The foundation that doesn't burn — matrix multiplication has outlasted every tech wave
via A CS Professor on Why Slow Learning Wins in the AI Era | CU Boulder, Tom Yeh
AI-generated writing.
Tom Yeh traces matrix multiplication through every “this changes everything” moment of the last 30 years: CGI after Jurassic Park, big data, machine learning engineering, today’s AI wave, and soon quantum computing. The field on the surface changes every few years; the math underneath doesn’t.
He uses the Gyeongbokgung palace in Seoul as the analogy: burned to the ground in the 1590s, rebuilt in the 1800s — on the exact same stone foundation that survived. If you have that foundation, you can rebuild every time the tools change. If you only know the tools, you have to start from scratch each cycle.
The implication for learning: ignore the surface features long enough to understand what’s load-bearing. DeepSeek was everywhere six months ago; it isn’t now. Transformers have been load-bearing for years and still are. Invest accordingly.