需要我們致力於構建合作關係、打造生態系統，並培育你們所代表的那種人才與能力。
6. 我認為，作為政府，我們做得並不差。有哪些指標可以說明這一點？去年年底，我有幸出席了一場專為非技術類公務員舉辦的全政府提示詞工程競賽。競賽的最後一項任務是為虛構的"新加坡2100年奧運會"搭建一個功能完整的網站。參賽者須在八分鐘內解決有關交易、訂票及宣傳推廣的問題。最終贏得比賽的，是一名身著 SCDF 制服、剛完成24小時輪班的在職消防員。
7. 僅憑我能講述這個故事——有一個生態系統在支援公共部門官員提升能力、交付真實產品——我認為，這本身就使我們與眾不同。
8. 我們認為，另一個令我們與眾不同之處在於我們的國家 AI 戰略。2023年，我們釋出了第二版，在第一版經驗的基礎上，同時結合大型語言模型的發展——正是這些發展改變了我們今天所探索的部分機遇。
9. 在整個過程中，我們始終與私營部門攜手合作。我們並沒有打算將一切都由政府包辦的想法。我們工作的大部分是與私營部門合作，匯聚應用場景與機遇，並最終確保與私營部門所代表的專業人才建立合作關係。
10. 加速器計劃是其中的重要組成部分。我們已與多家科技企業開展此類合作，同時也嘗試自身扮演加速器的角色。我們設有 Lorong AI——一個連線政府與私營部門的空間。人們不僅可以在此分享運用 AI 解決問題的方法，政策制定者和政府人員也能直接從參與者處瞭解他們如何應對 AI 產品開發與部署相關的問題。
11. 我們還支援各類平臺匯聚並展示來自整個區域的最佳創意。各方參與者以截然不同的方式應對某些問題，這對我們的學習非常有價值。其中一個例子，當然就是 Simon 所提到的 Meta AI 加速器。全新的 Llama 孵化器是這段旅程的新篇章，是我們在 AI 領域工作的可喜補充，並帶來了幾項重要特色。
12. 首先，是為參與的中小企業、初創公司和政府機構設立專屬賽道的理念。這一安排認識到，這三個賽道各自面臨不同的挑戰，在如何培育技能與開發產品方面也可能有著不同的需求。
13. 初創公司賽道幫助初創企業將 AI 融入業務運營，提供諸如完善 MVP 概念、驗證原型以及市場進入策略等方面的指導。對中小企業而言，則聚焦於如何建立並整合定製化 AI 解決方案，以提升運營效率與競爭力。因此，這是對"沒有放之四海而皆準的方案"這一現實的認可。
14. 第二個重要特色，是融入計劃的"負責任 AI 設計"理念。這一理念認為，AI 安全不僅是政策層面的基本概念，更是為客戶群建立產品信任的根本所在。
15. 據我瞭解，孵化器計劃將藉助 IMDA 的 AI 治理工具，包括對若干安全關鍵應用場景的深度探討。
16. IMDA 擁有這些工具和框架，再次印證了公共部門官員長期以來的努力——以不制約創新、不阻礙新加坡豐富生態系統發展的方式思考 AI 安全。多年來，他們與國際夥伴及私營部門攜手合作，深入探索並理解如何在使用者群體中建立信任。
17. 為此，他們開發了治理測試框架及 AI Verify、Project Moonshot 等開源工具。今天晚些時候我們將聽到相關介紹，希望大家能夠使用這些工具，藉助其所代表的經驗，將你們的產品與創意推向更高層次。
18. 當然，Simon 所提到的第三個重要特色，是 Llama 所依託的開源技術。開源方式的核心成果在於，即便不是天然的使用者或開發者，任何人都能以極低的門檻即時進入開發流程。這種開源方式讓新的群體得以實驗並受益於你們的 AI 平臺。
19. 正是這種賦能產業、研究和政府創新的能力，是由成熟 AI 工具所發展出的開源方式帶來的重要成果。這是將生成式 AI 民主化、惠及更廣泛利益相關者群體的體現。
20. Llama 是我們政府通過政府商業雲平臺所使用的模型之一。我們的公共部門官員不僅藉助該平臺進行實驗與探索，更以此開發產品，並向整個公共部門推广部署。
21. 以 Llama 為核心的孵化器與新加坡對開源方式的支援高度契合。因此，我們的機構確實一直支援該孵化器的發展。令人振奮的是，Meta 匯聚了來自我們生態系統各方的廣泛合作伙伴——AI Singapore、SG Innovate、e27 和 Deloitte。這凸顯了新加坡與 Meta 在 AI 領域日益深化的合作關係。
22. 歸根結底，這一切都植根於新加坡對 AI 使命的特定願景——推動公共利益。在眾多可能的路徑中，我們必須深思熟慮，明確自身的重點所在。當然，我們尋求合作伙伴關係、商業機會、人才培育與應用場景，但歸根結底，這一切都必須聚焦於一個戰略願景。我們希望向新加坡、向世界證明，使用最優秀的技術能夠、也應該服務於公共利益。這是我們的願景。我們所有的合作與夥伴關係都是為了推進這一願景，我們非常高興擁有志同道合的夥伴與合作者，他們對 AI 能夠實現的目標抱有相似的願景。
23. 因此，我們期待在 Demo Day 看到你們眾多的解決方案，共同慶祝你們的成功。我們期待你們的 AI 應用場景在新的受眾群體中廣泛傳播，也期待你們現有的受眾意識到——許多他們從未以為自己能做到的事，如今因為你們開發了最頂尖的前沿工具而成為可能。
24. 我們期待你們的探索，並希望藉此激勵更多人嘗試 AI、開發工具，成為通過 AI 改變我們的社會與世界的催化力量。
25. 非常感謝大家。
英文原文
MDDI 官網原始記錄 · 抓取日期: 2026-06-21
Opening remarks by Dr Janil Puthucheary, Senior Minister of State for Digital Development and Information, at the launch of Meta’s Llama Incubator Programme in Singapore (13 March 2025)
1. Welcome to the launch of Meta’s Llama Incubator programme here in Singapore.
2. I am heartened by the enthusiasm of this programme, having received over 100 applications with many very interesting proposals.
3. To the teams here, you will embark on a meaningful journey with mentors and partners over the next six months and work with leading AI tools to develop solutions that address real world challenges. It will be an exciting and meaningful journey!
4. Just like you, the government in Singapore has been embarking on a journey with AI. We started this years ago as an extension of what was then our Smart Nation journey. The latest iteration considers the technologically driven transformation of our society, economy, and government. We recognise that the use of AI can potentially transform what we do here in Singapore and allow us to overcome the very real, physical constraints that we have had to deal with for the last 60 years.
5. As that pace of technologically driven development increases, we have had to work to make sure that we're at the forefront. And that has required work to build partnerships and build an ecosystem and have the talent and capabilities that you represent.
6. I think we have not been doing too badly as a government. What are some of the markers of this? Well, towards the end of last year, I had the pleasure of attending a government-wide prompt engineering competition specifically for non-technical officers. The last task in the competition was to build a functioning website for a fictional Singapore 2100 Olympic Games. Participants were given eight minutes to solve issues regarding transactions, bookings and publicity. The person who won the competition was a serving fireman who turned up in his SCDF uniform after a 24-hour shift.
7. So just the fact that I can tell this story – that there's an ecosystem out there supporting public sector officers in capability development and delivering real world products - I think that sets us apart.
8. Another way we think we are set apart is our National AI Strategy. We launched the second version in 2023, building on what we had learned from the first version, and of course, considering the developments in Large Language Models that have transformed some of the opportunities we are working on today.
9. Throughout all this, we have worked with the private sector. We do not have this idea that we are going to do everything within the government. The bulk of what we do is in partnership with the private sector, crowding in use cases and opportunities, and ultimately making sure that there are partnerships with the skilled talent that the private sector represents.
10. Accelerator initiatives are an important part of this. We have done this with various tech players, but we have also tried to be an accelerator ourselves. We have Lorong AI, a space where we connect the government with the private sector. People can share not just the way in which they solve problems with AI but have policy makers and the people in government to understand directly from the participants how they're addressing some of the issues associated with the development and deployment of AI-based products.
11. We've also supported platforms to gather and showcase the best ideas from across the region. The players are approaching some problems in very different ways, and that's very useful for us to then learn from. And one of these examples is, of course, Meta’s AI Accelerator that Simon talked about. The new Llama Incubator is a new chapter in this journey. It's a very welcome addition to what we're doing in the AI landscape, and it brings a few important features to the table.
12. Firstly, is the idea of dedicated tracks for participating SMEs, startups, and government agencies. There’s a recognition that each of these three tracks has a different set of challenges and perhaps a different need around how they develop skill sets and products.
13. The startup track helps startups integrate AI into business operations, providing guidance on things like refining MVP concepts, validating prototypes, and strategies for entering markets. For SMEs, it's how to create and integrate custom AI solutions to boost operational efficiency and competitiveness. So, it's a recognition that there's no one size fits all.
14. The second important feature is the ‘Responsible AI by Design’ that's built into the programme. It’s the understanding that AI safety is not just a fundamental concept at the policy level, but it's fundamental in promoting trust in these products for the customer base.
15. I understand the Incubator is looking to leverage on IMDA’s AI Governance tools, including deep dives on a few of these safety critical use cases.
16. The fact that IMDA has these tools and frameworks, again, is a testament to the work that public sector officers have been doing for some time - thinking about AI safety in a way that has not constrained innovation and the development of a rich ecosystem here in Singapore. And they have been doing this for years, working with international partners and the private sector to develop a deep interest and understanding of what it takes to build trust in the user base.
17. And so, they've developed governance testing frameworks and open-source tools such as AI Verify and Project Moonshot. We will be hearing about those later today, and we hope that you will use them and use the experience that they represent to then take some of your products and ideas forward.
18. And of course, the third important feature Simon spoke about is the open-source technology that Llama is built around. And the key outcome with any open-source approach is that a person who is not a natural user or developer can have immediate access with a very low barrier to the development process. This open-source approach allows new groups of people to experiment with and benefit from your AI platform.
19. And it is that enabling of innovation in Industry, Research, and Government that is an important outcome from the open-source approach developed by sophisticated AI tools. It is the democratisation of Generative AI to a wider group of stakeholders.
20. Llama is one of the models used in our government, through our Government Commercial Cloud platform. Our public sector officers are leveraging the platform not just to experiment and tinker but develop products that then ship out to the rest of public sector and get deployed.
21. A Llama-focused incubator aligns very well with Singapore's support for open-source approaches. So, our agencies have indeed been supporting the development of this incubator. And it's encouraging that Meta has pulled together a broad set of partners from across our ecosystem - AI Singapore, SG Innovate, e27 and Deloitte. This highlights the growing partnership between Singapore and Meta on AI.
22. Ultimately, all of this is anchored on a particular vision that we have in Singapore around what we want AI to do, and that is to drive the public good. We must think through carefully, out of all the many ways in which we can approach this, what is our focus going to be? Of course we look for partnerships, business opportunities, talent development, and use cases, but ultimately, it must focus and come together around a strategic vision. And what we want to demonstrate here in Singapore and to the world is that the use of best-in-class technology can and should be for the public good. And we have that as a vision. All our collaborations and partnerships are to move to that vision, and we are very glad that we have partners and collaborators who have a similar vision of what AI can do.
23. So, we look forward to seeing all your many solutions and celebrating all your successes at Demo Day. And we look forward to a wide proliferation of your AI use cases amongst new audiences, and for your existing audiences to realise there are many things they never thought they could do, but which they now can because you have developed the best in class cutting edge tool.
24. We look forward to your approach, and to inspire many others to experiment with AI, develop tools, and become a catalyst through using AI to transform our society and our world.
25. Thank you all very much.