AISG Train-the-Trainer and teaching AI to the community

On 31 March 2026, I attended the AI Singapore Train-the-Trainer session at the NUS i4.0 building. It was a good change of pace from building AI systems for technical users, because teaching AI to the community requires a different kind of clarity.

One part that stayed with me was not technical at all. It was about managing different student behaviours in a classroom. There is the quiet learner who may understand but does not want to speak. The outspoken learner who can help the room but may also dominate it. The over-achiever who wants to go far beyond the lesson. And the person who is not interested, but still deserves a chance to find one useful thing from the session.

I learnt that teaching well means adjusting to these groups without making the topic too shallow or too intimidating. Sometimes the best thing to do is to invite a quiet person in gently. Sometimes it is to give the over-achiever a harder extension task so they stay engaged without pulling the whole class away.

The session also covered how to teach prompting techniques. Modern LLMs are much better at guessing what we want, so prompting can look less important than before. But I still think it matters. If we do not provide the full picture, the model can still go in the wrong direction and return something polished but not actually useful.

The parts I want learners to remember are simple: give context, explain the intent, set the tone, and say what constraints matter. A model can infer a lot, but it should not have to guess the most important parts of the task.

Footnote: Ported over from my personal blog. Initially posted on 31 March 2026.