MDDI 演講稿 · 2025-05-29
高階政務部長陳杰豪在ATxSummit第二天的開幕主旨演講
要點
- • 由 AI Verify 基金會和 IMDA 發起的全球 AI 保證試點已吸引來自 8 個司法管轄區、10 個行業的逾 30 家企業，在實際條件下測試真實 GenAI 應用場景。
- • 試點結束後釋出的公開報告總結了四條實踐經驗：風險因應用場景而異、測試資料須專項構建、流水線層面測試洞察更深，以及基於大語言模型的評估需審慎設計並保留人工監督。
- • 新加坡今日釋出《GenAI 應用測試入門工具包》並公開徵詢意見，以自願性指引形式就何時測試、測試什麼和如何測試提供指導，旨在降低負責任採用 GenAI 的門檻。
- • 入門工具包的七項基線測試已在 Project Moonshot 平臺上線，企業可自主執行負責任 AI 評估，後續將持續新增測試專案。
- • 新加坡 AI Singapore 與聯合國開發計劃署將簽署諒解備忘錄，將「AI for Good」計劃推向國際，首批試點覆蓋東南亞、加勒比地區和太平洋島國，致力彌合發展中國家的 AI 素養鴻溝。
完整譯文（繁體中文）
MDDI 英文原文譯文 · 翻譯日期: 2026-06-21
陳杰豪高階政務部長（數字發展與信息部）在ATxSummit第二天（2025年5月29日）的開幕主旨演講
各位閣下、各位嘉賓、女士們、先生們
我很高興再次回到新加坡亞洲科技大會（ATxSG）——這裡匯聚了科技、商業和政策領域的精英，共同塑造我們數字經濟的未來。
人工智慧的快速發展——尤其是生成式AI（GenAI）——開啟了巨大的潛力，但也引發了迫切而重要的問題：它將如何影響就業、重塑經濟，以及對我們的社會產生怎樣的影響。
世界各地的人們都在問：AI對他們的生計和未來意味著什麼。
各國政府和企業正爭相釋放AI的潛力，同時也在管控風險，引導其朝著惠及所有人的方向發展。
新加坡正在採取深思熟慮的舉措——不僅僅是為了跟上步伐，更是為了主動塑造AI的應用方式，以強化經濟、增強韌性、提升人民福祉。
我們的選擇至關重要，因為它將決定我們的長期競爭力，以及我們傳遞給下一代的數字未來。
作為一個互聯互通的小國，我們也希望為全球AI生態系統作出貢獻，以支援開放數字經濟——一個值得信賴、具有包容性且可互操作的生態系統，建立在我們堅實的數字基礎和信任文化之上。
因此，我們正在打造一個讓AI蓬勃發展的家園，不僅是為了我們自身，更是作為更廣泛國際社會中一個值得信賴的節點。
為實現這一目標，我們需要在實踐中檢驗信任的具體面貌，以便能夠更有把握地加以推廣。
正因如此，AI Verify基金會與IMDA於今年早些時候聯合推出了全球AI保障試點專案。
該舉措將擁有真實GenAI應用場景的企業與專業AI測試人員配對，使其能夠超越理論框架，親眼看到GenAI應用在實際條件下的表現。
短短數月內，來自8個司法管轄區、涵蓋人力資源、醫療保健和金融等10個行業的30餘家企業參與了這一舉措。
這表明企業希望正確推進GenAI的採用，因為信任是一項商業風險——當客戶等利益相關方產生顧慮時，競爭力便會受損。
隨著全球AI保障試點專案的圓滿結束，我們整理了一份報告，從這些合作中提煉出切實可行的經驗，每一條都有助於我們瞭解企業真正需要什麼，才能充滿信心地構建和使用GenAI。
第一，風險具有情境特異性。
你所面臨的風險取決於你的應用場景。
最有效的測試始於清晰瞭解哪些內容與你的應用場景相關，哪些不相關。
第二，有用的測試資料很少是現成的。
大多數企業並沒有現成完備的資料集。
生成真實且涵蓋邊界情況的測試場景，需要我們在機器輔助下付出周全的努力。
第三，超越輸出結果本身。
有時，問題隱藏在流程的更深處。
測試系統內部發生的情況，能夠提供更有價值的洞察——以及更高的保障。
第四，大型語言模型（LLM）可以協助評估，但需審慎使用。
它們速度快、可擴充套件，但仍需經過周全的設計、校準和人工監督。
在某些情況下，更簡單的方法同樣有效。
這些不僅僅是技術層面的觀察。
它們反映了協作與實踐的交匯之處，以及洞察轉化為行動的關鍵所在。
對於希望借鑑這些經驗的人——無論您是在商業場景中應用GenAI、構建測試工具，還是參與政策制定——該報告今日起作為資源向公眾開放。
報告以更詳盡的方式記錄了相關洞見，併為那些希望從原則走向落地實施的人提供了實用要點。
在全球AI保障試點（Global AI Assurance Pilot）的基礎上，我們正在降低企業採取行動的門檻——讓企業能夠按照自身條件和節奏推進，無需從零開始。
這將有助於我們培育一個AI生態系統，讓企業在其中得到支援與賦能，併成為更廣泛信任文化的一部分。
作為下一步舉措，我們開發了《GenAI應用測試入門套件》（Testing Starter Kit for GenAI Applications），該套件將於今日向公眾開放徵求意見。
入門套件是一套自願性指南，其本質是降低那些希望負責任地採用GenAI但可能不知從何著手的企業的門檻。
該套件汲取了全球AI保障試點的洞見，藉助從業者的實踐經驗，確保指導內容切實可用。
它旨在實現兩個目標：
第一，彙集GenAI應用測試的新興最佳實踐與方法論，幫助企業瞭解優質測試的標準。
第二，就如何開展測試提供實操指引——何時測試、測試什麼以及如何測試。
為確保指引具有可操作性，入門套件將輔以測試工具，供企業和開發者自行使用並開展評估。
作為起步，入門套件中的七項基準測試已在Project Moonshot上開放，使企業能夠輕鬆將負責任的AI實踐融入其運營。
我們將逐步通過Project Moonshot開放更多測試專案。
入門套件還被設計為可隨技術演進、新風險出現及應用場景日趨複雜而持續更新迭代。
無論您是試點聊天機器人的初創企業，還是大規模部署AI應用的大型企業，目標都是讓負責任的創新更加觸手可及、切實可行。
這在實踐、工具與政策之間形成反饋閉環，使治理保持靈活、務實且有利於創新。
更重要的是，這使企業能夠在構建可信GenAI方面發揮主導作用——以共同標準、開放框架以及致力於安全負責任創新的社群為後盾。
綜合來看，全球AI保障試點助我們積累經驗，入門套件讓更多人得以應用，而Project Moonshot則提供了規模化的手段。
但僅靠賦能工具並不足以構建一個蓬勃發展的AI生態系統。
要充分釋放AI的潛力，我們還需要賦能每一個人——從日常與AI互動的普通使用者，到與AI並肩工作的員工，再到能夠決定AI使用方式的領導者。
正如我們在技術層面持續投入，我們也必須在人力能力建設上加大投入。
這意味著為人們創造學習、適應和參與的空間——使AI不再是強加於他們的事物，而是他們能夠塑造、理解並從中受益的力量。
我很高興地宣佈，新加坡人工智慧（AI Singapore，簡稱AISG）與聯合國開發計劃署（United Nations Development Programme，簡稱UNDP）將簽署一份諒解備忘錄，旨在彌合AI素養鴻溝，推動發展中國家社群的轉型。
這將把AISG於2024年啟動的、旨在提升國家AI能力的成功專案"AI for Good"（AI4Good）從亞洲推向國際舞臺，以支援聯合國可持續發展目標4。
目標是幫助個人、組織和企業更好地理解AI的機遇、風險與倫理層面。
我們將與教育工作者共同開發AI教學資源，並通過有針對性的外展活動觸達弱勢群體，確保在我們作為全球社群共同前進的過程中不讓任何人掉隊。
AISG與UNDP將在東南亞、加勒比海地區和太平洋島國探索AI4Good試點專案，以便共同支援更具包容性的AI驅動增長參與。
這份諒解備忘錄代表著共同的承諾——讓AI惠及所有人，不僅限於AI得以發展和推進的地方，更要延伸至最需要它的地方。
在許多情境下，AI的潛力更難以兌現——受制於在獲取渠道、基礎設施或準備程度上的差異。
通過這一夥伴關係，我們致力於彌合這些差距，為更多人自信參與數字經濟創造機會，從新加坡出發，延伸至我們所在的區域乃至更廣闊的世界。
這一努力體現了同樣的精神，正是這種精神指引著我們更廣泛的工作——培育一個能在最關鍵之處產生切實影響的AI生態系統。
展望未來，我們必須牢記：AI領域的實質性進步，不僅在於規模或速度，更在於我們能否將其與民眾、企業和社群的需求緊密契合。
新加坡將繼續跨行業、跨國界開展合作，因為我們相信，一個值得信賴、具有包容性且切實有用的AI未來，必須攜手共建。
我們誠邀懷有共同願景的合作伙伴與我們攜手，共同打造這一AI家園，使其成為向善的力量。
非常感謝。
英文原文
MDDI 官網原始記錄 · 抓取日期: 2026-06-21
Opening Keynote by Mr Tan Kiat How, Senior Minister of State for Digital Development and Information, at ATxSummit Day 2 (29 May 2025)
Your excellencies, distinguished guests, ladies and gentlemen
I am happy to be back at AsiaTech Singapore, or ATxSG – where some of the best from technology, business, and policy come together to shape the future of our digital economies.
The rapid developments in AI – especially in generative AI, or GenAI – have opened up immense potential, but also surfaced urgent, important questions about how it will impact jobs, reshape economies, and influence our societies.
Around the world, people are asking what AI means for their livelihoods and their future.
Governments and businesses are racing to unlock AI’s potential, while also managing risks and guiding its impact in a direction that benefits all.
Singapore is taking deliberate steps – not just to keep pace, but to shape how AI is used to strengthen our economy, build resilience, and uplift our people.
What we choose to do matters, because it will shape our long-term competitiveness and the kind of digital future we pass on to our next generation.
As a small, connected nation, we also want to contribute to a global AI ecosystem in support of open digital economies – one that is trusted, inclusive, and interoperable, building on our strong digital foundation and culture of trust.
We are therefore building a home where AI can thrive, not just for ourselves but as a trusted node in a wider international community.
To realise this, we need to test what trust looks like in practice, so that it can be scaled with greater confidence.
That is why the AI Verify Foundation and IMDA launched the Global AI Assurance Pilot earlier this year.
The initiative pairs businesses bringing real-world GenAI use cases with specialist AI testers, so that they can move beyond theoretical frameworks and see how their GenAI applications perform under practical conditions.
In just a few months, more than 30 companies across 8 jurisdictions and 10 sectors, including HR, healthcare and finance, have participated in this initiative.
This shows that businesses want to get GenAI adoption right, because trust is a business risk – when stakeholders like customers hesitate, competitiveness suffers.
With the Global AI Assurance Pilot now concluded, we have put together a report that distils practical lessons from these collaborations , each one helping us understand what businesses really need to build and use GenAI with confidence.
One, risks are context specific.
Your risks depend on your use case.
The most effective testing starts with a clear understanding of what is relevant to your use case, as well as what is not.
Two, useful test data rarely comes ready-made.
Most businesses do not have the prefect dataset sitting on the shelf.
Generating realistic and edge case test scenarios requires thoughtful effort from us, with support from machines.
Three, go beyond outputs.
Sometimes, the issue lies deeper in the pipeline.
Testing what happens inside the system can offer more useful insights – and greater assurance.
And four, Large Language Models, or LLMs, can help with evaluation, but only with care.
They can be fast and scalable, but still require thoughtful design, calibration, and human oversight.
In some cases, simpler methods work just as well.
These are not just technical observations.
They reflect where collaboration meets practice, and where insights translate into action.
For those keen to draw on these lessons for your own interests – whether you are applying GenAI in a business setting, building testing tools, or shaping policies – the report is available today from today as a resource.
It captures the insights in greater detail and offers practical takeaways for those looking to move from principles to implementation.
Building on the Global AI Assurance Pilot, we are making it easier for businesses to take action – on their own terms and at their own pace – without having to start from scratch.
This will help us nurture an AI ecosystem where businesses are supported, empowered, and become part of a broader culture of trust.
As a next step, we have developed the Testing Starter Kit for GenAI Applications , which will be released to the public for comments today.
The Starter Kit is a set of voluntary guidelines that, in essence, lowers the barriers for businesses that want to adopt GenAI responsibly but may not know where to start.
It draws on insights from the Global AI Assurance Pilot, tapping on the experience of practitioners to ensure the guidance is practical and useful.
It aims to do two things:
First, it pulls together emerging best practices and methodologies for testing GenAI applications, so that businesses can understand what good testing looks like.
Second, it offers practical guidance on how to go about it – when to test, what to test, and how to test.
To ensure that the guidance is actionable, the Starter Kit will be complemented by testing tools that businesses and developers can use and run assessments on their own.
As a start, seven baseline tests from the Starter Kit are made available on Project Moonshot , enabling businesses to easily integrate responsible AI practices into their operations.
We will progressively make more tests available through Project Moonshot.
The Starter Kit is also designed to evolve as technologies shift, new risks might emerge, and use cases will grow more complex.
Whether you are a startup piloting a chatbot or a large enterprise deploying AI applications at scale, the aim is to make responsible innovation more accessible and achievable.
This creates a feedback loop between practice, tools, and policy that keeps governance agile, grounded, and innovation friendly.
More importantly, it allows businesses to take the lead in building trusted GenAI – backed by shared standards, open frameworks, and a community committed to safe and responsible innovation.
Together, the Global AI Assurance Pilot helped us learn, the Starter Kit enables more to apply, and Project Moonshot provides the means to scale.
But enablers alone do not make a thriving AI ecosystem.
To unlock the benefits of AI in full, we need to empower people too – from everyday users who interact with AI, to employees who work alongside it, and leaders who can make decisions about how it is used.
Just as we invest in the technical side, we must also invest in human capacity.
This means creating space for people to learn, adapt, and engage – so that AI is not something done to them, but something they can shape, understand and benefit from.
I am pleased to share that AI Singapore, or AISG, and the United Nations Development Programme, or UNDP, will be signing a Memorandum of Understanding to close the AI literacy divide and transform communities in developing countries .
This will extend AISG’s successful AI for Good (AI4Good) programme – launched in 2024 to bolster national AI capabilities – from Asia to an international scale, in support of the United Nations Sustainability Development Goal 4.
The aim is to enable individuals, organisations, and businesses to better understand the opportunities, risks, and ethical dimensions of AI.
We will co-develop resources to teach AI with educators, and reach underrepresented groups through targeted outreach, so that no one is left behind as we advance together as a global community.
AISG and UNDP will explore initial AI4Good pilots in Southeast Asia, the Caribbean, and the Pacific Islands, so that we can support more inclusive participation in AI-driven growth together.
This MOU represents a shared commitment to make AI work for everyone – not just where it can be developed and advanced but where it could be needed most.
In many contexts, the promise of AI is harder to realise – shaped by differences in access, infrastructure, or readiness.
Through this partnership, we seek to close these gaps and open up opportunities for more to participate confidently in digital economies, starting here in Singapore, and extending to our region and beyond.
This effort reflects the same ethos that has guided our broader efforts to nurture an AI ecosystem that delivers impact where it matters most.
As we look ahead, we must remember that meaningful progress in AI is not just about scale or speed, but how well we align it with the needs of our people, our businesses, and our communities.
Singapore will continue to partner across sectors and borders, because we believe that a trusted, inclusive, and useful future for AI must be built together.
We invite partners who share this vision to work with us so that, together, we can build this home for AI to become a force for good.
Thank you very much.