Wonjun Chang

Researcher

I am a researcher at KAIST AI, where I work under the supervision of Seung Wook Kim. My research focuses on building a multi-agent world model as a foundation for physical AI. While my current work focuses on video diffusion models for physical AI, I believe world models have enormous potential across many fields. My ultimate goal is to build a world model that can fully capture and simulate the workings of the human body, particularly the brain.

Dream

Philosophy

  • The Pareto Principle
    • I believe 20% of things matter most, which makes it possible to represent everything in the world through abstraction.
      • e.g. work, code, writing, people
    • We can also use this principle to identify if we correctly understand something.
      • If we cannot explain an idea in a single sentence, we probably do not understand it well enough.
      • If an action requires many different reasons to justify it, we may not have identified the real issue.
  • The Power of Compounding
    • I value what endures, accumulates, and transfers to other fields. This would make the compounding effect.
    • This approach might look slow at the cold start, but it can be extremely powerful in the long term.
    • e.g.
      • Writing, Reading, Speaking
      • Logical thinking, First principle thinking
      • Engineering skills
      • Mathematics, Physics
  • Engineer mindset
    • To me, engineering is essentially optimization, and the engineer mindset refers to the will to consider the possiblity of optimization on anything.
      • One of the most important optimization is about cognitive ability.
      • We can maximize the arithmetic intensity of our brain by good abstraction or automation.
    • This mindset can be applied to any kind of work to accelerate process.

Axioms

Who am I