Reading the game
I have been a games person for as long as I can remember. It started young and eventually turned serious: I played League of Legends at a professional level, and Blizzard's StarCraft II more casually on the side. Both are real-time strategy games: you make decisions in real time under fog of war, reading an opponent you can barely see and choosing between thousands of possible moves every second. What always pulled me in was the AI inside these games. Not the scripted enemies that follow fixed rules, but the harder question underneath them: could a machine ever play the way a great human plays, reading a situation instead of following a script?
The moment it clicked
In 2019, Google DeepMind, one of the world's leading AI labs, built AlphaStar, an AI that played StarCraft II. This was a genuinely hard problem. Chess and Go are complex, but you can see the whole board. StarCraft hides half the game from you and asks you to act in real time across a near-infinite space of moves. Most people assumed it was years away.
Then DeepMind put AlphaStar on stage against TLO, one of the best players in the world, in a live showmatch, and it won. Not a lucky game, a clean series. You can still read DeepMind's write-up of how it worked.
I remember exactly how that felt. A game I had spent years being amazed by, mastered by something that had learned it from nothing. I did not want to build StarCraft players. I wanted to build the thing that could learn like that. That was the moment AI stopped being a background interest and became what I wanted to work on.
Down the rabbit hole
So I went looking for how we got here. That search pulled me back through the whole story of AI in games. In 1997, Deep Blue beat Garry Kasparov, the reigning world chess champion, but it did so mostly by brute force, searching millions of positions a second using rules that human experts had written by hand. Then, decades later, AlphaGo beat Lee Sedol at Go, a game with more possible positions than there are atoms in the universe, far too many to brute force. And AlphaZero went further still: handed nothing but the rules, it taught itself chess, Go, and shogi from scratch by playing against itself, until it was superhuman. The machine was no longer executing our strategies. It was discovering its own.
The more I read, the more I wanted to build these things, not just admire them. That led me to DigiPen Institute of Technology, where I stopped playing systems and started building them from the engine up, the layer of software that makes a game actually run. Games were the way in, but computer science was the real subject, and I kept aiming it at AI.
Where I am today
These days games are just a hobby, something I play with friends when I have a rare free evening. My real pursuit is to be as good as I can be in my field. I work as an AI engineer in Singapore, and I get to point it at technology for public good, the kind of work I care about most.
AI right now feels like the frontier, the way the early web did for the people building Amazon and Google in the 1990s: an era-changing technology, still being figured out, where the ground shifts every few months. I am genuinely happy to be part of it, and to be building it for the years ahead rather than watching from the sidelines. It is the same pull I felt watching AlphaStar, only now I am the one at the keyboard.