Opening Address by SMS Tan Kiat How at the National Translation Committee (NTC) Conference 2026: Reimagining Translation In The Age Of AI

1. Distinguished guests, esteemed speakers, and members of our translation community,

2. Good morning, 大家早上好, selamat pagi, vanakkam.

3. Two years ago, when I spoke at this conference, I shared how NTC’s work could be anchored on three pillars: Technology, Talent and Community.

4. In just two years, AI has changed translation dramatically.

5. Today, almost anyone can put a sentence into an AI tool and get a translation in seconds.

6. This opens up tremendous possibilities. But it also raises fundamental questions.

7. What does good translation mean when machines can generate translations almost instantly?

8. What happens to the role of translation professionals?

9. And how do we ensure that technology strengthens, rather than weakens, our multilingual identity?

10. And for Singapore, these answers matter.

11. We are a multilingual society. Our languages connect us to our cultures, our heritage and one another as fellow Singaporeans.

12. So our ambition cannot simply be to translate more words, more quickly.

13. It must be to use technology to help people understand one another better.

14. And that requires us to reimagine our three pillars of Technology, Talent and Community.

Technology – Can AI understand Singapore?

15. Let me start with Technology.

16. Today, the Government manages at least 230,000 public-facing webpages. Yet not all are available in all four official languages.

17. For Singaporeans who are more comfortable reading in Chinese, Malay or Tamil, language should not be a barrier to accessing important Government information. And when that information comes from the Government, it is important that the translation is not only timely, but also accurate, consistent and trustworthy.

18. AI can help us close this gap.

19. Content that was previously too costly or time-consuming to translate can increasingly be made available in multiple languages at scale.

20. But there is a catch.

21. AI can understand language. But can it understand Singapore?

22. Take a simple example.

23. Ask a general-purpose AI model to translate 火锅 into English, and it may give you “hotpot”.

24. It is technically correct.

25. But many Singaporeans would simply say this is “steamboat”.

26. The same applies to familiar places.

27. In Singapore, Singapore Botanic Gardens is known as “Kebun Bunga Singapura” in Malay and “பூமலை (Poo-Ma-Lay)” in Tamil.

28. These may seem like small differences, but they reflect something important: Good translation is not just about getting the words right. It is about getting the meaning — and the context — right.

29. And that is the thinking behind SG Translate 2.0.

30. It combines the language resources and expertise we have built over the years with the capabilities of Large Language Models or LLMs and other AI technologies.

31. Our aim is not simply to build another translation engine.

32. We want to build technology that understands Singapore.

33. Early results are encouraging. SG Translate 2.0 has delivered four times the performance of a general-purpose LLM for selected-localised terms.

34. Let me show you what this means in practice. On the screen is a short paragraph about a Singapore family, filled with terms we encounter in everyday life - ROM, BTO, CPF, HDB, CDA, PSLE, MRT, CDC vouchers, CC, so on and so forth.

35. To many of us in this room, these terms are immediately familiar. But for a general-purpose translation engine, getting them right and understanding what they mean in the Singapore context is not always so straightforward.

36. As you can see on the screen, SG Translate 2.0 is translating the paragraph into Chinese, Malay and Tamil, while recognising these Singapore-specific terms in context.

37. That is what we mean when we say that SG Translate 2.0 is built for Singapore.

38. But don’t just take my word for it. Scan the QR code on the screen and try the engine for yourself. Compare the results with a general-purpose LLM and see how they measure up.

39. If you like what you see, please bookmark it and make it your go-to translation tool. We have also placed QR codes at the foyer outside, so do check them out, during the break.

40. SG Translate 2.0 is also being used to power tools such as Polyglot for Government websites. Polyglot is an AI-powered widget that automatically translates websites into our four official languages. Currently, we are using it on MDDI and Smart Nation webpages and more will come onboard soon.

41. It’s a very useful function, and we want to roll out this functionality to all public-facing websites, so that all Singaporeans can access government information, important information, in the language that you’re most comfortable with.

42. But importantly, besides these functionalities, we are not building it alone.

43. Our public officers and Citizen Translators have also been testing it and helping us improve it.

44. Today, I invite all of you to join us.

45. Use it. Test it. Challenge it.

46. Tell us when it gets things right; and when you think, “That’s not quite how a Singaporean would say it” – let us know too, so that we can keep refining it and improving it.

47. Ultimately, this technology is built with Singaporeans, for Singaporeans.

48. But better technology alone will not give us a better translation.

49. We still need people.

Talent – When AI gets better, judgment matters more

50. I know there is a question on the minds of many practitioners:

51. What does AI mean for me?

52. Some translation tasks will change. Work done manually today will increasingly be assisted or automated by AI.

53. We should be upfront about that.

54. But that does not mean human expertise becomes less important.

55. As AI gets better at translation, human judgement becomes more valuable, not less.

56. AI can generate a translation in seconds.

57. But is it accurate? Does it capture the intended meaning? Is the tone right? Does it understand the cultural context?

58. And in healthcare, law, finance, public communications and other domains of consequence — what happens if it gets something wrong?

59. These are questions of judgement.

60. The translator of the future should therefore not try to compete with AI on speed.

61. The opportunity is to focus on areas where humans do better than AI - understanding context, appreciating nuances, and exercising sound judgement.

62. Translation professionals will increasingly work with AI — using it, evaluating its output, applying specialised expertise, and knowing when human judgement must remain decisive.

63. Our universities are already adapting their courses to prepare practitioners for this changing environment.

64. NTC has also established the Translation Industry Workgroup, or TIWG, to examine how the wider ecosystem should evolve.

65. The Workgroup has studied approaches overseas, from Australia’s interpreter accreditation framework to China’s rapidly developing AI-enabled translation ecosystem.

66. We have also engaged stakeholders, both domestically and from around the world, including the European Union and Malaysia.

67. Some of them are here today, to join the conversation.

68. But we should not simply import someone else’s model.

69. We need an approach that works for Singapore.

70. That is why we are consulting practitioners, language communities, service providers, employers and buyers of translation services.

71. The work has already begun.

72. We will be conducting online surveys and industry consultations with the wider community at large.

73. This includes all of you here. We want to hear from you.

74. If you would like to be part of these conversations, do scan the QR code to join our mailing list and stay updated on upcoming opportunities to contribute. Your voices matter and together we are looking at how we can shape the broader translation ecosystem. Not just for practitioners and professionals but for many who are passionate about this area and want to contribute.

75. And I’m very heartened because, among all of you here today and those who could not join us, there are many Citizen Translators, Members of Public, retirees, students, who feel that this is something they are passionate about.

76. And the message that we are trying to drive is clear: We are not going to be passive bystanders. The future of translation should be shaped by all of us, together.

Community – Keeping our languages alive

77. That brings me to my third pillar, Community.

78. Today, our Citizen Translator community has grown to around 3,000 members.

79. And many of them are in this room today. They have spotted translation errors, supported community translations, and helped seniors record their nation-building stories in the vernacular languages for the upcoming Founder’s Memorial.

80. Did you know that the subtitling and translation for NDP 2025 were vetted by our Citizen Translators, so please give them a round of applause. It is not easy. On their own time, they spent many nights, going through the rehearsals, going through the translation to make sure it is context-specific to Singapore.

81. And why do they do that? Of course, their passion for language and wanting to contribute. They know that National Day Parade is very meaningful and symbolic. It reminds us of the formation of our country, what independence means for us, and what “home” stands for – Not just a physical location but a sense of identity and pride.

82. And when we have songs and performances, we want to make sure that the translated terms in the vernacular languages reflect that value and sense of pride of what it means to be Singaporean and that’s why they spend their own time going through that.

83. And they’ve done many more projects with us. Members of Public taking time to contribute – and that is what Community is about.

84. Technology is also creating entirely new ways for Singaporeans to contribute. Take Hari, a 16-year-old student who is studying Higher Tamil at Umar Pulavar Tamil Language Centre.

85. He was inspired to build his own AI translation bot and used it to add Tamil-language content about Singapore’s national monuments and heritage sites to Wikipedia.

86. Hari is unable to join us today as he is preparing for his O-Level examinations. We wish him all the best for his exams!

87. Think about that.

88. In his own way, a young Singaporean is using a new technology to keep our oldest language relevant for a new generation.

89. His story illustrates how technology can give languages new possibilities rather than replace them.

90. Hari’s story brings our three pillars together.

91. Technology gave him capabilities.

92. His knowledge of language gave the technology meaning and context.

93. And he used both to contribute back to the community.

94. That is the future we should build together.

Conclusion

95. AI will continue to advance.

96. Translation will become faster, cheaper and increasingly seamless.

97. We should welcome this. It can help more people access content in the language they are most comfortable with.

98. But we should also remain clear about what we want technology to achieve.

99. Our goal is not simply more translated words.

100. Our goal is to use the best words to forge better understanding between people.

101. Technology that understands Singapore.

102. Talent whose expertise and judgement are amplified by AI.

103. And a Community that takes ownership of our multilingual society.

104. Because ultimately, our languages are not a problem for technology to solve.

105. They are part of who we are – As an individual, as a community and as a nation.

106. If we get this right, AI will not make our languages less important.

107. It can help more people use them, engage with them and connect through them and appreciate our language, culture and heritage better.

108. So let us embrace what AI can do — without losing sight of what only people can do.

109. Technology can translate words.

110. But it is people who give words their meaning. 

111. In conclusion, 3 pillars – Technology, Talent, community. We are reimagining what language looks like in the age of AI. Let me conclude by saying a few words in Chinese and Malay.

112. Chinese: 人工智能或许能让翻译变得更快、更有效率，但真正赋予文字意义的，始终是人。关键在于我们如何善用AI，并结合自身的专业判断，生成不仅准确，而且说得对、听得懂、听得顺的译文。

113. Malay: AI boleh menterjemah dengan lebih pantas. Tetapi, penterjemah memberikan konteks, nuansa dan makna. Gabungan kedua-duanya dapat menghasilkan terjemahan yang tepat dan sesuai.

114. Thank you, 谢谢大家, terima kasih, நன்றி (Nan-dri).