The Outsider
I dug out an old deck this week.
Learning to Learn in this New Age
A talk I gave years ago at NUS. A diverse group of community leaders. I didn’t really have anything specific to share. So I explained my learning journey. (Link to old post with deck)
I’m preparing a new version. For traders this time. “Learning to Learn Risk Management in the Age of AI.” Same obsession, different audience. Not sure how it’s going to go down.
The original deck ended on a quiet slide. Just five words on a watercolour blob.
Nothing we learn is wasted
I believe that more now than when I wrote it. And this week reminded me why. Interestingly, it also relates to an AI concept.
Read on to find out why.
Nine conversations
Chinese New Year. Half my usual. But still interesting.
A consulting partner explaining why data engineering is still going strong. A diplomat on sabbatical obsessing about AI and building AI communities. Someone heading an AI safety and testing outfit. Another one in the same space who is between jobs. Someone I met a while back, who is trying to build something that helps law firms assess AI vendors. A blockchain association head. Two ex-MAS colleagues. One in manpower development. Another one steeped in AML. A friend of 2 decades who went from the public service to a governance institute.
Wildly different areas. Each deeply fluent in their own domain. Some commonalities.
Communities. The person building AI communities. The testing and assurance folks knew each other from an AI testing community. The person trying to build something for the legal community. And she also built a community to support her efforts. They were all part of communities.
Pilates and AI safety. One of the folks in the AI safety space I met is also a pilates instructor. I mentioned another student of my PhD supervisor - a professor - who is also active in the space. Then I recalled that there was once when the professor and I had coffee and wanted to ask our PhD supervisor along, but he could not join as he was taking a pilates class. Intersecting bubbles, but no one knows each other.
Parallel worlds, close enough to touch, but not quite connecting. I already saw this in previous weeks but did not recognise it then.
The Perpetual Outsider
And I was the outsider to all of these communities. Some have invited me in.
But being an outsider is an extremely familiar feeling.
Engineering into arts policy without an arts background. Arts into banking regulation. Regulation into quant finance. Quant finance into investment risk. Investment risk into a PhD in Computer Science at 42. Then AI risk management. And now, an outsider in this whole new world outside. No constraints. But which party to gatecrash?
The old deck on “Learning to Learn in this New Age” reminded me how I learned to be comfortable as an outsider.
Malcolm Gladwell’s 10,000 hours. David Epstein’s case for breadth over early specialisation. Josh Waitzkin’s neural pathways, carved slowly through deliberate practice. Peter Brown’s retrieval and interleaving. Different authors, different books, but the core idea is quite similar if you squint.
What you learn somewhere doesn’t disappear when you cross into a new domain. Some of it passes over the threshold. Some of it builds the foundation for the next.
Transfer learning
Same for AI. We call it transfer learning. Pre-train on one domain, adapt faster to another. Don’t start from zero if you don’t have to. That was the original insight.
But the idea kept evolving. Fine-tuning. Take what’s pre-trained and adapt it precisely to a new domain. Few-shot learning. Generalise from just a handful of new examples. Meta-learning. Learn how to learn itself, so that each new domain requires less effort than the last. Then in-context learning, adapt mid-conversation, no retraining required at all.
You don’t need to start from zero if you already know how to learn. The model gets more efficient at transfer with every generation. So do people, if they let themselves get comfortable with this feeling of being an outsider.
So my 5 words made some sense. I think.
Nothing we learn is wasted.
Being an outsider over and over again can sometimes help. An outsider generalises.
Strangely, this - being an outsider - for the past two decades has helped me to make friends that are as diverse as can be.
For that, I’m grateful.
Not a bad week.
#LearningToLearn #AI #AIRiskManagement #Transitions